Treatise on water science, 4-Vol. set

EDITOR-IN-CHIEF Peter Wilderer Technische Universitaet Muenchen, Institute for Advanced Study, Munich, Germany

vii

EDITORS Peter Rogers Harvard School for Engineering and Applied Sciences, Cambridge, MA, USA Stefan Uhlenbrook Department of Water Engineering, UNESCO-IHE, Delft, The Netherlands

Keisuke Hanaki The University of Tokyo, Tokyo, Japan Tom Vereijken European Water Partnership, Grontmij, The Netherlands

Fritz Frimmel Karlsruhe Institute of Technology, Karlsruhe, Germany

vii

THE IMPORTANCE OF WATER SCIENCE IN A WORLD OF RAPID CHANGE: A PREFACE TO THE TREATISE ON WATER SCIENCE The world in which we live is currently undergoing rapid changes, triggered by outstanding advances in natural sciences, medicine, and technology. As a result, the human population grows to levels never known before. Innovative communication and transportation means permit globalization of economy and urban lifestyle. Cities and city life exert an unprecedented pull. More than half of the world’s population already live in urban settings – the tendency is rising. Cities meet the expectations of immigrants, citizens, and businesses only when served by an appropriate infrastructure. Unfortunately, in many parts of the world cities grow faster than the required infrastructure can be planned, financed, and installed. In many cases, installation of water distribution networks and sewer systems, waterworks, and wastewater treatment plants is often lagging far behind schedule – be it because of the lack of financial resources or because higher priority is given to other infrastructural projects, roads, and highways, for instance. At a larger scale, the water demand of agriculture and industry is growing overproportionally with respect to population size as people shift preference to products requiring particularly high volumes of water during the growth season or during the fabrication process, respectively. Two examples underline this statement – the shift toward meat consumption and the preference of clothing made of cotton fibers. The consumers are often unaware of the water required to raise cattle, swine, and poultry, and to keep cotton fields productive particularly when such fields are located in arid regions as is the case in Uzbekistan, for instance. Although the water demand is increasing, worldwide, the capacity of local water resources is not. It is even decreasing in very many areas of the world, resulting from pollution of water bodies and soil, from over-abstraction of water, and from effects caused by climate change. Water deficits in municipal, industrial, and agricultural settings are the result. In many cases, urban and agricultural areas developed in regions where ab initio freshwater is scarce. Drought situations caused by global warming and climate change amplify the deficit between water demand and water availability. Overabstraction of groundwater to meet the local water demand is a common but unsustainable solution to the problem of water shortage. In areas close to the ocean, over-abstraction causes seawater intrusion and subsequent increase of the salinity of groundwater. Rising sea level caused by melting of shelf ice intensifies the intrusion of seawater not only in aquifers but in estuaries as well. In addition, deterioration of ground- and surface water is caused by excess usage of fertilizers and pesticides, and by uncontrolled dumping of solid and liquid wastes onto land. Aggravation of water deficits in municipal, industrial, and agricultural environments is the result. In the nineteenth and twentieth centuries, health problems and eutrophication caused by pollution of surface- and

groundwater were recognized and solved by legal frameworks and enforcement of regulations, and by investing large amounts of money in the development and implementation of infrastructural concepts and technologies. In high-income countries, design engineers and operators of water distribution and sewer systems, water works, and wastewater treatment plants are well trained, nowadays – a major prerequisite of proper functioning of technical installations. In the mediumand low-income countries, however, responsible management of water resources and effective operation and maintenance of water technology are often foreign words. In the twenty-first century, we are confronted with a comparably much larger and much more complex problem of water management compared to the years past. A new approach to water management and water technology is required in response to the rapid increase at the demand side, and rapid loss of capacity and quality at the supply side. A paradigm shift appears to be urgently necessary. The old paradigm was the answer to the conditions prevailing in the highly industrialized and water-rich regions of the world. Over the past decades, considerable time was available to develop, implement, and upgrade measures capable of solving the specific local and regional problems. This, however, is not the situation we have to deal with today and in the years to come. In future, we have to support people with effective and robust water and wastewater services even if the capacity of the local water resources is critically short. To avoid evolvement of economic and societal instabilities, we are obliged to develop techniques and management concepts which can be implemented in virtually no time. We have to serve people, industry, and agriculture alike while keeping the function of aquatic and terrestrial ecosystems preserved. We need methods which are adjustable to the changing climatic boundary conditions. We need well-educated water professionals in academia, water services, and water authorities who understand the local environmental, economic, and societal framework conditions, to draw appropriate decisions and take responsible action. We need methods which are financially affordable. These methods are to be safe with respect to public health. Moreover, they must guarantee ecosystems to exert their generic life-supporting function. The task to solve the complex issue of water-related problems caused by urbanization and lifestyle changes is challenging because of the speed of change at both the demand and the supply side, and also because of the limitations at the financial side. Business as usual is not a tolerable approach. In the course of a shifting paradigm, we should realize that sectoral approaches (as they were usually taken in the past) are to be overcome. We need to understand that the water quantity and quality issue are inextricably linked to the issue of energy and food supply, and with the issue of land management as well. What we need is a holistic approach. Measures

xv

xvi

The Importance of Water Science in a World of Rapid Change: A Preface to the Treatise on Water Science

are to be taken which permit solution of the energy, water, and food crisis in conjunction with measures which enable restoration of the self-regulating capacity of terrestrial and aquatic ecosystems in harmony with the human demand for land. Scientists and engineers are called to take up the task of problem solving as a challenge and as a chance. Solutions have to be found on the basis of the existing portfolio of knowledge and experience, but open minded with respect to the very local conditions in rapid transition. The Treatise of

Water Science is to be considered as a platform on which innovative research and development may proceed. It summarizes the contemporary state of knowledge in the field of water science and technology and paves the way toward a new horizon. Serving humanity with safe water while keeping the self-regulating capacity of the aquatic ecosystems intact – this has to be our common goal. Peter Wilderer

Preface – Management of Water Resources PP Rogers, Harvard University, Cambridge, MA, USA & 2011 Elsevier B.V. All rights reserved.

1 The Water Crisis 2 Why Studying Water Is So Important 3 Current Global Water Balance 4 Establishing Water Policy 5 Predicting Future Demands for Water 6 Drivers of Socioeconomic Growth 7 Transboundary Conflicts 8 River Basin Politics 9 The Contents of Volume I Acknowledgments References

The Greek philosophers gave us a physical world composed of four elements: land, water, fire, and air. Two thousand five hundred years later we are still focused upon these elements, now conventionally referred to as ecosystem (land), water (water), energy (fire), and atmosphere (air), more aware than ever that these elements are essential for all life on Earth. As human populations have multiplied 700-fold since the ancient Greeks, we are facing major crises with each of these elements. At different stages of human development, each has risen to prominence; control of fire was one of humankind’s earliest and fundamental scientific discoveries. With fire under control, land took on increased salience, humankind’s numbers soared, and we managed to inhabit the entire planet. The other two elements, air and water, were always essential; however, until recently, they were considered so abundant that we would never have to worry about depleting them. However, by the end of the nineteenth century when the globe’s land frontiers were closing and filling in, it was then that we as a species began to notice problems with having contaminated the air and water and that it was becoming difficult to find clean air to breathe safely and unpolluted water to drink. In addition, by the end of the twentieth century we discovered that our profligate use of fossil-fuel energy was in danger of changing the atmosphere in ways that were threatening the survival of our species by causing global warming. We find ourselves now on the threshold of the twenty-first century struggling to survive as a species. As a result, the two most salient global issues now facing humankind are energy and water. How do we manage our survival transition into the twenty-first century and beyond? The intertwined developing resources crises lead us to three critical questions pertaining to the global water situation: 1. Will we have enough water to grow food to feed ourselves in the twenty-first century? By far the largest quantities of freshwater are, and will be, those used in agriculture. Currently, agriculture uses about 4000 km3 of freshwater each year to feed approximately 7 billion people. Even though population growth has slowed down globally, we will still face a population of 9 billion by 2050. The demand for agricultural water is complicated by the fact that as people

1 2 2 2 2 3 4 5 5 6 6

become wealthier their dietary tastes change, moving away from grains toward animal products. The same amount of water that provided food for 10 people subsisting on grains previously now only satisfies the agricultural needs of one person who has moved up the food chain toward animal products. 2. How will we provide water and sanitation for an additional 3 billion urban dwellers? Since 2007 the urban population has exceeded the rural population. This has major implications for sustaining the water and sanitation for cities. For example, China’s urban population is expected to reach 1 billion by 2030. Urbanites typically are wealthier than their rural compatriots, and have radically different water demands, more appliances, washing machines, bathtubs, showers, and flush toilets. Even though the absolute magnitude of their demands is much smaller than the demand of agriculture, water plays an important role in urban public health which cannot be ignored. This is particularly the case in the large cities of Asia and Africa where already there are huge unserved populations demanding water and sanitation services. One study estimates that as much as $22 trillion is needed by 2030 just to meet the demands for water and sanitation services. 3. How should we address the future climate uncertainties? One issue that water engineers always prided themselves on was that they could make robust forecasts of the future, at least good enough to be able to build reservoirs, dams, and embankments that would function well enough under a wide range of actual future outcomes. The very existence of the possibility of climate change seriously challenges our ability to rely upon our forecasts. The shift from using stationary time series as the basis for future forecasting is seriously undermined when faced with the possibility of nonstationarity in the time series. There is a need for creative adaptation strategies that would help avoid rapid collapse of engineered and social systems. This first volume of the Treatise on Water Science has 11 chapters dealing with how to address these questions. It is about managing our water resources and will, hence, focus on water. Without water, life, as we know it, would disappear from the

1

2

Preface – Management of Water Resources

planet. Water plays an extremely important role in maintaining a sustainable life on this planet for all species including Homo sapiens. Only those few exotic species that have managed to survive in environments which do not require potable water would survive. We fool ourselves, however, if we focus only on water and ignore the connections of water to the overall use of global resources because each of them has a critical role to play in supporting human life on planet Earth. This is the major concern of modern studies of water resources.

1 The Water Crisis Since 1900 the world’s population has tripled, but its water demand has risen sixfold (FAO, 2009). These two facts have forced the global community to focus on the management of global water resources. The major emphasis has been on making integrated water resources management (IWRM) a reality (Global Water Partnership, 2000). This is the global water crisis whose dimensions are daily beginning to manifest themselves to international agencies, national and local governments, and particularly to individual citizens. Unlike the fear of using up energy, which has occurred very rapidly, sometimes seemingly almost overnight as with the 2007 and early 2008 petroleum price rises, the water crisis is a slower crisis – but a crisis nevertheless. The situation changes imperceptibly from day to day – we do not see doubling of prices over periods of months, but like melting glaciers, it is an inexorable slow burn toward scarcity. The time frame is years rather than months, but every day there are more of us, each making demands on this global resource. Although we can find replacements for fossil fuels to power our cars and heat our homes, there is no alternative to replace water. For most important water uses, such as irrigation and drinking water, there is no substitute. Water, however, is influenced by geophysical and geochemical processes which are highly influenced by climatic change on both the supply and demand sides. On the supply side, drying up lakes and melting glaciers can reduce water availability locally, and on the demand side increased temperatures will increase demands for irrigation of food crops, air conditioning, etc. All of these changes will have to be dealt with under a fairly constant global supply of water. The great irony here is that fossil fuels are usually described as nonrenewable resources – they have a fixed amount and could be exhausted – whereas water is a renewable resource of an essentially fixed amount and is used by everybody on the globe, and cannot be used up in the sense that petroleum can be because it is a renewable resource, but access to it by growing populations overtaxes its availability.

2 Why Studying Water Is So Important Water resources have been studied for millennia. Starting even before the ancient Greek philosophers, Plato (428–348 BC), Aristotle (384–322 BC), and Archimedes (287–212 BC), the Egyptians and the Assyrians had planned, designed, and built major water resource infrastructures throughout the Middle East. The Romans took the Greek concerns about water and

public health and expanded them up to a global scale throughout the Roman Empire. The city of Rome with its 16 major aqueducts was a marvel of both engineering and water management, with a per capita water availability equivalent to current European standards. Over the succeeding centuries, we have theorized, analyzed, and prioritized water in myriad ways. The great scientists and engineers from Renaissance Europe through the end of the nineteenth century, Galileo (AD 1564–1642), da Vinci (AD 1452–1519), Torricelli (AD 1608–47), Pascal (AD 1623–62), Daniel Bernoulli (AD 1700–82), and Darcy (AD 1803–58), just to mention a few of the major contributors, made major breakthroughs which still govern management of water in all its forms today.

3 Current Global Water Balance The International Water Management Institute (IWMI), located in Sri Lanka, has adopted the blue–green water paradigm suggested by Falkenmark and Rockstro¨m (2004), in which the water accounting is done according to whether it is due to evaporation (coded green) or due to the residual surface and groundwater runoff (coded blue). Of the total annual terrestrial rainfall, called the renewable freshwater resources, of 110 000 km3, 56% evaporates by biological processes, forest products, grazing land, and biodiversity; 4.5% is evaporated from rainfed agriculture (crops and livestock); a further 0.6% of the green water is evaporated from irrigated agriculture along with 1.4% from runoff sources (these are called blue water); and an additional 1.3% is evaporated from open water storages from man-made reservoirs and lakes. Cities and industry demand only 0.1% of the total and 36% returns to the ocean. The 110 000 km3 of precipitation on the terrestrial landscape is extremely small in comparison with the total resource base and the amounts evaporated in producing food and fiber. Of course, it should be recalled that the actual withdrawal of water from the ecosystem for cities and industries could be several times larger, but that about 85% of these uses (albeit contaminated) return to the runoff account of the terrestrial system. These issues are explored in greater detail in Volume II of this treatise.

4 Establishing Water Policy It is a commonplace fact that if a resource has little or no value then it will be overused. Therefore, one of the major issues in water resources planning and management is to identify the value of water. The value of water has been pondered by scholars for millennia. Plato observed that ‘‘only what is rare is valuable, and water which is the best of all things y is also the cheapest, as quoted by Hanemann (2006) based on Bowley (1973) from Plato’s Euthydemus. Two thousand years later, in considering the difference between the market price of commodities and their economic value, the eighteenth-century economist Adam Smith compared the value of diamonds and the value of water. In his book The Wealth of Nations (1776), Smith made the distinction between value in use and value in exchange. Water, which has great value in use, often has little value in exchange, whereas diamonds, which have little value

Preface – Management of Water Resources

in use, have enormous value in exchange. Both Plato and Smith pointed out that the market price of an item did not always represent its true value. In order to predict water demand how we value and price water is very important.

5 Predicting Future Demands for Water Predicting future demands for any resource is fraught with difficulties, but the complexity of water and its singular issue of finiteness make it particularly difficult to forecast. For example, how much should we worry about climate change and global warming? Global warming – one of the great scientific debates of the twentieth century – has now the opportunity to become the political debate of the twenty-first century. A few scientists in the nineteenth century warned of the effect on the atmosphere of excessive release of carbon dioxide into the atmosphere due to the burning of fossil (carbon-based) fuels. It was not, however, until the 1950s that serious comprehensive CO2 measurements were made. Since then large amounts of research funds have been expended in the field of climate science. Ultimately, in 1990, the United Nations established the Intergovernmental Panel on Climate Change (IPCC). The IPCC has produced four assessment reports so far dealing with the effects of changing greenhouse gas concentrations in the atmosphere. The IPCC has been incredibly successful in raising the status of the scientific understanding of climate change. The action in the United Nations (UN), however, has now shifted away from scientific research toward political action which will promote mitigation and adaptation strategies that could seriously curtail the increase of CO2 in the atmosphere by the end of the twenty-first century. In its Fourth Assessment Report in 2007, the IPCC identified five key impacts of increasing global average temperature: water, ecosystems, food, coasts, and health. A closer reading of the text shows that many of the most serious impacts on the nonwater areas are, in fact, mediated via water. Therefore, for instance, impacts on food are largely due to hydrological changes; aridity has major impacts on food, ecosystems, and human health. Thinking about the relative issues involved in climate change, Mike Mueller (2007) of the Global Water Partnership (GWP) said, ‘‘if it’s mitigation then the focus is rightfully energy, and if it’s adaptation, it will be water resources!’’ By this, he implied that the bulk of the mitigation strategies deal with handling the use, and development, of new energy resources, and the adaptation strategies will be mostly driven by water concerns. Hence, we need to focus on the water-adaptation strategies, bearing in mind that adaptation for other sectors may include many of the same, or similar, strategies. The pivotal role of water impacts, and hence water’s importance to adaptation, is also stressed in the Stern Review (2006). The local and regional effects on water however are inconsistent among the climate models, often predicting large regional differences in magnitude, variability, and direction of change for the most important hydrology parameter, the precipitation. However, whichever of these models one endorses, there is still a question as to what to do in meeting the future water demands. If we are interested in adaptation to global warming and climate change, it is largely irrelevant

3

which of the models we accept, because operationally there will be small differences among the adaptation strategies that one should follow, but major differences would arise if we were following a mitigation strategy. Surprisingly, even accepting the most conservative scenario leaves one in a strikingly similar situation – how to plan for the future under highly uncertain outcomes. The issue boils down to how do we deal with uncertainty in making decisions about water planning and management? Water engineers and hydrologists are supposedly expert at making such forecasts in a very uncertain world. Much of the focus has been on changes in the physical parameters, such as precipitation, stream flows, and evaporation, but rising global populations, coupled with rising incomes, and a concomitant increase in per capita consumption, will inexorably lead to serious consequences for the water resources in many areas of the globe, regardless of what happens to climate change. It is the old Malthusian population/ resources debate from the early 1960s; only now we have India and China moving into the middle classes in a big way. Keyfitz (1976) pointed out many years ago that it is the increasing middle class and their consumption patterns that were going to be the major problem for environmental sustainability. How we can adapt to meet these demands will be the major struggle for the remainder of this century. In planning for the future, we must also be aware of unintended consequences of our actions. One example of this is the current US attempt to mitigate climate change by reducing consumption of carbon-based fossil liquid fuels. During 2007–08, fueled by record crude oil prices, we rushed headlong toward a biomass-based liquid-fuel cycle. Because of their huge demand for cropland, water, and agricultural chemicals, the widespread development of biomass fuels turned out to be a disaster for the poor people of the world whose food budgets could not compete with the middle classes’ love affair with their automobiles. This means that water planners and managers need to worry a great deal about climate change. The consequences of the climate change will become apparent only if the planners work within a holistic framework to ensure that all of the consequences of climate change can manifest themselves.

6 Drivers of Socioeconomic Growth Among the earliest modern commentators on the drivers of socioeconomic growth and decline were Adam Smith, Edward Gibbon, Thomas Malthus, David Ricardo, and Karl Marx. Adam Smith, a Scottish economist, published his Wealth of Nations in 1776, which became the great classic of capitalist economic thinking. Gibbon, an English historian, combed the history of the Roman Empire for clues for these drivers in his Decline and Fall of the Roman Empire (1776–89). Malthus, an English country parson and economist, focused on the relationship between population growth and agricultural productivity in his seminal Essays on Population (1798). Ricardo, an English businessman and economist, focused on the declining economic returns from all forms of production and the increasing costs faced by industry over time. Finally, Karl Marx, a German sociologist and progenitor of Marxism, saw growth

4

Preface – Management of Water Resources

coming initially from capitalist accumulation and later from the labor of the proletariat. From their writings we see a concern about running out of resources as long ago as the eighteenth and early nineteenth centuries, well before the sixfold increase in population and 40-fold increase in per capita wealth arrived in the early twenty-first century. Malthus and Ricardo were particularly prescient about the roles of population, food, and energy resources. Malthus postulated a geometric rate of growth (like compound interest on a bank deposit) of population and an arithmetic growth (simple interest on a bank account) of land being brought under cultivation and, hence, arithmetic rate of growth of food production. Regardless of where they start, these curves will always intersect after a period of a couple of decades, and Malthus predicted widespread famine or violent conflicts to bring food and population into alignment with each other by ‘misery, war, pestilence, and vice’. Ricardo articulated ‘declining returns’ on investments in resources (coal and iron ore in his time; water, oil, and gas in our time) whereby the best (least-cost) resources are used first, followed by the next best, and so on. Increasing demand for the resource leads to price increases that will continue to rise until the resource becomes too expensive to use. These two nineteenth-century concepts can be used to explain our current water resources crisis and suggest pathways to end the crisis. We see these two concepts at work, for instance, in the case of New Delhi the population growth rate clearly exceeds the rate of possible increase in the water supplies (Malthus). On the other hand, in suburban Los Angeles (LA) as the cheapest sources of water are fully exploited, we see the Ricardo effect of increasing costs at work. When LA was developing in the 1930s, water was available at a reasonable cost, but as more and more people demanded more water the cost of supply was also increasing (the best projects had already been built). Without any technical breakthroughs, this means that the cost per unit of water keeps on increasing as time goes by. Of course, these constraints were also at work over previous centuries, even before they were articulated by Malthus and Ricardo, but Homo sapiens were able to avoid them by expanding our resource base through annexation and colonization, to bring in cheaper resources and food; by finding substitutes for scarce resources; and by improving our technology so that the same amounts of land and resources could be used more efficiently. Examples of these effects are seen in the British response to its nineteenth-century rapid population increases. More food was produced, not in England with its limited land and climate resources, but by Australia and other colonies such as Canada and India. This meant that the agricultural land was no longer a constraint on feeding the increasing population. So, Malthus’ limits and Ricardo’s increasing costs were avoided for the time being; however, since the globe is now pretty much filled up and most of the easiest available water is in use, there are few opportunities to expand the physical supply. The only option available to us now is improving the efficiency of water-use technology, but this is where we run into Ricardo’s increasing cost problem. The real question facing the globe at the start of the twenty-first century is whether we can keep on improving our technologies, or finding cheaper supplies or substitutes. However, just because these adaptations worked well over the past 200 years does not

mean that they will necessarily continue to work. This is the crux of the problem facing global water resources.

7 Transboundary Conflicts In historical times, control of water was the source of major conflicts among users often leading to skirmishes and minor wars. Peter Gleick (2009) tracked the history of water conflicts from 3000 BC to AD 2009. Historically, these have ranged from minor to major conflicts, but in recent times since the 1940s there has been less direct conflict and more attempts to resolve water issues by negotiation. He wrote, ‘‘There has been a lot of discussion about ‘water wars,’ a term that sounds great, but to which I do not subscribe: wars start and are fought for many reasons and while water has often been a target, tool, or objective of violence, it is certainly hard to ascribe the primary reason for any war to water alone’’ (Peter Gleick, 2009). However, the lack of availability and access to water may have been one of the conditions leading to many wars. The lack of access to water can have major impacts on the health and wealth of nations; major occupations, such as fishing and farming, cannot flourish, and the growth of cities will be limited. With the development of nation-states in the sixteenth and seventeenth centuries, the lack of access by downstream users and the control by the upstream populace was firmly established. This meant that, without a treaty, the downstream users were essentially cut off from use of the flowing river. The Industrial Revolution brought serious pollution to the rivers which also impacted the downstream users. The UN’s International Law Commission spent 26 years from 1971 to 1997 drafting the UN Convention on the Law of the Non-Navigational Uses of International Watercourses (1997). As of 2010, it has not yet been ratified by the UN General Assembly by the requisite 35 countries needed for it to come into force. The existence of such a treaty is a good indication of the international community’s intentions to improve the nature of collaboration among the riparians in international and transboundary rivers; however, the inability to ratify the Convention says a great deal about the wishes of upstream countries not to cede sovereignty to a supranational body. Despite the nonexistence of a clear set of laws and treaties, customary international laws have used many principles such as prior consultation, avoidance of significant injury, equitable apportionment, nondiscrimination and nonexclusion, and provision for settlement of disputes embedded in the UN treaty. Moreover, the fact that it has not yet come into force has not hindered the resolution of many smaller water conflicts relying on the common-sense ideas presented above. Moreover, even when ratified, the Convention lacks an effective enforcement mechanism and will thus still rely largely upon the goodwill of upstream parties, or the hegemonic strength of the downstream countries. Water conflicts seem to arise every time a river crosses a boundary. For instance, in the Colorado Basin, despite the existence of the Federal Interstate Colorado Compact, there are still serious water conflicts among the seven US basin states and Mexico. In India, we see similar conflicts regarding the Ganges River, both domestically and internationally, with Nepal and Bangladesh sharing access. Transboundary water

Preface – Management of Water Resources

conflict is one area in which the conflicts among the parties really emphasize the need for clear and transparent rules for cooperation. The present seems to be one of the periods of great interest in international rivers, as experts estimate that there are over 145 countries with at least participation in one or more of the 261 international river basins on the Earth. Over time, there have been as many as 300 river-sharing agreements in Europe since the Treaty of Versailles in 1815. However, almost all of these treaties dealt with regulating in-stream use for navigation, hydropower, fishing, and pollution disposal, all of which did not involve the large-scale diversions of water which now regularly occur with irrigation developments. Large withdrawals typically create very difficult water-allocation problems for the downstream countries and, in history, were typically resolved with violence or threats of violence. In our times, we would rather resort to negotiations than war. The previous period of great concern about transboundary river conflicts was in the 1950s and early 1960s. This period culminated in a successful treaty on the Indus Basin, brokered by the World Bank and signed by India and Pakistan. The accord fueled optimism for resolving other major water conflicts. At that time, basins such as the Ganges–Brahmaputra, the Mekong, and the Nile (and even the tiny Jordan River) were subject to detailed analysis, even to the extent of creating river basin commissions in an attempt to avoid conflict among the parties. Unfortunately, this era of concern came up short. Of these large rivers, only the Indus was eventually successfully developed. Currently, we are experiencing a resurrection of conflict, fueled by the shortage of water caused by rapid development and huge population growth, and possibly global warming. To allocate – or reallocate – the flows of a river is always a political decision. No matter how detailed the technical, economic, and social studies are, hard choices have to be made among the various users who stand to gain and lose from such accords. This is true whether the river is a national river or crosses international borders. However, transboundary rivers imply a level of political decision making that goes beyond local and national interest groups. It requires the ability to negotiate between sovereign nations. All rational planners recognize the value of cooperation on river-sharing issues, from sociocultural terms to trade and economic ones. What is not clear, however, is how to put a value on cooperation; in other words, just how valuable is cooperation?

8 River Basin Politics The problem with purely political decisions is the lack of predictive behavior on which they reside. Thus, many politically inclined decisions have led to a deviation from the scientific–technical-based analysis, which accounts for the quantitative benefits from sharing resources among the coalitions of competing groups. Political considerations are sometimes heavily influenced by noneconomic factors outside the technical analysis. They are often pursued separately and apart from economic objectives, with different personnel and rituals. Political approaches tend to be more descriptive and idiosyncratic than the analogous models in the sciences.

5

Ultimately, foreign policy is the most influential determinant of a country’s position on international rivers. Linkage of the river settlement to other outstanding economic and social issues between and among countries is important, as is achieving reciprocity for one’s actions either in the linkage of issues or in sharing benefits that are only achievable through international cooperation. In addition, the climate for agreement is a prime political basis for sharing water resources; this comes about when countries have common or shared technical perception of the problems, networking and contacts at the transgovernmental levels, and the need to be seen as being collaborative as a nation. A wide range of solutions are possible in most negotiations, while the net benefits are not the only consideration; many political issues dominate in shaping the decisions on the locations of the investments which might not necessarily be in the interest of the best technical planning. Managing common property resources is a very difficult endeavor, and the added complexities of transboundary water are no exception. An interesting phenomenon is that river flows have both negative and positive externalities typically working only in one direction, that is, downstream. This pervasive unidirectional feature of water use means that resolution of basin conflicts through mutual control of external effects that work reciprocally is generally ruled out. However, downstream countries can also benefit from some positive external effects of upstream use. Aside from the water allocation problems that arise from the physical sharing of a common resource, there are also many water-quality problems that can arise downstream as an effect of upstream use. Natural processes such as floods and droughts can also cause major downstream effects and are sometimes mistaken for man-made externalities, and thus lead to further mistrust and tensions among the riparian states.

9 The Contents of Volume I In presenting a discussion on water resources, this volume has been constrained by the width of the definitions of what constitutes the field. We have presented just 11 chapters which while they cover a broad range of concerns, they do not, by any means, cover the full range of concerns. We do, however, present materials dealing with the three questions outlined at the start of this chapter: feeding the global population, providing water supply and sanitation to the ever-increasing population, and some approaches to dealing with the huge uncertainties associated with potential global climate change. The first three chapters deal with the broad frameworks of IWRM, governance, and water as an economic good. The chapters attempt to lay the groundwork for dealing with water as a fundamental resource for development. In Chapter 1, Roberto Lenton explores the history and evolution of the concept of IWRM and reports on various assessments and critiques of the concept. In particular, the critics have focused on definitions that tend to be narrowly focused such as a country having a national water policy, or a water law, or the river basin as the focus of planning, or participatory management. Lenton also provides a set of criteria by which IWRM could be viewed in practice. These are a sensible set of criteria

6

Preface – Management of Water Resources

and if followed would make a major difference in the sustainability of water-sector decisions. In Chapter 2, Edella Schlager takes up the issues of water governance. She emphasizes a bottom-up approach working from the water users up through multiple layers of governance. She emphasizes the fact that the modern approaches to water development and management are mostly based on methodology developed by experts for experts. She claims that the challenge now is to design and fit water governance organizations into complex multiscale and intergovernmental and watershed systems. It is no longer a game for experts controlled and run by experts. In Chapter 3, John Briscoe deals with how to value water in its different uses. His major split is between urban water and water used for irrigation, or the use of water as urban infrastructure and public health and water in its productive mode for food production. In both cases, he describes how to evaluate the direct and indirect benefits associated with water use. He argues that the indirect benefits associated with water use can be as large as, or larger than, the conventionally measured direct benefits. He concludes with an appeal to move away from the conventional formulaic applications of benefit/cost analysis and attempt to identify critical supplementary investments to use more fully the multiplier effects of large infrastructure projects. The next three chapters deal with the practical socioeconomic issues of forecasting the demand for water, the pricing of water and sanitation services, and how to know if interventions in water supply really have the benefits attributed to them. In Chapter 4, Benedykt Dziegielewski and Duanne Baumann point out that credible long-term forecasts of water demand are essential to planning for the long-lived water infrastructure. They show that such forecasts must be based on a high level of disaggregation of demand; the uses of econometric models grounded in economic theories of production and consumption, considerations of potential climate change, and must, above all, provide explicit and plausible assumptions. Dale Whittington in Chapter 5 reviews the role of economic pricing approaches to managing water and sanitation services. In this chapter, he cautions against some of the enthusiasm for investments in social overhead capital expressed by Briscoe in Chapter 4, with the potential for oversubsidization of large projects at the expense of smaller ones. Following up on this theme, Alix Zwane and Michael Kremer, in Chapter 6, examine the evidence whether community-level rural water infrastructure successfully reduces diarrheal disease and conclude that the evidence does not support it. However, from their review of the literature they found evidence that sanitation and hygiene are more important than water quality. The next set of three chapters cover the role of groundwater in providing water resources for many different types of water services, managing water for agriculture, and managing the aquatic ecosystem to provide adequate protection of the environment. In Chapter 7, Lopez-Gunn, Llamas, Garrido, and Sanz assess the development of groundwater over the past half century. They very broadly review the assessment of the total resource available, the economics of groundwater use, institutions and governance of groundwater, and the future sustainability of the resource. They conclude that groundwater may be the most important water source under the more extreme climate-change scenarios, in particular for irrigation in

low-latitude countries. They stress the need for better governance structures for groundwater management and that a much higher level of user participation will be required for sustainable use of the resource. In Chapter 8, Jorge RamirezVallejo makes a comprehensive review of all aspects of managing water for agriculture and concludes that the major challenge in this area is to reverse the serious failure of institutional arrangements at the national and local levels to deal with water correctly. The concluding chapter in this section by Max Findlayson on managing aquatic ecosystems recognizes the interdependence of people and their environment and focuses on the management of water to support the ecosystem and the environment. He concludes with a strong support for the Millennium Ecosystem Assessment as the best approach to managing wetlands and their aquatic systems. He points to the need in the coming decades to address the trade-offs among current and future uses of wetland resources, importantly inbetween agricultural production and aquatic diversity. The final two chapters return to some political and social issues of water resources management. In Chapter 10, David Moreau reviews the problems with implementing ambiguous water policy. He uses the case of the experience in the US of implementing the Clean Water Act especially under the federal system where the states are left to implement national policy. He shows how there are few ambiguities in dealing with point sources of pollution, but many in dealing with nonpoint sources which has led to the Balkanization of the implementation with the individual states essentially ignoring downstream states when setting goals for total maximum daily loads (TMDLs). Fittingly, the volume concludes with a chapter by Casey Brown on risk assessment, risk management in the context of potential climate change. He develops an approach to risk management that attempts to reconcile traditional approaches with our growing knowledge of uncertainty that mark the hydrologic records. He concludes that the water community has focused primarily on the means to reduce the uncertainty related to hydrologic events, but little effort has been devoted to reducing hydrologic risk to society or to communicate risk to promote risk-reducing behavior.

Acknowledgments The editor wishes to thank the following persons who helped in the review of the manuscripts in Volume 1: Chris Basso, John Briscoe, Robert Brumbaugh, Ximing Cai, Torkil JonchClausen, Line Gordon, Chuck Howe, Annette Huber-Lee, Roberto Lenton, Tom Maddock, Suzanne Ogden, Margaret Owens, Cliff Russell, Mike Shapiro, and Richard Vogel.

References Bowley M (1973) Studies in the History of Economic Theory Before 1870. London: Macmillan. Falkenmark M and Rockstro¨m J (2004) Balancing Water for Humans and Nature: A New Approach in Ecohydrology. London: Earthscan. FAO (2009) AQUASTAT 2009: Water and Food Security. http://www.fao.org/nr/water/ aquastat/main/index.stm (accessed September 2010). Gibbon E (1776-89/1989) The History of the Decline and Fall of the Roman Empire, vols. I-VI. New York: St. Martin’s Press.

Preface – Management of Water Resources

Gleick PH (2009) Water brief 4: Water conflict chronology. In: The World’s Water 2008-2009: The Biennial Report on Freshwater Resources, pp. 151. Washington, DC: Island Press. Global Water Partnership (2000) Integrated Water Resources Management, Technical Advisory Committee, Background Paper No. 4. Hanemann WM (2007) The economic concept of water. In: Rogers P, Llamas MR, and Martinez-Contina L (eds.) Water Crisis: Myth or Reality? ch. 4. London: Taylor and Francis. Intergovernmental Panel on Climate Change (2007) IPCC Fourth Assessment Report: Climate Change. Cambridge: Cambridge University Press. http://www.ipcc.ch/ publications_and_data/publications_and_data_reports.htm (accessed September 2010). Keyfitz N (1976) World resources and the world middle class. Scientific American 235(1): 28--35.

7

Malthus TR (1798) An Essay on the Principle of Population. London: St. Paul’s Church Yard (printed for J. Johnson). Mueller M (2007) Policy Brief 5: Climate Change Adaptation and Integrated Water Resources Management – An Initial Overview, Global Water Partnership, Technical Advisory Committee. Smith A (1776/1976) An Inquiry into the Nature and Causes of The Wealth of Nations In: Cannan, E (ed.). Chicago: University of Chicago Press. Stern Sir N (2004) The Stern Review on the Economics of Climate Change. Cambridge: Cambridge University Press. United Nations (1997) Convention on the Law of the Non-navigational Uses of International Watercourses, adopted by the General Assembly on 21 May 1997. Not yet in force. http://untreaty.un.org/ilc/summaries/8_3.htm and http:// untreaty.un.org/ilc/texts/instruments/english/conventions/8_3_1997.pdf (accessed September 2010).

1.01 Integrated Water Resources Management R Lenton, The Inspection Panel, The World Bank, Washington, DC, USA & 2011 Elsevier B.V. All rights reserved.

1.01.1 1.01.2 1.01.3 1.01.4 1.01.5 1.01.6 1.01.7

Introduction IWRM at the Watershed Level: Watershed Management IWRM at the Water-Use Systems Level: Agricultural Water Management IWRM at the Water-Use Systems Level: Water Supply and Sanitation Services IWRM at the Basin Level IWRM at the National Level: Policies and Governance IWRM at the Transnational and Global Level: Information Sharing, Cooperation, and Technical and Financial Assistance 1.01.8 IWRM as a Meta-Concept 1.01.9 History and Evolution of the Concept of IWRM 1.01.10 Assessments and Critiques of the Concept of IWRM Acknowledgments References

1.01.1 Introduction This chapter is about the planning and management of water resources. It aims to provide a comprehensive look at the practices and approaches that have come to be known as integrated water-resources management (IWRM). Following GWP Technical Committee (2009), the chapter defines IWRM as the way in which water can be managed to achieve the objectives of sustainable development, and an approach that reflects the need to achieve a balance among economic efficiency, social equity, and environmental sustainability. The chapter begins with several sections that analyze integrated approaches to water-resources management at different levels (from small watersheds to basins, agricultural systems, and national and global policymaking), which have been practiced for some time and about which there is now a considerable body of knowledge. Building on these analyses, IWRM is not to be seen as a single approach but as a wide range of approaches to manage water and related resources – a meta-approach or meta-concept, as it were, which both transcends the various levels of decision making and recognizes the importance of integrating decision making at each level. The chapter then provides some historical perspective on the evolution of the concept of IWRM, concluding with a summary of recent assessments and critiques of the concept. Although the concept of IWRM is applicable to a variety of contexts, this chapter focuses on the management of water in the context of development, that is, on the management of water resources to advance sustainable development and reduce poverty. This means that the chapter examines the management of water through the lens of the major development and environment issues that are currently challenging countries across the world and which are intrinsically interconnected to water resources in one form or another. In particular, it means that the goals of water management addressed in the chapter relate to development, and the kinds of water challenges emphasized in the chapter are those most often found in the context of development – such as how to allocate more water to generate rural livelihoods and grow food in order to reduce income poverty and hunger.

9 9 10 12 13 14 16 17 18 18 19 20

The following six sections provide an overview of integrated approaches to water-resources management at different levels, from small watersheds to basins, agricultural systems, and national and global policymaking. Management at each level shares a common focus on managing water and related resources, to achieve multiple objectives, which has been practiced for some time, and has developed its own discipline, vocabulary, body of knowledge, networks of interested people, and global institutions. Each of the six sections therefore focuses on management at a specific level – explaining the main characteristics of management at this level, summarizing the literature on the subject, describing some of the networks and institutions working at this level, and providing examples of good practices. We begin with the watershed level.

1.01.2 IWRM at the Watershed Level: Watershed Management Perhaps the most salient features of management at the watershed level are (1) the crucial role of land as well as water management and (2) the strong relationships between watershed management and downstream impacts. As a result, in watershed management the need for a close integration of land and watermanagement activities and upstream/downstream considerations is imperative. Reflecting these key features, the World Bank, in Darghouth et al. (2008), has defined watershed management as ‘‘the integrated use of land, vegetation and water in a geographically discrete drainage area for the benefit of its residents, with the objective of protecting or conserving the hydrologic services which the watershed provides and of reducing or avoiding negative downstream or groundwater impacts.’’ In many developing countries, watershed management is a crucial part of rural-development efforts to generate rural livelihoods and increase incomes. The concepts and practice of watershed management have evolved in the last 30 or 40 years, in parallel with the evolution of the concept of IWRM as a whole. Importantly, from an initial emphasis on technology and engineering, driven primarily by downstream

9

10

Integrated Water Resources Management

environmental protection considerations, they have evolved toward more integrated approaches driven by upstream poverty reduction and livelihood generation. As the above discussion suggests, there are several features of watershed management that are important in the IWRM context. First, the management of watersheds involves management of a range of resources, including soil resources, crops, forests, livestock systems, and water in the form of overland flow, streams, and soil moisture. Second, the goals of watershed management are generally economic, social, and environmental in nature, and relate both to communities in the watershed itself (principally increased productivity and incomes) and to downstream (principally flood management and drought mitigation, through more controlled water resources). Third, the instruments involved in watershed management have usually involved a combination of institutional, economic, and environmental measures. Participatory approaches have proved to be particularly valuable. The economic returns from the use of water for productive purposes within watersheds have usually proved to be a hugely important catalyst and incentive for collective environmental-preservation activities. A crucial feature of many watershed management efforts, as illustrated by the Sukhomajri experience (see Box 1), is the interplay among environmental, economic, and institutional approaches. Darghouth et al. (2008) have noted that ‘‘where communities could see the economic benefits and were empowered, they were willing to invest in long term conservation.’’ By now, there is a significant body of knowledge and experience in watershed management, which draws on both the range of watershed management experiences in many countries and a growing body of research and evaluation carried

out by institutions such as the World Bank, the International Water Management Institute (IWMI), the World Agroforestry Centre, and the Food and Agriculture Organization (FAO). These and other institutions have established some important watershed management programs and networks. The World Agroforestry Centre, for example, has established the Rewarding Upland Poor for Environmental Services (RUPES) program in Southeast Asia, which aims to develop ‘‘mechanisms for rewarding the upland poor in Asia for the environmental services they provide.’’ FAO has also been active in the subject area, in terms of both publications and networks, playing a key role, for example, in the Latin American Technical Cooperation Network on Watershed Management (RDLACH), which was created in 1980 to promote watershed management in Latin America and the Caribbean. The World Bank has been an active player in watershed management, providing finance for important initiatives and evaluating and drawing lessons from many watershed-management programs in different parts of the world (see e.g., World Bank, 2003, 2004a, 2004b, 2004c, 2005). A recent World Bank discussion paper (Darghouth et al., 2008) summarizes these experiences and lessons learned. The IWMI has carried out important research in this area, as illustrated by Sharma et al. (2005). The Comprehensive Assessment of Water Management in Agriculture (CA), which was spearheaded by IWMI, has also sponsored research on watershed management. One study under the CA (Joshi et al., 2005) carried out an in-depth analysis of the impacts of watershed management programs in India, evaluating 311 case studies of watershed programs in terms of economic efficiency, equity, and sustainability, and concluding that these programs yielded an average internal rate of return of 22%. Other

Box 1 Watershed Management in Sukhomajri, India. From Lenton R and Walkuski C (2009) A watershed in watershed management: The Sukhomajri experience. In: Lenton R and Muller M (eds.) Integrated Water Resources Management in Practice: Better Water Management for Development. London and Sterling, VA: Earthscan. Sukhomajri is a small village of about 450 people on the edge of the Shivalik mountain range near Chandigarh in India. By the 1970s, over a century of heavy logging in the area and the overgrazing of cattle, sheep, and goats in open forest lands had severely degraded the area surrounding Sukhomajri. Its people were impoverished and survived primarily by raising rainfed crops and keeping goats that foraged in the denuded hills. The Sukhomajri program came about because the citizens of Chandigarh, whose Sukhna Lake had lost nearly 70% of its storage capacity due to siltation by the early 1970s, asked the nearby Central Soil and Water Conservation Research and Training Institute (CSWCRTI) for assistance in solving this problem. The Institute soon found that the lake’s siltation was caused by soil erosion in the hills in and around Sukhomajri, and developed a program to improve soil and water conservation in the watershed in consultation with the local community. While initially the people of Sukhomajri were not very interested in a project principally designed to protect Chandigarh’s lake, their attitude changed dramatically when the Institute built a small dam to control runoff and the villagers found they had a reliable source of water relatively close at hand. A cooperative effort began, focused on community participation in decision making and management, incentives for villagers to graze their animals outside of the watershed, and – crucially – an equitable system of water allocation that would benefit all villagers equally, with water rights granted to all villagers whether or not they owned land. Over time, the villagers organized and formed a water-users association and later the Hill Resource Management Society, to manage and distribute irrigation water. All these actions underscore the Sukhomajri program’s balance of economic efficiency, social equity, and environmental considerations. Over time, the impact of the Sukhomajri program has been considerable. Annual household incomes, for example, rose from around US$230 in 1979 to about US$1360 in the 2000s – more than double the per capita income of the state of Haryana, which itself is one of the highest in India. In addition, tree density in the area rose a 100-fold from 13 ha 1 to around 1300 ha 1 between 1979 and 1995, underscoring its environmental regeneration benefits. Socially, the Sukhomajri program has had important equity impacts because of the focus on landless people and equal distribution of irrigation water. More broadly, Sukhomajri’s approach to community-integrated watershed management has become a model for watershed-development programs elsewhere in India. Importantly, watershed management in Sukhomajri has undergone many changes, not all of which have been positive, since the program began. For example, the check dam led to a rise in groundwater levels, which led an increasing number of villagers to build shallow tubewells and use these tubewells rather than the check dam for irrigation water; this development lowered the incentives for limiting watershed grazing and participating in communal activities. In addition, changes in taxation led to significant declines in the income for the Hill Resource Management Society. This fall in revenues, coupled with the lower incentives for villagers to participate in watershed conservation, led to silting dams and deteriorating pipelines.

Integrated Water Resources Management

important works on watershed management include Farrington et al. (1999), Heathcote (1998), and Sharma et al. (2005). An excellent example of watershed management in the context of IWRM is the Sukhomajri program in Northwest India, which has been extensively documented and analyzed by Seckler (1986), Agarwal and Narain (1999), Kerr (2002), Khurana (2005), CSE (1994), CSE (1998), CSE (2002), and CSE (2007). This case is summarized in Box 1.

1.01.3 IWRM at the Water-Use Systems Level: Agricultural Water Management The key feature of IWRM at the agricultural-systems level, commonly referred to as agricultural water management (AWM), is that it is defined by a particular type of water use – agriculture – which in most countries is by far the largest consumer of water. AWM thus has a common purpose, generally defined in terms of increasing and sustaining agricultural production. Management itself takes place at several levels, from a single farmer’s field to small farmer-managed systems to large publicly operated irrigation systems, usually with policy and other support at higher levels as well. Reflecting these features, AWM may be viewed as encompassing the range of structural and nonstructural measures at various levels to harness, control, and manage surface water, groundwater, and rainwater to improve and sustain agricultural production. (This definition draws on the definition used in World Bank (2006), but has been broadened to encompass both large-scale and small-scale, structural and nonstructural interventions at a variety of levels.) Structural measures include combinations of irrigation, drainage, and flood control, water conservation and storage, on-farm water management, and soil-moisture conservation, while nonstructural measures include institutions and policies to improve physical and financial sustainability, and user operation and management . AWM involves increasing access to reliable and affordable water supplies, improving management of rainwater, soil moisture, and supplemental irrigation, finding ways to gain higher yields and value from the same water amounts, and enhancing management of the resource as a whole. The interventions involved vary significantly from level to level. While at the farm level, AWM interventions might involve investments in irrigation or soil management, at the farmermanaged irrigation system level, they also include community mobilization, and at the large-system level, the operation of canals and the governance of resources. Decision making at national policy levels is also a fundamentally important aspect of AWM, as discussed later in this section. AWM, like water-resources management as a whole, is best viewed as an integrated, holistic process. Indeed, AWM is an aspect of water-resources management, and should be understood as such. Water has several characteristics that impact on its management and use in agriculture. For example, water has many competing uses outside of agriculture, and it is required in relatively large quantities to produce yield increases, but is heavy, bulky, and costly to transport in comparison to other inputs.

11

While earlier it was noted that AWM is defined by the common objective of increasing and sustaining agricultural production, in fact, AWM generally has broader objectives beyond enhancing agricultural productivity, such as improvements in livelihoods and incomes, reductions in risk, and long-term sustainability of the resource. Increasing the numbers of small holders with access to reliable and affordable water provides more regular employment and livelihood opportunities to landless people as well as small holders, and enhances the prospects of ensuring access to domestic water supply and sanitation in rural areas. Efficient AWM creates opportunities for farmers to improve livelihoods, leveraging investment in other productive inputs such as improved seed and fertilizer, while also helping to ensure longterm sustainability of both surface water and groundwater resources. Finally, good AWM aims to reduce the risks that farmers and countries experience from variable rainfall, which can have a powerful impact on growth. All this reinforces the notion that efficient AWM, like IWRM as a whole, generally has a broad range of economic, social, and environmental goals. Importantly, AWM generally involves the integrated management of both blue water – water withdrawn from rivers, reservoirs, lakes, or aquifers for irrigation purposes – and green water – rainfall stored in soil moisture. While blue water is visible and its role in irrigated agriculture is clearly understood, green water and its crucial role in rainfed agriculture often goes unrecognized. Green water management measures to improve agricultural productivity can encompass soil management, crop choices and practices, and water capture. Blue water irrigation management, on the other hand, can involve water storage, lift, transportation, delivery, application, and reuse at various levels. However, average yields from rainfed agriculture using green water are much lower than those from irrigated agriculture using blue water, and, as a result, only half of the world’s food is produced under rainfed conditions practiced by the majority of the world’s farmers. Better green water management can reach relatively large numbers of farmers at relatively low cost, but the productivity gains are relatively small; improved blue water irrigation management, on the other hand, can achieve higher productivity gains, but reach relatively smaller numbers of farmers and with a relatively high cost per farmer. Within blue water, AWM generally involves the management of both groundwater and surface water. Like green water, groundwater is less visible and frequently overlooked, but enables farmers to exercise much greater control over the amount and timing of water applied to their crops than those forms of irrigation that depend on unreliable surface-water supplies, and is an effective way for small holders to improve crop production. In South Asia and North China, in particular, groundwater irrigation over the last 30 or 40 years has grown considerably, and played an ever-greater role in efforts in these regions to improve productivity, food security, livelihoods, and incomes. Indeed, according to the CA (CA, 2007), the rapid growth of groundwater irrigation in South Asia and the North China plains between 1970 and 1995 was at the heart of the agrarian boom in these two regions. Nevertheless, this boom has come at the cost of decreasing water tables that threaten its long-term sustainability (Shah, 2009).

12

Integrated Water Resources Management

There are numerous examples of best practices in AWM that have adopted an integrated approach. One such example is summarized in Box 2. While the physical management of water for agriculture takes place at the agricultural-systems level, decision making at national policy levels is a fundamentally important aspect of AWM. Some of these policy issues relate specifically to irrigation, such as the balance of large-scale and small-scale irrigation, the balance of government-led and community-based interventions, and the role of the private sector. While the institutional, technological, and environmental problems that need to be tackled to improve irrigation performance are usually system specific, the use of economic instruments to increase the efficiency of irrigation is usually a matter of national policy. Other national policy issues relate to agricultural water as a whole. In many countries, for example, a major policy question is the scope for growth in food production in rainfed and irrigated agriculture, from which combination of the two should future food production lie, and what are the wateravailability implications of each of these strategies. A related topic is the extent to which trade, as against domestic food production, should be used as a strategy for sustainable food security, taking into account not only the virtual water transfers that are embodied in food trade but also the risks and uncertainties associated with trade in food supplies and the role of domestic food production in providing a source of

income and livelihoods to rural populations. A third topic is the water and energy nexus. In countries in which both energy and water resources are scarce, a key policy question is whether scarce water supplies should be allocated for biofuels or for hydropower, and if so what should be the crop choice. There is a vast literature on AWM. While much of it is narrowly focused on specific technical issues of irrigation, there is a growing literature that takes an integrated approach and addresses some of the broader policy issues identified above. A recent review commissioned by the Technical Committee of the Global Water Partnership (GWP) (GWP Technical Committee (2009)) showed the wide range of available literature on subjects such as water–food interactions, including trade in virtual water, biofuels and their implications for water, and water efficiency and productivity, in the works of Rosegrant et al. (2002); de Fraiture et al. (2008), and Hellegers et al. (2008). The most recent comprehensive review of AWM, however, is the CA, a recent multi-institutional assessment of the current state of knowledge on how to manage water resources for agriculture. This assessment is summarized in Box 3. Several global institutions have played a major role in supporting national efforts to improve AWM. The World Bank has long been a major source of financial and technical assistance in AWM, investing some US$13.2 billion in 56 countries in the 10-year period between 1994 and 2004 alone (World Bank, 2006). FAO has had an active program in AWM

Box 2 Irrigation reform in Mali. From Barry B, Namara R, and Bahri A (2009) Better rural livelihoods through improved irrigation management: Office du Niger (Mali). In: Lenton R and Muller M (eds.) Integrated Water Resources Management in Practice: Better Water Management for Development. London and Sterling, VA: Earthscan. The Office du Niger in Mali was formed in the 1930s as a centralized public enterprise to produce irrigated cotton and rice. Starting in the 1990, the Office du Niger was significantly revamped through a process involving measures such as physical rehabilitation and modernization, farming-systems intensification, improvements in land-tenure security, the creation of pro-farmer support services, the establishment of farmer organizations, and the introduction of innovations in agricultural technology, along with macroeconomic and cereal market reforms. All these have led to dramatic gains in rice production and farm incomes as well as reductions in rural poverty. The case shows that changing agricultural water management requires a supportive macro-policy environment, appropriate institutional changes, and infrastructural investments. Equally important, it shows that those reforms may need to precede improvements in water management. Moreover, in aid-dependent low-income countries, reform cannot occur unless both government and donors concur on the need for change. Finally, the case drives home that improving water management is a continuing process; gains to date in economic efficiency and (to a lesser extent) equity in the Office du Niger now need to be matched by improvements in environmental sustainability.

Box 3

The Comprehensive Assessment of Water Management in Agriculture. From http://www.iwmi.cgiar.org.

The Comprehensive Assessment of Water Management in Agriculture (CA, 2007) was a critical evaluation of the benefits, costs, and impacts of the past 50 years of water development, the water management challenges communities face today, and the solutions people have developed around the world, aimed at assessing the current state of knowledge and stimulating ideas on how to manage water resources to meet the growing needs for agricultural products, to help reduce poverty and food insecurity, and to contribute to environmental sustainability. The assessment was produced by a broad partnership of practitioners, researchers, and policymakers and organized through the CGIAR’s Systemwide Initiative on Water Management (SWIM). SWIM was convened by the International Water Management Institute (IWMI), which initiated the process and provided a secretariat to facilitate the work. The CA’s scope was water management in agriculture, including fisheries and livestock, and the full spectrum of crop production from soil tillage through supplemental irrigation and water harvesting to full irrigation in a sustainable environment context. The review covered a range of topics, from managing water in rainfed agriculture and groundwater use in agriculture to agricultural use of marginal-quality water resources and integrating water and livestock development. The assessment was originally framed by 10 questions, later expanded as interest grew, and included the overarching question: How can water in agriculture be developed and managed to help end poverty and hunger, ensure environmentally sustainable practices, and find the right balance between food and environmental security? Some of the questions addressed by the CA included: What are the options and their consequences for improving water productivity in agriculture? What are the options for better management of rainwater to support rural livelihoods, food production, and land rehabilitation in water-scarce areas? What are the options for integrated water resources management in basins and catchments? What policy and institutional frameworks are appropriate under various conditions for managing water to meet the goals of food and environmental security?

Integrated Water Resources Management

for several decades, focusing on knowledge generation and dissemination, policy and technical advice, and preparation, investment, and implementation of projects. On the research and development side, the IWMI has played a major role in shaping thinking and action on AWM. IWMI’s evolution from a narrow blue-water focus on the management of irrigation at the systems level to its current broad mission to ‘‘improve the management of land and water resources for food, livelihoods and the environment,’’ which encompasses work at all levels, mirrors the evolution of the discipline of AWM as a whole.

1.01.4 IWRM at the Water-Use Systems Level: Water Supply and Sanitation Services While agriculture consumes the lion’s share of the world’s water resources, AWM is not the only example of a level of water management defined by a particular type of water use. Water supply and sanitation (WSS) can also be considered as a type of water management defined by its use, with a common purpose generally understood in development circles as increasing sustainable access to basic sanitation and safe water supplies for domestic purposes. As with AWM, management takes place at several levels, from the household to large cities, with policy and other support at higher levels as well. Unlike AWM, however, the amounts of water used for WSS are relatively small. The key feature of WSS is thus that the management of resources other than water – financial, human, and institutional resources in particular – is often more important than the physical management of water itself. Despite or perhaps because of this key difference, water and sanitation-services management has followed a course similar to AWM, in that it has evolved from a somewhat narrow focus on technologies to a much broader understanding of the political, institutional, and financial dimensions of improving access to water and sanitation at all levels. As a result, it is generally agreed today that increasing access to water and sanitation services by the unserved will require a broad integrated approach, at many levels, involving increased political commitment, institutional and technological innovation, and increased financial allocations, with attention not only to the supply side but also to the demand side (Lenton et al., 2005). Several global institutions have played a major role in shaping the thinking in this field as well as supporting national efforts to improve access to water and sanitation services. These include the World Bank (both through its investment program and its WSS program), the World Health Organization, and the United Nations International Children’s Education Fund (UNICEF), the Water Supply and Sanitation Collaborative Council, and several major international nongovernmental organizations, such as WaterAid.

1.01.5 IWRM at the Basin Level The key feature of water management at the basin level is that the basin is the basic unit for integrating the supply side of water-resources management, that is, for integration within the natural system. This means that the basin is well suited for

13

integrated management of land and water management, green and blue water, surface and groundwater management, water quantity and water quality, spatial and temporal variability, and upstream and downstream interests. The chapter on freshwater in Agenda 21, the key document resulting from the 1992 Earth Summit (UNCED, 1992), states that ‘‘Integrated water resources management, including the integration of land- and water-related aspects, should be carried out at the level of the catchment basin or sub-basin.’’ Likewise, Article 26 of the Johannesburg Plan for Implementation, the key outcome of the 2002 World Summit for Sustainable Development in Johannesburg (WSSD, 2002), states that ‘‘the river (or water) basin should be used as the basic unit for integrating management.’’ Water management at the basin level has thus become the central focus of much of the advances in thinking about IWRM. Nevertheless, while basin boundaries provide a useful way of delimiting the supply side of the equation, they are not necessarily the best means to integrate the demand side, especially since basin boundaries usually do not coincide with political or administrative boundaries. Integrating natural and human systems therefore generally requires work at other levels beyond the basin. Basins can be characterized by the degree of pressure placed on a basin’s water resources, that is, by the relationship between the requirement for freshwater resources and its availability in the basin, taking into account both quantity and quality considerations and variability over time and space. Basins in which requirements (including environmental requirements) exceed availability and where additional water needs cannot be met without reallocating water from other users, or by improving water-use efficiency, are called closed systems. Open systems, by contrast, are those in which water availability exceeds current requirements and where there is, therefore, still room for expanding water use without environmental damage. Needless to say, management of closed or closing water basins is significantly more challenging than that of open basins. Importantly, integrated management at the basin level does not necessarily imply the need for a basin organization. An analysis by the CA, conducted together with GWP and International Network of Basin Organizations (INBO) (CA, 2008), concluded that ‘‘adaptive, multilevel, collaborative governance arrangements’’ are best able to deal with the complexities of issues that arise at the basin level. Noting that solutions need to be driven by contextual realities and that what works in one basin may not work in another, the CA emphasized that ‘‘Not all water-related problems can or should be solved at the river basin level. Some problems are best addressed at the sub-basin or local level. Others have solutions beyond the basin itself and even outside the water sector, for example in national or federal agricultural policies.’’ Numerous basin-management organizations have been established in different parts of the world. The significance of basin management is reinforced by the INBO, which was established to facilitate exchange of experience and expertise among organizations interested in river-basin management and to promote the principles and means of sound water management. Similarly, there is a vast literature that addresses water management at the water-basin level, particularly at the

14

Integrated Water Resources Management

river-basin level. Perhaps the most influential and ambitious publication on the subject is Hooper (2005), which focuses on integrated approaches to river-basin management and develops an integration framework for river-basin management based on principles of natural-resources management and planning. A recent handbook on basin management has been published by GWP and INBO (2009).

There are numerous examples of good practices at the basin level. Boxes 4 and 5 illustrate the different ways in which integrated approaches have been applied at a basin level. Box 4 focuses on the Lerma–Chapala Basin in Mexico, which has faced extreme pressure on its water resources because of many factors, and can therefore be considered a closed system. The Lerma–Chapala Basin has been analyzed

Box 4 The Lerma-Chapala Basin in Mexico. From Hidalgo J and Pen˜a H (2009) Turning water stress into water management success: Experience in the Lerma–Chapala river basin. In: Lenton R and Muller M (eds.) Integrated Water Resources Management in Practice: Better Water Management for Development. London and Sterling, VA: Earthscan. Mexico’s Lerma–Chapala basin is one of world’s most over-committed basins, where population growth and industrial and agricultural development have led to a serious imbalance between water availability and water use. It is a classic example of a closed basin, with no natural outflow. The Lerma–Chapala basin is part of the hydrological system formed by the rivers Lerma and Santiago and covers part of the states of Guanajuato, Jalisco, Mexico, Michoaca´n, and Queretaro in central Mexico. Six large cities with more than 1 million inhabitants compete for water in and outside of the basin, and several large industrial corridors lie in the basin and discharge a high concentration of pollutants to the river and water bodies. Over 50% of the land in the basin is used for agriculture, of which a significant amount is irrigated. Water pollution and soil degradation are serious problems, as evidenced by declining soil fertility and increasing soil erosion. The current imbalance in water use began to develop in the 1940s, driven by population growth, heavy construction of water infrastructure, significant industrialization, and large increases in irrigated agriculture. As a result of intensive water use in the middle and lower reaches of the Lerma River, water stopped flowing naturally from Lake Chapala into the Santiago River in the early 1980s. The shrinking volumes of water in Lake Chapala – which dropped 90% in the two decades from 1981 to 2001 – clearly indicate the growth of water demand in the upper basin. While drought undoubtedly contributed to this drop, the main problem appears to be that between the 1940s and 1980s when the federal water authority provided too many water concessions, which led to overexploitation of water sources and a severe water imbalance. Indeed, during the 1990s, less water flowed into Lake Chapala than flowed out through evaporation and water withdrawals, yielding an annual mean deficit of 400 million ha m. To make matters worse, untreated wastewater discharges have degraded the lake’s water quality. To address this situation, the federal government first established the Lerma–Chapala Basin Regional Management, which in the 1980s made efforts to collect more information, improve water plans, define better institutional roles, and involve basin stakeholders in decision making. This led to the first coordination agreement between the federal government and the governments of the five basin states in 1989. A Control and Evaluation Advisory Council (CEAC), a technical working group, designed new rules for surface water reallocation that were approved by consensus by the main stakeholders in 1991. At the same time, a first stage of a water-treatment program was initiated. The CEAC was transformed into the Lerma–Chapala Basin Council (LCBC) in 1993 – the first basin council in Mexico. In 2004, a revised National Water Law strengthened the basin councils, giving them more responsibilities and an organizational structure, and also transformed the river-basin regional management offices into basin organizations, recognizing them as the regional water authorities. At present, the Lerma–Chapala Basin Organization (LCBO) and the LCBC share responsibilities for basin-water governance. Within the context of these institutional changes, a range of instruments have been used over time to tackle the problems caused by water over-use in the basin, including a basin plan and a program of water-resources assessment and use. As a result, Lake Chapala is beginning to recover its natural level, and the Lerma River’s water quality is improving.

Box 5 Basin Planning in the Rio Colorado, Argentina. From Major DC and Lenton RL (1979) Applied Water Resource Systems Planning. Englewood Cliffs: Prentice Hall. The Rı´o Colorado is a relatively small river in southern Argentina that rises near Argentina’s border with Chile, and flows for some 1100 km through a largely arid and sparsely populated area to the Atlantic Ocean, with only minor tributaries. Irrigation is the principal use of water in the basin, although power generation is also important to widely dispersed rural communities. While at present the river waters are used only sparingly, principally for irrigation and hydroelectric power, over the years there have been many proposals for expansion of irrigation and hydroelectric power and for water exports to wine-growing areas north of the basin. The Colorado basin comprises five provinces – Mendoza, Rio Negro, Neuque´n, La Pampa, and Buenos Aires – with different interests in the use of the river’s waters. In the early 1970s, on the initiative of the Argentine government and the five riverine provinces, the Rio Colorado was the site of a pioneering scientific study that aimed to provide a scientific basis for reaching agreement on the allocation of water from the river. The study involved the first complete use of innovative new mathematical modeling techniques for river-basin planning, and indeed the first large-scale application of the approach outlined by the Harvard Water Program (Maass et al., 1962). The study focused on the analysis of alternative development plans, with multiple objectives, with a strong emphasis on the use of mathematical models to analyze such alternatives. As described in Major and Lenton (1979), the study presented a phased set of investment possibilities to provincial and national authorities, including an initial stage of development focused on irrigation, with a reservoir site in the lower basin that has since been developed; a second stage with complementary power; and a third stage with essentially full development of the basin. Importantly, the study formed the basis for a successful 1977 accord on water, energy, and irrigation among the five provinces and the central government of Argentina that has endured for more than 30 years. Since water availability in the Rio Colorado basin (both today and at the time of the study) significantly exceeds requirements, the study’s conclusions and thus the agreement to which it led may not be resilient in the face of possible future changes in water demand and supply. Global changes in agricultural trade, for example, may increase demands for the agricultural and livestock outputs of the Rio Colorado, and increased economic development could lead to increased demand for hydroelectric power. On the supply side, river flows could well decrease as a result of changes in rainfall patterns and the melting of the Andean glaciers. Over time, therefore, increased pressure on the basin’s water resources could force decision makers to consider the prospects for water reallocation and/or water-use efficiency improvements.

Integrated Water Resources Management

extensively, including by Hidalgo and Pen˜a (2009), Dau and Aparicio (2006), Guitro´n et al. (2003), and Wester (2008). Further information can be obtained from Comisio´n Nacional del Agua, 2001. Box 5 focuses on the Rio Colorado Basin in Argentina, an open basin where pressures on the basin’s water resources are fairly low and in which there is still room for meeting additional water needs without reallocating water from existing users.

1.01.6 IWRM at the National Level: Policies and Governance The key feature of water-resources management at the national level is that it focuses not on the physical management of water itself but on the overall governance of the resource and the policies and institutions that facilitate and support management at other levels. Management at the national level therefore deals with strengthening the enabling environment, through measures such as establishing water-resources policies and strategies, improving water legislation, enhancing institutional roles, and strengthening processes for stakeholder participation. The allocation of financial resources (e.g., for investments in water infrastructure) is also clearly a vital task of management at the national level. National-level action can also support the development and use of better instruments for water management, such as those for water-resources assessment. Importantly, national policies and strategies must facilitate and support integrated action at lower levels. While much of what might be considered natural-system integration typically takes place at the basin level, much of what might be called human-system integration takes place at the national level. This includes ensuring that governmental policies take account of water-resource implications and considering water-resource policy within national economic and sectoral policies (GWP Technical Advisory Committee, 2000). Integrating water-resource policy with national economic and sectoral policies, and ensuring that decisions of

15

economic-sector actors are water sensitive, forms another vital set of policymaking at the national level. The need for strengthening water policy as a major issue for sustainable development was among the major accomplishments of the Rio Earth Summit, whose Agenda 21 included a full chapter on freshwater resources and which called for the application of integrated approaches to the development, management, and use of water resources. Attention to national water policies and governance has increased significantly in recent years, and countries the world over have worked to find ways to strengthen their water policies to advance overall sustainable development goals. In Chile, for example, national water policy has played a major role in the country’s overall development since the 1970s. This experience, which has been well documented by Pen˜a et al. (2004), Bauer (2004), GWP Technical Committee (2006), and Pen˜a (2009), is summarized in Box 6. Chile’s example illustrates the importance of aligning water-resource planning and management strategies to overall economic development models. South Africa provides a further example of a country which made a major effort to align its water policies and strategies with national development goals, following its transition to democracy in 1994. This process has been described in Muller (2009), De Coning (2006), and DWAF (2004). In contrast to Chile, however, South Africa attempted to address social, environmental, and economic needs simultaneously, taking advantage of the opportunities to improve water management provided by the country’s broader political-change process (Muller, 2009). There has been a growing literature on the subject of policies and water governance. One landmark publication in this area is Rogers and Hall (2003), who define water governance as ‘‘the range of political, social, economic and administrative systems that are in place to develop and manage water resources, and the delivery of water services, at different levels of society.’’ Recent literature has also examined what might be termed the triggers of water-policy change, and the steps that might be taken to accelerate and sustain positive change. GWP Technical Committee (2009) emphasizes that change is a negotiated

Box 6 National water policy and development: The case of Chile. From Pen˜a H (2009) Taking it one step at a time: Chile’s sequential, adaptive approach to achieving the three Es. In: Lenton R and Muller M (eds.) Integrated Water Resources Management in Practice: Better Water Management for Development. London and Sterling, VA: Earthscan. Chile stretches along a narrow strip of land 4200 km long, between the Andes and the Pacific Ocean. The Northern half of the country is arid, while in the Southern half, water is plentiful. Irrigated agriculture accounts for 85% of total consumptive water demand, while domestic uses account for 5%, and mining and industrial uses, around 10% (DGA, 1999). There is significant pressure on existing water resources in the Northern and Central regions of the country, where water requirements are high and water availability is low. At the end of the 1970s, Chile embarked on a new policy of opening up the economy to international trade, promoting the export of products in which the country was competitive. Importantly, nearly all of these export products involved significant water use as part of the production process and were located in areas where water is scarce; for example, copper production was located in the Atacama Desert. As a result, good water-resource management became a prerequisite for the success of the overall export model. Good management in turn required changes in public policies related to water. As described in Pen˜a et al. (2004), Bauer (2004), and GWP Technical Committee (2006), water policy in Chile has evolved considerably since that time, in line with the evolving overall governance structure. Initially, during Chile’s authoritarian period, a water law issued in 1981 with a strong market orientation called for private participation in water supply and sewerage services. A later irrigation law left irrigation initiatives to the private sector, with some partial financing through government subsidies. These steps led to increased efficiency in the use of water in various production processes, as well as increased private investment in Chile’s sewerage and water-supply sector. With the arrival of the democratic period, more attention was given to social, environmental, and regulatory considerations through a range of new laws. In particular, in 2005, the water law was reformed, which called for a more balanced consideration of economic, environmental, and social dimensions and reaffirmed the State’s regulatory role (Pen˜a et al., 2004).

16

Integrated Water Resources Management

Table 1

Factors contributing to a conducive environment for change

Particulars

Australia

Chile

Morocco

Namibia

South Africa

Sri Lanka

Water scarcity/conflicts Financial crisis Draughts/salinity Macroeconomic reforms Political reforms Social issues Donor pressures Internal/external agreements Institutional synergy/pressures

**

*

**

**

*

*

*

**

**

***

*

***

***

–

***

*

**

–

***

**

***

–

–

***

–

***

–

***

***

*

*

–

*

**

**

–

–

*

**

*

–

***

***

–

–

*

*

–

**

***

*

*

*

*

The number of *s signifies the relative importance of the factors in the context of each country. ‘‘–’’ means the aspect in question is ‘‘not applicable’’ or ‘‘not evaluated’’. From Saleth RM and Dinar A (2005) Water institutional reforms: Theory and practice. Water Policy 7: 1–19.

political process that is informed by history, public perception, development challenges, and social and economic context; while there are no universally applicable solutions, lessons from experience can provide practical guidance to those involved in change processes. Saleth and Dinar (2005) have identified a number of factors that contribute to a conducive environment for change, as depicted in Table 1. The GWP, the IWMI, and the Stockholm International Water Institute have compiled an extensive set of literature on policy-change processes, the link to which is available in the section titled ‘Relevant websites’. There has also been increasing attention to water in the forums that foster intergovernmental discussions and agreements on sustainable development issues, including in particular the UN Commission on Sustainable Development (UNCSD), established following the Earth Summit in 1992. Recognizing the important role of national policies and strategies in fostering and enabling the development of more integrated management approaches at all levels as called for at the Earth Summit, the UNCSD paid early attention to waterpolicy issues and laid the ground for the World Summit on Sustainable Development (WSSD) in Johannesburg in 2002, which called for all countries to prepare national IWRM and water-efficiency plans by the end of 2005. Subsequently, a significant number of countries embarked on such plans, both in response to the call and as part of national efforts to meet development goals and address specific water and development challenges. Special literature on the subject also emerged at that time, such as Catalyzing Change: A Handbook for Developing Integrated Water Resources Management (IWRM) and Water Efficiency Strategies (GWP Technical Committee, 2004). A growing number of global institutions have focused on fostering better policies and governance at the national level. These include in particular the GWP, which supports countries in the sustainable management of their water resources, and the World Water Council, which promotes international dialog via periodic world water forums and other means. Within the UN system, the United Nations Development Program (UNDP) has given priority to strengthening water governance at the national level, and has established a water-governance facility in Stockholm for this purpose. United Nations Environment Program (UNEP), GWP, and others have played a role in assisting countries to prepare IWRM plans and strategies.

Importantly, the Johannesburg call for the preparation of national IWRM and water-efficiency plans reflected the significant worldwide experience in the development and implementation of national water-resource management policies and strategies. As emphasized by Lenton and Muller (2009), such an experience strongly suggested that national waterresources planning and management must be linked to a country’s overall sustainable development strategy. At the same time, it reinforced the notion that local-level management needs a supportive national-policy framework and support at higher levels.

1.01.7 IWRM at the Transnational and Global Level: Information Sharing, Cooperation, and Technical and Financial Assistance The salient feature of water resources at the transnational and global level is that it focuses on the management of water resources, or a product or resource related to water resources, which crosses national borders. Management at this level thus involves stakeholders in more than one country. The words used to describe transnational and global water management therefore tend to include information sharing, cooperation, negotiation, and technical and financial assistance. Importantly, transnational management complements, rather than replaces, decision making at national and subnational levels. Individual project decisions, for example, are usually taken at national levels, while information gathering and assessment exercises may involve shared efforts. Water resources and products and resources related to water can cross national borders or have an impact in other countries in at least five ways. This translates into at least five different forms of transnational and global water-resources management: 1. Trans-boundary river-basin management. Trans-boundary river-basin management is perhaps the most common form of transnational-level management. It is also the form most often cited in the literature and discussed in international conferences. A very large number of the world’s rivers cross national borders, and many of these involve some degree of trans-boundary management or

Integrated Water Resources Management

2.

3.

4.

5.

cooperation. The Lower Mekong Basin, for example, has witnessed several decades of cooperation in management through the Mekong River Commission and its predecessor, the Mekong Committee (Bruhl and Waters, 2009). Trans-boundary management in the Mekong involves a combination of decision making at different levels. While the Mekong River Commission facilitates information gathering and assessment, individual project decisions are taken by the member countries. The key management challenge here is therefore to manage information sharing, cooperation, negotiation, and conflictresolution activities in ways that increase benefits and/or reduce costs to all parties. Trans-boundary aquifer management. Groundwater resources frequently cross borders in the form of transnational aquifers (see IGRAC, 2009, for a world map of transboundary aquifers). However, despite the large number of transnational aquifers around the world, they have received much less attention in the literature and in international deliberations than trans-boundary river basins (Bourne, 1992). The key management challenge here is generally to ensure the sustainability of resource use by all parties, with assessment and information sharing as major tools. Exports or imports of water resources. Sometimes water resources are exported from one country to another, usually for drinking-water purposes, in pipelines or by other means. Malaysia, for example, exports water resources to Singapore, which in fact depends on imported water for most of its supplies. Since transporting water in bulk over large distances is costly, such imports and exports of water only take place in particular circumstances where other supply options are not feasible or too costly. The key management challenge here usually goes significantly beyond water itself, requiring integration with foreign policy and trade considerations. International trade. Water resources are routinely exported from one country to another via trade in virtual water, which is the term coined to denote the water used in the production of a good or service. In Chile, water-resources management has played a predominant role in the country’s export-oriented economic growth. As noted earlier, the concept of virtual water is much discussed in the literature on trade versus domestic food production. The management challenge here is to ensure that decision making on trade issues takes full account of a country’s water-resource endowments and the impacts that imports or exports of virtual water might have on national water-resource availabilities and requirements. Knowledge management and technical and financial assistance. Managing knowledge and information about waterresources management, as well as technical, institutional, and financial support for water management, generally entails a significant degree of management at regional and global levels. Indeed, much of the focus of the growing number of international forums on water management, such as the UNCSD or the World Water Forums, is on technical assistance, knowledge sharing, capacity building, and financial support. The key challenge here is for countries to coordinate global and regional efforts for

17

knowledge management and technical and financial assistance in ways that best support national policies and strategies to improve water management at all levels. Over the last couple of decades, a number of global institutions have taken on aspects of this challenge. A range of UN organizations assists countries to manage water resources through technical support and capacity building, knowledge sharing, and global monitoring and analysis. The work of these agencies has recently been reinforced by a strengthening of UN-Water, the body which brings together the water activities of all the UN systems and which also oversees the World Water Development Report. Several of the global institutions mentioned earlier are involved in knowledge and information and/or technical, institutional, and financial support for water management, such as the GWP and the World Water Council. The UN Secretary General’s Advisory Board on Water and Sanitation (UNSGAB) helps highlight and mobilize action on global issues that need urgent attention.

1.01.8 IWRM as a Meta-Concept The above-discussed six sections provided an overview of integrated approaches to water-resources management at different levels and helped to unpack the meaning of IWRM at each level. This overview of integrated approaches to waterresources management at different levels suggests that it is best to view IWRM not as a single approach but as a wide range of approaches to manage water and related resources – a metaapproach or meta-concept, as it were, that both transcend the various levels of decision making and recognizes the importance of integrating decision making at each level. Viewing IWRM in this way reinforces the need to understand integration not only in horizontal terms but also in vertical terms, that is, across the different levels of decision making. Clearly, actions at one level should seek to reinforce and complement actions at other levels, within the generally agreed principle that decision making on water resources should be taken at the lowest appropriate level. Importantly, at each level, the approach to integrated management that applies has usually been referred to as an IWRM approach and given a name specifically tailored to that level, such as watershed management or basin management. Some researchers have likewise suggested alternative names to give emphasis to particular attributes, such as adaptive water management (Pahl-Wostl, et al., 2005), which – drawing on concepts from ecosystem management – stresses the need for management approaches that increase adaptive capacity. As we have seen, the way an IWRM approach is given expression differs from level to level. The term ‘integrated’ therefore takes on many forms, depending on the level. Rather than being taken literally as requiring that everything needs to be connected to everything else, the term is best understood as an adjective used to reinforce the notion that intelligent and broad-based management is needed to achieve the goals of sustainable development. In other words, the term integrated is a symbol to describe an approach that goes well beyond integration as such and could be described using other

18

Integrated Water Resources Management

adjectives such as sound, intelligent, broad based or holistic, or systemic. In addition, integration in itself is a multidimensional concept, with the relative importance of each of these dimensions varying from level to level. For example, some aspects of integration within the natural system (e.g., green and blue water, surface and groundwater management) are more relevant at the watershed and basin levels, whereas some aspects of human-system integration (such as considering water-resource policy within national economic and sectoral policies) are more relevant at the national level. Water and development processes also take place at different spatial and temporal scales, which means that integration must also be understood as bridging different spatial and temporal scales, the latter being especially important in the context of climate change. Beyond these differences, however, there are at least four common elements that apply to sound management at all levels and that embody the essence of what an overall IWRM approach is all about: 1. The approach recognizes that water is both a social and an economic good and has multiple uses. 2. The approach seeks to balance multiple objectives that at their core relate invariably to economic efficiency, social equity, and environmental sustainability. 3. The approach entails a broad, holistic, and integrated perspective relevant to the given level of decision making. 4. The approach requires the appropriate involvement of users at the given level of decision making. The body of research and practice to date at all the different levels suggests that there are some elements common to the way in which the approach has evolved at each level as well. For example, while it is now recognized that water-resource management at all levels requires an appropriate combination of hard and soft components – that is, economically, environmentally, and socially sound infrastructure coupled with effective institutions and governance – at most levels, this was not necessarily the starting point. The early stages of watershed management, for example, emphasized engineering approaches to control soil erosion and thus downstream effects, while much of the initial literature on water management at the national level focused on governance and policy issues, at the same time neglecting the need for investments in infrastructural development. Gradually, however, a more balanced approach began to gain prominence at all levels. Similarly, while in the initial stages there was often a tendency toward packages of practices, with a more blueprint approach, in more recent years, a more pragmatic, sequenced, and contextdriven approach has prevailed. At all levels, calls for more integrated forms of management are emphasizing the need to view these as an approach rather than a formal methodology or prescription.

One force identified by Lenton and Muller (2009) was the technical and methodological advances in dealing with complex water systems that began in earnest with the Harvard Water Program carried out at Harvard University from 1955 to 1960. This large multidisciplinary program resulted in the publication of Design of Water Resources Systems: New Techniques for Relating Economic Objectives, Engineering Analysis, and Governmental Planning (Maass et al., 1962). While the program focused principally on developing planning and design methodologies for complex multipurpose water-resource systems at the river-basin level, the approach advanced by the Harvard Water Program combined economic and engineering analysis, drew on techniques of mathematical efficiency models and computer simulation for river systems, and emphasized the role of political processes in decision making. A second force identified by Lenton and Muller (2009) was the recognition of the impact of human activities on the natural environment as crystallized in the Bruntland Report (World Commission on Environment and Development, 1987), and the full expression of these ideas and concepts at the Rio Earth Summit and its far-reaching action document, Agenda 21. Indeed, Chapter 18 of Agenda 21 – the longest chapter in the document – already embodied the concept in its title, which was ‘Protection of the quality and supply of freshwater resources: Application of integrated approaches to the development, management and use of water resources.’ It made a strong case for more integrated approaches than heretofore had been the norm, stating unequivocally that ‘‘the widespread scarcity, gradual destruction and aggravated pollution of freshwater resources in many world regions, along with the progressive encroachment of incompatible activities, demand integrated water resources planning and management.’’ Viewing IWRM as a meta-concept, however, suggests a third and crucial force, which is the set of conceptual and practical advances in management at different levels that have evolved over the last 30 years. Importantly, these advances came not from purely academic centers, but rather from the world of practice and of practical policymaking, reinforced by the findings of field-based research efforts. They have included

•

•

• •

1.01.9 History and Evolution of the Concept of IWRM

• The discussion in the previous section highlights the fact that the concept of an IWRM approach has been an evolving one. Several forces have given impetus to this concept.

Advances in watershed management, led by practical efforts to manage watersheds in a variety of contexts across the globe, the documentation and evaluation of many of these initiatives by organizations such as the World Bank, and the field-based research efforts of institutions such as IWMI and the World Agroforestry Centre. Advances in basin management, led by practical efforts in a number of countries such as France that set up river-basin organizations, field-based research by initiatives such as the CA, and the knowledge-exchange efforts of INBO and others. Advances in AWM, in which the field-based research of IWMI has played a major role. Advances in WSS, fostered by the work of the World Bank’s Water and Sanitation Program and the Water Supply and Sanitation Collaborative Council. Advances in water policy and governance, spearheaded principally by innovative efforts at the national level but supported by organizations such as the GWP and its technical committee.

Integrated Water Resources Management

Importantly, while advances at each level focused on the management issues critical to that level, they all seemed to have some important common elements. In particular, the best practices advocated as appropriate at each level tended to balance multiple objectives – economic efficiency, social equity, and environmental sustainability. They tended to start from a broad, holistic, and integrated perspective relevant to the given level of decision making. In addition, they tended to find a way to involve users at the given level of decision making. Not surprisingly, these common elements are those now recognized as being fundamental to an IWRM approach. While these fairly distinct drivers can be identified, the concept and practice of IWRM evolved over time in a nonlinear way, as is so often the case. While it began with best practices, advances in analytical tools, and advances in specific areas, IWRM as such was formally adopted as a concept at the Rio Earth Summit in 1992.

1.01.10 Assessments and Critiques of the Concept of IWRM As indicated earlier, there is an extensive literature analyzing and critiquing integrated approaches to water-resources management at each of the different levels at which these approaches have evolved in the last several decades – in watershed management, for example, or AWM or basin organizations. In recent years, however, this literature has been supplemented by a significant set of publications on the concept of IWRM as a whole. Initially, much of this literature on IWRM approaches emanated from the GWP, which in 1998 began producing the GWP Technical Committee (TEC) Background Paper series, which address key conceptual issues related to water-resources management (see e.g., Falkenmark, 2003; GWP, 2003, 2008; Rees, 2002, 2006; Rees et al., 2008; and Rogers et al., 1998). Later, the GWP began producing shorter technical and policy briefs on a variety of aspects relating to IWRM, including several that were designed to support countries in their efforts to prepare IWRM and water-efficiency strategies or plans, as well as a set of publications associated with the GWP IWRM ToolBox. Importantly, a range of other authors and groups associated directly or indirectly with GWP also took on the task of explaining and further articulating the concept of IWRM. Examples of these publications include the IWRM tutorial published by CapNet (CapNet, 2008), which provides

19

a brief introductory tutorial of the basic principles of IWRM, and a recent book by Soncini-Sessa et al. (2007), which focuses on the tools available for integrated and participatory water-resources management. Particularly in the last 4 or 5 years, a growing number of researchers and practitioners have taken a harder look at the concept of IWRM, sparking a lively and continuing debate on the subject. Some of these works examine IWRM in a specific geographic context, while others analyze aspects of IWRM and its theoretical underpinnings or critique the IWRM approach and its practical application. Antao and Walkuski (2008, Literature Survey on IWRM, Global Water Partnership, unpublished document) have prepared a comprehensive bibliography on the subject. A comprehensive book edited by Warner (2007) examined multi-stakeholder platforms and their role in integrated management. One important set of publications analyzes aspects of IWRM, including its scientific and theoretical underpinnings. A special issue of the Journal of Contemporary Water Research and Education published in 2006 aimed to provide new insights into the widespread experience of IWRM (Hooper, 2006). In 2007, a special issue of The Geographical Journal focused on what the journal described as Integrated Water Management (IWM); many of the papers in this special issue emphasized that IWM is largely influenced by contextual conditions. Some recent literature (e.g., Pahl-Wostl et al., 2005; Pahl-Wostl and Sendzimir, 2005; and Timmerman et al., 2008) focuses on the concept of adaptive water-resources management, which as explained earlier emphasizes the need for more adaptive systems of management, drawing on ecosystem-management theory. Other recent literature looks at the practical aspects of IWRM. The CA (CA, 2008) notes that in developing countries, what is usually passed-off in the name of IWRM has tended to have a very narrow blueprint package focus – national water policy, a water law and regulatory framework, recognition of the river basin as the unit of planning and management, treating water as an economic good, and participatory management – that represents a big shift from current paradigms and that makes these IWRM initiatives ineffective or counterproductive. Biswas (2004), Biswas and Tortajada (2004), and Biswas (2008) see IWRM as a somewhat fixed approach and raise questions about whether such an approach is practical given the wide range of conditions under which the approach might be applied. While acknowledging some of the above concerns, Lenton and Muller (2009) examine a range of practical examples to

Box 7 Key lessons from IWRM in practice. From GWP Technical Committee (2009) Lessons from Integrated Water Resources Management in Practice, Policy Brief 9. Stockholm: Global Water Partnership. *

* * *

*

IWRM is not a one-size-fits-all prescription and cannot be applied as a checklist of actions. Pragmatic, sensibly sequenced institutional approaches that respond to contextual realities have the greatest chance of working in practice. Water-resource planning and management must be linked to a country’s overall sustainable development strategy and public administration framework. Water management must ensure that the interests of the diverse stakeholders who use and impact water resources are taken into account. Approaches to water-resources management will evolve as the pressures on the resource and social priorities change. The challenge is to support the development of institutions and infrastructure that can meet the challenges of new circumstances. While the river basin is an important and useful spatial scale at which to manage water, there are often circumstances where it is appropriate to work at smaller sub-basin scale or at a regional multi-basin level.

20

Integrated Water Resources Management

see how the principles embodied in the concept of IWRM have been applied at different scales, from very local experiences to reform at the national level and beyond, and give evidence of the positive results obtained. The lessons learned from looking at these examples have been summarized in GWP Technical Committee (2009), and reproduced in Box 7. Taken as a whole, the examples in ‘IWRM in practice’ suggest that, although IWRM has been criticized as a theory that is very difficult to put into practice, it is more like a practice around which it has been very difficult to build a good theory.

Acknowledgments The chapter draws on, and expands, the analytical framework used by the author and his colleague Mike Muller in Integrated Water Resources Management in Practice: Better Water Management for Development (Lenton and Muller, 2009), which greatly facilitated the preparation of this chapter and for which he is very grateful to Mike Muller. The author would also like to acknowledge, with thanks, the very many discussions on the IWRM approach held with colleagues on the Technical Committee of the Global Water Partnership from 2003 to 2006. These discussions, as well as the many background papers and briefs emanating from the Technical Committee during this period, have substantially contributed to the conceptual thinking and information contained in this chapter.

References Agarwal A and Narain S (1999) Community and household water management: The key to environmental regeneration and poverty alleviation. Presented at EU UNDP Conference. Brussels, February 1999. Barry B, Namara R, and Bahri A (2009) Better rural livelihoods through improved irrigation management: Office du Niger (Mali). In: Lenton R and Muller M (eds.) Integrated Water Resources Management in Practice: Better Water Management for Development, pp. 71–87. London and Sterling, VA: Earthscan. Bauer C (2004) Siren Song: Chilean Water Law as a Model for International Reform. Washington, DC: Resources for the Future. Biswas AK (2004) Integrated water resources management: A reassessment. Water International 29: 248--256. Biswas AK (2008) Integrated water resources: Is it working? Water Resources Development 24(1): 5--22. Biswas AK and Tortajada C (eds.) (2004) Appraising the Concept of Sustainable Development: Water Management and Related Environmental Challenges. Oxford: Oxford University Press. Bourne C (1992) The International Law Commission’s draft articles on the law of international watercourses: Principles and planned measures. Colorado Journal of International Environmental Law and Policy 3: 65--92. Bruehl H and Waters M (2009) Transboundary cooperation in action for integrated water resources management and development in the Lower Mekong Basin. In: Lenton R and Muller M (eds.) Integrated Water Resources Management in Practice: Better Water Management for Development, pp. 189–204. London and Sterling, VA: Earthscan. CA (Comprehensive Assessment of Water Management in Agriculture) (2007) Water for Food, Water for Life: A Comprehensive Assessment of Water Management in Agriculture. London: Earthscan; Colombo: International Water Management Institute. CA (Comprehensive Assessment of Water Management in Agriculture) (2008) Developing and managing river basins: The need for adaptive, multilevel, collaborative institutional arrangements, Issue Brief No. 12, International Water Management Institute and Global Water Partnership. CapNet (2008) IWRM Tutorial. http://www.archive.cap-net.org/iwrm_tutorial/ mainmenu.htm (accessed March 2010).

Comisio´n Nacional del Agua (2001) Programa Hı´drico de la cuenca Lerma-Santiago 2001–2006. Me´xico. CSE (Centre for Science and Environment) (1994) Partners in prosperity. Down to Earth, 15 February 1994. CSE (Centre for Science and Environment) (1998) Sukhomajri at the crossroads. Down to Earth, 15 December 1998. CSE (Centre for Science and Environment) (2002) Foisting failure Down to Earth, 31 August 2002. CSE (Centre for Science and Environment) (2007) Saga of two villages Down to Earth, 15 November 2007. Darghouth S, Ward C, Gambarelli G, Styger E, and Roux J (2008) Watershed Management Approaches, Policies, and Operations: Lessons for Scaling Up, Water Sector Board Discussion Paper Series, Paper No. 11. Washington, DC: The World Bank. Dau FE and Aparicio MJ (eds.) (2006) Acciones para la recuperacio´n ambiental de la cuenca Lerma-Chapala. In: Comisio´n Estatal de Agua y Saneamiento, Gobierno del Estado de Jalisco, 126 pp. Me´xico: Comisio´n Estatal de Agua y Saneamiento de Jalisco, Guadalajara. De Coning C (2006) Overview of the water policy process in South Africa. Water Policy 8: 505--528. de Fraiture C, Giordano M, and Liao Y (2008) Biofuels and implications for agricultural water use: Blue impacts of green energy. Water Policy 10(supplement 1): 67--81. Direccio´n General de Aguas (DGA) (1999) Repu´blica de Chile, Polı´tica Nacional de Recursos Hı´dricos. DWAF (2004) National Water Resource Strategy, 1st edn. Pretoria: Department of Water Affairs and Forestry. Falkenmark M (2003) Water Management and Ecosystems: Living with Change, TEC Background Papers No. 9. Stockholm: Global Water Partnership. Farrington J, Turton C, and James AJ (1999) Participatory Watershed Development: Challenges for the Twenty-First Century. New Delhi, India: Oxford University Press. Guitro´n A, Hidalgo J, Aparicio J, and Aldama A´ (2003) A water crisis management: The Lerma-Chapala basin case. In: Brebbia CA (ed.) Water Resources Management II, pp. 345--354. Southampton: WIT Press. GWP (Global Water Partnership) (2003) Integrated Water Resources Management Toolbox, Version 2. Stockholm: GWP Secretariat. GWP (Global Water Partnership) and INBO (International Network of Basin Organizations) (2009) A Handbook for Integrated Water Resource Management in Basins. GWP Technical Advisory Committee (2000) Integrated Water Resources Management, TAC Background Papers No. 4. Stockholm: Global Water Partnership. GWP Technical Committee (2004) Catalyzing Change: A Handbook for Developing Integrated Water Resources Management (IWRM) and Water Efficiency Strategies. Stockholm: Global Water Partnership. GWP Technical Committee (2006) Water and Sustainable Development: Lessons from Chile, Catalyzing Change Series Policy Brief 2. Stockholm: Global Water Partnership. GWP Technical Committee (2009) Lessons from Integrated Water Resources Management in Practice, Policy Brief 9. Stockholm: Global Water Partnership. Heathcote IW (1998) Integrated Watershed Management: Principles and Practices. p. 441. New York, NY: Wiley. Hellegers P, Zilberman D, Steduto P, and McCornick P (2008) Interactions between water, energy, food and environment: Evolving perspectives and policy issues. Water Policy 10(supplement 1): 1--10. Hidalgo J and Pen˜a H (2009) Turning water stress into water management success: Experience in the Lerma–Chapala river basin. In: Lenton R and Muller M (eds.) Integrated Water Resources Management in Practice: Better Water Management for Development, pp. 107–120. London and Sterling, VA: Earthscan. Hooper B (2005) Integrated River Basin Governance: Learning from International Experiences. London: IWA Publishing. Hooper B (2006) Integrated water resources management: Governance, best practice, and research challenges. Journal of Contemporary Water Research and Education 35: 1--7. IGRAC (International Groundwater Resources Assessment Centre) (2009) Transboundary Aquifers of the World Map – Update 2009. Joshi PK, Jha AK, Wani SP, Joshi L, and Shiyani RL (2005) Meta-analysis to assess impact of watershed programme and people’s participation. In: Comprehensive Assessment of Water Management in Agriculture, Research Report 8. Colombo: International Water Management Institute. Kerr J (2002) Sharing the benefits of watershed management in Sukhomajri, India. In: Pagiola S (ed.) Selling Forest Environmental Services: Market-Based Mechanisms for Conservation and Development. London: Earthscan. Khurana MR (2005) Common property resources, people’s participation and sustainable development: A study of Sukhomajri. Panjab University Research Journal (Arts) XXXII(18.2). April–October.

Integrated Water Resources Management

Lenton R and Muller M (2009) Integrated Water Resources Management in Practice: Better Water Management for Development. London and Sterling, VA: Earthscan. Lenton R and Walkuski C (2009) A watershed in watershed management: The Sukhomajri experience. In: Lenton R and Muller M (eds.) Integrated Water Resources Management in Practice: Better Water Management for Development. London and Sterling, VA: Earthscan. Lenton R, Wright A, and Lewis K (2005) Health, Dignity and Development: What Will It Take?, pp. 17–28. London: Earthscan. Maass A, Hufschmidt MM, Dorfman R, Thomas HA Jr, Marglin SA, and Fair GM (1962) Design of Water-Resource Systems; New Techniques for Relating Economic Objectives, Engineering Analysis, and Governmental Planning. Cambridge: Harvard University Press. Major DC and Lenton RL (1979) Applied Water Resource Systems Planning. Englewood Cliffs: Prentice Hall. Muller M (2009) Attempting to do it all: How a New South Africa has harnessed water to address its development challenges. In: Lenton R and Muller M (eds.) Integrated Water Resources Management in Practice: Better Water Management for Development, pp. 169–185. London and Sterling, VA: Earthscan. Pahl-Wostl C, Downing T, Kabat P, et al. (2005) Transition to Adaptive Water Management: The NeWater Project. NeWater Working Paper 1, Institute of Environmental Systems Research, University of Osnabru¨ck. Pahl-Wostl C and Sendzimir J (2005) The Relationship between IWRM and Adaptive Water Management. NeWater Working Paper 3. Institute of Environmental Systems Research, University of Osnabru¨ck. Pen˜a H (2009) Taking it one step at a time: Chile’s sequential, adaptive approach to achieving the three Es. In: Lenton R and Muller M (eds.) Integrated Water Resources Management in Practice: Better Water Management for Development, pp. 153–168. London and Sterling, VA: Earthscan. Pen˜a H, Luraschi M, and Valenzuela S (2004) Water, Development and Public Policies. South American Technical Advisory Committee (SAMTAC), Economic Commission for Latin America and the Caribbean (ECLAC) and Global Water Partnership (GWP). Rees JA (2002) Risk and Integrated Water Management, TEC (formerly TAC) Background Papers No. 6. Stockholm: Global Water Partnership. Rees JA (2006) Urban Water and Sanitation Services: An IWRM Approach, TEC Background Papers No. 11. Stockholm: Global Water Partnership. Rees JA, Winpenny J, and Hall AW (2008) Water Financing and Governance, TEC Background Papers No. 12. Stockholm: Global Water Partnership. Rogers P, Bhatia R, and Huber A (1998) Water as a Social and Economic Good: How to Put the Principle into Practice, TAC Background Papers No. 2. Stockholm: Global Water Partnership. Rogers P and Hall A (2003) Effective Water Governance, TEC Background Papers No. 7. Stockholm: Global Water Partnership. Rosegrant M, Cai X, and Cline S (2002) World Water and Food to 2025: Dealing with Scarcity. Washington, DC: IFPRI. Saleth RM and Dinar A (2005) Water institutional reforms: Theory and practice. Water Policy 7: 1--19. Seckler D (1986) Institutionalism and agricultural development in India. Journal of Economic Issues XX(4). December.

21

Shah T (2009) Taming the Anarchy: Groundwater Governance in South Asia. Washington, DC: Resources for the Future. Colombo: International Water Management Institute. Sharma BR, Samra JS, Scott CA, and Wani SP (eds.) (2005) Watershed Management Challenges: Improving Productivity, Resources, and Livelihoods. New Delhi, India: International Water Management Institute. Soncini-Sessa R, Cellina F, Pianosi F, and Weber E (2007) Integrated and Participatory Water Resources Management – Practice, Volume 1b. Developments in Integrated Environmental Assessment Series. Amsterdam: Elsevier. Timmerman JG, Pahl-Wostl C, and Moltgen J (2008) The Adaptiveness of IWRM: Analysing European IWRM Research. London: IWA Publishing. UNCED (United Nations Conference on Environment and Development) (1992) Agenda 21, Report of the United Nations Conference on Environment and Development. http://www.un.org/esa/sustdev/documents/agenda21 (accessed March 2010). Warner J (ed.) (2007) Multi-Stakeholder Platforms for Integrated Water Management Studies in Environmental Policy and Practice. Aldershot, UK: Ashgate Publishing. Wester P (2008) Shedding the Waters: Institutional Change and Water Control in the Lerma-Chapala Basin, Mexico. PhD Dissertation, Wageningen University, Wageningen, The Netherlands. World Bank (2003) Tunisia Northwest Mountain Areas Development Project Performance Assessment Report. Washington, DC: World Bank. World Bank (2004a) Peru Sierra Project Implementation Completion and Results Report. Washington, DC: World Bank. World Bank (2004b) Tajikistan Community Agriculture and Watershed Management Project, Project Appraisal Document. Washington, DC: World Bank. World Bank (2004c) Turkey Eastern Anatolia Project, Project Performance Assessment Report. Washington, DC: World Bank. World Bank (2005) China Loess II Project Implementation Completion and Results Report. Washington, DC: World Bank. World Bank (2006) Water Management in Agriculture: Ten Years of World Bank Assistance, 1994–2004. Washington, DC: Independent Evaluation Group (IEG)/The World Bank. World Commission on Environment and Development (1987) Our Common Future (Brundtland Report). Oxford: Oxford University Press. WSSD (World Summit on Sustainable Development) (2002) Johannesburg Plan of Implementation. http://www.un.org/esa/sustdev/documents/WSSD_POI_PD/ English/POIToc.htm (accessed March 2010).

Relevant Websites http://www.rlc.fao.org Food and Agricultural Organization of the United Nations; Regional Office for Latin America and the Caribbean; Network on Watersheds Management. http://www.gwpforum.org Global Water Partnership. http://rupes.worldagroforestry.org Rewards for, Use of and shared investment in Pro-poor Environmental Services.

1.02 Governing Water: Institutions, Property Rights, and Sustainability E Schlager and C Bauer, The University of Arizona, Tucson, AZ, USA & 2011 Elsevier B.V. All rights reserved.

1.02.1 1.02.2 1.02.3 1.02.3.1 1.02.3.2 1.02.4 1.02.5 References

Introduction International Organizations and Water Policy Debate Governing Water from the Ground Up Local Communities, Property Rights, and Water Linking Water Uses and Administration across Multiple Scales and Jurisdictions Courts: Hiding in Plain View Conclusion: Reconceptualizing Water Governance

1.02.1 Introduction What is water governance and why is it included in this treatise? Water governance is one of those elastic terms that has something for everyone – everyone agrees that it is important and many disagree about what it means (‘institutions’ is another example). In a general way, governance refers to how people make decisions and govern themselves, whether in organizations or at the larger scale of societies. Thus, governance is about social, political, and economic processes and how they interact over time. This of course includes the realm of formal governmental institutions; however, it goes beyond this realm as well to include other aspects of social and political life. Is governance any different from politics? It depends on what one thinks politics means. They are not easily distinguishable, especially since both terms tend to be used with a great deal of abstraction and generality. In much international debate about water policy, water governance has become a black box, where we put important but complicated issues that we are not sure how to think about or talk about. Some of these issues are so deeply political that they are dangerous to discuss in public, and governance is soothing and blurs the edges. Our goal in this chapter is to open the lid of the black box of water governance and look more closely at what is inside. Where is the basic political economy, the distribution of wealth, and power in society? Who gains and who loses in water governance? What are people talking about when they talk about governance? What analytical tools can we bring to bear that would allow us to flesh out and make sense of governance?

1.02.2 International Organizations and Water Policy Debate We begin by reviewing the positions on water governance of several prominent international organizations. A good place to start is Peter Rogers and Alan Hall’s paper on water governance for the Global Water Partnership, or GWP (Rogers and Hall, 2003; building on Rogers’ work for the Inter-American Development Bank (Rogers, 2002)). The GWP is an international organization that was established in 1996 to promote integrated water resources management (IWRM) around the

23 23 25 25 27 30 31 32

world. IWRM is an international catch phrase that refers to the idea that water should be managed in a holistic, comprehensive, and multidisciplinary way – a way that does justice to the hydrologic cycle. This usually means focusing on the relationships between water quality and quantity, between surface water and groundwater, integrating across different water-using sectors at the level of river basins. Since the GWP aims to be the official, mainstream voice of IWRM around the world, its position on water governance is worth looking at (Bauer, 2004; Conca, 2006; GWP, 2000b). According to the GWP, ‘‘IWRM is a process which promotes the coordinated development and management of water, land and related resources, in order to maximize the resultant economic and social welfare in a equitable manner without compromising the sustainability of vital ecosystems’’ (GWP, 2000a: 22). Rogers and Hall begin by underlining the intensely political nature of allocating water, and then state that ‘‘governance is about effectively implementing politically achieved allocations’’ (Rogers and Hall, 2003: 4). Governance also ‘‘broadly embraces the formal and informal institutions by which authority is exercised’’ (Rogers and Hall, 2003: 7). They quote contemporary definitions of governance by the United Nations Development Program and the GWP that are breathtaking in their sweep: Governance is the exercise of economic, political and administrative authority to manage a country’s affairs at all levelsyit comprises the mechanisms, processes and institutions through which citizens and groups articulate their interests, exercise their legal rights, meet their obligations and mediate their differences. (UNDP, 2001)

Water governance refers to the range of political, social, economic, and administrative systems that are in place to develop and manage water resources, and the delivery of water services, at different levels of society. (GWP, 2002)

On such a wide-open playing field, Rogers and Hall run through a series of major issues, approaches, and schools of thought in political science and political and social theory. Examples include the incentives for legislators’ behavior, the relationship between the state and civil society, and the tension between markets and centralized hierarchies. Where they focus on water governance in particular, they emphasize the legal foundations and different forms of property rights and

23

24

Governing Water: Institutions, Property Rights, and Sustainability

institutional arrangements. They argue that water governance is affected by factors both internal and external to the water sector itself. Finally, they make the link from water governance to the need for water policy reforms and IWRM, along the lines expressed by the GWP mission. The GWP has published other documents that have shaped the international water policy debate. In its framing paper for the 2nd World Water Forum in 2000, the GWP famously declared ‘‘The water crisis is mainly a crisis of governance.’’ Governance here means conflict resolution: ‘‘The present threat to water security lies in the failure of societies to respond to the challenge of reconciling the various needs for and uses of watery. And governance lies at the center of the tension and delicate balance between different water uses and their management’’ (GWP (2000b: 23), emphasis in original; see also Cosgrove and Rijsberman (2000)). From this angle, that is, to deal with conflict, the GWP argues for a series of policy reforms to improve water governance. One key reform is to put IWRM into practice, although here too the GWP underlines the political nature of ‘‘integrating the seemingly incompatible goals, beliefs, interests, and knowledge of every water user’’ (GWP, 2000b: 25). Another key reform is to strengthen institutions and management, which will clearly require the organization and exercise of political power. Two other reforms to improve governance are to promote transparency and participation and to take a more economic approach to valuing water and designing price incentives. In short, the GWP position on water governance revolves around recognizing political conflicts and developing the capacity – that is, power – to handle them. As befits an international organization, the GWP documents are silent or vague about any specific political alliance or opposition in a given country (GWP, 2000a). In the year 2000, the other landmark event in global water governance (besides the 2nd World Water Forum) was the final report of the World Commission on Dams, or WCD (World Commission on Dams, 2000). The WCD was an unusual organization in many ways, a pioneering example of building multi-stakeholder and international consensus about complex and conflictive issues. The story and process of the WCD are as significant as the conclusions. The WCD was created through negotiations between the World Bank, several national governments, and a group of international environmental nongovernmental organizations (NGOs), and its task was to evaluate the world experience of large dams. The commissioners, once named, had to hire staff and decide how to carry out their tasks, including how to build legitimacy and authority for their results and recommendations. The Commission itself had no regulatory power and it was dissolved after disseminating its final report (the details are well delineated in Dubash et al. (2001) and Conca (2006)). The WCD report looked at water governance as a matter of social equity and justice above all: conflict is understood as driven by who wins and who loses. The report’s broad scope is announced in the second sentence: ‘‘The debate about dams is a debate about the very meaning, purpose and pathways for achieving development.’’ The WCD proposes a new framework for decision making, describing an approach based on recognizing different rights and assessing different risks involved with dams. This rights-and-risks approach aimed to broaden

the range of who counted as legitimate stakeholders in negotiating problems about dams, as well as broadening the issues on the agenda. The WCD argument is in part a moral argument. The Commission’s concrete impacts remain uncertain. Let us conclude this brief review with the World Bank. In the World Bank’s last two major documents about water policy, in 1993 and 2004, the authors use the term management instead of governance, but they are writing about the same issues. Both documents use the language of contemporary IWRM, while adding or strengthening the promarket stamp that has come with the World Bank in recent decades (e.g., World Bank (2004) adopts the GWP image of the comb to illustrate IWRM: the teeth of the comb represent different water-using sectors and the handle holds them together). However, the World Bank’s position on water politics and political economy has changed since the early 1990s, with significant implications for water governance (even if the World Bank calls it management). The World Bank’s 1993 Water Policy Paper had something for almost everyone: a great degree of emphasis on markets, privatization, and pricing, coupled with arguments in favor of strong government regulation and strengthened institutional arrangements (World Bank, 1993: 40). The paper spoke in grand generalities about IWRM and a so-called comprehensive analytical framework for resolving water problems, but the concrete meaning was vague and the political tone was muted. A decade later, in contrast, the World Bank’s key water experts were reasserting their authority and perspective after years of defending the World Bank from outside criticism. The 2004 Water Resources Sector Strategy focused on the implementation of the World Bank’s ideas and policies, which means a more down-to-earth and pragmatic approach to how reforms play out in the real world. The key is recognizing that water resources management is intensely political and that reform requires the articulation of prioritized, sequenced, practical and patient interventions. To be a more effective partner, the Bank must be prepared to back reformers and to pay more explicit attention in design and implementation to the political economy of reform. (World Bank, 2004: 3, emphasis added)

What does that mean? Later in the document the ‘‘political economy of water management and reform’’ is described as placing ‘‘particular emphasis on the distribution of benefits and costs and on the incentives that encourage or constrain more productive and sustainable resource use.’’ (World Bank, 2004: 13, emphasis added) This bare-knuckle approach means that ‘‘the World Bank will re-engage with high reward/high risk hydraulic infrastructure,’’ that is, dams (World Bank, 2004: 3). The World Bank’s maneuver is a notable dodge: having invoked political economy as the distribution of benefits and costs, the World Bank reframes the debate in terms of risks and rewards and then is silent about how the risks and rewards are distributed. If we want to know, we will have to find out for ourselves. The challenge for researchers is to identify specific reformers whom the World Bank has backed, and specific examples of water reforms, and then analyze the distribution of benefits and costs: Who gained and who lost? Also, how was that related to the reforms’ design and implementation?

Governing Water: Institutions, Property Rights, and Sustainability

The key international actors have recognized the centrality of politics in water governance, but have avoided tackling the more difficult issues of who participates in making collective decisions, the types of authority participants have to address problems, issues, and conflicts, and how benefits and burdens are distributed among people. While IWRM is often pointed to as the model for governance, it is also largely content free (Conca, 2006). Furthermore, IWRM possesses a distinct topdown bias (Kemper et al., 2007; Carlsson and Berkes, 2005). It assumes that a central government shares its power to make and enforce decisions with other, lower-level governments, and that civil society – in whatever form that takes: water user associations, nonprofit organizations, etc. – is invited into the decision-making processes. In the following section, we propose an alternative starting point, water users, and a form of policy analysis that attempts to diagnose the politics of specific water settings before prescribing institutional reforms.

1.02.3 Governing Water from the Ground Up 1.02.3.1 Local Communities, Property Rights, and Water A major research program in the social sciences focuses on the study of common pool resources, such as rivers, streams, and groundwater basins and the riparian and aquatic habitat such water sources support (Blomquist, 1992; Ostrom et al., 2002). One of the defining lessons from more than two decades of this research is the centrality of local-level resource users. Sustainable use of resources requires active participation of local water users in water governance. Fundamentally, common pool resources present two challenges to resource users and others engaged in management and governance – exclusion and use (Ostrom et al., 1994). Since it is challenging to realize and sustain exclusion, and since subtractability means resource units harvested by one user are not available for another user, common pool resources tie users together. Realizing exclusion and limiting the number of resource units subtracted from a common pool resource require the cooperation of most resource users. The benefits and costs resource users achieve in using a common resource depend on the actions of all. If sustainable use of common pool resources is to be realized, then both exclusion and subtractability must be carefully considered and managed. For instance, a common approach to address declining groundwater tables among states in the western US is to adopt well moratoria. Such a policy tool directly addresses exclusion: only those resource users with wells are allowed to access groundwater basins and no new resource users are allowed in. This policy, however, does not address problems raised by subtractability. If a well moratorium is not matched with limits on the amount of water each well may pump, a groundwater basin may be mined by existing users. Conversely, placing limits on how much water may be withdrawn from a groundwater basin, and by what means, without addressing exclusion, also exposes the basin to mining. For instance, the Arizona Groundwater Management Act, as originally written, limited water use by municipal water utilities to a specified number of gallons per person per day.

25

While gallons per person per day mandates directly addressed subtractability issues, they did not address exclusion issues. In effect, any person could move to a municipality governed by the groundwater act and through the municipal water provider access the groundwater basin. No limits were placed on the number of people that water utilities could serve, thus allowing access to anyone to the groundwater basin. Until relatively recently, a common belief among policy analysts and policymakers was that local-level resource users could not adequately address exclusion and subtractability issues, that is, they could not sustainably govern their shared water resources. Rather, the expectation was that most common pool resource settings were characterized by the rule of capture and, consequently, a race to harvest as many resource units as possible before they were captured by others (Hardin, 1968; Olson, 1965; Ostrom, 1990). The result was over-harvesting at best, and severe degradation or destruction of a common pool resource at worst, and the need for external intervention to save resource users from the race to harvest. It turns out, however, that the models predicting overuse and degradation were too simple, failing to adequately capture key features of many common pool resource settings, such as the ability of resource users to communicate with one another, to share experiences and knowledge of the resource; or the capacity of resource users to develop norms of reciprocity and sharing; or the values that resource users place on the resource and on the opportunity to have their children and grandchildren use the resource; or the experience resource users have in governing other areas of their lives that they can transfer to governing a common pool resource, and so on and so forth (Ostrom, 2007). For the past two decades, hundreds of cases of resource users sustainably managing water resources have been documented and published (Digital Library of the Commons, 2009). In a number of instances, local-level resource users have developed governing arrangements that perform better than government regulation and management. For instance, in an early study comparing farmer-managed irrigation systems and government-managed irrigation systems, Tang (1994) found that among irrigation systems that performed well, rules that govern water allocation and maintenance activities are better crafted to the specific conditions of each irrigation system. High-performing systems, which were more likely to be farmer managed, were associated with multiple rules that adequately limited access to the system and that fairly allocated water among the irrigators. In other words, farmers paid careful attention to exclusion and subtractability. Poorly performing irrigation systems, which were more likely to be government managed, were characterized by a single simple rule set or by no rules at all. Access to the irrigation systems was not adequately regulated and water allocation rules often did not work well. Monitoring and enforcement systems also differ between irrigator-owned systems and government-owned systems. Government-owned systems relied on full-time, paid guards. Farmer-owned systems relied on unpaid part-time guards (Tang, 1994: 241). However, guards in farmer-owned systems were much more likely to impose sanctions on rule breakers than were guards in government-owned systems. Furthermore, rule-following behavior was much more common in

26

Governing Water: Institutions, Property Rights, and Sustainability

farmer-owned systems than in government-owned systems, whether guards were present or not (Tang, 1994: 241). Farmers who participate in devising their own irrigation rules are much more likely to follow and actively monitor and enforce their rules. The means by which water users address exclusion and use issues is through devising, following, monitoring and enforcing institutional arrangements. By institutional arrangements we refer to strategies, norms, rules, and property rights systems (Ostrom et al., 1994; Ostrom, 2007). Property rights define relationships among people in relation to things, such as a common pool resource and the resource units it produces. For every right an individual holds, rules authorize or forbid specific actions in exercising the right (Schlager and Ostrom, 1992: 250). For instance, among the irrigators studied by Tang (1992, 1994), those irrigators in farmer-managed systems exercised their rights of access to and withdrawal of water from the irrigation system in substantially different ways than irrigators who were part of a government-managed system. Access rights to government-managed systems were operationalized by a single rule – ownership of land in the irrigation command area (Tang, 1994: 231). Access rights to farmer-managed systems were operationalized by a variety of rules, such as purchasing or leasing shares in the irrigation system, becoming a member in an irrigation organization, or paying fees for access to water (Tang, 1994: 231). Similarly, withdrawal rights in government-managed systems were also operationalized by a single rule, whereas farmer-managed systems were characterized by multiple rules. Furthermore, irrigators in farmer-managed systems also exercised rights of management and exclusion (Schlager and Ostrom, 1992: 251), that is, irrigators possessed the right to regulate how water would be allocated and how maintenance activities would take place (management rights), as well as to determinine who would hold rights of access and how such rights would be exercised (exclusion rights). Irrigators in government-managed systems did not exercise rights of management and exclusion; rather those rights were exercised by government officials. Rights of management and exclusion are collective-choicelevel rights (Schlager and Ostrom, 1992). Holders of such rights are authorized to develop rules that define how rights of access and withdrawal may be exercised. Who holds rights of management and exclusion and their relationship to the common pool resource affect the types of rules devised. Government officials, even as they exercise rights of management and exclusion, will not exercise rights of access and withdrawal. Since they are not directly subject to the irrigation rules they devise, they face few incentives to design rules that ensure the effective operation of irrigation systems. Instead, they may devise rules for other purposes, for instance, to increase their political support or lighten their administrative burdens. Conversely, because farmers in farmer-managed irrigation systems directly experience the consequences of their rule-making decisions, they confront incentives to craft the rules to the particular situation that they face (Tang, 1992). What common pool resource studies have demonstrated is the diversity and the multi-dimensionality of institutional arrangements devised by water users. In any given common pool resource setting, resource users are likely to hold different

bundles of property rights and exercise them in different ways, depending on the rules that specify different actions. In other words, legal pluralism, that is, multiple property rights systems, is likely to be the rule and not the exception across water settings. For instance, irrigation ditches and districts in eastern Colorado often control a portfolio of water that consists of several different types and sources, each governed by a different set of property rights and rules. Water from streams and rivers is governed differently from federal project water, and federal project water is governed differently from irrigation district reservoir water (Blomquist et al., 2004). Rivers and streams are governed by the prior appropriation doctrine, federal project water by the project’s enabling legislation, and district reservoirs by the rights and rules developed by the districts’ governing boards. In addition, the institutional arrangements devised by water users are multidimensional. Resource users give consideration to the transactions costs of implementing and administering property rights and rules, the allocation of benefits and costs realized by the rights and rules, how the rights and rules allocate risk, and whether the rights and rules are likely to dampen conflict among resource users. In other words, whether rules are effective is evaluated along multiple dimensions. A classic illustration of the multidimensionality of institutional arrangements involves the irrigation systems built and managed by farmers, located in Ilocos Norte, the Philippines, as reported by Coward (1979). Farmers gain entry into the systems by purchasing shares. A share entitles a household to a rich set of property rights – access, withdrawal, exclusion, and management. The rules the farmers have devised for exercising their property rights and engaging in irrigated farming are sophisticated. A share entitles a household to several plots of land dispersed along a canal so that all farmers have land located closer to the more desirable head of the canal and land located closer to the less desirable tail end of the canal. Not only do such rules dampen conflict that often emerges among farmers in different locations of an irrigation system, but they also allow the farmers to spread the risk of water shortages. During extremely dry periods, a portion of the irrigation system may be shut down, and farmers forbidden from irrigating plots located in the closed sections. However, farmers still have use of plots located in open sections. The shares allocate risks, benefits, and costs in a proportionate manner. Water and work obligations for maintaining the systems are allocated based on the proportion of land encompassed by shares. More land translates into not only more water but also greater work obligations. Finally, attention is paid to monitoring of water use. Water monitors are selected from among farmers and they are paid for their services by grants of land at the tail end of canals. Whether those plots receive water depends in part on how well the irrigators on the canal are monitored. The Filipino irrigation systems are just one example of many by which resource users have devised institutional arrangements that are self-reinforcing, that is, the decisions and trade-offs made along one dimension (e.g., allocating plots of land across an irrigation command area), supporting and reinforcing decisions made along another dimension (e.g., allocating the risk of water shortages), and so on. Many other such irrigation examples may be pointed to, such as Bali

Governing Water: Institutions, Property Rights, and Sustainability

(Lansing, 1991), Nepal (Lam, 1998), and Spain (Maass and Anderson, 1986). Cases such as these illustrate the institutional artisanship that resource users are capable of. As Ostrom (1990, 1999, 2007) has repeatedly noted, there is no single best set of property rights nor single rule set that support sustainable uses of water. Rather, long-enduring institutional arrangements appear to share some common features (Ostrom, 1990). According to Ostrom (1990), the most important feature is exclusion of nonowners. Exclusion is critical if water users are to commit to following a set of institutional arrangements over time and investing in modifying them as circumstances warrant. Water users must be assured that they will capture the benefits of their actions. Exclusion, however, while critical, is insufficient to ensure long-term commitment to property rights and rules. The institutional arrangements must be appropriate, crafted to the exigencies of the situation, and as the situation changes, the resource users must have the ability to modify the rules. Accountable monitors and graduated sanctioning maintain water users’ commitment to institutional arrangements. Finally, conflict-resolution mechanisms and at least a minimal recognition of the right to organize prevent these institutional arrangements from unraveling due to internal strife or invasion from external governmental authorities (Ostrom, 2007). While it is clear that water users are capable of designing relatively sophisticated and resilient institutional arrangements, there are just as many instances of water users failing to develop or sustain governing arrangements for many reasons. Resource users may not have the capacity, experience, or incentives to overcome collective action problems and devise governing arrangements that allow them to sustainably use their water resources (see Ostrom, 1999, 2000, 2007). Alternatively, certain types of resources or resource problems are extremely difficult for local resource users to address (Schlager and Blomquist, 2005; Schlager, 2005). For instance, groundwater basins present extraordinarily difficult challenges that resource users struggle with. Boundaries of basins are not easily identified, if at all, by resource users, neither is the structure of the resources. Groundwater pumpers, even if they are in close proximity of one another, may not know whether they are pumping from the same basin, or, if they are, whether they are pumping from the same aquifer within the basin. They are also unlikely to know whether and to what extent their basin is connected to other basins, or whether and in what ways their basin is connected to surface water sources. In other words, groundwater pumpers cannot easily sketch out the boundaries and dimensions of the resource or the boundaries of the resource users (Schlager and Blomquist, 2005). In addition, a groundwater basin, even if its boundaries and basic structure are well specified, is not just a bathtub filled with water. It consists of different pieces, such as layers of aquifers, connected surface water sources, recharge areas, flow paths and impediments, and so forth, structured, connected, and disconnected by varying geological formations. The pieces are characterized by different temporal and spatial scales. For instance, depending on soil characteristics, water percolating from the land surface will reach different portions of an aquifer at differing rates. Some water quality or water supply

27

impacts will be felt quickly and others gradually, some throughout an aquifer and others in localized zones. The uncertainties around boundaries of the groundwater basins and the identities of groundwater pumpers combined with the complexity of the resource create substantial barriers for groundwater users to organize and develop institutional arrangements that adequately address exclusion and subtractability issues. Encouraging resource users to collectively limit their harvesting activities will be difficult if they are not assured that they will reap the benefits of conservation (Schlager and Blomquist, 2005). Finally, even if resource users do develop governing arrangements, those arrangements may rest on values or center on goals in conflict with those of the society at large. Segments of communities may be excluded from a water resource or from participating in decision-making processes based on ethnicity, gender, or class (Ilahiane, 1999). Allocation rules may favor one group of users over another (Tang, 1992), and enforcement of rules may rest on questionable practices (Rose, 2002). Each of these issues, a lack of local self-governance, the inability to tackle particular types of problems, and the pursuit of questionable values or goals, draws attention to the larger context in which local governance operates. What types of resources, authorities, and institutional ties link resource users with governments and organizations at the regional, national, or international levels that may be drawn upon to support investment in local governance capacity, or assist local resource users in addressing regional problems, or allow higherlevel officials to intervene to address particularly inequitable outcomes? These are not easy questions to address. Local knowledge and contextual information critical for designing workable rules and policies are primarily centered among resource users. As a result, workable rules and policies require the active participation of resource users in governance. Empowering them, however, is fraught with challenges. Providing aid and assistance to local communities to bolster their conservation activities is not a straightforward process. Communities are not homogeneous political, social, and economic groupings that can be treated as a single unit (Agrawal and Gibson, 1999). External interventions, if not carefully crafted to the setting, may result in tragic unintended consequences.

1.02.3.2 Linking Water Uses and Administration across Multiple Scales and Jurisdictions How to characterize and think about the ties and linkages among governments, organizations, and groups, in order to gain traction for understanding and addressing practical problems is an ongoing struggle. The chasm between the dynamics of local-level, self-governing arrangements characterized by dense social networks and webs of norms, and central governments characterized by rent seeking, interest group politics, and corruption seems too vast to bridge. Conversely, the chasm between a relatively well-operating national government capable of exercising appropriate authority and deploying necessary resources and poorly designed local governments, or resource users who are unorganized and trapped in a race to harvest and who actively resist outside

28

Governing Water: Institutions, Property Rights, and Sustainability

intervention appears to be insurmountable. However, the divide must be bridged to realize the promise of watershed governance. How do watershed and river basin governing systems get assembled, and assembled in ways that allow for the balancing of different values, accounting for externalities across users and jurisdictions, and the resolution of conflict in productive ways? Carlsson and Berkes (2005) propose a productive way of engaging with complex governing systems that span multiple scales: to focus initially on the functional aspects of governance, rather than the structural design of the governing institutions and organizations. More specifically, they propose a number of steps for analyzing problem-solving settings that draw heavily on the Institutional Analysis and Development Framework (IAD) developed by Ostrom and others (Kiser and Ostrom, 1982; Ostrom et al., 1994; Ostrom, 2007). First, what types of problems do water resources exhibit, or what problems and conflicts are water users experiencing? Second, once the social–ecological system has been identified and problems specified, including conflicting conceptualizations of problems, the policy analyst should turn to identifying who the participants are, how they are organized, and how they relate to and affect the essential management tasks implicated by the problems. As Carlsson and Berkes (2005: 73) explain, ‘‘The logic is that we start from the ‘bottom’, in the activities themselves, and try to figure out how management is organized, if power is shared, if rights and duties are contracted out, and if State authorities have ‘a finger in the pie’.’’ Third, how the actors and participants are linked, the types of organizational ties and types of authority they exercise that make them relevant to others in the situation, needs to be identified. Fourth, only then can remedies be sketched out that include strengthening governing capacity as well as more specific policy changes that address identified problems. The analytic steps that Carlsson and Berkes (2005) spell out overlap and interact; they do not constitute a strict linear process, they do not promise a particular set of outcomes, nor do they result in a specific configuration of institutional arrangements. Furthermore, if engaged in carefully, they require the policy analyst to grapple with a wide range of governance levels and issues, from the constitutional choice level, where the terms and conditions of governance are specified, to the collective-choice level where decision-making authority is exercised, and implementation and monitoring occur, to the operational level where actions and activities around water resource use take place (see also Ostrom, 2007). The value of using a problem-solving approach that incorporates different levels of governance is illustrated by comparing across the states of USA and the ability of water users, local jurisdictions, and state governments to assemble workable watershed-scale governing arrangements. At a first glance, it appears that among states whose constitutions grant local jurisdictions and water users considerable autonomy and decision-making authority, the authority is used to develop a variety of new associations, organizations, and governments at multiple scales that solve different problems. In addition, in investing in new organizations and governments, resource users expand their capacity to govern at the watershed scale. For instance, Landre and Travis (1998) describe the development of new forms of governance around the Keuka Lake and

watershed in the Finger Lakes region of New York. The lake’s water quality was slowly deteriorating, threatening a highly valued resource. The lake was used for drinking water and recreation (swimming, fishing, and boating). An association of homeowners surrounding the lake spearheaded the effort to protect the lake. Drawing on expertise from a local university and watershed planning grants from the state, the association launched a research and educational campaign, and developed a forum for public officials from the dozen or so surrounding towns. It was the public officials who had the authority to develop and implement water quality regulations. New York state permits local jurisdictions to enter into memoranda of agreement (MOA) by which local governments can jointly govern a shared resource or address a common problem. The local jurisdictions used the MOA to establish common septic tank regulation that applied to all homes, especially those surrounding the lake, and a septic tank inspector, to ensure that septic tanks were properly installed and maintained, thereby protecting the lake from pollution. The state could have intervened to address the water quality problem if local jurisdictions failed to act; however, local residents and governments chose to develop their own arrangements that they are responsible for and that are accountable to them (Keuka Lake Association, 2009). Much the same type of process has emerged among a number of watersheds in southern California, another state that grants citizens and local jurisdictions considerable authority and discretion to govern their water resources. As Schlager and Blomquist (2008) describe, the San Gabriel Basin covers much of Los Angeles County, one of the most highly urbanized counties in the country. Over 100 local political jurisdictions are located in the basin. The basin ties together these multiple jurisdictions through the San Gabriel River, the Rio Hondo River, three interconnected groundwater basins, and one groundwater basin not hydrologically connected to the others (Blomquist, 1992). Several water resource management problems have arisen in the San Gabriel River watershed, owing to the combined effects of the region’s limited water supplies, its extensive agricultural and then urban development, and the hydrogeology of the watershed itself. Each of these problems has been multi-jurisdictional in scope. Water users responded to each by developing new institutional arrangements. The arrangements are fitted together through a system of interorganizational and intergovernmental relationships. For instance, one of the initial problems addressed by local jurisdictions was importing water to meet the needs of rapidly growing populations and industries. A number of jurisdictions participated in the formation of the Metropolitan Water District in the 1920s to import water from the Colorado River Basin. As additional jurisdictions sought membership in the Metropolitan Water District and access to imported water supplies, they banded together and formed water districts whose purpose was to bring imported water to their member jurisdictions. Later, these districts participated in solving groundwater overdraft problems. Districts and larger municipalities spearheaded efforts to adjudicate rights in groundwater. To ensure representation of interests not adequately covered by municipalities and districts, water associations that encompassed the major water users within each basin were

Governing Water: Institutions, Property Rights, and Sustainability

formed and participated in developing agreements for allocating groundwater. Later still, the municipalities, districts, associations, and water masters, who monitor water rights, were the foundation on which water quality issues were addressed. Initially, water masters were given the task of developing water quality monitoring systems; however, as the discovery of water quality problems began to mount, and the projects and resources needed to remediate water supplies grew, a water quality authority was created to carry out the task of remediation. Thus, problem-by-problem water users and local jurisdictions assembled a San Gabriel River Basin governance system, sometimes granting new authorities to existing governments to address new problems, and sometimes creating new associations and governments to address problems. Watershed governance in both New York and California is predicated on allowing local jurisdictions and governments the authority to devise their own institutional solutions to shared problems. However, if water users and local governments did not want to participate for a multitude of reasons – they did not believe that (1) there was a problem, or (2) they were involved in causing the problem, or (3) they should pay to resolve the problem – they could not easily opt out or free ride off of the efforts of others. For instance, New York State could have intervened and established water quality standards for Keuka Lake and imposed regulations for meeting those standards. In southern California, local water user associations and districts, and municipal water providers regularly used state courts to bring all parties to the negotiating table to address water quantity problems. The California Health Department and the US Environmental Protection Agency, in setting water quality standards and in addressing highly polluted sites through the Superfund Program, called attention to and required action in relation to water quality problems. In other words, it is not just local autonomy that is important, but local autonomy and the larger institutional environment in which it is embedded that matters. Can local jurisdictions hold one another accountable, do they have access to conflict resolution mechanisms, and can they easily avoid participating in collective action to address water quantity or quality problems that they helped create? These are important considerations in watershed governance. The experience of other states suggests that it is no easy matter to go back and fill in institutional gaps in order to encourage, support, and require local jurisdictions to account for the impacts of their actions on others. For instance, until very recently, Nebraska water law and administration separated surface water from groundwater, even if the two are hydrologically linked (Schlager and Blomquist, 2008). Surface water, grounded in the prior appropriation doctrine, is administered by a state agency. Groundwater, grounded in the beneficial use doctrine, is administered by local natural resources districts, governed by groundwater users. The state had no legal authority to require natural resources districts to manage groundwater to account for externalities or environmental values. Beginning in the 1970s, with the listing of endangered species along the Platte River in south central Nebraska, planning and development of large surface water projects ceased (Aiken, 1999). By the 1990s, existing hydropower/irrigation projects were threatened, as the Federal

29

Energy Regulatory Agency required the projects to account for their impacts on endangered species habitat to gain license renewal. Part of the problem the projects faced was the effects of groundwater pumping on their surface water supplies. In addition, Kansas, a downstream neighbor of Nebraska, filed a suit before the US Supreme Court, claiming that Nebraska was failing to abide by its legal obligations to allow Kansas’ share of the Republican River to flow through to Kansas because of groundwater pumping. Finally, Nebraska surface water rights holders began filing lawsuits against groundwater pumpers to try protect surface water flows. In each instance, the state of Nebraska was helpless to respond because it had no authority to require natural resources districts to regulate groundwater pumping. After two decades of conflict, that included multiple lawsuits, several governorappointed commissions, and numerous public hearings and meetings, a law was adopted that allowed the state water agency to declare river basins overappropriated. Such a designation would immediately trigger a well moratorium and require the natural resources districts within the overappropriated basin to develop groundwater regulations sufficient to bring the basin to a fully appropriated status. This is just the first, and most critical, step in developing ties and linkages among multiple governments and jurisdictions within Nebraska that will allow the coordination of ground and surface water and, which, in turn, will allow the state to meet its water obligations to surrounding states. Nebraska state law granted groundwater users significant authority to govern groundwater supplies through natural resource districts, but provided for few accountability mechanisms. The above cases highlight the value of taking a problemsolving approach to analyzing water resources issues and taking into account the ties and relations among citizens, organizations, and governments at multiple scales. As Young (2002: 266) argues, ‘‘The extent to which specific environmental or resource regimes yield outcomes that are sustainable – much less efficient or equitable – is a function not only of the allocation of tasks between or among institutions operating at different levels of social organization but also of crossscale interactions among distinct institutional arrangements.’’ Young (2002) provides a useful set of analytic concepts to assist policy analysts in assessing the institutional ties among water actors. The three analytic concepts are competence, compatibility, and capacity. Competence refers to the political and legal authority to engage in and implement commitments. For instance, in both New York and California, local jurisdictions possess the legal authority to engage in binding agreements with each other to regulate the use of a water source. However, the state of Nebraska did not have the competence to implement its agreement to provide Kansas with a designated amount of surface water. It did not possess such competence because it did not have the authority to regulate groundwater pumping. Compatibility refers to the congruence of institutional arrangements among governments operating in a given resource setting, such as a river basin. In the San Gabriel Basin, governing arrangements for each of the linked groundwater basins are now largely compatible and in line with California water doctrines; however, it took decades for those governing arrangements to be developed, with the downstream basin,

30

Governing Water: Institutions, Property Rights, and Sustainability

which experienced problems first, taking the initial steps to regulate groundwater. As groundwater users in each of the upstream basins experienced problems and came to understand their own hydrologic ties, they too developed governing arrangements. The processes were adversarial and conflictual and took several decades to complete. In the end, each basin has its own groundwater rights and rules crafted to each setting, and the institutional arrangements are aligned so that complementarities and externalities are accounted for (Blomquist, 1992; Schlager and Blomquist, 2008). Capacity refers to the material resources and social capital needed to devise, implement, administer, and monitor rules and property rights systems. For instance, one of the challenges confronting the municipalities surrounding Keuka Lake was to provide sufficient resources to facilitate joint monitoring and enforcement of the septic tank rules adopted (Landre and Travis, 1998). A facilitator brokered an agreement that the costs would be shared equally among the jurisdictions and not be proportionate to the proximity of a jurisdiction to the lake. Each jurisdiction agreed to invest in the capacity of the MOA by making adequate resources available for its implementation. Competency, compatibility, and capacity further flesh out the problem-solving analytic approach that Carlsson and Berkes (2005) propose. One of the most important tasks in assessing a complex socio-ecological setting is identifying how the actors are linked, or not, and the quality of those linkages. Paying attention to the types and qualities of linkages also highlights the delicate balance that exists among different governments, associations, and organizations that constitute river basin or watershed governance. Just as it is challenging for regional and central governments to intervene and work with local communities of resource users without creating negative unintended consequences, so also is it difficult for actors to intervene in watershed governance systems to encourage, establish, and support productive linkages. However, what Carlsson and Berkes (2005), Young (2002), and many other scholars, who have attended to the governance of complex socio-ecological systems, have stressed is the importance of using a problem-centered approach for identifying capabilities and limitations of existing governance systems, and critically examining the competency, compatibility, and capacity of governments at different scales, rather than assuming that governments at specific levels are more competent or have greater capacity than governments at other levels (Larson, 2004; Ribot et al., 2006).

1.02.4 Courts: Hiding in Plain View As water users, water managers, and public officials attempt to link (or sever) ties among governments and organizations at different scales, they often contest the competency of a participant to engage in decision making; or they question the compatibility of different courses of action; or they seek ways to build the capacity of an organization, and in so doing they often turn to courts for assistance. Certainly, each of the US cases in the previous section include courts as an important actor in providing venues, shaping issues, and structuring solutions. Yet, in spite of talk about water governance revolving

around water conflicts and how to resolve them, the specific role of courts has been widely overlooked in both national and international contexts. In some countries, water experts recognize the practical importance of judicial decisions as features in the institutional landscape, but generally without thinking further about the courts’ significance. This is an unfortunate omission, because courts play a strategic and fundamental role in many countries’ political and economic systems. This is illustrated by a large academic literature, including comparative judicial politics (Jacob et al., 1996; Shapiro, 1981) and law and economics (e.g., Mercuro and Medema, 2006). The courts’ role is especially critical in situations of market-oriented policies and institutional frameworks. Market-driven governance aims to restrict state regulation, which requires strong judicial watchdogs. We return to this later in discussing the rule of law. Here, we intend to simply highlight a few key issues involving courts in the hope that people in water governance will be moved to investigate further. In the first place, courts and judges are one of the archetypal forms of conflict resolution. The judge sits apart and hears both sides of a dispute before deciding who wins and who loses. This is what political scientist Martin Shapiro has called the ‘‘logic of the triad in conflict resolution’’ (Shapiro, 1981: 1). Making that judicial decision means applying legal rules to specific fact situations, a process that requires reasoning, analysis, and interpretation. Particularly to people outside the legal profession, the judicial process is mysterious, and indeed how it works varies widely in different national and social contexts. In a classic book comparing different countries and legal traditions, Shapiro describes the conventional prototype of courts as growing out of conflict resolution. The prototype has four elements: ‘‘(1) an independent judge applying (2) preexisting legal norms after (3) adversary proceedings in order to achieve (4) a dichotomous decision in which one of the parties was assigned the legal right and the other found wrong’’ (Shapiro, 1981). Shapiro debunks each of these elements even in the context of conflict resolution, and he argues more broadly that courts also perform two other essential social functions: social control and lawmaking. The three functions often overlap. In the contexts of social control or lawmaking, however, the courts are farther away from their ‘‘basic social logic [and] perceived legitimacy.’’ Shapiro is worth quoting at length about the logic of courts: The basic social logic, or perceived legitimacy, of courts rests on the mutual consent of two persons in conflict to refer that conflict to a third for resolution. This basic logic is threatened by the substitution of office and law for mutual consent, both because one of the two parties may perceive the third as the ally of his enemy and because a third interest, that of the regime, is introducedy. When we move from courts as conflict resolvers to courts as social controllers, their social logic and their independence is even further undercut. For in this realm, while proceeding in the guise of triadic conflict resolver, courts clearly operate to impose outside interests on the parties. Finally, in the realm of judicial lawmaking, courts move furthest from their social logic and the conventional prototype because the rules they apply in the resolution of conflicts between two parties are neither directly consented to by the parties nor ‘preexisting.’ Instead, they are created by the third in the course of the conflict resolution itself. Thus, while the triadic mode of conflict resolution is nearly universal,

Governing Water: Institutions, Property Rights, and Sustainability

courts remain problematical in the sense that considerable tension invariably exists between their fundamental claims to legitimacy and their actual operations.’’ (Shapiro, 1981: 36–37, emphasis added)

The US is probably the most extreme example of a strong judicial role in water conflicts. The courts are major actors and decision makers in water policy and water rights and routinely combine the functions of conflict resolution, lawmaking, and social control. The courts are of course not the only important arena for addressing water conflicts, but they are a distinct arena with modes of reasoning and operation that are quite different from overtly political or economic approaches. Courts have been at the heart of one of the dominant slogans in international affairs in recent years: the rule of law. The idea of the rule of law shaped international debate about political and economic development in the 1980s and 1990s, as many countries went through historic processes of reform, generally toward democratization in the political sphere and toward markets in the economic sphere (Carothers, 1998; Dezalay and Garth, 2002; Thome, 2000; Trubek, 2006). This was approximately the same time that IWRM and sustainability came to dominate international debate about water. The rule of law means, in a nutshell, that laws apply to everyone, even powerful social actors and including government officials at all levels. Different views of the rule of law have put differing emphases on political versus economic issues. For many people, the rule of law refers to protection of human rights and democratic procedures of government. For others, it refers primarily to protection of economic and property rights from excessive government regulation. What all these views have in common is the goal of placing limits on the exercise of state power (Thompson, 1975; Whitlock, 2000). Moreover, all views agree that the ultimate watchdog is an independent judicial power, capable of challenging the legislative and executive powers of government. The courts are the foundation of the rule of law. The different political and economic aspects of the rule of law mean that courts are in a complicated position in many countries, and this bears directly on water conflicts. In political terms, the courts’ legitimacy is debatable because judges are not elected by the usual democratic processes and yet they sometimes overrule decisions made by popular sovereignty. In broader terms, social and economic as well as political, the courts are a critical arena for sorting through multiple values, rights, and interests. Many values are qualitatively different from each other and belong to different categories, and therefore cannot be quantitatively compared according to a single measure. Qualitative comparison is a task for legal and political institutions, and particularly courts in the case of conflicts. Law – whether made by legislators, executives, or judges – is vital to determining economic value, by creating the rules of the game and influencing how markets determine prices (Bromley, 1982; Commons, 1924; Whiteley et al., 2008).

1.02.5 Conclusion: Reconceptualizing Water Governance In an eye-catching headline ‘‘Stationarity is Dead: Whither Water Management?’’ a group of scientists argue that climate

31

change is leading to changing hydrologic cycles and no longer can engineers rely on stationarity as the foundation for managing river basins (Milly et al., 2008). Stationarity refers to ‘‘the idea that natural systems fluctuate within an unchanging envelope of variability’’ (Milly et al., 2008: 573). Instead, the scientists argue, ‘‘nonstationary, probabilistic models of relevant environmental variables’’ must be developed to optimize water management. Furthermore, considerable attention must be paid to the rapid flow of information between climate scientists and water managers to make the new modeling efforts relevant to policy and management. How to realize such a complete transformation of the foundation of the design and management of large water projects? Interestingly, the authors propose a program that reflects the spirit of the Harvard Water Program. The Harvard Water Program, begun in the 1950s, used newly emerging computer technology and large data sets to illustrate the value and the possibility of computer simulations to examine alternative choices among objectives (Reuss, 1992, 2003). It provided a tool, multi-objective planning, for explicitly incorporating and analyzing multiple values and goals in developing and managing large water projects and the river basins in which they are situated. To that point in time, engineering considerations and values largely drove the design, planning, and operations process for developing water projects (Reuss, 1992). In proposing a new Harvard Water Program, Milly et al., to their credit, realize that responding to changing hydrologic cycles is not just an engineering problem, rather it is a societal problem. How river basins should be governed in light of climate change will entail engaging many people, organizations, associations, and governments expressing and pursuing many values. What, then, would a contemporary Harvard Water Program look like in what has been called the epoch of watershed sustainability (Sabatier et al., 2006)? The epoch of watershed sustainability is characterized by features both unique to it and in contrast to earlier time periods. It is characterized by many different stakeholder groups (interest groups, government agencies, scientists, local resource users, and native peoples) actively engaged in face-to-face interactions searching for win–win solutions to complex socio-ecological problems, grounded in extensive and intensive information development processes. In contrast, the Harvard Water Program existed during a time, in the US at least, when one federal agency had primary jurisdiction in a watershed or riverbasin, ‘‘with other agencies and interest groups acting as supplicants’’ (Sabatier et al., 2006: 23). The primary agency was ‘‘principally concerned with fulfilling its statutory mandate’’ with limited consideration given to other values (Sabatier et al., 2006: 24). The revolutionary aspect of the Harvard Water Program was a new methodology developed by experts for experts that allowed for the consideration of a wider range of values in river basin and water project planning and management. For a new Harvard Water Program to be successful, it would have to develop credible science, salient for a wide range of stakeholders, in processes viewed by scientists and stakeholders alike as legitimate (Cash et al., 2003). The challenge would no longer be how to fit a broader range of values into the design and operation of models used to optimize

32

Governing Water: Institutions, Property Rights, and Sustainability

water development and management, as it was five decades ago, and as Milly et al. appear to envision it. Rather, the challenge would be how to design and fit organizations charged with the development of nonstationary, probabilistic hydrologic models that would be useful for water managers, policymakers, and citizens alike into complex, multiscale, intergovernmental, and organizational watershed governance systems. In other words, it would be a program that paid as much attention to institutional design (and all the disciplines that entails) as it would to the design of decision support systems (and all the disciplines that entails). IWRM and new modeling efforts around nonstationarity share much in common. Both recognize multiple values and the conflicts and politics that are likely to emerge as people fight, argue, contest, cooperate, and compete to ensure the realization of their cherished values. Rather than embracing such disorder, IWRM and new decision support models attempt to tidy things up by providing better information through models, or by bringing many interests and stakeholders to the table in search of common ground. While both approaches are useful, they fail to provide insight and guidance around water politics and governance as it unfolds in practice. What is instead needed and what we attempted to lay out in this chapter is an analytic approach for understanding complex social and ecological systems that recognizes a wide variety of institutional arrangements (including courts) and cross-scale linkages.

References Agrawal A and Gibson C (eds.) (1999) Communities and the Environment. New Brunswick, NJ: Rutgers University Press. Aiken JD (1999) Balancing endangered species protection and irrigation water rights: The Platte River cooperative agreement. Great Plains Natural Resources Journal 3: 119--158. Bauer C (2004) Siren Song: Chilean Water Law as a Model for International Reform. Washington, DC: RFF Press. Blomquist W (1992) Dividing the Waters: Governing Groundwater in Southern California. San Francisco, CA: ICS Press. Blomquist W, Schlager E, and Heikkila T (2004) Common Waters, Diverging Streams: Linking Institutions and Water Management in Arizona, California, and Colorado. Washington, DC: Resources for the Future. Bromley D (1982) Land and water problems: An institutional perspective. American Journal of Agricultural Economics 64(5): 834--844. Carlsson L and Berkes F (2005) Co-management: Concepts and methodological implications. Journal of Environmental Management 75: 65--76. Carothers T (1998) The rule of law revival. Foreign Affairs 77(2): 95--106. Cash D, Clark W, Alcock F, et al. (2003) Knowledge systems for sustainable development. Proceedings of the National Academy of Sciences of the United States of America 100: 8086--8091. Commons J (1924) Legal Foundations of Capitalism New York, NY: Macmillan. Conca K (2006) Governing Water: Contentious Transnational Politics and Global Institution Building. Cambridge, MA: MIT Press. Cosgrove W and Rijsberman F (2000) World Water Vision: Making Water Everybody’s Business. London: Earthscan. Coward EW (1979) Principles of social organization in an indigenous irrigation system. Human Organization 38: 28--36. Dezalay Y and Garth B (2002) Global Prescriptions: The Production, Exportation, and Importation of a New Legal Orthodoxy. Ann Arbor, MI: University of Michigan Press. Digital Library of the Commons (2009) Digital Library of the Commons Repository. http://dlc.dlib.indiana.edu/dlc (accessed April 2010). Dubash N, Dupar M, Kothari S, and Lissu T (2001) A Watershed in Governance? An Independent Assessment of the World Commission on Dams. World Resources Institute, Lokayan, and Lawyers Environmental Action Team.

GWP (Global Water Partnership) (2000a) Integrated Water Resources Management, TAC Background Paper No. 4. Stockholm: Global Water Partnership. GWP (Global Water Partnership) (2000b) Towards Water Security: A Framework for Action. Stockholm: Global Water Partnership. Hardin G (1968) The tragedy of the commons. Science 162: 1243--1248. Ilahiane H (1999) The ethnopolitics of irrigation management in the Ziz Oasis, Morocco. In: Agrawal A and Gibson C (eds.) Communities and the Environment: Ethnicity, Gender, and the State in Community-Based Conservation, pp. 89--110. New Brunswick, NJ: Rutgers University Press. Jacob H, Blankenburg E, Kritzer HM, Provine DM, and Sanders J (1996) Courts, Law, and Politics in Comparative Perspective. New Haven, CT: Yale University Press. Kemper KE, Blomquist W, and Dinar A (eds.) (2007) Integrated River Basin Management through Decentralization. Berlin: Springer. Keuka Lake Association (2009) About Keuka Lake Association. http:// www.keukalakeassoc.org/what/about_kla.php (accessed April 2010). Kiser L and Ostrom E (1982) The three worlds of action: A metatheoretical synthesis of institutional approaches. In: Ostrom E (ed.) Strategies of Political Inquiry, pp. 179--222. Beverly Hills, CA: Sage. Lam WF (1998) Governing Irrigation Systems in Nepal: Institutions, Infrastructure, and Collective Action. San Francisco, CA: ICS Press. Landre P and Travis L (1998) Collaborative watershed management in the Finger Lakes region, New York. Presented at the Conference of the International Association for the Study of Common Property. Vancouver, BC, 10–14 June 1998. http:// www.indiana.edu/˜iascp/Final/landre.pdf (accessed April 2010). Lansing JS (1991) Priests and Programmers: Technologies of Power in the Engineered Landscape of Bali. Princeton, NJ: Princeton University Press. Larson A (2004) Formal decentralization and the imperative of decentralization ‘from below’: A case study of natural resource management in Nicaragua. European Journal of Development Research 16(1): 55--70. Maass A and Anderson RL (1986) y and the Desert Shall Rejoice: Conflict, Growth, and Justice in Arid Environments. Malabar, FL: R.E. Krieger. Mercuro N and Medema S (2006) Economics and the Law, 2nd edn. Princeton, NJ: Princeton University Press. Milly PC, Betancourt J, Falkenmark M, et al. (2008) Stationarity is dead: Whither water management? Science 319: 573--574. Olson M (1965) The Logic of Collective Action: Public Goods and the Theory of Groups. Cambridge, MA: Harvard University Press. Ostrom E (1990) Governing the Commons: The Evolution of Institutions for Collective Action. New York, NY: Cambridge University Press. Ostrom E (1999) Coping with the tragedy of the commons. Annual Review of Political Science 2: 493--535. Ostrom E (2007) Understanding Institutional Diversity. Princeton, NJ: Princeton University Press. Ostrom E, Dietz T, Dolsak N, et al. (2002) The Drama of the Commons. Washington, DC: National Academies Press. Ostrom E, Gardner R, and Walker J (1994) Rules, Games, and Common Pool Resources. Ann Arbor, MI: University of Michigan Press. Reuss M (1992) Coping with uncertainty: Social scientists, engineers, and federal water resources planning. Natural Resources Journal 32: 101--135. Reuss M (2003) Is it time to resurrect the Harvard Water Program? Journal of Water Resources Planning and Management 129(5): 357--360. Ribot J, Agrawal A, and Larson A (2006) Recentralizing while decentralizing: How national governments reappropriate forest resources. World Development 34(11): 1864--1886. Rogers P (2002) Water Governance in Latin America and the Caribbean. Washington, DC: Inter-American Development Bank. Rogers P and Hall A (2003) Effective Water Governance, TAC Background Paper No. 7. Stockholm: Global Water Partnership. Rose C (2002) Common property, regulatory property, and environmental protection: Comparing community-based management to tradable environmental allowances. In: Ostrom E, Dietz T, Dolsak N, et al. (eds.) Drama of the Commons, pp. 263--292. Washington, DC: National Academies Press. Sabatier P, Weible C, and Ficker J (2006) Eras of water management in the United States: Implications for collaborative watershed approaches. In: Sabatier P, Focht W, Lubbell M, et al. (eds.) Swimming Upstream: Collaborative Approaches to Watershed Management, pp. 23--52. Cambridge, MA: MIT Press. Schlager E (2005) Getting the relationships right in water property rights. In: Bruns BR, Ringler C, and Meinzen-Dick R (eds.) Water Rights Reform: Lessons for Institutional Design, pp. 27--54. Washington, DC: International Food Policy Research Institute. Schlager E and Blomquist W (2005) Beneath the surface: Shared attributes of fisheries and aquifers and implications for institutional design. Prepared for Presentation at

Governing Water: Institutions, Property Rights, and Sustainability

Festschrift for Elinor Ostrom, Workshop on Political Theory and Policy Analysis, Indiana University, Bloomington, IN, 22–23 November 2005. Schlager E and Blomquist W (2008) Embracing Watershed Politics. Boulder, CO: University Press of Colorado. Schlager E and Ostrom E (1992) Common property and natural resources: A conceptual analysis. Land Economics 68: 249--252. Shapiro M (1981) Courts: A Comparative and Political Analysis. Chicago, IL: University of Chicago Press. Tang SY (1992) Institutions and Collective Action: Self-Governance in Irrigation. San Francisco, CA: ICS Press. Tang SY (1994) Institutions and performance in irrigation systems. In: Ostrom E, Gardner R, and Walker J (eds.) Rules, Games, and Common Pool Resources, pp. 225--246. Ann Arbor, MI: University of Michigan Press. Thome J (2000) Heading south but looking north: Globalization and law reform in Latin America. Wisconsin Law Review 2000: 691--712. Thompson EP (1975) Whigs and Hunters: The Origin of the Black Act. London: Allen Lane.

33

Trubek D (2006) The ‘rule of law’ in development assistance: Past, present, and future. In: Trubek DM and Santos A (eds.) The New Law and Economic Development: A Critical Appraisal, pp. 74--94. New York, NY: Cambridge University Press. Whiteley J, Ingram H, and Perry R (2008) Water, Place and Equity. Cambridge, MA: MIT Press. Whitlock W (2000) The rule of law. Wisconsin Law Review 2000: 723--742. World Bank (1993) Water Resources Management: A World Bank Policy Paper. Washington, DC: World Bank. World Bank (2004) Water Resources Sector Strategy: Strategic Directions for World Bank Engagement. Washington, DC: World Bank. World Commission on Dams (2000) Dams and development: A new framework for decision-making. The Report of the WCD. London: Earthscan. Young O (2002) Institutional interplay: The environmental consequences of cross-scale interactions. In: Ostrom E, Dietz T, Dolsak N, et al. (eds.) Drama of the Commons, pp. 263--292. Washington, DC: National Academies Press.

1.03 Managing Aquatic Ecosystems CM Finlayson, Charles Sturt University, Albury, NSW, Australia & 2011 Elsevier B.V. All rights reserved.

1.03.1 1.03.2 1.03.2.1 1.03.2.2 1.03.2.3 1.03.3 1.03.3.1 1.03.3.2 1.03.3.3 1.03.3.4 1.03.4 1.03.4.1 1.03.4.2 1.03.4.3 1.03.4.4 1.03.4.5 1.03.4.6 1.03.5 1.03.5.1 1.03.5.2 1.03.5.3 1.03.5.4 1.03.6 References

Introduction Key Concepts Wise Use of Wetlands Ecological Character The Ecosystem Approach Distribution and Classification of Aquatic Ecosystems Classification of Inland Aquatic Ecosystems Extent and Distribution Loss and Degradation of Inland Aquatic Ecosystems Loss of Species from Inland Aquatic Ecosystems Drivers of Change in Inland Aquatic Ecosystems Drainage, Clearing, and Infilling Modification of Water Regimes Invasive Species Overfishing Water Pollution and Eutrophication Climate Change Management Responses Integrated Management Processes International Cooperation and Action Restoration and Wise Use of Wetlands Supporting Local Community Involvement in Management Conclusions

1.03.1 Introduction The importance of aquatic ecosystems for people has been highlighted in recent years by the Millennium Ecosystem Assessment (Finlayson et al., 2005) and subsequent assessments such as the World Water Development Report (UNESCOWWAP, 2006), the Global International Waters Assessment (UNEP, 2006), the Global Biodiversity Outlook (Secretariat of the Convention on Biological Diversity, 2006) and the Global Environment Outlook (UNEP, 2007), and the Comprehensive Assessment of Water Management in Agriculture (Molden, 2007). However, despite many aquatic ecosystems being highly important for people and biodiversity, they have been degraded over many decades and many lost (Finlayson and D’Cruz, 2005). The continued degradation and loss of these highly valued ecosystems bring into question the effectiveness of current management practices and whether or not different approaches should be developed to ensure that they are managed wisely. The scope and definition of aquatic ecosystems have been widely discussed with many different definitions and classifications being used to cover a range of inland and coastal aquatic ecosystems or wetlands (Finlayson and van der Valk, 1995; Mitsch and Gosselink, 2007; Whigham, 2009). The term wetland has often been used to define a narrow range of inland aquatic systems, such as bogs, marshes, and swamps, while at other time it has been used to define a wider range

35 35 35 37 39 42 42 43 44 45 46 47 48 50 52 53 54 55 55 56 56 56 56 57

including rivers, lakes, reservoirs, and rice fields as well (Finlayson et al., 1999). Extensive information on wetland definition and delineation is available (e.g., Finlayson and van der Valk, 1995; Mitsch et al., 1994), but the failure to consider the different definitions completely that have been used around the world has resulted in confusion and inaccurate analyses on the extent and condition of these ecosystems (Finlayson and Spiers, 1999). The term aquatic ecosystem is used here in order to illustrate that the wider range of inland wetlands is being considered. This corresponds with the approach taken in the Millennium Ecosystem Assessment and encompasses the terms inland water systems, inland waters, or inland wetlands and includes marshes, swamps, lakes, and rivers, regardless of their size or whether they are permanent or temporary or saline or fresh (Finlayson and D’Cruz, 2005). As there is no clear boundary between inland and coastal aquatic ecosystems, the differentiation between them is indicative only. As there are strong interactions between inland and coastal aquatic ecosystems, a clear differentiation is not adopted. Coastal brackish and saline marshes, estuaries, and mangroves are not considered unless there is a strong connection between them and nearby freshwater aquatic ecosystems. Inland salt lakes are included, especially as many undergo flooding, whether periodic or irregular, with freshwater. The range of ecosystems considered is shown in Figure 1.

35

36

Managing Aquatic Ecosystems

Figure 1 Inland aquatic ecosystems.

The text below outlines some of the key concepts and approaches for managing inland aquatic ecosystems with a focus on water management. Key sources of information include material derived from the Ramsar Convention on Wetlands (Ramsar Convention Secretariat, 2006a), the Millennium Ecosystem Assessment (Finlayson et al., 2005), and the Wetland Handbook (Maltby and Barker, 2009). The importance of the Ramsar Convention in promoting the conservation and wise use of wetlands globally cannot be underestimated – the text of the Convention was signed in 1971 and in many ways was a forerunner of many subsequent developments in conservation, especially the shift from species preservation to sustainable use and integrated management of wetland ecosystems (Matthews, 1993). A summary of the key features of the Convention is provided in Box 1.

1.03.2 Key Concepts A number of key concepts associated with the management of inland aquatic ecosystems are outlined in the section below. These include the concepts of wise use of wetlands and ecological character of wetlands used by the Ramsar Convention on Wetlands, and the ecosystem approach used by the

Convention on Biological Diversity and promoted by the IUCN Commission on Ecosystem Management.

1.03.2.1 Wise Use of Wetlands Since its inception in 1971, the Convention on Wetlands (Ramsar, Iran, 1971) has promoted the wise use of wetlands (Matthews, 1993). The contracting parties to the Convention have subsequently accepted a mission statement that commits them to ‘‘y the conservation and wise use of all wetlands through local, regional and national actions and international cooperation, as a contribution towards achieving sustainable development throughout the world’’ (Ramsar Strategic Plan 2009–2015). The contracting parties have also agreed to deliver this mission through three streams of activities: (1) the wise use of all wetlands, (2) the designation and management of wetlands of international importance (Ramsar sites; see Figure 2), and (3) international cooperation. Wise use of wetlands was included in the text of the Convention under article 3.1 ‘‘The Contracting Parties shall formulate and implement their planning so as to promote y as far as possible the wise use of wetlands in their territory.’’ Matthews (1993) reported that from the outset the wise use of wetlands was seen as the maintenance of their ecological character, as a basis not only for nature conservation, but for

Managing Aquatic Ecosystems

37

Box 1 The Ramsar Convention on Wetlands. Information from The Ramsar Convention on Wetlands, http://www.ramsar.org (accessed September 2010). The Convention on Wetlands of International Importance, called the Ramsar Convention, is an intergovernmental treaty that provides a framework for national action and international cooperation for the conservation and wise use of wetlands and their resources. Unlike the other global environmental conventions, Ramsar is not affiliated with the United Nations system of Multilateral Environmental Agreements, but it works very closely with the other biodiversity-related treaties and agreements. The Convention was negotiated through the 1960s by countries and nongovernmental organizations that were concerned at the increasing loss and degradation of wetland habitat for migratory waterbirds; the treaty was adopted in the Iranian city of Ramsar in 1971 and came into force in 1975. It is the only global environmental treaty that deals with a particular ecosystem. The Convention’s mission is ‘‘the conservation and wise use of all wetlands through local and national actions and international cooperation, as a contribution towards achieving sustainable development throughout the world.’’ It uses a broad definition of the types of wetlands covered in its mission, including lakes and rivers, swamps and marshes, wet grasslands and peatlands, oases, estuaries, deltas and tidal flats, near-shore marine areas, mangroves and coral reefs, and human-made sites such as fish ponds, rice paddies, reservoirs, and salt pans. At the center of the Convention is the wise use concept which has at its heart the conservation and sustainable use of wetlands and their resources, for the benefit of human kind. The Convention has 160 contracting parties (12 December 2009) that have agreed to four main obligations: 1. designate at least one wetland at the time of accession for inclusion in the list of wetlands of international importance (the Ramsar List) and to promote its conservation, and in addition to continue to designate suitable wetlands within its territory for the list; 2. include wetland conservation in their national land-use planning and to promote, as far as possible, the wise use of wetlands in their territory; 3. establish nature reserves in wetlands, whether or not they are included in the Ramsar List, and to promote training in the fields of wetland research, management, and wardening; and 4. consult with other contracting parties about implementation of the Convention, especially in regard to transboundary wetlands, shared water systems, and shared species.

There are currently 1891 wetland sites (information as of 23 July 2010) designated for the list of wetlands of international importance covering a total surface area of 185 464 092 ha (Figure 2). Further information on the history and development of the Convention can be obtained from Matthews (1993).

Figure 2 Wetlands of international importance listed under the Ramsar Convention on Wetlands. From http://www.ramsar.org.

sustainable development also. With this background, wise use of wetlands was later defined as ‘‘y their sustainable utilisation for the benefit of humankind in a way compatible with the maintenance of the natural properties of the ecosystem.’’ With sustainable utilization it is defined as ‘‘y human use of a wetland so that it may yield the greatest continuous benefit to present generations while maintaining its potential to meet the needs and aspirations of future generations’’ (Davis, 1993).

Guidelines for the wise use of wetlands were formally adopted by the Convention in 1990 and emphasized the link between wetlands and people, and placed wetlands in a catchment or coastal zone context (Davis, 1993; Ramsar Convention Secretariat, 2006b; Maltby, 2009). The definition and guidelines remained in place until 2005 when wise use was redefined as ‘‘y the maintenance of their ecological character, achieved through the implementation of ecosystem approaches, within the context of sustainable development.’’

38

Managing Aquatic Ecosystems

The change in definition strengthened the link between wise use and the concept of sustainable development that had been promoted by the UN World Commission on Environment and Development (Brundtland, 1987). The change in definition also reflected an expansion in the wise use guidelines with the development and adoption of an increasing range of policy and technical guidelines known as the Ramsar Toolkit of Wise Use Handbooks (Ramsar Convention Secretariat, 2006c). Table 1 provides a summary of the material covered in the handbooks. The guidelines have also been mapped onto the conceptual framework developed for the Millennium Ecosystem Assessment which is consistent with the Ramsar concept of wise use (Finlayson et al., 2005). The framework provides a guide to the Ramsar wise use guidelines, and can be used to identify gaps (Figure 3). Many of the guidelines apply directly to ecosystem

services and the links between these services and the ecological components and processes that characterize wetlands, including those for describing and assessing the condition of the wetland. Others address interventions covering the direct drivers of change to ecosystems while two sets – those covering national wetland policies and on reviewing legislative and institutional frameworks – deal with the indirect drivers of change. Some, such as those on international cooperation, communications, education, and public awareness, apply to several parts of the framework. The guidelines are under continual review by the Convention and updated at regular intervals. It is anticipated that further attention will be devoted to guidelines that address policy issues and encourage greater integration between resource users and conservation interests – themes that have been evident in recent global assessments covering biodiversity and ecosystem services.

Table 1 Guidance available through the 3rd edition of the Ramsar Wise Use Handbooks and the relevant Resolutions agreed by the Contracting Parties to the Convention Handbook no.

Title

Content

Resolutions

1.

Wise use of wetlands

IX.1

2. 3.

National wetland policies Laws and institutions

A Conceptual Framework for the wise use of wetlands and the maintenance of their ecological character Developing and implementing National Wetland Policies

4.

Wetland CEPA

5.

Participatory skills

6.

Water-related guidance River basin management Water allocation and management Managing groundwater Coastal management Inventory, assessment, and monitoring Water allocation and management Impact assessment

7. 8. 9. 10. 11.

12.

Reviewing laws and institutions to promote the conservation and wise use of wetlands The Convention’s Programme on communication, education, and public awareness (CEPA) 2003–2008 Establishing and strengthening local communities’ and indigenous people’s participation in the management of wetlands An integrated framework for the Convention’s water-related guidance Integrating wetland conservation and wise use into river basin management Guidelines for the allocation and management of water for maintaining the ecological functions of wetlands Guidelines for the management of groundwater to maintain wetland ecological character Wetland issues in Integrated Coastal Zone Management An integrated framework for wetland inventory, assessment, and monitoring

14.

Designating Ramsar sites

15.

Addressing change in ecological character Managing wetlands

Guidelines for the allocation and management of water for maintaining the ecological functions of wetlands Guidelines for incorporating biodiversity-related issues into environmental impact assessment legislation and/or processes and in strategic environmental assessment The Strategic Framework and guidelines for the future development of the List of Wetlands of International Importance Addressing change in the ecological character of Ramsar sites and other wetlands Frameworks for managing Ramsar sites and other wetlands

International cooperation

Guidelines for international cooperation under the Ramsar Convention on Wetlands

13.

16. 17.

VII.6 VII.7 VIII.31 VII.8

IX.1 VII.19, IX.1, IX.3 VIII.1 IX.1 VII.21, VIII.4 IX.1

VIII.1 VII.16, VIII.9

VIII.10

V.4, VI.1, VII.24, VIII.8, VIII.16, VIII.20, VIII.22, IX.6 V.7, VI.1, VII.10, VIII.14, VIII.18, VIII.19, IX.4 VII.19

Reproduced from The Ramsar ‘Toolkit’, 3rd edn. (2007) The Ramsar Handbooks for the Wise Use of Wetlands. http://www.ramsar.org/cda/ramsar/display/main/ main.jsp?zn=ramsar&cp=1-30^21323_4000_0 (accessed August 2010).

Managing Aquatic Ecosystems

39

Global Regional Local Human well-being and poverty reduction

Indirect drivers of change

• Health security • Environmental security • Economical security • Cultural security • Equity

• Demographic • Economic (e.g., trade, subsides, markets) • Sociopolitical(e.g., governance, institutional and legal framework) • Science and technology • Cultural and religious (e.g., choices about what and how much to consume)

777

171 HB17: International cooperation

HB0: Water allocation and management

HB1: Wise use

HB3:Laws and institutions

HB4: Wetland CEPA HB7: River basin Management

171

HB16: Managing wetlands HB12: Wetland inventory

Ecosystem services

HB11: Inventory, assessment, monitoring

HB14: Designating rumuer sites HB5: Participatory skills

HB10: Central management

• Provisioning (e.g., food, fresh water, fuel, genetic resources) • Regulating (e.g., climate, water, natural hazard mitigation) • Cultural (e.g., spiritual, aesthetic) • Supporting (e.g., primary production, nutrient cycling)

Life on earth: biodiversity Strategies and Interventions

HB2: National wetlands policies

Direct drivers of change

HB13: Impact assessment

HB9: Groundwater

HB8: Water allocation and management

• Changes in local land use and land cover • Species removals and/or invasive introductions • Eutrophication and pollution • Hydraulic infrastructure development • Water abstraction • Climate change HBO: Water-related guidance

HB10: Coastal management

HB7: River basin management HB15: Change in cool character

777 No specific guidance

HBxx

Dark background: Handbooks include interventions into several red bars

Figure 3 A framework for the wise use of wetlands and the application of the guidelines in the Ramsar ‘Toolkit’ of Wise Use Handbooks, 3rd edn. Reproduced from Ramsar Convention Secretariat (2006b) Wise use of wetlands: A conceptual framework for the wise use of wetlands. In: Ramsar Handbooks for the Wise Use of Wetlands, 3rd edn., vol. 1. Gland, Switzerland: Ramsar Convention Secretariat, with permission from Ramsar; updated from Finlayson CM, D’Cruz R, and Davidson NJ (2005) Ecosystem Services and Human Well-Being: Water and Wetlands Synthesis. Washington, DC: World Resources Institute.

1.03.2.2 Ecological Character The concept of ecological character was introduced in the text of the Ramsar Convention and is now seen as a basis for the wise use of wetlands globally. Contracting parties to the Ramsar Convention are required to promote the conservation of all wetlands through the maintenance of their ecological character and to do this they are expected to establish management planning and monitoring mechanisms (Ramsar Convention Secretariat, 2006a). Ecological character is now defined as ‘‘y the combination of the ecosystem components, processes and benefits/services that characterise the wetland at a given point in time’’ (Ramsar Convention Secretariat, 2007). The previous definition considered ecosystem services separately to the ecological components and processes that were seen as comprising the ecological character of a wetland (Figure 4). Ecosystem services were incorporated into the definition of ecological character after the Convention adopted the findings of the Millennium Ecosystem Assessment as they applied to water and wetlands (Finlayson et al., 2005). This widened the concept of wetland conservation and provided an overt link with the management and uses of a wetland, as expressed through ecosystem services which provide benefits to people.

The description of the ecological character of a wetland provides baseline data that establish the range of natural variation in ecological components and processes and ecosystem services at each site within a given time frame, against which change can be assessed. To describe the ecological character of a wetland and ascertain when adverse change may have occurred, the Convention has established an integrated framework for wetland inventory, assessment, and monitoring (Ramsar Convention Secretariat, 2006c; Finlayson et al., 2005). In support of this approach, the Convention also adopted the concept of ecosystem services as provided by the Millennium Ecosystem Assessment (2003) with four categories of services (Figure 5): 1. provision services – products obtained from ecosystems; 2. regulating services – benefits obtained from regulation of ecosystem processes; 3. cultural services – nonmaterial benefits obtained from ecosystems; and 4. supporting services – services necessary for the production of all other ecosystem services.

40

Managing Aquatic Ecosystems

Provisioning Biological Regulating

Chemical Physical

Cultural

Supporting

Ecosystem components and processes

Ecosystem services

Figure 4 The ecological character of a wetland showing the relationship between the ecosystem components and processes and services that comprise the wetland.

Provisioning services

Regulating services

Cultural services

Goods produced or provided by ecosystems

Benefits from regulation of ecosystem processes

Nonmaterial benefits from ecosystems

• Food • Fuel wood • Fiber • Timber

• Water partitioning • Pest regulation • Climate regulation • Pollination

• Spiritual • Recreational • Aesthetic • Educational

Support services Factors necessary for producing ecosystem services • Hydrological cycle • Soil formation • Nutrient cycling • Primary production

Figure 5 Four categories of ecosystems services as outlined in the Millennium Ecosystem Assessment (2003). Reproduced from Millennium Ecosystem Assessment (2003) Ecosystems and Human Well-Being: A Framework for Assessment. Washington, DC: Island Press.

The relative importance of these services in different types of inland aquatic ecosystems is shown in Figure 6. The information in this figure was based on expert opinion given the absence of sufficient data to support a quantitative analysis. Given this situation the collection of further information on the extent of ecosystem services provided by wetlands is encouraged by the Convention as a basis for wise use.

1.03.2.3 The Ecosystem Approach In response to widely articulated weaknesses in sectoral approaches for wetland management (e.g., Hollis, 1992, 1998), a number of ecosystem approaches have been developed to promote the conservation and sustainable and equitable use of wetlands (Brown et al., 2005). The Convention on

Managing Aquatic Ecosystems

41

Alpine and tundra wetlands

Springs and oases

Geothermal wetlands

Underground wetlands including caves and groundwater systems

Seasonal lakes, marshes, and swamps including floodplains Forested wetlands, marshes, and swamps, including floodplains

Permanent lakes, reservoirs

Permanent and temporary rivers and streams

Comments and examples

Services

Scale is low •, medium , to high ; not known?; blank cells indicate that the service is not considered applicable to the wetland type. The information in the table represents expert opinion for a global average pattern for wetlands; there will be local and regional differences in relative magnitudes.

?

?

?

?

?

?

?

?

?

?

Inland wetlands Provisioning Food Fresh water

Fiber and fuel Biochemical products Genetic materials

Regulating Climate regulation

Hydrological regimes

Pollution control and detoxification Erosion protection

Natural hazards Cultural Spiritual and inspirational Recreational Aesthetic Educational Supporting Biodiversity Soil formation Nutrient cycling Pollination

Production of fish, wild game, fruits, grains, and so on Storage and retention of water; provision of water for irrigation and for drinking Production of timber, fuelwood, peat, fodder, aggregates Extraction of materials from biota Medicine; genes for resistance to plant pathogens, ornamental species, and so on

? ?

?

Regulation of greenhouse gases, temperature, precipitation, and other climatic processes; chemical composition of the atmosphere Groundwater recharge and discharge; storage of water for agriculture or industry Retention recovery, and removal of excess nutrients and pollutants Retention of soils and prevention of structural change (such as coastal erosion, bank slumping, and so on) Flood control; storm protection

?

Personal feelings and well-being; religious significance Opportunities for tourism and recreational activities Appreciation of natural features Opportunities for formal and informal education and training Habitats for resident or transient species Sediment retention and accumulation of organic matter Storage, recycling, processing, and acquisition of nutrients

?

Support for pollinators

Figure 6 Ecosystem services provided by different inland aquatic ecosystems. Reproduced from Finlayson CM, D’ Cruz R, and Davidson NJ (2005) Ecosystem Services and Human Well-Being: Water and Wetlands Synthesis. Washington, DC: World Resources Institute; and Millennium Ecosystem Assessment (2005) Ecosystems and Human Well-Being: Synthesis. Washington, DC: Island Press.

42

Managing Aquatic Ecosystems

Table 2 1. 2. 3. 4.

5. 6. 7. 8. 9. 10. 11. 12.

Principles of the ecosystem approach adopted by the Convention on Biological Diversity

The objectives of management of land, water, and living resources are a matter of societal choices Management should be decentralized to the lowest appropriate level Ecosystem managers should consider the effects (actual or potential) of their activities on adjacent and other ecosystems After recognizing potential gains from management, there is a need to understand the ecosystem in an economic context. Any ecosystem management program should: (a) reduce those market distortions that adversely affect biological diversity; (b) align incentives to promote sustainable use; and (c) internalize costs and benefits in the given ecosystem to the extent feasible Conservation of ecosystem structure and functioning, in order to maintain ecosystem services, should be a priority target of the ecosystem approach Ecosystems must be managed within the limits to their functioning The ecosystem approach should be undertaken at the appropriate spatial and temporal scales Recognizing the varying temporal scales and lag effects which characterize ecosystem processes, objectives for ecosystem management should be set for the long term Management must recognize that change is inevitable The ecosystem approach should seek the appropriate balance between, and integration of, conservation and use of biological diversity The ecosystem approach should consider all forms of relevant information, including scientific, indigenous, and local knowledge, innovations, and practices The ecosystem approach should involve all relevant sectors of society and scientific disciplines

Information derived from Ecosystem Approach, Principles, http://www.cbd.int/ecosystem/principles.shtml (accessed August 2010).

Biological Diversity has promoted the ecosystem approach by focusing on managing environmental resources and by promoting a balance between human needs and biodiversity. The wise use approach adopted by the Ramsar Convention as well as integrated coastal zone management and integrated catchment management are compatible with the ecosystem approach adopted by the Convention on Biological Diversity (Shepherd, 2004; Brown et al., 2005). The ecosystem approach adopted by the Convention on Biological Diversity was seen as a way of reaching a balance between the objectives of the Convention covering conservation, sustainable use, and the fair and equitable sharing of the benefits arising out of the utilization of genetic resource. It is based on a strategy for the integrated management of land, water, and living resources and promotes conservation and sustainable use in an equitable way (Shepherd, 2004). It places human needs at the center of biodiversity management and acknowledges that ecosystems perform multiple functions of importance to people both locally and further afield. The guiding principles and strategies adopted by the Convention are shown in Table 2. As the Convention’s ecosystem approach does not comprise a specific applicable method, it has been criticized for being too vague to be of practical value, while others have highlighted its flexibility (Brown et al., 2005). Taking note of the criticisms, Shepherd (2004) grouped the guiding principles into the following five steps, each involving a range of actions to encourage discussion, planning, and step-by-step action:

1. determine the main stakeholders, define the ecosystem area, and develop the relationship between them; 2. characterize the structure and function of the ecosystem, and set in place mechanisms to manage and monitor it; 3. identify the important economic issues that will affect the ecosystem and its inhabitants; 4. determine the likely impact of the ecosystem on adjacent ecosystems; and 5. decide long-term goals and flexible ways of reaching them.

Brown et al. (2005) have highlighted some of the constraints that are generally seen to characterize ecosystem approaches. These include: a failure to consider specific areas, resources, or species that may need a more targeted approach for their conservation; uncertainties and lack of guidance about how to balance conservation and sustainable use; and difficulties with establishing collaboration between stakeholders and negotiating trade-offs between them in a fair and equitable way. Even with these constraints the principles outlined in Table 2 have been widely accepted and, in one way or the other, now feature in many current approaches for managing aquatic ecosystems (Finlayson and D’Cruz, 2005; Maltby, 2009). The principles adopted by the Convention on Biological Diversity are generally applicable to other ecosystem approaches and have largely been addressed through the more detailed guidance for wise use developed by the Ramsar Convention (Ramsar Convention Secretariat, 2006a, 2006b). Ecosystem approaches generally provide for a broad, crosssectoral approach for managing aquatic ecosystems by addressing both direct and indirect drivers of change and considering the multiple benefits people derive from the ecosystem services that they provide.

1.03.3 Distribution and Classification of Aquatic Ecosystems The extent and distribution of inland aquatic ecosystems is poorly and unevenly known at the global and regional scales due to differences in definitions as well as difficulties in delineating and mapping ecosystems with variable boundaries due to fluctuations in water levels (Finlayson et al., 1999; Rebelo et al., 2009; Mackay et al., 2009). In many cases, comprehensive documentation of the extent and distribution of inland aquatic ecosystems at the regional or national levels also does not exist. The larger ecosystems, such as lakes and inland seas, have been mapped along with the major rivers, but for many parts of the world smaller ecosystems are not well mapped or delineated. As a consequence, assessment of

Managing Aquatic Ecosystems

the extent of and change in these ecosystems at the continental level is compromised by the inconsistency and unreliability of the data.

1.03.3.1 Classification of Inland Aquatic Ecosystems The classification of inland aquatic ecosystems or wetlands has consumed an inordinate amount of time and controversy with many systems developed and used in different countries (see summary in Finlayson and van der Valk (1995)). While a wetland inventory can be undertaken without recourse to an agreed classification, given that a standardized and logical process of data collection or collation is undertaken, it is likely to become necessary to classify them at some stage during the assessment phase of wetland management, especially where it is useful or necessary to make comparisons between different wetlands. At this stage an agreed set of terms is not only desirable but possibly mandatory to ensure conformity of comparisons and hence decisions. Thus, the importance of classification cannot be overstated, but it equally needs to be Table 3

remembered that classification is a tool within a larger set of tools that are designed to provide an adequate information base for the wise use, conservation, and management of all wetlands. The Ramsar wetland definition is supported by a classification scheme with 42 categories that is purposefully simple and global in scope, and readily compatible with other classifications that may be preferred regionally, nationally, or locally (Scott and Jones, 1995). The classification divides wetlands into three main types: marine/coastal (12 wetland types), inland (20), and human-made (10), with several categories, according to vegetation, soil/rock, inundation, water quality (freshwater to saline water), and landform, within each type (Ramsar Convention Secretariat, 2006d). The characteristics of the inland wetland types are shown in Table 3. While used in a general sense by contracting parties to the Convention, it has also been augmented by more specific classifications. The broad wetland categories included in the classification have been particularly useful when comparing between countries or regions, but less so when more specific

Characteristics of inland wetland types contained within the Ramsar wetland classification

Wetland type

Wetland characteristics

1.

Permanent rivers/streams/creeks

2. 3. 4.

Permanent inland deltas Freshwater springs; oases Seasonal/intermittent/irregular rivers/ streams/creeks

5. 6. 7. 8.

Permanent freshwater lakes Permanent freshwater marshes/pools Seasonal/intermittent freshwater lakes Seasonal/intermittent freshwater marshes/ pools on inorganic soils

Lakes and pools

9. 10. 11. 12.

Permanent freshwater marshes/pools Shrub-dominated wetlands Freshwater, tree-dominated wetlands Seasonal/intermittent freshwater marshes/ pools on inorganic soils

Marshes – inorganic soils

Permanent Permanent/seasonal/ intermittent Seasonal/intermittent

Herb dominated Shrub dominated Tree dominated Herb dominated

13. 14.

Nonforested peatlands Forested peatlands

Marshes – peat soils

Permanent

Nonforested Forested

15. 16.

Alpine wetlands Tundra wetlands

Marshes – inorganic or peat soils

High altitude Tundra

17. 18.

Permanent saline/brackish/alkaline lakes Seasonal/intermittent saline/brackish/ alkaline lakes and flats Permanent saline/brackish/alkaline marshes/pools

Lakes

Permanent Seasonal/intermittent

19.

43

20.

Seasonal/intermittent saline/brackish/ alkaline marshes/pools

21.

Geothermal wetlands

22.

Karst and other subterranean hydrological systems

Freshwater

Flowing water

Permanent

Season/intermittent

Permanent Seasonal/intermittent

Saline, brackish, or alkaline water

Rivers, streams, creeks Deltas Springs, oases Rivers, streams, creeks 48 ha o8 ha o8 ha o8 ha

Permanent Marshes and pools Fresh, saline, brackish, or alkaline water

Seasonal/intermittent

Geothermal Subterranean

Derived from Ramsar Convention Secretariat (2006c). Inventory, assessment, and monitoring: An integrated framework for wetland inventory, assessment, and monitoring. In: Ramsar Handbooks for the Wise Use of Wetlands, 3rd edn., vol. 11. Gland, Switzerland: Ramsar Convention Secretariat.

44

Managing Aquatic Ecosystems

or accurate classification was required. Semeniuk and Semeniuk (1997) pointed out that the Ramsar classification for inland wetlands was not entirely systematic and promoted the adoption of a more systematic classification based on the hydro-geomorphology of the wetland. Nevertheless, except for the occasional addition of wetland types the Ramsar classification has not been substantially changed and still forms a readily available general model for many purposes.

1.03.3.2 Extent and Distribution Estimates of the global extent of inland aquatic ecosystems differ greatly and highly depend on the definition of wetlands used and on the methods for delineating wetlands (Finlayson et al., 1999; Mitsch and Gosselink, 2007; Whigham, 2009). The Global Review of Wetland Resources and Priorities for Wetland Inventory conducted on behalf of the Ramsar Convention estimated the extent of all wetlands from national inventories as approximately 1280 million hectares, which was considerably higher than previous estimates (Finlayson et al., 1999). This included inland (including lakes, rivers, swamps, and marshes), coastal (including lagoons, swamps, and estuaries), near-shore marine areas (tidal flats and marine areas to a depth of 6 m below low tide), and human-made wetlands (such as reservoirs and rice paddies). Nevertheless, this figure is considered an underestimate, especially for southern America and for certain wetland types (such as intermittently flooded inland wetlands, peatlands, and artificial wetlands) where data were incomplete or not readily accessible. The most recent attempt to ascertain the extent and distribution of inland aquatic ecosystems systems (Lehner and Doll, 2004) based on analysis of existing data sets derived largely from the Earth observation is shown in Figure 7. As with previous estimates, these data contain many inaccuracies and gaps. For example, intermittently inundated inland aquatic ecosystems are not included, and there are many inaccuracies because of the problems of scale and resolution.

Nevertheless, these data and the mapping products are considered incredibly useful even while requiring updating and refinement (Finlayson et al., 2005; Whigham, 2009). Lehner and Doll (2004) reported the extent of inland aquatic ecosystems as 917 million hectares, comprising Africa with 131, Asia 286, Europe 26, Neotropics 159, North America 287, and Oceania with 28 million hectares. Inventory and mapping of inland aquatic ecosystems have been undertaken in many parts of the world, but the level of detail varies from region to region with some regions and ecosystem types considered to be under-represented in the data given above (Finlayson and D’Cruz, 2005). The latter includes rivers, lakes and reservoirs, peatlands, and rice paddies. Inventories of major river systems are available, but there is considerable variability between areal estimates, based on the method and definitions used. Information on the estimated 5–15 million lakes distributed globally is also highly variable and dispersed (WWDR, 2003). Large lakes have been mapped reasonably well, but issues of scale occur with many smaller lakes underestimated or not recorded. Reservoirs are also widespread with the number of large dams (415 m in height) in the world increasing from approximately 5000 in 1950 to more than 45 000 with 3–6 times the standing water held by natural river channels (WCD, 2000; Vo¨ro¨smarty et al., 2005). The overall number of dams globally is uncertain with an estimated 800 000 small dams and further investment in the construction of others, large and small (Vo¨ro¨smarty et al., 2005). The total area of peatlands is estimated as approximately 400 million hectare with the majority in Canada (37%), Russia (30%), USA (13%), and Indonesia (6–7%) (Joosten and Clarke, 2002). The global area of paddies has been estimated to be 130 million hectares (Aselmann and Crutzen, 1989) with almost 90% in Asia. Information on other humanmade wetlands is variable and lacking for some countries. Groundwater systems vary in size from small-scale alluvial sediment along rivers to extensive aquifers such as the 1.2

Lake Reservoir River Freshwater marsh, floodplain Swamp forest, flooded forest Coastal wetland Pan, brackish/saline wetland Bog, fen, mire Intermittent wetland/lake 50−100% wetland 25−50% wetland Wetland complex (0−25% wetland) Figure 7 Distribution of inland aquatic ecosystems described as large lakes, reservoirs, and wetlands. Adapted from Lehner B and Doll P (2004) Development and validation of a global database of lakes, reservoirs and wetlands. Journal of Hydrology 296: 4–22.

Managing Aquatic Ecosystems

million square kilometers of the Guarani aquifer located across parts of Argentina, Brazil, Paraguay, and Uruguay (Danielopol et al., 2003). Groundwater systems have many connections with surface waters although many of these are not well understood, although the karst systems of Slovenia cover nearly 8800 km2 and are well known for their great species biodiversity, while others are not known at all (Finlayson and D’Cruz, 2005; Vo¨ro¨smarty et al., 2005). Whigham (2009) provided a general description of the main features of inland aquatic ecosystems commonly referred to as marshes, peatlands, and swamps. Marshes are dominated by emergent herbaceous vascular plants and occur in areas that are frequently or continuously flooded and most often have mineral soils that do not accumulate peat. Dominant plant species include reeds, rushes, grasses, and sedges, although they can also contain a wide variety of plant species with many different life forms. Peatland is a generic term for inland aquatic ecosystems that have at some point accumulated partially decayed plant matter because of incomplete decomposition, usually to a depth less than 30 cm. Many terms have been used to describe peatlands, such as mires, fens, and bogs. Swamps are flooded intermittently or permanently and are dominated by trees or shrubs. They are diverse and occur from the temperate zones to the tropics with those associated with rivers typically having inorganic substrates while those with little or no connection to flowing streams may develop peat substrates. While these generic descriptions exist, there are many local variations and terms used to describe inland aquatic ecosystems.

1.03.3.3 Loss and Degradation of Inland Aquatic Ecosystems The loss and degradation of inland aquatic ecosystems have been reported from many parts of the world (Finlayson et al., 1992; Mitsch, 1998; Moser et al., 1996; Whigham, 2009), but there are few reliable estimates of the actual extent of this loss globally. Dugan (1993) speculated that about 50% of wetlands had been lost globally, but did not provide supporting evidence. As a reliable estimate of the extent of inland aquatic ecosystems, particularly peatlands and intermittently inundated wetlands in semiarid areas, is not available (Finlayson et al., 1999), it is not possible to ascertain the extent of wetland loss globally (Finlayson and D’Cruz, 2005). While there is considerable uncertainty surrounding the extent of wetland losses globally, there is a lot of information available from specific countries or regions. Some examples are given below, taken from a collation provided in the Millennium Ecosystem Assessment (Finlayson and D’Cruz, 2005). The information available on the loss of inland aquatic ecosystems is far better for North America than for many other parts of the world with systematic monitoring of wetlands, excluding lakes and rivers, in the United States showing a loss of 116 000 ha yr 1 from the mid-1970s to the mid-1980s decreasing to 23 700 ha yr 1 from 1986 to 1997 (Dahl and Johnson, 1991; Dahl, 2000). Most of this loss was from the conversion or drainage of wetlands for urban development and agricultural purposes with an estimated 42.7 million hectares of wetlands remaining out of the 89 million hectares estimated to have been present at the time of European

45

colonization (Dahl, 2000). The decline in the rate of loss was attributed largely to the successful implementation of wetland policies and programs that promoted the restoration, creation, and enhancement of wetlands, as well as incentives to deter draining of wetlands. This is shown by a net gain of about 72 870 ha of wetland after 1997 and a 47 000 ha increase in the area of lakes and reservoirs (Dahl, 2000). While the data are not as illustrative for other parts of the world, Finlayson and D’Cruz (2005) concluded from the literature that much of the loss of wetlands in the northern temperate zone occurred during the first half of the twentieth century. Since the 1950s, many tropical and subtropical wetlands, particularly swamp forests, have also been lost or degraded, particularly as a consequence of agricultural expansion. The OECD (1996) estimated that by 1985, 56– 65% of available wetland had been drained for intensive agriculture in Europe and North America, 27% in Asia, 6% in South America, and 2% in Africa – a total of 26% loss to agriculture worldwide. This is still occurring, for example, in South America where peatlands linked with the Andean paramos ecosystems are being converted for agriculture, forestry, and peat mining (Hofstede et al., 2003; Blanco and de la Balze, 2004). In Southeast Asia large areas of the once-extensive tropical peat swamp forests have been degraded or lost over the last four decades mainly because of logging for timber and pulp and more recently by clear-felling and conversion to oil palm plantations (Glover and Jessup, 1999; Page et al., 1997; Rieley and Page, 1997). The most dramatic loss of peatlands to agriculture has been in northern Europe in countries, including Finland, Estonia, Denmark, the United Kingdom, and the Netherlands (once one-third peatland) which has lost virtually all of its natural peatlands (Brag et al., 2003; Joosten, 1994). Despite the absence of national data, there are many welldocumented examples of large inland aquatic ecosystems that have been degraded or lost. These include the impacts of water diversions to the Aral Sea in Central Asia (Lemly et al., 2000) and the Mesopotamian marshes in Iraq (Richardson et al., 2005), the Murray-Darling Basin in Australia (Kingsford and Johnson, 1998), the Everglades in the United States (Richardson, 2008), Donana in Spain (Bartolome and Vega, 2002), and the Hadejia-Nguru wetland complex in Nigeria (Lemly et al., 2000). The Aral Sea in Central Asia represents one of the most extreme cases in which water diversion for irrigated agriculture has caused severe environmental degradation of an inland water system with detrimental impacts on human well-being (see summary in Finlayson and D’Cruz, 2005; Box 2 and Figure 8). The extent of adverse change is so severe that the Aral Sea is considered by Falkenmark et al. (2007) as an example of human modification of an inland aquatic system having gone too far.

1.03.3.4 Loss of Species from Inland Aquatic Ecosystems The extent of species loss from inland aquatic ecosystems has been documented through several programs in recent years (Revenga et al., 2000; Finlayson and D’Cruz, 2005; Dudgeon et al., 2005; Loh et al., 2005). Information from these documents has been used to summarize the extent of species losses. The following excerpt from Revenga et al. (2005) serves as an

46

Managing Aquatic Ecosystems

Box 2 The Aral Sea. Based on information derived from Finlayson CM and D’Cruz R (2005) Inland Water Systems Millennium Ecosystem Assessment, Volume 2: Conditions and Trends. Washington, DC: Island Press and Falkenmark M, Finlayson CM, and Gordon L (2007) Agriculture, water, and ecosystems: Avoiding the costs of going too far. In: Molden D (ed.) Water for Food, Water for Life: A Comprehensive Assessment of Water Management in Agriculture, pp. 234–277. London: Earthscan. The Aral Sea is one of the most prominent examples of how unsustainable water management has led to a large-scale and possibly irreversible ecological and human disaster. Drastically reduced water flow into the sea has impaired human livelihoods and health, affected the local climate, and reduced if not decimated much of the biodiversity. Since 1960 the volume of water in the basin that surrounds the Aral Sea Basin has been reduced by 75% mainly as a consequence of the development of almost 7 million hectares of irrigation (UNESCO, 2000; Postel, 1999). It is considered unlikely that the ecological and social changes that have occurred as a consequence will be successfully restored, despite efforts to rehabilitate the northern part of the sea.

1973

1986

1999

2001

Figure 8 Changes in the Aral Sea 1973–2001. Reproduced from UNEP (2005).

introduction to the loss of species from inland aquatic ecosystems, including mention of the drivers, the relative risk of extinction compared to other ecosystems, and the inadequate level of assessment and information: Human activities have severely affected the condition of freshwater ecosystems worldwide. Physical alteration, habitat loss, water withdrawal, pollution, overexploitation and the introduction of nonnative species all contribute to the decline in freshwater species. Today, freshwater species are, in general, at higher risk of extinction than those in forests, grasslands and coastal ecosystems. For North America alone, the projected extinction rate for freshwater fauna is five times greater than that for terrestrial fauna – a rate comparable to the species loss in tropical rainforest. Because many of these extinctions go unseen, the level of assessment and knowledge of the status and trends of freshwater species are still very poor, with species going extinct before they are even taxonomically classified.

As with inland aquatic ecosystems the data on the condition and trends of freshwater species are, for the most part,

poor at the global level, although some countries have reasonable inventories. Key conclusions from Revenga and Kura’s (2003) assessment of the level of knowledge of the distribution and condition of inland water biodiversity were: fish and waterbirds were by far the best-studied groups, although with considerable regional differences; aquatic plants, insects, freshwater mollusks, and crustaceans were poorly known in most parts of the world, with fragmentary information; and that every group of organisms considered, including aquatic plants, invertebrate, and vertebrate animal species, contained examples of extinct, critically endangered, endangered, and vulnerable taxa. This contrasts with the importance of inland aquatic ecosystems which were reported by McAllister et al. (1997) to be species rich relative to other ecosystems and to support a disproportionately large number of species of some taxonomic groups, for instance, some 40% of known species of fish and about 25–30% of all vertebrate species diversity (Leveque et al., 2005). The living planet index developed by WWF and UNEPWCMC (Loh and Wackernagel, 2004) provides a measure of the trends in more than 3000 populations of 1145 vertebrate species around the world. The 2004 freshwater species population index, which took the trend data into account for 269 temperate and 54 tropical freshwater species populations (93 of which were fish, 67 amphibians, 16 reptiles, 136 birds, and 11 mammals), showed that freshwater populations declined consistently and at a faster rate than the other species groups assessed, with an average decline of 50% between 1970 and 2000 (Figure 9). Over the same period, both terrestrial and marine fauna decreased by 30% (Figure 9). A summary of the status of separate groups of species is given in Box 3. Revenga et al. (2005) reported that a review by the World Resources Institute of the status and trends of inland water biodiversity for the Convention on Biological Diversity (Revenga and Kura, 2003) drew the following conclusions:

1. freshwater fishes and waterbirds were by far the beststudied groups of species from inland aquatic ecosystems, although there were considerable regional differences; 2. aquatic plants, insects, freshwater mollusks, and crustaceans were poorly known or assessed in most parts of the world, with only fragmentary information available; and 3. in every group of organisms considered there were examples of extinct, critically endangered, endangered, and vulnerable taxa, making it clear that inland aquatic ecosystems were among the most threatened of all environments.

Managing Aquatic Ecosystems

120

100

Index 100 in 1970

Terrestrial 80 Marine 60 Freshwater 40

20

0 1970

1975

1980

1985

1990

1995

2000

Figure 9 Trends in freshwater, marine, and terrestrial living planet indices, 1970–2000. From Finlayson CM, D’Cruz R, and Davidson NJ (2005) Ecosystem Services and Human Well-Being: Water and Wetlands Synthesis. Washington, DC: World Resources Institute; and Millennium Ecosystem Assessment (2005) Ecosystems and Human Well-Being: Synthesis. Washington, DC: Island Press.

They further concluded that in general the information on species of inland aquatic ecosystems was poor, even for economically important groups, such as fish, and pointed out that many inventories tended to be organized by taxonomic groups and not by ecosystem types which made it hard to assess the condition of aquatic ecosystems. These conclusions have been largely supported by the analyses undertaken in the Millennium Ecosystem Assessment (Finlayson and D’Cruz, 2005; Finlayson et al., 2005) and more generally by other assessments, such as the Global Environment Outlook (Arthurton et al., 2007).

1.03.4 Drivers of Change in Inland Aquatic Ecosystems Analyses over the past two decades have identified a suite of common drivers of change in inland aquatic ecosystems (e.g., Revenga and Kura, 2003; Whigham et al., 1993; Mitsch, 1994; Finlayson and D’Cruz, 2005; Dudgeon et al., 2005). Many of these previous reviews have focused primarily on biophysical pressures that directly affect the ecological condition of these ecosystems, as depicted in Figure 10. Also, the importance of addressing the indirect drivers of change has been increasingly recognized with the Millennium Ecosystem Assessment (2003) and the Global Environment Outlook (Arthurton et al., 2007) providing comprehensive overviews. The indirect drivers are derived primarily from the following: demographics such as population size, age, gender structure, spatial distribution, and migration; economic forces such as national and per capita income, macroeconomic policies, international trade, and capital flows; sociopolitical

47

processes such as governance, institutional and policy frameworks, and the roles of women and wider civil society; scientific and technological developments including rates of investments in research and development and the rates of adoption of new biotechnologies and information technologies; and cultural and religious choices individuals make about what and how much to consume and what they value. These drivers are not static – they can change rapidly and over long periods as shown by fluctuations in the global economy and the increasing global population, for example. Changes in these drivers are also expected to increase the demand for food, fiber, energy, and freshwater (Vo¨ro¨smarty et al., 2005; Molden et al., 2007). The direct drivers of change in inland aquatic ecosystems are interconnected with the indirect drivers, and include: changes in land use as a consequence of clearance, drainage, and infilling; the spread of infrastructure for urban, tourism and recreation, aquaculture, agriculture, and industrial purposes; the introduction and spread of invasive species; the regulation and fragmentation of rivers; abstraction of surfaceand groundwater; over fishing and in places unsustainable hunting; chemical pollution, salinization, and eutrophication; and more recently the impacts of global climate change. In some cases, these drivers act synergistically or cumulatively. Multiple interacting drivers can cause changes in aquatic ecosystems and their species and ecosystem services that may not be readily attributable to one or the other driver. There are many interdependencies between and among the indirect and direct drivers of change, and, in turn, changes in ecosystems can lead to feedbacks on the drivers of change. When addressing the complex scenarios of multiple drivers, it is necessary to have a clear understanding of the nature of the changes and their likely causes before implementing management responses; risk and vulnerability assessments can help ascertain the nature of change and guide management. The Millennium Ecosystem Assessment (2003) outlined the interactions between drivers and ecosystems in a framework that linked the consequences of indirect and direct drivers with changes in the biodiversity and services provided by ecosystems and the consequences for human well-being. The direct drivers of change in inland aquatic ecosystems are described below – it draws heavily on the assessment undertaken by Finlayson and D’Cruz (2005) as part of the Millennium Ecosystem Assessment and provides an update.

1.03.4.1 Drainage, Clearing, and Infilling It has been well established that clearing or drainage for agricultural expansion is the principal cause for wetland loss worldwide. Agriculture, including rangelands, now occupies roughly 40% of the world’s terrestrial surface and is a major contributor to global environmental change (Foley et al., 2005), with cropping occurring on more than 50% of the land area in many river basins in Europe and India and more than 30% in the Americas, Europe, and Asia (Millennium Ecosystem Assessment, 2005). In this respect, inland aquatic ecosystems are subject to many of the same pressures from agriculture as are other ecosystems – agriculture is recognized as a major driver of change in ecosystems globally (Millennium Ecosystem Assessment, 2005).

48

Managing Aquatic Ecosystems

Box 3 Status of animals from inland aquatic ecosystems. Based on information derived from Finlayson CM and D’Cruz R (2005) Inland water systems Millennium Ecosystem Assessment, Volume 2: Conditions and Trends. Washington, DC: Island Press. Invertebrates The conservation status of most aquatic invertebrates has not been comprehensively assessed, except for regional assessments of certain taxonomic groups. Clausnitzer and Jodicke (2004) assessed the global status of dragonflies and damselflies and reported that while some 130 species were previously listed as threatened that many more were now also threatened. IUCN (2003) reported that 130 freshwater species of aquatic insects, 275 species of freshwater crustacean, and 420 freshwater mollusks were globally threatened, although no comprehensive global assessment has been made of all the species in these groups. For the United States, one of the few countries to assess freshwater mollusks and crustaceans comprehensively, 50% of known crayfish species and two-thirds of freshwater mollusks are at risk of extinction, and at least one in 10 freshwater mollusks are likely to have already gone extinct (Master et al., 1998). Freshwater fish A number of regional overviews of the status of freshwater fish are available, yet many of the existing overviews underestimate the number of species, as there are still many species to be described and assessed. There is, therefore, a high level of uncertainty about the status of fish in many inland waters with estimates of the number of freshwater fish in Latin America varying from 5000 to 8000; in tropical Asia and Africa, there are estimated 3000 species on each continent (Revenga and Kura, 2003), although these figures are almost certainly underestimates. It is estimated that in the last few decades more than 20% of the world’s 10 000 described freshwater fish species have become threatened or endangered or are listed as extinct (Moyle and Leidy, 1992). In the 20 countries for which assessments are most complete, an average of 17% of freshwater fish species are globally threatened (IUCN, 2003). Amphibians The recent Global Amphibian Assessment (IUCN et al., 2004) reported that the decline in conservation status of freshwater amphibians was worse than that of terrestrial species listed with 964 of 3908 freshwater species listed as threatened. Species associated with flowing water were found to have a higher risk of extinction than those associated with still water. Salamanders and newts have an even high level of threat (46% globally threatened or extinct) than frogs and toads (33%) and Caecilians (2%, although knowledge of these is poor, with only one-third assessed). Basins with the highest number of threatened freshwater amphibians include the Amazon, Yangtze, Niger, Parana, Mekong, Red, and Pearl in China, Krishna in India, and Balsas and Usumacinta in Central America, all of which have between 13 and 98 threatened freshwater species. Reptiles Van Dijk et al. (2000) reported that of the 200 species of freshwater turtles, 51% of the species of known status have been assessed as globally threatened, and the number of critically endangered freshwater turtles more than doubled in the four years preceding and that of the 90 species of Asian freshwater turtles and tortoises, 74% are considered globally threatened. Of the 23 species of crocodilians, which inhabit a range of wetlands including marshes, swamps, rivers, lagoons, and estuaries, four are critically endangered, three are endangered, and three are vulnerable (IUCN, 2003). There is very little information on the conservation status of aquatic snakes although IUCN (2003) reported that some semiaquatic snakes are vulnerable. Waterbirds Many waterbird species are globally threatening and the status of both inland and marine/coastal waterbirds is deteriorating faster than those in other habitats (Davidson and Stroud, 2004). Of the 35 bird families with species that are entirely or predominantly coastal/marine or inland wetland dependent, 20% of the 1058 species for which assessment data exist are currently globally threatened or extinct. Waterbirds dependent on freshwater ecosystems, especially those that also use marine and coastal ecosystems, have deteriorated in status faster than the average for all threatened species, but at similar rates for other migratory bird species. Shorebirds, many of which also use freshwater ecosystems, are declining worldwide with 48% of populations with a known trend declining. Other waterbirds in decline include cranes with 47% of populations with a known trend declining, rails (50%), skimmers (60%), darters (71%), ibis and spoonbills (48%), storks (59%), and jacanas (50%). Only gulls (18%), flamingos (18%), and cormorants (20%) appear to have a relatively healthy status. Mammals Although most mammals depend on freshwater for their survival, and many feed in rivers and lakes or live in close proximity to freshwater ecosystems, only a few are considered aquatic or semiaquatic mammals. Revenga and Kura (2003) provided an analysis of the status of aquatic and semiaquatic mammals. Some 37% of inland water-dependent mammals are globally threatened, compared with 23% of all mammals and includes otters (50% of species of known status threatened), seals (67% threatened), manatees (100% threatened), river dolphins and porpoises (100% threatened), and wetland-dependent antelopes (29% threatened) (Revenga et al., 2005).

Peatlands in particular have for centuries been converted for agriculture in many parts of the world, particularly in Europe, but also more recently in the highlands of South America and in parts of China, Southeast Asia, and Africa. In Southeast Asia, large areas of the once-extensive tropical peat swamp forests have been heavily degraded, and large extents have been lost over the last four decades as a consequence of logging for timber and pulp and conversion to oil palm plantations. The peatlands of Malaysia and Indonesia are especially threatened by drainage and forest clearing that then makes them susceptible to fire. Land clearing and subsequent uncontrolled fires in 1997 severely burned about 5 million

hectares of forest and agricultural land on the Indonesian island of Borneo (Page et al., 2002, 2009). Irrigated agriculture is another major driver of the loss and degradation of inland aquatic ecosystems with water withdrawals for irrigation worldwide resulting in major changes in river flows (Revenga et al., 2000) – flows that are essential for sustaining the ecosystem services and species that occur in inland aquatic ecosystems. The global extent of irrigated agricultural land has increased from approximately 138 million hectares in 1961 to 271 million hectares in 2000, and currently accounts for an estimated 40% of total food production even though it represents only 17% of global

Managing Aquatic Ecosystems

49

Dams

Overharvesting of wild resources,

interrupt the connectivity of river systems, disrupting fish spawning and migration. Dams with large reservoirs alter seasonal flood regimes and retain sediment needed to maintain the productivity of floodplain agriculture.

especially fish, is driven both by the subsistence needs of a growing population and by unsustainable commercial exploitation, threatening future food security and livelihoods.

River channelization and dredging for navigation reduces riverine habitat and alters flood patterns.

Large-scale irrigation and river diversions after natural flow regimes, reduce downstream water availablity for agriculture, and contribute to salinization through saltwater intrusion in the coastal zone.

Agricultural expansion is often achieved by converting natural inland water systems, reducing aquatic biodiversity and natural flood control functions, and increasing soil salinity through evaporation. When accompained by intensive use of agrochemicals, off-site pollution effects can be extensive.

Forest clearing in permanently or seasonally inundated zones, often motivated by unsustainable aquaculture production, dramatically reduces habitat for wild aquatic organisms. In the coastal zone, it also make the landscape much more susceptible to erosion.

Roads and flood-control infrastructure often interrupt wetland connectivity, disrupting aquatic habitat, reducing the function of wetlands to remove pollutants and absorb floodwaters, and potentially increasing the losses when high floods do occur.

Urban and industrial pollution, when released untreated into aquatic environments, reduces water quality, affecting the diversity and abundance of aquatic organisms as well as human health.

Figure 10 Pictorial representation of some of the direct drivers of change in inland and coastal aquatic ecosystems. Invasive species, climate change, and land conversion to urban or suburban areas affect all components of the catchment and coastal zone and are therefore not represented pictorially. From Millennium Ecosystem Assessment; adapted from Ratner BD, Ha DT, Kosal M, Nissapa A, and Chanphengxay S (2004) Undervalued and Overlooked: Sustaining Rural Livelihoods through better Governance of Wetlands, Studies and Review Series. Penang, Malaysia: World Fish Centre.

cropland area (Wiseman et al., 2003). In this respect, it seems to have a disproportionate negative impact on inland aquatic ecosystems relative to the land area involved. Possibly more significantly, around 66% of all water withdrawn for direct human use is now being used for agriculture (Scanlon et al., 2007). The problem of increasing abstraction of freshwater from aquatic ecosystems is exacerbated by the loss of much of this water from the immediate landscape – very little of the water returns as runoff to the rivers as most of it is evaporated or transpired (Falkenmark and Lannerstad, 2005). There are many well-documented examples where diversion of water for agriculture has caused a decline in the extent and degradation of inland aquatic ecosystems and their species richness (Revenga et al., 2000; Finlayson and D’Cruz, 2005; Dudgeon et al., 2005). Lake Chad provides an interesting example with major ecosystem change being due to

both human-induced and natural changes, with the loss of many species and ecosystem services as the lake shrank from about 25 000 km2 in surface area to one-twentieth of its size over 35 years at the end of the twentieth century as a consequence of a drier climate and high agricultural demands for water (Coe and Foley, 2001). This example illustrates the complexity that can arise when multiple drivers of change impact on an aquatic ecosystem, especially in areas of high climate variability.

1.03.4.2 Modification of Water Regimes Agricultural development and the diversion of water for irrigation, and increasingly for urban purposes, have modified the water regime in many inland aquatic ecosystems. Modifications include the construction of river embankments to

50

Managing Aquatic Ecosystems

Table 4

Alteration of inland freshwater systems worldwide

Alteration

Pre-1990

1900

Waterways altered for navigation (km) Canals (km) Large reservoira Number Volume (km2) Large dams (415 m high) Installed hydro-capacity (MW) Hydro-capacity under construction (MW) Water withdrawals (km3 yr 1) Wetlands drainageb (km3)

3125 8750

8750 21 250

41 14 – – – – –

581 533 – – – 578 –

1950–60 – – 1 105 1 686 5 749 o290 000 – 1 984 –

1985 4500 000 63 125 2768 5879 – 542 000 – 3200 160 000

1996–98

– 2836 6385 41 413 660 000 126 000 3800 –

a

Large reservoirs are those with a total volume of 0.1 km3 or more. This is only a subset of the world’s reservoirs. Includes available information for drainage of natural bogs and low-lying grasslands as well as disposal of excess water from irrigated fields. – Data not available. From Revenga C and Kura Y (2003) Status and Trends of Biodiversity of Inland Water Ecosystems, Technical Series No. 11. Montreal: Secretariat of the Convention on Biological Diversity. b

improve navigation, drainage of wetlands for agriculture, construction of dams and irrigation channels, and the establishment of interbasin connections and water transfers. Revenga and Kura (2003) provided data on the extent of alteration to inland freshwater systems worldwide (Table 4). These changes have had many beneficial outcomes for people through the provision of local flood control and hydropower, improved fisheries and increased agricultural output (Molden et al., 2007), but at the same time there have been many negative ecological effects on inland aquatic ecosystems (Revenga et al., 2000). Rivers have been disconnected from their floodplains and wetlands; seasonal changes in water flows have disrupted fish and bird migration and breeding; greater runoff in rivers has increased the likelihood and severity of flooding; and links with groundwater systems have been disrupted, and, in some coastal regions, enabled saline water to intrude on freshwater systems. They have also transformed many rivers through (1) the construction of large reservoirs, such as those on the Volta and Zambezi Rivers in Africa, or along the Volga River in Russia; (2) the embankment and channelization of rivers such as that along the Mississippi and Missouri rivers in the United States; or (3) significantly reduced flows to floodplains and downstream ecosystems, including deltas such as the Indus in Pakistan, or the lakes at the mouth of the Murray River in Australia. Even the large inland seas are not safe from the impacts of river regulation and diversion of water away from terminal water bodies. The Dead Sea located in the Syrina-African rift valley at the southern outlet of the Jordan River and at 417 m below sea level is the world’s saltiest large water body. It is threatened by excessive withdrawal of water from the river to support industrial, agricultural, and tourism development (ILEC and UNEP, 2003). The annual historical flow of the Jordan River to the Sea was about 1285 million cubic meters in the 1950s compared to 505 in mid-1970s, 275 in 2000s, and a projected 170 million cubic meters in the mid-2020s (Courcier et al., 2005). In addition to the reduction in water flows, the diversion of water has resulted in the development of a complex socioeconomic system (Figure 11) with undoubted economic and political ramifications associated with any proposals for

further development and regulation of water flows, including management of wastewater and irrigated agriculture (Courcier et al., 2005). The Mesopotamian marshlands also provide another example of a complex social–political scenario associated with river regulation and the restoration of an intermixed social and ecological system. The marshes have been severely affected by river regulation in recent decades with the original areas of 15 000–20 000 km2 before being reduced by drainage and dam construction along the Tigris and Euphrates rivers to less than 400 km2 (Partow, 2001). The total capacity of the reservoirs along these rivers exceeds the annual discharge of both rivers, drastically reducing the supply of flood waters that were so important for delivering sediments and nutrients to the marshlands. In addition to regulating the flows along the rivers by the construction of dams in the upstream reaches, attempts were made in the early 1990s to drain water away from the marshes through large canals. The combined effect of these moves was to reduce the extent of the marshes and threatened the culture and biodiversity that depended on the annual flooding regime. More recently, there have been partially successful but still insufficient efforts to restore parts of the marshes by breaking banks and flooding some 20% of the original area of marsh (Richardson et al., 2005). While these efforts have indicated the potential for further successful restoration, Falkenmark et al. (2007) have cautioned that attempts to return water to the central areas of the marshes upstream of the confluence of the Tigris and Euphrates could generate adverse impacts on aquatic ecosystems further downstream. That is, without an increase in the amount of water available for flooding the marshes, simply returning the water to upstream areas may not be enough to restore them and could further reduce the flow of water to downstream areas and possibly further reduce the flow to the Persian Gulf. Richardson et al. (2005) have also reported the construction of a dike along the Iraq/Iran border that could further reduce the flow of water into the marshes. The construction of large dams has doubled or tripled the residence time of river water with impacts on suspended sediment and carbon fluxes, waste processing, and aquatic

Managing Aquatic Ecosystems

1950

2000s

51

Mid-1970s

Mid-2000s

Figure 11 Changes in water resources and their allocation in the lower Jordan basin from the 1950s to the mid-2020s. Reproduced from Courcier R, Venot J-P, and Molle F (2005) Historical Transformations of the Lower Jordan River Basin (in Jordan): Changes in Water Use and Projections (1950– 2025). Colombo, Sri Lanka: International Water Management Institute.

52

Managing Aquatic Ecosystems

habitat, and has resulted in fragmentation of the river channels with 37% of 227 river basins around the world strongly affected by fragmentation and altered flows, 23% moderately affected and 40% unaffected (Revenga et al., 2000). Small dams can also have major effects on the ecological condition of inland aquatic ecosystems. The debate about the construction of dams is ongoing (WCD, 2000). The effects of modification of flow regimes on fish migrations have been reviewed by Revenga and Kura (2003) with direct impacts on diadromous fish species such as salmon being well known and increasingly recognized, whereas the indirect impacts of flow alteration, such as the reduction of floods and loss of lateral connections on floodplains, are not always as evident. In many cases, construction of dams has resulted in the disappearance of fish species adapted to river systems and the proliferation of species adapted to lakes. Changes in the fish are indicative of many changes in the biodiversity of regulated rivers and associated aquatic ecosystems, although the extent of data and information about the wider biodiversity is often inadequate or lacking (Revenga et al., 2005).

1.03.4.3 Invasive Species Despite the current concern about invasive species in inland aquatic ecosystems their importance has not always been as widely appreciated (Finlayson, 2009). This was in part because the problem of invasive species in these ecosystems was seen largely as one for developed countries, despite the paradoxical occurrence of well-documented cases of invasive species in African wetlands and lakes (e.g., Salvinia molesta and Nile perch – Lates nilotica). The reasons for this situation are not clear, although they probably included insufficient awareness and information about the impacts of these species and ways of controlling them. However, several recent assessments (such as the Millennium Ecosystem Assessment (2005)) and initiatives (such as the Global Invasive Species Programme) have demonstrated that the loss of wetland species as a consequence of invasion by invasive species is now much more of a concern globally (Revenga et al., 2005; Finlayson, 2009). The spread and establishment of non-native invasive species in inland aquatic ecosystem have caused many changes to the native biota and are likely to become even more common

Box 4

with the further development of aquaculture, interbasin transfers of water, and shipping and global commerce. There are many documented and well-known examples of plant species that have successfully invaded and established in inland aquatic ecosystems, including the pan-tropical weeds salvinia (Salvinia molesta) and water hyacinth (Eichhornia crassipes) that originated in South America but are now widely established in many countries. In many instances though, the occurrence of alien plant species may not be seen as undesirable, as shown by the establishment of the alien species Egeria densa in the Rio Cruces wetland in Chile where it was considered to be an ecological engineer and thrived and provided the mainstay to support a population of blacknecked swans (Cygnus melancoryphus) which was highly appreciated and valued by local residents (Yarrow et al., 2009). The angst in this case came not from the establishment of the alien invasive plant but from its decline when its biomass crashed suddenly in 2004 and the swans dispersed to other wetlands, leaving an acrimonious debate about the cause of the population crash (Delgado et al., 2009). The example of Canadian pond weed (Elodea canadensis) outlines many of the dilemmas raised by introduced species (Sculthorpe, 1967). It originated in North America and invaded the waterways of Europe in the late nineteenth century where it grew rapidly and spread vegetatively to reach a maximum population density within a period of a few months to 4 years. This population level was maintained for up to 5 years but then declined to levels that were not considered a nuisance. The reasons for the rapid increase and decline were not determined. Many animal species, both large and microscopic, have also invaded inland aquatic ecosystems, such as those outlined for European lakes (Box 4). The larger invasive animals include the cane toad (Bufo marinus), bullfrog (Rana catesbeiana), European domestic pig (Sus scrofa), carp (Cyprinus carpio), and zebra mussel (Dreissena polymorpha) that have become established outside of their native range and disrupted the inland water systems that they have invaded. Many fish species have been spread beyond their native ranges often in response to demands for aquaculture and aquarium species. Fish introductions have usually been done to enhance food production and recreational fisheries or to control pests such as mosquitoes and aquatic weeds. The spread of trout and salmon

Invasive species in European rivers. Based on information supplied by H. Ketelaars.

The construction of canals between rivers and other water bodies in Europe over the past two centuries has provided channels for the migration of many aquatic species, whether they migrated themselves or were carried by on the hulls of ships or in their ballast water. The Volga–Baltic Waterway, reconstructed in 1964 to connect the Caspian with the Baltic enabled the translocation of many aquatic species, including copepods, rotifers, the onychopod Bythotrephes longimanus, and several fish species to the Volga basin. The Main-Danube Canal, officially opened in 1992, is another that allowed many Ponto-Caspian invertebrate species to reach the Rhine basin and from there to disperse to other basins, mainly in ballast water. Intentional introductions of aquatic species have occurred mainly in the past two centuries. The North American amphipod Gammarus tigrinus was deliberately introduced in 1957 to Werra and Weser rivers in Germany where the local gammarid fauna had disappeared due to excessive chloride pollution. The mysid Mysis relicta was been introduced to many Scandinavian lakes to stimulate fish production. Three North American introduced crayfish species have established themselves in many European waters and with the introduced crayfish plague (Aphanomyces astaci) have almost eliminated the native crayfish (Astacus astacus). At least 76 freshwater fish species have been introduced into European fresh waters, with approximately 50 establishing self-sustained populations. When introductions between areas within Europe are also considered, the number of introduced fish species is more than 100. The numerically most important families are cyprinids and salmonids, of which grass carp (Ctenopharyngodon idella), silver carp (Hypophthalmichthys molitrix), rainbow trout (Oncorhynchus mykiss), and brook char (Salvelinus fontinalis) are now widely distributed.

Managing Aquatic Ecosystems

species for sports fishing is well known. The introduction of alien fish species has though often resulted in major ecological change, including the collapse of native fish populations. Finlayson and D’Cruz (2005) provided a summary of impacts of some invasive species, including the adverse impact of salmonids on the genetic diversity of wild stocks in many countries and the spread of tilapia species into Central and Southern America and parts of Asia. Herbivorous and omnivorous species, such as Indian, Chinese, and common carp, account for the majority of introductions in tropical Asia (Revenga and Kura, 2003). In many cases, the impact of invasive species on the native fish has not been documented. Finlayson (2009) noted that while there are some obvious examples of species that have established outside their native range, it should not be assumed that all newly established species have been transplanted by human activities or are invasive. Recent concerns over global climate change and variability provide a scenario where it may no longer be possible to attribute the occurrence of new species to natural fluctuations versus human activity.

1.03.4.4 Overfishing Inland fisheries are a major source of protein for a large proportion of the world’s population with the global production of fish and fishery products from inland waters in 2002, amounting to 32.6 million tonnes with 8.7 tonnes from wild capture and the rest from aquaculture (FAO, 2004). While inland fisheries have increased, FAO (1999) also reported that most inland fisheries that relied on natural reproduction of fish stock were overfished or being fished at their biological limit. Arthurton et al. (2007) also reported that inland fish stocks were subjected to a combination of direct pressures, including habitat alteration, and loss, altered flows, and habitat fragmentation due to dams and other infrastructure, and also faced problems from pollution, exotic species, and overfishing. With much of inland fisheries catches destined for subsistence consumption or local markets, food demand for growing populations is a major factor driving exploitation levels in inland waters. Monitoring of the extent of fishing in inland aquatic systems also seems to be underreported by a factor of 2 or 3, due to the large volume of harvest that is consumed locally, and remains unrecorded (Allan et al., 2005). Even with discrepancies in the data, it is well established that inland fisheries are extremely important in Asia and Africa and in 2002 accounted for 90% of the inland fish catch (FAO, 2004). China alone accounted for at least one-quarter of the inland catch, followed by India (9%), Bangladesh (8%), and Cambodia (4%) (FAO, 2004). The importance of aquaculture as a component of inland fish supply is shown by continued growth at an average rate of nearly 9% per year since 1970 – a much higher rate than that for capture fisheries (B1%) (FAO, 2004). Almost 58% of this came from China with an average annual increase of 11% between 1970 and 2000, compared with 7% for the rest of the world (FAO, 2004). However, many aquaculture operations, depending on their design and management, can contribute and have contributed to habitat degradation, pollution,

53

introduction of exotic species, and the spread of diseases through the introduction of pathogens (Naylor et al., 2000). There is ample evidence that overfishing is a significant factor in the decline of numerous species and fisheries, and is of global importance as a threat to inland water biodiversity (Allan et al., 2005). There are two main types of overfishing with intensive fishing of a targeted species leading to marked declines in catch per unit effort and size of individuals captured, while assemblage or ecosystem overfishing leads to sequential declines of species and depletion of individuals and species of large size, especially piscivores, and declines in the mean trophic level of the assemblage and changes in the responsiveness of populations to environmental fluctuations. The historic influence of overharvesting of fish is shown by the decline of the Murray cod of the Murray-Darling river system in Australia, some sturgeon stocks of Eurasia, the tilapiine species Oreochromis esculentus and Oreochromis variabilis of Lake Victoria, and perhaps the Pacific salmon of the Columbia River while the decline of the Mekong giant catfish and the Nile perch of Lake Victoria provide contemporary cases (Allan et al., 2005; Box 5). The consequences of eliminating fish species from inland aquatic ecosystems are likely to be numerous and of varying severity with the progressive reduction in assemblage diversity meaning that fewer species are available to perform critical functions in the ecosystem with dire consequences following the loss of a species with a disproportionately strong influence on nutrient, habitat, or assemblage dynamics (Allan et al., 2005).

1.03.4.5 Water Pollution and Eutrophication Vo¨ro¨smarty et al. (2005) noted that attempts to summarize patterns and trends in the quality of inland waters, particularly at a global scale, encompassed an array of challenges that included basic definitional problems, a lack of worldwide monitoring capacity, and an inherent complexity in the chemistry of both natural and anthropogenic pollutants. Furthermore, despite improvements in analytical methods the capacity to monitor trends in water quality is limited in terms of the spatial coverage, frequency, and duration of monitoring data. Data comparability was yet another constraint while the monitoring of groundwater was more problematic than surface water. The Millennium Ecosystem Assessment (2005) highlighted changes in the global nitrogen cycle with the loading of reactive nitrogen to the landmass having doubled from 111 million to 223 million tonnes per year with the greatest increases in North America, continental Europe, and South and East Asia (Green et al., 2004). This has caused the transport of dissolved inorganic nitrogen in rivers to increase from about 2–3 million tonnes per year from preindustrial times to about 15 million tonnes today, especially in drainage basins that are heavily populated or supporting extensive industrial agriculture. While rivers and wetlands can assimilate some nitrogen, the self-purification capacity is not unlimited and the water quality in many has deteriorated, resulting in eutrophication, harmful algal blooms, and high levels of nitrate in drinking water (Malmqvist and Rundle, 2002). Jorgensen et al. (2001) reported that eutrophication was a widespread problem in lakes and reservoirs and also one of the most

54

Managing Aquatic Ecosystems

Box 5 Nile Perch invasion in Lake Victoria, East Africa. Based on information from Howard (2009) Case study 1 Nile Perch invasion in Lake Victoria. In: Finlayson CM (2009) Biotic pressures and their effect on wetland functioning. In: Maltby E and Barker T (eds.) The Wetlands Handbook, pp. 674–676. Oxford: Wiley-Blackwell. The Nile perch (Lates niloticus) is a large predatory fish that can grow to 1.8 m in length and weigh as much as 200 kg. It is native to the White Nile River system in Uganda, Sudan, Ethiopia, and Egypt and was introduced into Lake Victoria in the 1950s and 1960s. At the time the lake contained around 350 species, including 300 endemic cichlids of the subfamily Haplochromiinae that occupied many niches in the lake and its wetlands. The Nile perch was introduced to Lake Victoria to boost its fishery, but was hardly seen in catches until the late 1970s with the entire lake yielding less than 25 000 tonnes in 1981 but rising to 363 000 tonnes in 1993. At the same time, the total catch of all species rose from around 100 000 tonnes in 1979 to about 500 000 tonnes in 1989 with the proportion of Nile perch rising from less than 0.1% in 1974 to more than 50% 20 years later. Over the same period the lake lost as many of 50% of its species of haplochromines and became dominated by three species, Nile perch, the introduced Tilapia nilotica, and a native cyprinid (sardine) Rastrineobola argentea. As other changes occurred in the Lake at the same time, such as the advent of the invasive alien water hyacinth (Eichhornia crassipes), it is difficult to attribute changes in diversity of the haplochromines solely to the Nile perch. There are a number of probable causes of the dramatic population increase of the Nile perch: 1. 2. 3. 4. 5.

lack of other wide-ranging predators (competitors); a fast growth rate and reproductive potential; a great range of body size during development which permits exploitation of various habitats in the lake; changes in the lake ecosystem resulting from human activities (e.g., eutrophication); and adaptability of the perch to different sources of food.

It is still not clear though whether or not the population of Nile perch and the wider fish structure in Lake Victoria has reached (or is anywhere near) stability.

difficult to abate, and cyanobacteria blooms have increased and are a major problem worldwide. The extent of water pollution from point sources is well known with an estimated 90% of wastewater in developing countries being discharged directly to rivers and streams without any waste processing treatment, and in some locations both surface- and groundwater have been so polluted that they are unfit even for industrial use (WMO, 1997). The agricultural sector also contributes a large amount, although this is usually from diffuse sources (Verhoeven et al., 2006). It is also well known that pollution from point sources such as mining has had many devastating impacts on inland waters in many parts of the world; for example, the spillage in 1998 of an estimate 5.5 million cubic meters of stored tailings (mine wastes) from the Aznalcollar mine nearly 50 km from the Don˜ana National Park in Spain spread over 46 000 km of downstream habitat with fatal consequences for much of the biota (Bartolome and Vega, 2002). The cost of removing the tailings and contaminated soil reached about 3.8 billion Euros. Meybeck (2003) provided an overview of water pollution problems for inland waters (Table 5). This showed that in industrial countries fecal contamination has been largely eliminated, while new problems, particularly from agriculture runoff, were increasing. In other countries, this was not the case and fecal contamination was still a major problem with urban and industrial pollution sources increasing faster than wastewater treatment. Contamination by pesticides has increased rapidly since the 1970s, with many different substances being involved. Vo¨ro¨smarty et al. (2005) concluded that since the 1990s the water-quality situation in most developing countries and countries in transition was likely to be worse in terms of overall water quality. In Eastern Europe, Central and South populated Americas, China, India, and populated Africa, it was probably worse for metals, pathogens, acidification, and organic matter, while there were slight improvements for the

Table 5 Major water quality issues in inland aquatic ecosystems at the global scale Issue

Rivers Lakes Reservoirs Groundwaters

Pathogens Suspended solids Decomposable organic matter Eutrophication Nitrate Salinization Trace metallic elements Organic micropollutants Acidification

XXXX XXX XXXX

XX NA XX

XX XX XXX

XXX NA XX

XX XX XX XXX

XXX X X XXX

XXXX X XX XXX

NA XXXX XXXX X

XXXX XX

XXX XX

XXX XXX

XXXX X

XXXX, severe or global deterioration observed; XXX, important deterioration; XX, occasional or regional deterioration; X, rare deterioration; NA, not applicable. Information from Meybeck M (2003) Global analysis of river systems: From Earth system controls to Anthropocene syndromes. Philosophical Transactions of the Royal Society London B 358: 1935–1955.

same issues in Western Europe, Japan, Australia, New Zealand, and North America. Nitrate though was generally still increasing everywhere, as it has since the 1950s. In the former Soviet Union there seems to have been an improvement in water quality as a consequence of the decline of industrial activities, whereas in Eastern Europe there have also been some improvements, such as those in the Danube and the Elbe basins. A few rivers, such as the Rhine, have seen a stabilization of nitrate loads after 1995. Arthurton et al. (2007) reported that water-quality degradation from human activities continued to degrade inland aquatic ecosystems and affected the health of many people. Pollutants of primary concern included microbial pathogens and excessive nutrient loads with the latter leading to eutrophication of downstream and coastal waters, and loss of

Managing Aquatic Ecosystems

55

Box 6 Waterbirds and climate change. Based on information reported in Finlayson CM, Gitay H, Bellio MG, van Dam RA, and Taylor I (2006) Climate variability and change and other pressures on wetlands and waterbirds – impacts and adaptation. In: Boere G, Gailbraith C, and Stroud D (eds.) Water Birds around the World, pp. 88–97. Edinburgh: Scottish Natural Heritage. While the general nature of the impacts of climate change on waterbirds can be described there is less certainty when it comes to identifying the extent, intensity, and time frames for such changes. It is difficult to predict with great certainty as the models used for global climate change projections are still very coarse and the ecological relationships between waterbirds and climate and aquatic ecosystems is insufficiently known. The most severe effects and those most likely to occur earliest include: 1. the loss of intertidal areas and increased salinity of coastal freshwater wetlands caused by rising sea levels; 2. a reduction in the extent of wetlands and duration of flooding in arid and semiarid areas from changes in rainfall; and 3. the loss of wetland breeding areas in the Arctic and sub-Arctic areas caused by increasing temperatures, expanding boreal forests and fires.

The extent of loss of intertidal habitats and its effects on coastal waterbirds, many of which also frequent inland aquatic ecosystems, will depend on the ability of coastal environments to migrate inland as sea level rises. The effects of rising temperatures on plant communities will be particularly strong in the Arctic with an expected expansion of the boreal forest into the tundra areas where two-thirds of all goose and 95% of all Calidrid sandpipers breed. The impacts of habitat loss could be offset to some extent by rising temperatures increasing productivity and breeding success; however, these may also be affected by an increase in loss of nests and chicks to predation as predators such as the Red Fox (Vulpes vulpes) expand their range. This example illustrates both the complexity of the changes that may occur as well as the complexity of identifying the many interactions that may occur within an ecosystem or between species. As changes in global circulation patterns will result in changes to rainfall patterns, with some areas experiencing increases and others decreases. The latter in particular may be extremely detrimental to waterbirds in areas that are already dry and subject to drought, such as parts of Australia, Asia, and Africa. As wetlands and waterbirds in these areas are already highly stressed from the impacts of agriculture, reduced water flows, pollution, and increasing salinization, they may be highly vulnerable to changes in the climate. As reduced rainfall will increase the intervals between flooding events and shorten their duration there could be reduced breeding success and recruitment of waterbird species that formerly depended on flooding events of sufficient duration to stimulate breeding and enable fledging. Reduced rainfall and flooding across large areas of arid land will particularly affect bird species that rely on a network of wetlands that are alternately or even episodically wet and fresh or drier and saline. While the exact nature of changes cannot be confirmed there will almost certainly be great regional variation, with some areas experiencing increases in waterbird populations and others, decreases. The fragmentation of rivers and wetlands or the disruption or loss of migration corridors will affect the manner in which waterbirds (and other species) will be able to respond and adapt.

beneficial human uses. Pollution from diffuse land sources such as agriculture and urban runoff was also of concern.

1.03.4.6 Climate Change It is increasingly expected that global climate change will increase the pressure on inland aquatic ecosystems in many locations both directly, especially through increased temperatures, changes in snowmelt and runoff, and in places severely decreasing rainfall, as well as indirectly by interacting with existing pressures and drivers of change (Revenga and Finlayson, 2009; Finlayson et al., 2006; Finlayson and D’Cruz, 2005). Revenga and Finlayson (2009) anticipated that most pronounced impacts from climate change will come from increased temperatures and changes in precipitation, and these will not affect all wetlands in the same way. As a consequence some aquatic ecosystems and their catchments will be drier, while others will experience more rainfall and storms, or even more intense and fewer storms. High-altitude wetlands seem to be particularly vulnerable as the annual and previously predictable glacier-melt decreases; freshwater systems near the coast are susceptible to rise in sea level and salinization. Many combinations of temperature increase and precipitation changes will affect the frequency, duration, and timing of peak floods or base-flows in rivers with subsequent impacts on aquatic species that are sensitive to changes in water flow for migration, breeding, and feeding. Further assessment is needed to ascertain the vulnerability of particular ecosystems and species. The latter has been summarized for waterbirds by

Finlayson et al. (2006) and outlined in Box 6. However, in many instances, the certainty with which we can attribute cause and effect of climate change is undermined by the extent of existing data and knowledge. There is some confidence that many inland aquatic ecosystems are vulnerable to climate change with those at high latitudes and altitudes, such as Arctic and sub-Arctic bog communities, or alpine streams and lakes being highly vulnerable (Gitay et al., 2002; Finlayson et al., 2006; Revenga and Finlayson, 2009), as well as those that are isolated or are low-lying and adjacent to coastal wetlands (Bayliss et al., 1997; Pittock et al., 2001). Danielopol et al. (2003) expected groundwater ecosystems to change as recharge of aquifers is affected by rainfall and runoff. Besides changes in waterbird populations, the warming of inland waters could affect chemical and biological processes, reduce the amount of ice cover and dissolved oxygen in deep waters, alter the mixing regimes, and affect the growth rates, reproduction, and distribution of organisms and species (Gitay et al., 2002). The rise in sea levels will affect freshwater ecosystems in low-lying coastal regions and plant species that are not tolerant to increased salinity or inundation could be eliminated. Changes in the vegetation will affect both resident and migratory animals, especially if these result in a major change in the availability of staging, feeding, or breeding grounds for particular species. The impacts on the distribution of fish species could be profound with cold-water fish being further restricted in their range, and cool- and warm-water fish potentially expanding their range, and even moving poleward. These examples illustrate the statements made at the outset of this discussion that the extent of change in inland aquatic

56

Managing Aquatic Ecosystems

ecosystems due to the climate change should not be addressed in isolation of other drivers of change, as many of the adverse effects of the above-mentioned drivers of change will be exacerbated by climate change.

• •

1.03.5 Management Responses The information on management responses is largely paraphrased from that provided by Finlayson and D’Cruz (2005) in their assessment of the condition of inland waters for the Millennium Ecosystem Assessment. As their assessment was drawn from the published literature available at the time, it provides a widespread opinion as well as various options and opportunities for sustainable use and, where necessary, rehabilitation of inland aquatic ecosystems. Particular technical responses for specific wetlands or many drivers of change outlined in the text above are not provided – these are on the whole sufficiently well known, or could be developed through the application of the processes outlined below.

1.03.5.1 Integrated Management Processes The management of inland aquatic ecosystems worldwide has often been based on sectorally based decision-making mechanisms that have not included sufficient consideration of the wider implications or outcomes of specific actions. The information provided in the text above illustrates many adverse outcomes of past sectorally based management decisions. In many instances, these decisions have not adequately considered the trade-offs between the multiple uses and values of inland aquatic ecosystems and have too often resulted in the degradation of these ecosystems. The development of more multi-sectorally based responses and decisions is strongly encouraged as way to reverse the past loss and degradation of inland aquatic ecosystems and the decline in the ecosystem services that they deliver.

1.03.5.2 International Cooperation and Action The past loss and degradation of inland aquatic ecosystems have been recognized through international conventions and treaties. The Ramsar Convention on Wetlands has provided leadership and worked collaboratively with many other organizations, both informally and through formal agreements, to develop multisectoral approaches to stop and reverse the loss and degradation of wetlands. This includes working collaboratively to reduce the rate of loss of biodiversity and restore degraded wetlands. The Mediterranean wetland (MedWet) program is an example of a collaborative initiative that has supported actions to halt and reverse the loss and degradation of wetlands. The declaration behind this initiative was made in February 1991 and contained many recommendations that are still important today. These covered:

• •

identification of priority sites for wetland restoration and rehabilitation, and the development and testing of techniques for their complete rehabilitation; evaluation of existing and proposed policies to determine how they affect wetlands;

• •

increased institutional capacity to conserve and effectively manage wetlands through vigorous education and training programs; integrated management of all activities concerning wetlands, their support systems, and the wider area surrounding them carried out by properly funded and wellstaffed multidisciplinary bodies with active participation of representatives of government, local inhabitants, and the scientific and nongovernmental communities; open consultation and free flow of information when managing wetlands; and adoption and enforcement of national and international legislation for better management.

These recommendations have since been repeated or extended in many forums and with widespread acceptance, although the sentiment behind these have not always been transferred to on-ground actions and outcomes.

1.03.5.3 Restoration and Wise Use of Wetlands The concept of replacing lost wetlands has received increasing support in recent decades and more attention is now directed toward wetland restoration worldwide. However, current rates of restoration are inadequate to offset the rate of wetland loss in many regions – more is required even as efforts are undertaken to stop further loss. In support of these efforts, the Ramsar Convention has provided a suite of guidance for the wise use of wetlands covering national wetland policies; laws and institutions; river basin management; participatory management; wetland communication, education, and public awareness; management planning; international cooperation; wetland inventory, assessment and monitoring; water allocation and management; coastal management; and management of peatlands. One of the key barriers in preventing further loss and degradation of wetlands is the seeming unwillingness of parties to the above-mentioned collaborative initiatives and declarations to undertake effective actions. Sufficient knowledge is generally now available to know what actions are required to stop further loss and degradation, although there seems to be an inadequate adoption and understanding of ecosystem approaches for managing inland aquatic ecosystems, especially when dealing with water allocations. Ongoing dialog about the allocation of water for environmental outcomes in rivers and associated wetlands is still needed – these also need to address the trade-offs that are needed to support equitable outcomes and support many services that inland aquatic ecosystems provide to wider society and the inordinate costs associated with reinstating these once they have been lost.

1.03.5.4 Supporting Local Community Involvement in Management There has been increased interest in the development of mechanisms to support the capacity of local communities to contribute to the management of inland aquatic ecosystems. This can particularly be important where local knowledge and experience can be directly applied to local management issues, but can also support wider strategic planning. Recognition of the beneficial outcomes that can occur when local people are

Managing Aquatic Ecosystems

involved in the management of inland waters and their services has long underpinned efforts by the Ramsar Convention and its partners to encourage best management practices for wetlands. Participatory management and the involvement of local communities in management planning are implicit in the guidance provided by the Convention covering policy and legal instruments, economic and social interactions, and technical tools. The challenge for the Convention and others is to ensure that such instruments and tools are used effectively and as often as possible. This can be done by adopting an adaptive management approach which incorporates active learning mechanisms, the involvement of key stakeholders, and the balancing of vested interests.

1.03.6 Conclusions In drawing the above to a close reference is again made to the Ramsar Convention on Wetlands and the Millennium Ecosystem Assessment. For over 35 years, the Convention has recognized the interdependence of people and their environment. It has promoted the wise use of wetlands as a means of maintaining their ecological character – the ecosystem components and processes that comprise the wetland and underpin the delivery of ecosystem services, such as freshwater and food – and strongly supported the Millennium Ecosystem Assessment with a set of key messages about wetlands and their importance for people. These messages are shortened and paraphrased below as a way of concluding the above review of the condition and management of inland aquatic ecosystems:

•

•

•

•

•

•

Wetlands deliver a wide range of ecosystem services that contribute to human well-being, such as fish and fiber, water supply, water purification, climate regulation, flood regulation, coastal protection, recreational opportunities, and, increasingly, tourism. A priority when making decisions that directly or indirectly influence wetlands is to ensure that information about the full range of benefits and values provided by different wetland ecosystem services is considered. The degradation and loss of wetlands is more rapid than that of other ecosystems. Similarly, the status of both freshwater and coastal wetland species is deteriorating faster than those of other ecosystems. The primary indirect drivers of degradation and loss of inland wetlands have been population growth and increasing economic development. The primary direct drivers of degradation and loss include infrastructure development, land conversion, water withdrawal, eutrophication and pollution, overharvesting and overexploitation, and the introduction of invasive alien species. Cross-sectoral and ecosystem-based approaches to wetland management – such as river (or lake or aquifer) basin-scale management – that consider the trade-offs between different wetland ecosystem services are more likely to ensure sustainable development than many existing sectoral approaches. Major policy decisions in the next decades will have to address trade-offs among current uses of wetland resources

•

•

57

and between current and future uses. Particularly important trade-offs involve those between agricultural production and water quality, land use and biodiversity, water use and aquatic biodiversity, and current water use for irrigation and future agricultural production. The adverse effects of climate change will lead to a reduction in the services provided by wetlands. Removing the existing pressures on wetlands and improving their resiliency are the most effective methods of coping with the adverse effects of climate change. The Millennium Ecosystem Assessment conceptual framework for ecosystems and human well-being provides a framework that supports the promotion and delivery of the Ramsar Convention’s wise use concept. This enables the existing guidance provided by the Convention for the wise use of all wetlands to be expressed within the context of human well-being and poverty alleviation.

References Allan JD, Abell R, Hogan Z, et al. (2005) Overfishing of inland waters. BioScience 55: 1041--1051. Arthurton R, Barker S, Rast W, et al. (2007) Water. In: United Nations Environment Programme (ed.) Global Environment Outlook 4 – Environment for Development, pp. 115--156. Nairobi: United Nations Environment Program. Aselmann I and Crutzen PJ (1989) Global distribution of natural freshwater wetlands and rice paddies, and their net primary productivity, seasonality and possible methane emissions. Journal of Atmospheric Chemistry 8: 307-358. Bartolome J and Vega I (2002) Mining in Donana: Learned Lessons. Madrid: WWF Spain. Bayliss BL, Brennan KG, Eliot I, et al. (1997) Vulnerability assessment of the possible effects of predicted climate change and sea level rise in the Alligator Rivers Region, Northern Territory, Australia. Supervising Scientist Report 123. Canberra: Supervising Scientist. Blanco DE and de la Balze VM (eds.) (2004) Los Turbales de la Patagonia: Basespara su inventario y la conservacion de su biodiversidad, Publication No. 19. Buenos Aires, Argentina: Wetlands International. Brag O, Lindsay R, Risager M, Silvius M, and Zingstra H (2003) Strategy and Action Plan for Mire and Peatland Conservation in Central Europe. Wageningen: Wetlands International. Brown K, Mackensen J, and Rosendo S (2005) Integrated responses. In: Chopra K, Leemans R, Kumar P, and Simons H (eds.) Ecosystems and Human Well-Being: Policy Responses, vol. 3, pp. 425--465. Washington, DC: Island Press. Brundtland GH (1987) Our Common Future: Report of the World Commission on Environment and Development. Oxford: Oxford University Press. Clausnitzer V and Jodicke R (2004) Guardians of the watershed: Global status of dragonflies: Critical species, threat and conservation. International Journal of Ordonatology 7: 1388--7890. Coe MT and Foley JA (2001) Human and natural impacts on the water resources of the Lake Chad Basin. Journal of Geophysical Research (Atmospheres) 106: 3349--3356. Courcier R, Venot J-P, and Molle F (2005) Historical Transformations of the Lower Jordan River Basin (in Jordan): Changes in Water Use and Projections (1950– 2025). Colombo, Sri Lanka: International Water Management Institute. Dahl TE (2000) Status and Trends of Wetlands in the Conterminous United States 1986–1997. Washington, DC: US Department of the Interior, Fish and Wildlife Service. Dahl TE and Johnson CE (1991) Status and Trends of Wetlands in the Conterminous United States, Mid-1970’s to Mid-1980’s. Washington, DC: US Department of the Interior, Fish and Wildlife Service. Davidson NC and Stroud DA (2004) African–Eurasian flyways: Current knowledge, population status, and future challenges. In: Boere GC, Galbraith CA, and Stroud DA (eds.) Proceedings of the Waterbirds around the World Conference., pp. 63–73. Edinburgh, April 2004. London: Her Majesty’s Stationary Office. Davis TJ (ed.) (1993) Towards the Wise Use of Wetlands. Gland, Switzerland: Ramsar Convention Bureau.

58

Managing Aquatic Ecosystems

Danielopol DL, Griebler C, Gunatilaka A, and Notenboom J (2003) Present state and future prospects for groundwater ecosystems. Environmental Conservation 30: 104--130. Delgado LE, Marı´n VH, Bachmann PL, and Torres-Gomez M (2009) Conceptual models for ecosystem management through the participation of local social actors: The Rı´o Cruces wetland conflict. Ecology and Society 14(1): 50. http:// www.ecologyandsociety.org/vol14/iss1/art50/ (accessed August 2010). Dugan P (ed.) (1993) Wetlands in Danger. New York: Oxford University Press. Dudgeon D, Arthington AH, Gessner MO, et al. (2005) Freshwater biodiversity: Importance, threats, status and conservation challenges. Biological Review 81: 163--182. Falkenmark M and Lannerstad M (2005) Consumptive water use to feed humanity – curing a blind spot. Hydrology and Earth System Sciences 9: 15--28. Falkenmark M, Finlayson CM, and Gordon L (2007) Agriculture, water, and ecosystems: Avoiding the costs of going too far. In: Molden D (ed.) Water for Food, Water for Life: A Comprehensive Assessment of Water Management in Agriculture, pp. 234--277. London: Earthscan. FAO (1999) Review of the State of World Fishery Resources: Inland Fisheries, 1999. FAO Fisheries Circular, No. 942. Rome: FAO. FAO (2004) The State of the World Fisheries and Aquaculture 2004. Rome: FAO Fisheries Department. Finlayson CM (2009) Biotic pressures and their effect on wetland functioning. In: Maltby E and Barker T (eds.) The Wetlands Handbook, pp. 667--688. Oxford: Wiley-Blackwell. Finlayson CM and D’Cruz R (2005) Inland Water Systems Millennium Ecosystem Assessment, Volume 2: Conditions and Trends. Washington, DC: Island Press. Finlayson CM, D’Cruz R, and Davidson NJ (2005) Ecosystem Services and Human Well-Being: Water and Wetlands Synthesis. Washington, DC: World Resources Institute. Finlayson CM, Davidson NC, Spiers AG, and Stevenson NJ (1999) Global wetland inventory – status and priorities. Marine and Freshwater Research 50: 717--727. Finlayson CM, Gitay H, Bellio MG, van Dam RA, and Taylor I (2006) Climate variability and change and other pressures on wetlands and waterbirds – impacts and adaptation. In: Boere G, Gailbraith C, and Stroud D (eds.) Water Birds around the World, pp. 88--97. Edinburgh: Scottish Natural Heritage. Finlayson CM, Hollis GE, and Davis TJ (eds.) (1992) Managing Mediterranean Wetlands and Their Birds, IWRB Special Publication No. 20. Slimbridge: International Waterbird and Wetlands Research Bureau. Finlayson CM and Spiers AG (eds.) (1999) Global review of wetland resources and priorities for wetland inventory. Supervising Scientist Report 144. Canberra: Supervising Scientist. Finlayson CM and van der Valk AG (eds.) (1995) Classification and Inventory of the World’s Wetlands, Advances in Vegetation Science 16. Dordrecht: Kluwer. Foley JA, DeFries R, Asner GP, et al. (2005) Global consequences of land use. Science 309: 570--574. Gitay H, Suarez A, Watson R, et al. (2002) Climate Change and Biodiversity. IPCC Technical Paper V. Geneva: Intergovernmental Panel on Climate Change. Glover D and Jessup T (eds.) (1999) Indonesia’s Fires and Haze: The Cost of Catastrophe. Singapore and Ottawa: Institute of South East Asian Studies and IDRC. Green P, Vo¨ro¨smarty CJ, Meybeck M, Galloway J, and Peterson BJ (2004) Preindustrial and contemporary fluxes of nitrogen through rivers: A global assessment based on typology. Biogeochemistry 68: 71--105. Hofstede R, Segarra P, and Mena V (2003) Los Pa´ramos del Mundo. ProyectoAtlas Mundial de los Pa´ramos. Quito: Global Peatland Initiative/NC-IUCN/Eco-Ciencia. Hollis GE (1992) The causes of wetland loss and degradation in the Mediterranean. In: Finlayson CM, Hollis GE, and Davis TJ (eds.) Managing Mediterranean Wetlands and Their Birds, IWRB Special Publication No. 20, pp. 83–90. Slimbridge: International Waterbird and Wetland Research Bureau. Hollis GE (1998) Future wetlands in a world short of water. In: McComb AJ and Davis JA (eds.) Wetlands for the Future, pp. 5--18. Adelaide: Gleneagles Publishing. Howard G (2009) Case study 1 Nile Perch invasion in Lake Victoria. In: Finlayson CM (2009) Biotic pressures and their effect on wetland functioning. In: Maltby E and Barker T (eds.) The Wetlands Handbook, pp. 674–676. Oxford: Wiley-Blackwell. ILEC and UNEP (International Lake Environment Committee Foundation and United Nations Environment Programme) (2003) World Lake Vision: A Call to Action, 39pp (ISBN 4–9901546–0–6). IUCN (2003) 2003 IUCN Red List of Threatened Species. http://www.redlist.org (accessed April 2010). IUCN, Conservation International, and Nature Serve (2004) Global Amphibian Assessment. http://www.globalamphibians.org (accessed April 2010). Joosten H (1994) Turning the tides: Experiences and perspectives of mire conservation in the Netherlands. In: Grunig A (ed.) Mires and Man: Mire Conservation in a

Density Populated Country – The Swiss Experience, pp. 300--310. Birmensdorf, Switzerland: Swiss Federal Institute for Forest, Snow and Landscape Research. Joosten H and Clarke D (2002) Wise Use of Mires and Peatlands – Background and Principles Including a Framework for Decision-Making, International Mire Conservation Group and International Peat Society. Totnes, Devon: NHBS. Kingsford RT and Johnson W (1998) Impact of water diversions on colonially nesting waterbirds in the Macquarie Marshes of arid Australia. Colonial Waterbirds 21: 159--170. Jorgensen SE, de Bernardi R, Ballatore T, and Muhandiki V (2001) Lake Watch 25: Changes in the World’s Lakes. Draft report. Kusatsu: International Lake Environment Committee. Lehner B and Doll P (2004) Development and validation of a global database of lakes, reservoirs and wetlands. Journal of Hydrology 296: 4--22. Lemly AD, Kingsford RT, and Thompson JR (2000) Irrigated agriculture and wildlife conservation: Conflict on a global scale. Environmental Management 25: 485--512. Leveque C, Balian EV, and Martens K (2005) An assessment of animal species diversity in continental water systems. Hydrobiologia 542: 39--67. Loh J, Green RE, Ricketts T, et al. (2005) The Living Planet Index: Using species population time series to track trends in biodiversity. Philosophical Transactions of the Royal Society B 360: 289--295. Loh J and Wackernagel M (eds.) (2004) The Living Planet Report 2004. Gland: World Wide Fund for Nature. MacKay H, Finlayson CM, Fernandez-Prieto D, Davidson N, Pritchard D, and Rebelo LM (2009) The role of Earth Observation (EO) technologies in supporting implementation of the Ramsar Convention on Wetlands. Journal of Environmental Monitoring 90: 2234--2242. Malmqvist B and Rundle S (2002) Threats to the running water ecosystems of the world. Environmental Conservation 29: 134--153. Maltby E (2009) The changing wetland paradigm. In: Maltby E and Barker T (eds.) The Wetlands Handbook, pp. 3–42. Oxford: Wiley–Blackwell. Maltby E and Barker T (eds.) (2009) The Wetlands Handbook. Chichester: WileyBlackwell. Master LL, Flack SR, and Stein BA (eds.) (1998) Rivers of Life: Critical Watersheds for Protecting Freshwater Biodiversity. Arlington, VA: The Nature Conservancy. Matthews GVT (1993) The Ramsar Convention on Wetlands: Its History and Development. Gland, Switzerland: Ramsar Convention Bureau. McAllister DE, Hamilton AL, and Harvey B (1997) Global freshwater biodiversity: Striving for the integrity of freshwater ecosystems. Sea Wind-Bulletin of Ocean Voice International 11: 1--140. Meybeck M (2003) Global analysis of river systems: From Earth system controls to Anthropocene syndromes. Philosophical Transactions of the Royal Society London B 358: 1935--1955. Millennium Ecosystem Assessment (2003) Ecosystems and Human Well-Being: A Framework for Assessment. Washington, DC: Island Press. Millennium Ecosystem Assessment (2005) Ecosystems and Human Well-Being: Synthesis. Washington, DC: Island Press. Mitsch WJ (ed.) (1994) Global Wetlands: Old World and New. Amsterdam: Elsevier. Mitsch WJ (1998) Protecting the world’s wetlands: Threats and opportunities in the 21st century. In: McComb AJ and Davis JA (eds.) Wetlands for the Future, pp. 19--31. Adelaide: Gleneagles Press. Mitsch WJ and Gosselink JG (2007) Wetlands, 4th edn. New York: Wiley. Mitsch WJ, Mitsch RH, and Turner RE (1994) Wetlands of the Old and New Worlds: Ecology and management. In: Mitsch WJ (ed.) Global Wetlands: Old World and New World, pp. 3--56. Amsterdam: Elsevier. Molden D (ed.) (2007) Water for Food, Water for Life: A Comprehensive Assessment of Water Management in Agriculture. London: Earthscan. Molden D, Frenken K, Barker R, et al. (2007) Trends in water and agricultural development. In: Molden D (ed.) Water for Food, Water for Life: A Comprehensive Assessment of Water Management in Agriculture, pp. 57--89. London: Earthscan. Moser M, Prentice C, and Frazier S (1996) A global overview of wetland loss and degradation. In: Papers, Technical Session B, Vol. 10/12B, Proceedings of the 6th Meeting of the Conference of Contracting Parties, pp. 21–31. Brisbane, Australia, 19–27 March 1996. Gland, Switzerland: Ramsar Convention Bureau. Moyle PB and Leidy RA (1992) Loss of biodiversity in aquatic ecosystems: Evidence from fish faunas. In: Fielder PL and Jain SK (eds.) Conservation Biology: The Theory and Practice of Nature Conservation, Preservation and Management, pp. 127--169. New York: Chapman and Hall. Naylor RL, Goldburg RJ, Primavera JH, et al. (2000) Effect of aquaculture on world fish supplies. Nature 405: 1017--1024. OECD (1996) Guidelines for Aid Agencies for Improved Conservation and Sustainable Use of Tropical and Subtropical Wetlands. Paris: Organisation for Economic CoOperation and Development.

Managing Aquatic Ecosystems

Page S, Hosciło A, Woosten H, et al. (2009) Restoration ecology of lowland tropical Peatlands in Southeast Asia: Current knowledge and future research directions. Ecosystems 12: 888--905. Page SE, Siegert F, O’Reiley J, von Boehem H-D, Jaya A, and Limin S (1997) The amount of carbon released from peat and forest fires in Indonesia during 1997. Nature 420: 61–65. Page SE, Siegert F, O’Reiley J, von Boehm H-D, Jaya A, and Limin S (2002) The amount of carbon released from peat and forest fires in Indonesia during 1997. Nature 420: 61--65. Partow H (2001) The Mesopotamian Marshlands: Demise of an Ecosystem, 46pp. Nairobi, Kenya: UNEP. Pittock B, Wratt D, Basher R, et al. (2001) Australia and New Zealand. In: Climate Change 2001. Working Group II of the Intergovernmental Panel on Climate Change: Impacts, Adaptation and Vulnerability, ch. 12. Cambridge: Cambridge University Press. Postel S (1999) Pillar of Sand: Can the Irrigation Miracle Last? New York: WW Norton. Ramsar Convention Secretariat (2006a) The Ramsar Convention Manual: A Guide to the Convention on Wetlands (Ramsar, Iran, 1971), 4th edn. Gland, Switzerland: Ramsar Convention Secretariat. Ramsar Convention Secretariat (2006b) Wise use of wetlands: A conceptual framework for the wise use of wetlands. In: Ramsar Handbooks for the Wise Use of Wetlands, 3rd edn., vol. 1. Gland, Switzerland: Ramsar Convention Secretariat. Ramsar Convention Secretariat (2006c) Inventory, assessment, and monitoring: An integrated framework for wetland inventory, assessment, and monitoring. In: Ramsar Handbooks for the Wise Use of Wetlands, 3rd edn., vol. 11. Gland, Switzerland: Ramsar Convention Secretariat. Ramsar Convention Secretariat (2006d) Designating Ramsar sites: The Strategic Framework and guidelines for the future development of the List of Wetlands of International Importance. In: Ramsar Handbooks for the Wise Use of Wetlands, 3rd edn., vol. 1. Gland, Switzerland: Ramsar Convention Secretariat. Ramsar Strategic Plan 2009–2015. http://www.ramsar.org/pdf/key_strat_plan_2009_e.pdf (accessed August 2010). Rebelo L-M, Finlayson CM, and Nagabhatla N (2009) Remote sensing and GIS for wetland inventory, mapping and change analysis. Journal of Environmental Management 90: 2144--2153. Revenga C, Brunner J, Henninger N, Kassem K, and Payne R (2000) Pilot Analysis of Global Ecosystems: Freshwater Systems. Washington, DC: World Resources Institute. Revenga C, Campbell I, Abell R, de Villiers P, and Bryer M (2005) Prospects for monitoring freshwater ecosystems towards the 2010 targets. Philosophical Transactions of the Royal Society B 360: 397--413. Revenga C and Finlayson CM (2009) Wetlands and climate change State of the Wild. New York: World Conservation Society. Revenga C and Kura Y (2003) Status and Trends of Biodiversity of Inland Water Ecosystems. Technical Series No. 11. Montreal: Secretariat of the Convention on Biological Diversity. Richardson CJ (2008) The Everglades Experiment: Lessons for Ecosystem Restoration. New York, NY: Springer. Richardson CJ, Reiss P, Hussain NA, Alwash AJ, and Pool DJ (2005) The restoration potential of the Mesopotamian Marshes of Iraq. Science 307: 1307--1311. Rieley JO and Page SE (eds.) (1997) Biodiversity and Sustainability of Tropical Peatlands. Cardigan: Samara Publishing. Secretariat of the Convention on Biological Diversity (2006) Global Biodiversity Outlook 2. Montreal: Convention on Biological Diversity. Scanlon BR, Jolly I, Sophocleous M, and Zhang L (2007) Global impacts of conversions from natural to agricultural ecosystems on water resources: Quantity versus quality. Water Resources Research 43: W03437 (doi:10.1029/ 2006WR005486). Scott DA and Jones TA (1995) Classification and inventory of wetlands: A global overview. In: Finlayson CM and van der Valk AG (eds.) Classification and Inventory of the World’s Wetlands, Advances in Vegetation Science 16, pp. 3–16. Dordrecht: Kluwer.

59

Sculthorpe CD (1967) The Biology of Aquatic Vascular Plants. London: Edward Arnold. Semeniuk V and Semeniuk CA (1997) A geomorphic approach to global classification for natural wetlands and rationalization of the system used by the Ramsar Convention – a discussion. Wetlands Ecology and Management 5: 145--158. Shepherd G (2004) The Ecosystem Approach: Five Steps to Implementation. Gland, Switzerland and Cambridge: IUCN. The Ramsar ‘Toolkit’, 3rd edn. (2007) The Ramsar Handbooks for the Wise Use of Wetlands. http://www.ramsar.org/cda/ramsar/display/main/ (accessed May 2010). UNEP (2002) Vital Water Graphics – an Overview of the State of the World’s Fresh and Marine Waters. Nairobi: United Nations Environment Programme. UNEP (2006) Challenges to International Waters – Regional Assessments in a Global Perspective. Nairobi: United Nations Environment Programme. UNEP (2007) Global Environment Outlook 4 – Environment for Development. Nairobi: United Nations Environment Programme. UNESCO (2000) Water Related Vision for the Aral Sea Basin for the Year 2025. Paris: United Nations Educational, Scientific, and Cultural Organization. UNESCO-WWAP (2006) Water: A Shared Responsibility. The United Nations World Water Development Report 2. Paris and New York: United Nations Educational, Scientific, and Cultural Organization and Berghahn Books. van Dijk PP, Stuart BL, and Rhodin AGJ (2000) Asian Turtle Trade: Proceedings of a Workshop on Conservation and Trade of Freshwater Turtles and Tortoises in Asia Chelonian Research Monographs, No. 2. Lunenburg: Chelonian Research Foundation in association with WCS, TRAFFIC,WWF, Kadoorie Farm and Botanic Gardens US Fish and Wildlife Service. Verhoeven JTA, Arheimer B, Yin C, and Hefting MM (2006) Regional and global concerns over wetlands and water quality. Trends in Ecology and Evolution 21: 96--103. Vo¨ro¨smarty CJ, Le´veˆque C, and Revenga C (2005) Fresh water. In: Hassan R, Scholes R, and Ash N (eds.) Ecosystems and Human Well-Being: Current State and Trends: Findings of the Condition and Trends Working Group. Washington, DC: Island Press. WCD (2000) Dams and development: A new framework for decision-making. The Report of the World Commission on Dams. London: Earthscan. Whigham DF (2009) Global distribution, diversity and human alterations of wetland resources. In: Maltby E and Barker T (eds.) The Wetlands Handbook. Chichester: Wiley-Blackwell. Whigham DF, Goode RE, and Kvet J (eds.) (1993) Wetland Ecology and Management – Case Studies. Dordrecht: Kluwer. Wiseman R, Taylor D, and Zingstra H (eds.) (2003) Proceedings of the Workshop on Agriculture, Wetlands and Water Resources: 17th Global Biodiversity Forum, 122pp. Valencia, Spain, November 2002. New Delhi, India: National Institute of Ecology and International Scientific Publications. WMO (1997) Comprehensive Assessment of the Freshwater Resources of the World. Stockholm: World Meteorological Organization and Stockholm Environment Institute. WWDR (2003) Water for People, Water for Life. United Nations World Water Assessment Programme. Paris, France: UNESCO/Berghahn Books. Yarrow M, Marı´n VH, Finlayson M, Tironi A, Delgado LE, and Fischer F (2009) The ecology of Egeria densa Planchon (Liliopsida: Alismatales): A wetland ecosystem engineer? Revista Chilena de Historia Natural 82: 299--313.

Relevant Websites http://www.gisp.org Global Invasive Species Programme. http://www.ramsar.org Ramsar: The Ramsar Convention on Wetlands. http://www.ramsar.org Ramsar: The Ramsar Convention on Wetlands; A brief history of the Ramsar Convention.

1.04

Water as an Economic Good: Old and New Concepts and Implications for Analysis and Implementation

J Briscoe, Harvard University, Cambridge, MA, USA & 2011 Elsevier B.V. All rights reserved.

1.04.1 1.04.2 1.04.3 1.04.3.1 1.04.3.2 1.04.3.3 1.04.3.4 1.04.4 References

Introduction Challenge One: Revisiting the Old Issue of the Indirect Effects of Investments in Major Water Projects Challenge Two: Managing Water as a Scarce Resource Issue One: The Radically Different Markets in which Irrigation and Urban Water Operate Issue Two: How Appropriate Pricing Is Understood by Economists and by Users and the Implications for Practice Issue Three: The Crucial Distinction between Financial Costs and Opportunity Costs, and the Implications for Practice Issue Four: The Political Economy of Change Conclusions

1.04.1 Introduction This chapter addresses two conceptual and operational challenges which are of major importance for the management and development of water resources. Challenge one relates to the economic impact of large water infrastructure projects and suggests, first, that conventional economic analytic tools are of little value and, second, that newly emerging tools can be of substantial practical use. Challenge two relates to management and describes both fallacies and emerging approaches for ensuring that the economic productivity of water is maximized.

1.04.2 Challenge One: Revisiting the Old Issue of the Indirect Effects of Investments in Major Water Projects During the era of rapid economic growth, the political leaders of now-rich countries invested heavily in much major water (and other) infrastructure because they believed that these investments would transform the regional economies in which the projects – such as the Tennessee Valley Authority (TVA), Hoover Dam, and Grand Coullee projects in the United States – were located. Decades after these projects were completed (and delivered these large indirect benefits), the US Office of Management and Budget in the 1950s declared that under conditions believed to be prevailing in the United States (full employment and mobile factors of production) these indirect benefits should not be taken into account in investment decisions. As described in detail elsewhere (Briscoe, 2008), this was greeted with incredulity by political leaders who asked ‘‘if we had followed this advice, what infrastructure would ever have been built?’’ (with the implicit answer ‘‘very little’’). Although this issue faded from the sight of politicians, as most of the major infrastructure in the US had already been built, it did not fade from the bible of economists and was incorporated into conventional economic wisdom and standard appraisal practices of institutions such as the World Bank.

61 61 62 62 62 63 64 64 65

A recent set of detailed analyses (Bhatia et al., 2008) of the Bhakra Dam in northwest India, the Sao Francisco dams in Brazil, and Aswan Dam in Egypt confirm similar findings from earlier studies – namely that such projects have major backward linkages (for inputs into agriculture) and forward linkages (for processing of agricultural products, for instance). In all cases, not only the indirect effects were as large as the direct effects (as had been demonstrated in other analyses of the Muda projects in Malaysia and Grand Coullee in the United States) but also these projects had stimulated precisely the regional development which politicians had hoped for (and which economists now said, ‘‘don’t count’’). Equally important, where the data were available (as in the Bhakra case; Bhatia et al., 2008) it turned out that the biggest proportional beneficiaries were not the landlords but the landless, as a result of the sharp increase in the demand for labor. For decision makers in the real world, the conclusion is that these indirect impacts are large, and that such projects can, indeed, be the basis for regional development. It is true, nevertheless, that there is a serious analytic challenge and an even more serious practical challenge. The analytic challenge is that these studies are all ex post. There is no reliable ex ante method for assessing the indirect impacts. The practical challenge is that there is no established methodology for deciding on what packages of complementary public investments are needed in order to maximize the likelihood that the unquantifiable-but-very-important indirect benefits do, in fact, materialize. On the latter, there is a ray of light, from the original work of Harvard economist Ricardo (e.g., Hausmann and Klinger, 2008) on development patterns as defined by the ‘product space’. This work represents a major intellectual departure from the normative and mechanical work embodied in classic cost–benefit analysis. The approach starts not with principles, but with the collection of data and the use of network approaches to describe revealed patterns of economic development paths in hundreds of economies over time. In the case of major water projects, this would mean the following. First, planners would describe ‘‘where the region is’’ (in our case,

61

62

Water as an Economic Good: Old and New Concepts and Implications for Analysis and Implementation

after the building of the major infrastructure, the emergence of new configurations of energy generation, agriculture, industry, and transport). Second, taking account of the new regional reality, assess the opportunities this starting point affords by examining development paths which have evolved from similar endowments, thus identifying the paths that are most promising (and those which are no more than pie-in-the-sky fantasies). Armed with this X-ray (for the example of Pakistan, see Hausmann and Klinger (2008)), planners of large water infrastructure can then decide in a systematic and informed way on what complementary investments are needed to maximize the likelihood that multipliers will develop. The development of such a methodology is a high priority for developing countries (and there are many), who are in the early stages of investing in major water infrastructure, and offers an escape from what have become ritualistic and uninformative standard cost–benefit procedures.

1.04.3 Challenge Two: Managing Water as a Scarce Resource Many countries face multiple concerns regarding the growing scarcity of water, the associated conflicts among users, and ways of transferring water from low-value to high-value uses. Prominent and well-informed commentators often state that having users pay the full cost of water would solve these problems (recent examples include the CEO of Nestle (Brabeck-Letmathe, 2008) and The Economist (2008)). Experience has shown that the situation is considerably more complex and nuanced, and that it is not enough to just extol the virtues of pricing. This chapter outlines a different approach – one of principled pragmatism. Principled because economic principles such as ensuring that users take financial and resource costs into account when using water are very important; and pragmatism because solutions need to be tailored to specific, widely varying natural, cultural, economic, and political circumstances, in which the art of reform is the art of the possible. The general arguments are illustrated by focusing on two major users – farmers and cities. Here, four issues are addressed. This chapter draws on the World Bank’s Water Resources Sector Strategy (World Bank, 2003): 1. the quite different economic environments that pertain to these two sectors; 2. the crucial distinctions between the perspective of economists and the perspective of users on what constitutes appropriate pricing, and some of the implications of these distinctions for practice; 3. the critical distinction between the financial cost of providing a service and the opportunity cost of the resource itself, and the implications of this distinction; and 4. a review of some good practice developments, and the implications for a country-specific, practical, sequenced approach to dealing with these crucial issues.

1.04.3.1 Issue One: The Radically Different Markets in which Irrigation and Urban Water Operate The first, fundamental distinction is between the markets in which urban water supply and irrigation operate.

In the case of urban water supply, the product can largely be considered as a local, nontradable good. The price charged for water in Helsinki is entirely immaterial to the price charged in Timbuktu. More specifically, if Helsinki chooses to subsidize its water users, that is of no relevance to water users in Timbuktu. In the case of irrigation, where the end products are agricultural goods that trade on a global market, the situation is radically different. If the government of a developed country chooses to subsidize water (and other inputs and outputs) of its farmers, this has an impact on world prices, and thus a direct impact on producers in developing countries. As the magnitude of the agricultural subsidies from OECD countries (OECD, Organization for Economic Cooperation and Development) is huge (about $350 billion/year, to the detriment of consumers in developed countries and producers in developing countries), this has a major impact on the prices of agricultural products in developing countries and on the economic returns from farming. These distortions reinforce the demands of farmers in developing countries with regard to subsidies for water, energy, and other inputs, usually causing further harm to both the economy and the environment. This crucial fact makes the political economy of water pricing reform especially complex (in both theory and practice) for irrigation. Experience suggests that the appropriate approach is to acknowledge the need for subsidies and to document the existing levels. Then it is possible – for example, as has been done in Mexico (Gonzalez, 1997) – for the government and farmers to agree upon a subsidy-neutral transformation from a package of perverse subsidies (of fertilizers, pesticides, and water, for instance) to a package of virtuous subsidies (such as for improving land quality and for more efficient technology).

1.04.3.2 Issue Two: How Appropriate Pricing Is Understood by Economists and by Users and the Implications for Practice Economists have long had a sound theoretical basis for assessing the resource implications of pricing, namely charging users for the marginal cost of producing the next unit of input. This rule is clear and correct, because that is the signal which will cause users to take into account the cost of the next unit of production when they consider using another unit of the resource. Unfortunately, even sound theory does not always translate into rules that can easily be understood and applied in practice. The first reason for this is that ordinary users understand a price as a payment for a service rendered. When the supplier is a monopoly (and prices are set outside of the market), this means that the legitimate price in the eyes of users is that which it costs an efficient producer (usually a public utility) to produce the service. In economic terms, this means that users consider average, not marginal, cost to be legitimate. Two more questions arise from this: What is included in cost and what happens if the service provider is not efficient? Costs that users consider legitimate certainly include, in all cases, the costs of operating and maintaining the existing infrastructure. Moreover, with some explanation and communication, experience (Langford et al. (1999) describe the

Water as an Economic Good: Old and New Concepts and Implications for Analysis and Implementation

Australian case) shows that users see the costs of replacement as legitimate costs. However, even under the most advantageous of settings, users vigorously resist the notion that they should pay for sunk costs which, in their eyes, have already been paid for by taxes or other assessments. The issue of the efficiency and accountability of the service provider is critical. ‘‘Why should I pay the costs of the Water Department when it is overstaffed, corrupt, and does not maintain our systems?’’ is a frequent and legitimate complaint from consumers and farmers. An illustration of the lower bound of these inefficiencies comes from the state of Victoria in Australia. Before reform, irrigation services were provided by a government department with well-trained and wellperforming staff, and there was little corruption. When reform took place, and farmers had to pay the full costs of operation and maintenance, increased scrutiny of the supply agency led to a 40% reduction in these costs. In most developing countries, the inefficiency is much greater and the users’ resistance to paying for these services is correspondingly higher. Exhortations to increase cost recovery without addressing these fundamental accountability questions are a major part of the reason why cost recovery has been so poor in many countries. A review by the World Bank’s Operations Evaluation Department (2003) shows that, despite the fact that the World Bank has been by far the most constant and insistent advocate of cost recovery for decades, ‘‘there is no evidence of better cost recovery or of covenant compliance either.’’ The bottom line, then, is that in most urban and irrigation systems cost recovery is critical for the supply of good services. The road to cost recovery does not lie in conditionalities imposed by aid agencies, however, but in realigning the institutional arrangements so that suppliers are accountable to users, and so that charges become a principal tool used for ensuring the mutual obligations of suppliers and users.

1.04.3.3 Issue Three: The Crucial Distinction between Financial Costs and Opportunity Costs, and the Implications for Practice User payments for the financial costs of services rendered is a fundamental requirement for any financially sustainable water supply system – this is very important. However, the claims for pricing typically go beyond that of maintaining and operating infrastructure, and suggest that if ‘‘the prices are right, allocation will be optimal.’’ Proceeding from the viewpoint of users (as one must when considering political economy of reform rather than theoretical elegance), it is vital to distinguish between two radically different types of costs. First, there are the costs that any user can understand, namely the financial costs associated with pumps, treatment plants, and pipes. Second is the far more subtle concept of the opportunity cost of the resource itself. There have been many proposals for doing sophisticated calculations of this opportunity cost, and charging users for this ‘‘to ensure appropriate resource allocation.’’ This has not worked in practice for three fundamental reasons: first, because it is impossible to explain to the general public (let alone to angry farmers) why they should pay for something that does not cost anything to produce; second, because opportunity costs vary widely by place and time and could not be

63

accurately calculated by even the most sophisticated of regulatory agencies; and, third, because those who have implicit or explicit rights to use of the resource (correctly) argue that they have already paid for the (implicit or explicit) rights and argue (appropriately) such proposals to be the confiscation of property. An added, and highly relevant, factor is that the ratio between financial and opportunity costs is often radically different for different sectors (Briscoe, 1996). Although everything in water (like politics) is local, there are two broad patterns. It costs a lot (per unit of water) to operate the dams, water and wastewater treatment plants, and pumps and pipes that provide households with the modest amount of water they use (and the sewage that is removed). Alongside these large financial costs, the opportunity cost of the resource itself (as measured by the value of the raw water in its next best use, often irrigation) is typically quite low. For municipal and industrial water, therefore, financial costs generally dominate opportunity costs. For irrigation, the situation is almost exactly the opposite. It costs relatively little (per unit of water) to build, operate, and maintain the usual gravity systems that provide very large quantities of water. However, the opportunity cost of the water (for cities and, increasingly, for high-value agricultural uses) is, in situations of scarcity, often much higher (typically at least an order of magnitude higher) than the financial cost of supplying the water. These numbers (remembering, of course, that every place is different) have profound implications. They mean that, from the point of view of ensuring that users take into account the cost of the resources they are using, the emphasis must be on financial costs for municipal supplies, and on opportunity costs for irrigation. (It is worth emphasizing that this does not mean that cost recovery does not matter for irrigation. Cost recovery for irrigation remains very important for infrastructure sustainability, but not for efficiency in the allocation or use of water.) The great challenge for irrigation, in light of these theoretical and practical realities, is how to have farmers take account of the opportunity cost of water. In most parts of the world where water is scarce, informal water markets have arisen, in which those who have (implicit) rights sell water to those who need it. In some cases, the practice has existed for hundreds of years and has been formalized (as in the Water Court of Valencia, Spain, which has managed transfers among users for a 1000 years). In many other cases (such as western India; Shah, 1993), these markets are extensive, sophisticated, and illegal. Throughout the arid western United States, water rights have long been legal property and, under different rules in different states, allowed for approved transfers between willing buyers and willing sellers. As other parts of the world have experienced scarcity, a number of countries facing water stress have turned toward formal, legal, managed water markets. This took place in recent decades in Chile, Australia, and Mexico. The Australian case shows the benefits – the adverse impact of water reductions on the regional economy – are reduced by two-thirds when there is both intra- and interstate water trading (Productivity Commission, 2004).

64

Water as an Economic Good: Old and New Concepts and Implications for Analysis and Implementation

From the perspective of the present discussion on how to ensure that users take account of opportunity costs, these market-based arrangements have a unique virtue. Once users have clear, transferable property rights, then they automatically consider whether they wish to forego a particular use of water in exchange for compensation from another user who may place a higher value on the water. Reallocating water then becomes a matter of voluntary and mutually beneficial agreements between willing buyers and willing sellers, and not a matter of confiscation or an endless search for new sources of supply. This is not to suggest that the establishment of water markets is simple or a panacea. The operation of such systems is demanding in terms of rules for establishing initial rights (including those for the environment and informal customary rights); the plumbing required to measure and move water; the regulatory institutions that are essential to protect the rights of other water users and the environment and to ensure that the public interest is represented; and the information and management systems. Many consider these prerequisites so onerous that they cannot be made to work in most developing countries. In addition, many point to early problems that all countries have faced in making such changes. Without in any way minimizing these challenges, three observations are germane. First, the prerequisites are really prerequisites for any form of well-managed allocation system and the absence of such prerequisites is a problem for all allocation systems, including the administrative allocation systems practiced in most countries. (As with everything in water management, the choice is not between the first and the second best, but between imperfect and even more imperfect.) Second, one of the many virtues of a market-based system is that, once started, there is a strong demand for better measurement, transparency, regulation and information. Third, all such established systems are working, often after initial adjustments, reasonably well. In none of the countries that have adopted such systems is there any thought to reverting to the previous allocation procedures.

1.04.3.4 Issue Four: The Political Economy of Change The implications for practitioners are clear. First, from the point of view of financial cost recovery, the key is an institutional framework whereby service providers are accountable and efficient. When this materializes, and when users see that their payments are being used to improve the quantity and quality of services, they can and will pay. Here (as discussed earlier), watchwords are competition, regulation, transparency, benchmarking, and accountability. In the urban water supply and energy sectors, these ideas are now accepted in most parts of the world. In the irrigation sector, there is a gradual, albeit still far too slow, acceptance of these principles. Building on the historic experiences in countries such as Spain and the United States, a number of countries (including Australia, Chile, Mexico, and, more recently, the provinces of Punjab in Pakistan (Government of Punjab, Pakistan, 2008) and Maharashtra in India (Government of Maharashtra, 2003)) have moved toward systems which (1) charge irrigators for the cost incurred in providing services and (2) have clarified and made transparent water entitlements which will,

slowly and inexorably, lead to trading and the revelation of opportunity costs. In all settings, a critical element of this approach is to develop innovative mechanisms for breaking out of the typical low-level equilibrium, in which services are poor, users will not pay, service quality declines, etc. In one good example of such innovation, the World Bank helped the government of Guinea Conakry break the circle by guaranteeing a new, accountable operator a declining proportion of reasonable costs over a 5-year period (World Bank, 1993). In the first year, then, the operator had sufficient revenues (mostly from the International Development Association (IDA) credit, but some from users) to improve the operation of the system. As the level of service improved, users were informed that they would be charged for the new, improved service and that, eventually, they would pay the full costs of the service. The art of reform is less one of articulating a vision than of tracing a path for making improvements, for applying generic principles in a way that takes account of the very widely varying historical, cultural, natural, social, and economic conditions which govern water management (Briscoe, 1997). An analysis of experiences of successful reforms suggests that this means, inter alia: ‘‘picking the low-hanging fruit first,’’ for instance, by starting with temporary trading in well-defined systems where good infrastructure is in place; ‘‘not making the best the enemy of the good,’’ by having a well-defined, sequenced, prioritized, and patient approach for moving toward improvement, not seeking to attain perfection in one fell swoop; and ‘‘keeping one’s eyes peeled,’’ by understanding that it is broader reforms outside of the water sector (often relating to overall economic liberalization and fiscal and political reform) which will provide the preconditions for making the critical first steps. Recent reviews of water reforms in Pakistan and India (Briscoe and Qamar, 2007; Briscoe and Malik, 2006) describe, in considerable detail, what the application of these principles might be in practice.

1.04.4 Conclusions There is growing understanding that there are broad benefits – for the economy, users, and for the environment – if water is developed and managed as an economic good and a growing search for a new set of analytic and operational tools. In recent years, there has been a subtle but important change in discussion of economic policy. The landmark Growth Commission (Spence et al., 2008), written by several Nobel prize laureates and many eminent development practitioners, draws lessons from the history of successful growth experiences. The Commission discarded the rigid prescriptions so often advocated, and noted that there were a wide variety of different, successful, experiences. What they did conclude was that there were some common elements – for example, a disciplined examination of, and adherence to, comparative advantage – and then application of economic principles in a sequenced, nuanced manner appropriate to particular cultural and economic circumstances. Application of this less rigid approach has major implications for water.

Water as an Economic Good: Old and New Concepts and Implications for Analysis and Implementation

In terms of the development of infrastructure, it means getting away from what have become uninformative, formulaic analyses of internal rates of return, to an approach which uses new tools to identify critical supplementary investments needed to maximize the multiplier effects of major investments. In terms of management, it means moving away from the tired phrase of ‘get the prices right’ (and everything will be okay) that has been repeated for years, with little impact on the ground. It means paying much more attention to incentives and to opportunity costs, and to creating an enabling environment in which users will make much better use of limited water, or transfer the right to use that water to others who can use it more productively. Finally, it also means giving greater attention to the political economy of change. This chapter advocates (as does the 2003 World Bank Water Resources Strategy) a path of principled pragmatism – in which the principles of sound economic management are well defined and respected, but in which they are applied in a pragmatic and sequenced way which takes account of local circumstances and political economy, and in which the focus is on moving in the right direction, and on the art of the possible.

References Bhatia R, Cestti R, Scatasta M, and Malik RPS (eds.) (2008) The Indirect Economic Impact of Dams. New Delhi: The Academic Foundation. Brabeck-Letmathe P (2008) Global drying. Wall Street Journal, Asia 13 June 2008: A13. Briscoe J (1996) Water as an economic good: The idea and what it means in practice. In: Proceedings of the World Congress of the International Commission on Irrigation and Drainage. Cairo, Egypt, September 1996.

65

Briscoe J (1997) Managing water as an economic good: Rules for reformers. Water Supply 15(4): 153--172. Briscoe J (2008) How theory, practice, politics and time affects views on the indirect economic impact of water infrastructure. In: Bhatia R, Cestti R, Scatasta M, and Malik RPS (eds.) The Indirect Economist Impact of Dams. New Delhi: The Academic Foundation. Briscoe J and Malik RPS (2006) India’s Water Economy: Bracing for a Turbulent Future. New Delhi: Oxford University Press. Briscoe J and Qamar U (2007) Pakistan’s Water Economy: Running Dry. Oxford: Oxford University Press. Gonzalez F (1997). Water Reforms in Mexico. Water Week, World Bank. Government of Maharashtra (2003) Maharashtra Water Regulatory Authority Bill XIX. Mumbai. Government of Punjab, Pakistan (2008) Entitlements. http://irrigation.punjab.gov.pk/ Entitlement.aspx (accessed July 2010). Hausmann R and Klinger B (2008) Structural Transformation in Pakistan. Center for International Development, Harvard University. Langford KJ, Foster CL, and Malcolm DM (1999) Towards a Financially Sustainable Irrigation System, World Bank Technical Paper 413. Washington, DC: World Bank. Operations Evaluation Department (2003) Bridging Troubled Waters: Assessing the Water Resources Strategy Since 1993. Washington, DC: World Bank. Productivity Commission (2004) Modelling Water Trade in the Southern Murray Darling Basin, Staff Working Paper, Canberra. Shah T (1993) Groundwater Markets and Irrigation Development: Political Economy and Practical Policy. Bombay: Oxford University Press. Spence M, et al. (2008) The Report of the Growth Commission: Strategies for Sustained Growth and Inclusive Development. Washington, DC: World Bank. The Economist (2008) Running dry. The Economist 18 September 2008. United States Congress (1955) Discussion of Budget Bureau Circular A-47, Hearings before the Committee on Interior and Insular Affairs, Serial no. 5. World Bank (1993) Development and Environment: The World Development Report. Washington, DC: World Bank World Bank (2003) The World Bank’s Water Sector Strategy. Washington, DC: World Bank.

1.05 Providing Clean Water: Evidence from Randomized Evaluations A Ahuja, Harvard University, Cambridge, MA, USA M Kremer, Harvard University, Cambridge, MA, USA AP Zwane, Bill and Melinda Gates Foundation, Seattle, WA, USA & 2011 Elsevier B.V. All rights reserved.

1.05.1 1.05.2 1.05.2.1 1.05.2.2 1.05.3 1.05.3.1 1.05.3.2 1.05.4 1.05.4.1 1.05.4.2 1.05.4.3 1.05.4.4 1.05.5 1.05.6 1.05.6.1 1.05.6.2 1.05.6.3 1.05.7 References

Introduction Water Quantity Health Impacts Maintenance Solutions Water Quality Health Impacts Valuation Nonprice Determinants of Clean Water Adoption Information on Water Contamination Levels Gain versus Loss Framing and Other Behavioral Marketing Communal versus Individual Persuasion Personal Contact Potentially Scalable Approaches to Improving Water Quality Methods and Theory: Contributions of Randomized Evaluations of Domestic Water Survey Effects Valuation: Revealed Preference versus Contingent Valuation Combining Randomized Evaluations with Structural Modeling Conclusion

1.05.1 Introduction Some 1.6 million children die each year from diarrhea and other gastrointestinal diseases for which contaminated drinking water is a leading cause (Wardlaw et al., 2010). This chapter critically reviews experimental work on the provision of water and improved water quality for domestic use in developing countries, discussing both policy implications and methodological lessons. Earlier work has been reviewed in research using nonrandomized approaches (Zwane and Kremer, 2007). Holla and Kremer (2008) provide a summary of the literature on randomized evaluations related to pricing and access in health and education. Cardenas (2009), Pattanayak and Pfaff (2009), and Timmins and Schlenker (2009) provide reviews on related issues. Recent calls for investment in experiments in environmental economics include Greenstone and Gayer (2009) and Bennear and Coglianese (2005). Local public good investments for services such as water for domestic use are arguably typically best prioritized by local policymakers who know the preferences of the communities they serve. However, the sole quantitative environmental target in the United Nations Millennium Development Goals is the call to ‘‘reduce by half the proportion of people without sustainable access to safe drinking water.’’ In practice, efforts to meet this goal have translated into increased donor and national government funding for building local public goods such as wells and standpipes. By increasing the number of water points, this reduces the time to collect water and makes the task more convenient.

67 68 68 69 70 70 71 72 72 73 73 74 74 74 74 75 75 76 76

Health externalities that could cross jurisdictional boundaries could be an important part of the case for national or supra-national investments targeted specifically at the water sector. Diarrhea is an infectious disease. Distributional concerns provide another potential rationale for national or international policymakers to target aid to the water sector in particular. Outsiders may place more value on the consumption of child survival goods relative to consumption of other goods than the local household or other local decision maker does. One of the leading debates in the literature has been on the relative health impact of increases in water quantity versus improved water quality. Simply providing more convenient access to water, even without improving water quality, could potentially stimulate greater handwashing, which has been shown to be very important for health, and more washing of clothes and dishes. At this point, however, the limited evidence available from randomized studies does not demonstrate that increasing access to water without changing its quality improves health. In contrast, there is now abundant evidence that improved water quality reduces self-reported diarrhea. This evidence is an example of how randomized evaluations, which often yield quite different estimates of impact than nonexperimental analyses, can clarify questions that are difficult to resolve using analytic techniques that have difficulty separating causal effects of programs from selection bias. As seen below, there is some reason for external support for water quantity beyond what might be chosen by local decision makers, but much less than for water quality. Though there is

67

68

Providing Clean Water: Evidence from Randomized Evaluations

limited evidence from randomized trials for health (as opposed to a time use or convenience) benefits of water quantity, increasing water quantity disproportionately benefits women and thus there may be a distributional case for public policy to support increased water quantity. In contrast, there is strong evidence that improving water quality improves health by reducing infectious disease and thus builds a case for subsidizing water-quality treatment based on reducing externalities. Improving water quality disproportionately also helps an even more severely under-represented group: young children. As discussed below, the evidence of low valuation for improvements in water quality, in contrast to water quantity, is consistent with the hypothesis that households put very little weight on the health costs of dirty water in the case of young children. External funders or national policymakers may put more weight on child health. A challenge with prioritizing water-quantity investments for donor funding is that maintaining the associated hardware has traditionally been difficult. When infrastructure falls into disrepair, the stream of benefits associated with the investment may be lost. Randomized impact evaluations suggest that external contracting can perform better than community-based voluntary arrangements at least in some circumstances. Evidence from India suggests that women are more likely to invest in water infrastructure but evidence from Kenya suggests little effect of efforts to encourage selection of female user committee chairs on quality of water infrastructure maintenance. Water-quality interventions face a different challenge than hardware maintenance if their biomedical benefits are to be sustained over time. In contexts where treatment does not occur at a centralized treatment plant, as in most developing countries, individuals influence the level of diffusion and adoption of treatment technologies. Randomized impact evaluations have provided evidence on the determinants of uptake in such cases, including a steep demand curve for treatment products such as chlorine or water filters. In addition to shedding light on policy debates such as the investment decision regarding quality versus quantity, experiments can help researchers come up with new solutions to hurdles such as this technology adoption decision. Randomized evaluations have demonstrated that the demand curve can be shifted outward by providing information and making treatment easy and convenient, as well as local promotion of ongoing use. Combining evidence of low valuation with this other information about influencers of adoption has allowed for new approaches to the service delivery problem to be developed. In particular, providing dilute chlorine solution free at the point of water collection, together with a local promoter, can increase takeup of water treatment from less than 10% to more than 60%. Methodologically, randomized evaluations have provided evidence that the process of collecting data through surveys can itself affect behavior and that revealed preference estimates of willingness to pay for environmental interventions in developing countries are far smaller than stated preference estimates. Recent work also marries randomized evaluations with structural modeling to provide guidance on the potential impact of alternative policies and social norms. The remainder of this chapter is structured as follows: Section 1.05.2 summarizes evidence from randomized

evaluation on the impact of infrastructure investments to increase water-quantity improvements in developing countries and on the maintenance of these investments. Section 1.05.3 argues there is considerable evidence that water-quality improvements yield health benefits, but that many households are willing to pay very little for cleaner water. Section 1.05.4 assesses alternative means of shifting the demand curve for water quality, examining information provision, communal versus individual persuasion, and local promoters. Section 1.05.5 discusses cost-effective and potentially scalable approaches to water quality drawing on the lesson of Sections 1.05.2, 1.05.3, and 1.05.4. Section 1.05.6 reviews methodological contributions from randomized evaluations of domestic water interventions. Section 1.05.7 concludes this chapter.

1.05.2 Water Quantity 1.05.2.1 Health Impacts Identifying the aspects (quantity vs. quality) of improved water supply is important for policy because different interventions affect quality and quantity asymmetrically. For example, adding chlorine to water affects quality but not quantity. Providing household connections to municipal water supplies to households that currently use standpipes is likely to have a bigger effect on the convenience of obtaining water and thus on the quantity of water consumed than on water quality. There has been considerable debate over whether increasing the quality of water or increasing the quantity of water has a greater impact on disease. Much of the most convincing nonexperimental evidence on the health impact of water and sanitation makes it difficult to separate the impact of quantity and quality (Cutler and Miller, 2005; Watson, 2006; Galiani et al., 2005; Gamper-Rabindran et al., 2010) because the interventions that are studied both reduced the cost of collection and improved quality, making it unclear which route of disease transmission mattered the most in practice. In the 1980s and 1990s, nonrandomized studies were frequently cited as evidence that water-quantity interventions were more important for health impacts than water-quality interventions (Esrey, 1996; Esrey et al., 1991). Some argued that these results could be explained because, when water supplies are rationed, increased availability of and convenience of water facilitate more frequent washing of hands, dishes, bodies, and clothes, thus reducing disease transmission (Esrey, 1996; Esrey et al., 1991; Curtis et al., 2000). However, the question remained unsettled because it was difficult to assess causality in the absence of randomized evaluations or other convincing identification. In the past 10 years, a new body of evaluations has been developed to address this question. We discuss in Section 1.05.3 the numerous randomized evaluations that have shown impacts of improved water quality on health while also confirming the importance of hand washing in reducing disease transmission. There remains limited evidence on the question of water quantity, with one recent relevant randomized evaluation. Consistent with the early claims made about the relative importance of water quantity, health benefits from hand

Providing Clean Water: Evidence from Randomized Evaluations

washing have been shown in several settings. Luby et al. (2004) reported the results of a cluster-randomized trial in a large sample of households in Karachi, Pakistan, of a hand washing promotion campaign aimed at mothers. Infants and malnourished children under age 5 living in treatment households had 39% fewer days of diarrhea compared with the control group after 1 year of intervention and observation. Two other older randomized controlled trials of hand-washing interventions with more than two communities in their samples (Khan, 1982; Han and Hlaing, 1989) each had a relatively large sample size randomly divided into treatment and control groups and measured compliance by observing or weighing provided bars of soap as well as by tracking diarrhea cases. The studies report large effects of hand washing and soap provision programs on the incidence of diarrhea. Khan (1982) reported that the provision of either soap and water-storage containers or soap alone, along with initial instructions to increase the frequency of hand washing, reduced shigella reinfection by 67% in Bangladesh. Han and Hlaing (1989) reported a 40% reduction in diarrhea incidence among children under age 2 (though there was no reduction in incidence for older children) following hand washing education and the provision of soap to a random sample of mothers in Rangoon (Yangon). Although impacts may be heterogeneous across settings, and caution is warranted in drawing general conclusions, the one available randomized evaluation found that increasing the quantity of water while maintaining unchanged quality did not lead to significant health improvements. DeVoto et al. (2009) examined provision of piped connections to homes in urban Morocco previously served by standpipes. This increased the quantity of water used by the household, but did not improve water quality, since the alternative was chlorinated water from communal taps, which was of similar quality to the water received at home. As part of a planned piped water service extension in Tangier, Morocco, the authors randomly selected half the households eligible for a first connection to receive information about and an offer of credit toward a new connection, and administrative assistance in applying for credit. Takeup was 69% (as compared to 10% in the control group). The authors compare outcomes of those who received this treatment to those for households in the control group. They find that piped water provision in this urban Moroccan context had few health benefits. There is no evidence for an impact of treatment on a subjective ranking of health of the family or on diarrhea in children under age 6 (though baseline rates were relatively low, with the average child in the control group experiencing 0.27 days of diarrhea in the past week). Households in the treatment group report increasing their frequency of baths and showers: the number of times respondents in the treatment group washed themselves (baths, showers) during the last 7 days is 25% higher than in the control group. However, hygiene practices that require less water, such as hand washing, were not affected, according to self-reports. We would not conclude that increased water quantity never yields health benefits. The benefits of increased water quantity may be context specific and require further research to fully understand. In particular, understanding when and how

69

increased access to water leads to more hand washing is a research priority. Having a piped water connection had substantial private benefits, despite the lack of impacts on self-reported diarrhea, consistent with the evidence of private valuation. In particular, it saved time, which was used for leisure and social activities. Evidence of substantial willingness to pay for water quantity has been noted by other authors as well in observational studies. In Morocco, the intervention also improved measures of social integration and overall welfare for households. Consistent with this finding, households are willing to pay a substantial amount of money to gain access to a private tap at home: 1 year into the program, not only had the encouragement design resulted in high rates of takeup in the treatment group, but also, for these households, their average monthly water bill more than doubled, from 73 to 192 Moroccan dirhams (MAD), or US$9 to $24 a month (the previous cost came from households, who took water from their neighbors). The importance of a household visit in inducing adoption is something we discuss further in Section 1.05.4.4; additional evidence on personal contact has been generated in other settings as well. There is evidence from India that women particularly value water investments and that women’s involvement in investment decisions could result in improved water supply in situations where local governments set priorities among local public good investments. Women’s valuation of these goods can be part of a distributional argument in favor of additional external support for these investments as well. Chattopadhyay and Duflo (2004) found that a randomized policy change in India that increased the role of women in policy decision making led to more investment in water infrastructure. A 1993 constitutional amendment called for one-third of village council leader positions to be reserved for women. Rules ensured random assignment of the leadership reservations. Chattopadhyay and Duflo showed that village councils headed by women were significantly more likely to invest in public infrastructure for drinking water. These investments were borehole wells and other storage infrastructure that likely improve quality as well as the convenience of water collection. We takeup in the next section the question of whether women’s involvement in local decision making can also make these kinds of investments easier to maintain and keep up.

1.05.2.2 Maintenance Solutions In general, water-quantity investments, whether bundled with water-quality improvements or not, often require significant infrastructure investments. Water quality can be improved with virtually no investment in infrastructure (e.g., by leaving water in the sun or the addition of chlorine), though of course other water-quality interventions can also require hardware (as the first example in Section 1.05.3 illustrates). Along with infrastructure investments comes the challenge of maintenance, which has historically been a major problem in developing countries. The rural water sector in particular has a poor track record of maintaining infrastructure investments. For instance, a quarter of India’s water infrastructure is believed to be in need of repair (Ray, 2004). World Development Report 2004

70

Providing Clean Water: Evidence from Randomized Evaluations

(World Bank, 2003) estimates that more than a third of rural water infrastructure in South Asia is not functional. Miguel and Gugerty (2005) reported that nearly 50% of borehole wells dug in a large project in western Kenya in the 1980s, and subsequently maintained using a community-based maintenance model, had fallen into disrepair by 2000. Difficulties with maintaining water infrastructure, particularly in rural areas, reduce the cost effectiveness of these interventions relative to other measures that prevent diarrhea. There are two solutions frequently mentioned as potential elements of a solution to this infrastructure challenge: empowering women to manage water resources and including communities in participatory management schemes. As discussed above, women value water investments when capital expenditure decisions are made. Randomized evidence on these solutions suggests that neither may be as effective as contracting out services via maintenance contracts, however, as we review below. Kremer et al. (2008) provided evidence from a randomized field experiment in Kenya that, at least in that context-enhanced women’s involvement in infrastructure management, did not lead to better maintenance of water supplies. This evaluation studies the impact of female affirmative action policies on actual management outcome measures relevant for protected springs (e.g., time since storm drains or drainage trenches were cleaned). When protected springs were provided to 100 communities in rural Kenya, all communities formed water-user committees. In addition, onehalf of the communities received messages encouraging women to take leadership roles in their water-user committees. This encouragement intervention did result in more women representation in the treatment communities. Communities that received the female participation intervention were twice as likely to have women in the role of water committee chair. However, this did not lead to differences in the effectiveness of the user committees’ spring management as measured by the maintenance outcome variables. Thus, the authors conclude that advocacy for female participation can increase women’s involvement without any impact (either positive or negative) on project outcomes. This has a positive interpretation: empowerment goals can be met without attended costs to project outcomes, as well as a more negative one: including women in management cannot alone solve the water infrastructure maintenance challenge, even if these investments are priorities for women. In addition to increasing women’s participation and decision-making power, another standard model for maintaining donor-funded infrastructure projects, such as water schemes, is to establish user groups responsible for maintenance and management. This approach grew out of the widespread perception that centralized government maintenance was unsuccessful. Giving communities direct control or ownership over key project decisions was intended to improve the quality of public services and increase financial sustainability. There is little convincing empirical evidence, however, that local user-committee management of local public goods such as improved drinking water sources results in better quality service than other models ongoing centralized funding from public budgets. Collective action problems may be difficult to overcome, and voluntary committees tasked with collecting user fees may be difficult to sustain or empower. In a recent

comprehensive review of community-based development projects, Mansuri and Rao (2004) noted that existing research examining successful community-based projects does not compare these projects with centralized mechanisms for service delivery or infrastructure maintenance (e.g., city or state financed). This makes it difficult to determine whether alternative project designs would have had different results. The limited empirical evidence suggests that the impact of the community-based development approach on infrastructure maintenance is mixed at best. In addition to randomly assigning the gender empowerment encouragement intervention, in the same study as described above, the nongovernmental organization (NGO) randomly assigned communities to contracted maintenance and community-based management schemes. Kremer et al. (2008) compared payments to private contractors for spring maintenance and ongoing grants to user committees, with the outcomes of a control group, in which user committees received no grants. The traditional model, user committees without grants, performed worse than either alternative across a range of maintenance outcomes. Providing grants to user committees improves a measure of overall water source maintenance quality by around 30% of one standard deviation on average, while paying contractors to maintain water source leads to an average improvement in measured maintenance quality by around 50% of one standard deviation. This difference is significant at the 10% level. This evidence from spring protection maintenance, a relatively simple technology that seems favorable to communitybased management, suggests that contracting for private maintenance service may be a promising alternative to committee-based management schemes. Nonexperimental evidence from Argentina (Galiani et al., 2005) also suggests that contracted private provision of service can expand coverage and improve health outcomes at least in certain settings in middle-income countries. Certainly, further research is needed that transparently compares the counterfactual of subsidized public service provision and community-based management schemes. In summary, the health benefits of water-quantity interventions require further investigation. Increasing availability of water, even leaving quality unchanged, brings major nonhealth benefits; yet insofar as these seem unlikely to create externalities beyond the household, let alone cross-jurisdictional externalities, local governments may be the proper institution for allocating budgets between water and other public goods. There may be a distributional case for national or supra-national water investments, as these are valued by women, however. Whatever benefits water-quantity interventions do provide can quickly be lost if infrastructure falls into disrepair or is broken. Contracting models seem to hold promise for maintaining water infrastructure.

1.05.3 Water Quality 1.05.3.1 Health Impacts A body of randomized evaluations examines interventions that affect water quality without affecting water quantity. These yield strong evidence of reductions in reported diarrhea.

Providing Clean Water: Evidence from Randomized Evaluations

One study examines source water-quality improvements. Kremer et al. (2009a), in the first randomized evaluation of the provision of improved communal water infrastructure, estimated that protecting springs reduced fecal contamination as measured by the presence of E. coli bacteria by two-thirds in water at the source, but only by 25% for water stored at home. This is likely in part due to recontamination in transport and storage within the household, as well as to the use of alternative sources. Despite the incomplete pass through of the water-quality improvement, this led to a reduction in selfreported child diarrhea of about 25%. Other epidemiological evidence on water quality, manipulated via filtration or treatment rather than infrastructure, also suggests that there may be large health gains from investments in quality. The bulk of the evidence suggests that, with takeup rates on the order of 70% (achieved via frequent visits and reminders to subjects) household water treatment reduces child diarrhea by 20–40%. (One caveat is that the outcome measure in these studies is typically mothers’ reports of child diarrhea. Studies with objective outcomes, infrequently measured, would be desirable (Schmidt and Cairncross, 2009). Nonetheless, we believe that the weight of the evidence is strong enough to believe that reductions in diarrhea are real. To the extent that reporting bias lowers estimates of diarrhea in both the treatment and comparison groups, it may in fact make it harder to pick up reductions in diarrhea. The extent of reporting bias in treatment groups would have to be very large to explain the reported reductions in diarrhea associated with cleaner water. If the reductions in diarrhea are even a fraction as large as those estimated, water treatment would still be very cost-effective.) Comprehensive reviews of this literature are provided by Waddington and Snilstveit (2009), Fewtrell et al. (2005), Arnold and Colford (2007), and Clasen et al. (2006). Because water treatment can be extremely cheap, even a 20–40% reduction in diarrhea makes water treatment very low cost per disability adjusted life years saved. To get a sense of how cheap it is to treat water, note that 1.42 gallon generic bottle of bleach with approximately 6% sodium hypochlorite concentration sold in Walmart for $2.54 as of December 2009 has enough chlorine to treat 163 400 l of water. This corresponds to a price of $0.00002 per liter of water treated. Even making generous allowances for the fact that chlorine used for water treatment should be sold at lower concentrations and has to be transported, etc., if mortality reductions are proportional to reported morbidity reductions, the cost per DALY is similar to that of childhood vaccination, at under $40 per DALY.

1.05.3.2 Valuation Despite the evidence of health benefits associated with water quality, a number of papers suggest very little willingness to pay for this class of interventions. Moreover, there is little evidence that households with young children place substantial additional value on clean water, suggesting low valuation of child health. Kremer et al. (Spring cleaning: rural water impacts, valuation, and institutions, unpublished manuscript) exploited exogenous changes in the trade-off that households face when choosing between multiple water sources, some of which are

71

close but contaminated and others of which are far but clean. This variation in the distance/water quality trade-off is generated by the spring protection intervention discussed above that was randomly phased in to almost 200 communities in rural Kenya. (Spring protection reduces contamination by sealing off the eye of the spring so that it is no longer vulnerable to surface-water runoff.) The authors compare how many trips households make to protected springs and other sources, controlling for differences in the time it takes to walk to each source. The estimated mean valuation for spring protection is equivalent to 32.4 workdays. Based on household reports of trade-offs between walking time and money, this corresponds to approximately US$2.96 per household per year. Under additional assumptions, this translates into a willingness to pay $23.68 per DALY saved, which is well below the benchmark of $100–150 often assumed to be appropriate for health investments in developing countries. Kremer et al. (2009a) used randomly assigned discounts to investigate willingness to pay for dilute chlorine. They described behavior consistent with a steep demand curve for water treatment and found no evidence of higher valuation among households with vulnerable young children. In a set of impact evaluations that tested both price and nonprice interventions to increase takeup of chlorine, households were randomly assigned either to a comparison group or to treatment arms in which they received a free supply of individually packaged chlorine or coupons for half-priced chlorine that could be redeemed at local shops. Comparison households could buy WaterGuard through normal retail channels, at about $0.25 for a 1-month supply (roughly a quarter of the agricultural daily wage). Although 70–90% of households in the study region had heard of the local brand of point-of-use chlorine and roughly 70% volunteered that drinking dirty water is a cause of diarrhea, only 5–10% of households reported that their main supply of drinking water was chlorinated prior to the interventions. There were no significant differences between treatment and comparison groups at baseline. Access to free chlorine increased takeup rates to over 50%, whereas coupons for even a 50% discount hardly affected takeup relative to the comparison group. The point estimate suggests a four percentage point increase relative to the comparison group, but this is not statistically significant. This is evidence for very price elastic demand. Households with young children did not behave differently from other households (p value of 0.85 on the test of equality of means). This evidence for low valuation of child health can be part of a distributional argument for subsidizing water treatment. Preliminary results from DeVoto et al. (2009) on distribution of chlorine through clinics in Kenya and from Berry et al. (2008) on distribution of water filters in Ghana also suggest very steep demand curves for improved water quality. The authors conclude that high takeup rates can be achieved at sufficiently low prices, but that demand for chlorine among their rural samples is extremely sensitive to price even though the retail price of the product is still relatively low. Ashraf et al. (2009) used a two-stage price randomization that enables both measurement of willingness to pay for water treatment and also, under specific assumptions, allows for testing of whether higher prices induce a sunk-cost effect that

72

Providing Clean Water: Evidence from Randomized Evaluations

leads households who pay more for chlorine to use it more and/or screens out households less likely to use the product. In a door-to-door marketing campaign, roughly 1000 households in the study were first asked if they wanted to purchase a bottle of dilute chlorine at a randomized offer price. If a household agreed to purchase and was able to come up with the cash needed for the transaction, they were then offered an additional randomly assigned discount which determined the transaction price. Variation in offer prices is used to test for whether households who are willing to pay a higher price are more likely to use chlorine for water treatment, controlling for the price the household ultimately did pay; variation in transaction prices is used to test for a sunk-cost effect that might lead households who actually paid more for chlorine to be more likely to use it, controlling for willingness to pay. Approximately 2 weeks after the marketing campaign, the survey team reached almost 900 of the households who received the marketing intervention to test for the presence of chlorine in stored drinking water supplies and administered a follow-up study. The authors found that willingness to pay is low, and that many more households are willing to purchase chlorine at low prices. Consistent with other evaluations, Ashraf et al. did not find that charging higher price leads to more effective targeting to those households with higher potential health gains, those households with children under age 5 or pregnant women. There are at least two possible counter-arguments to the case for subsidies in water quality: (1) a sunk-cost effect, in which paying for the product makes people more likely to use it and (2) wastage of the product when people are given it but do not use it or use it for purposes not valued by the funder. Ashraf et al. (2009) found no evidence of a sunk-cost effect, finding that the actual transaction price does not affect propensity to use, controlling for offer price. Ashraf et al. (2009) did find that when the price is lowered, the marginal households induced to buy chlorine are less likely to show chlorine residual in their water 2 weeks later. The hypothesis that they favor is that these households start using the products for other off-label uses such as cleaning clothes or toilets. They presented evidence for this hypothesis drawn from a convenience sample. However, this is somewhat puzzling since, as they note, dilute chlorine sold for water treatment is considerably more expensive per unit of chlorine than commercially available bleach. Overall, it is difficult to distinguish the hypothesis that these households are not using the product for water treatment from the alternative hypotheses that they are storing the product or giving it away for water-treatment usage (or that they tried the product but did not like the taste; this would be only a short-run loss). These hypotheses have quite different policy implications as only the first is wastage that might reduce the social value of a program that supplied the product for free. Even if diversion to alternative uses is common, because chlorine is very cheap and can have a large impact on health, high levels of diversion to alternate uses are likely to be acceptable if this occurs as a result of a process that increases use for water treatment overall. Policymakers confronted with this evidence of low valuation of water quality and child health must determine whether subsidies for water-quality interventions such as chlorination are warranted. If governments or external donors

place more value on child health relative to other consumption compared to local households, the lack of valuation for water quality and child health provides a potential rationale for subsidies. Externalities from consumption provide another potential rationale for subsidies in some cases. Although there is no direct evidence on health externalities from water treatment in any of the papers reviewed here, to the extent that consumption reduces disease incidence for the user, it is also likely to reduce disease transmission from the user to others. In this case, eliminating prices for water-quality improvements is likely to be welfare maximizing due to these externalities. In fact, given the externalities combined with the low cost of water disinfectant, negative prices may be optimal.

1.05.4 Nonprice Determinants of Clean Water Adoption In this section, we review experimental evidence on several nonprice variables that could potentially affect household behavior regarding water quality. The emphasis in this section is on identifying potential mechanisms that could increase uptake of safe water rather than also judging their cost effectiveness or scalability. Section 1.05.5 discusses potentially scalable models drawing on the lessons of this and previous sections.

1.05.4.1 Information on Water Contamination Levels Several papers suggest that providing households with information about source water quality can change behavior, but that the effects of information are small relative to price and that people are responding not as Bayesian decision makers rationally processing information. Rather, information may be important because it increases the salience of water contamination. Jalan and Somanathan (2008) randomly assigned households in their urban Indian sample to receive information on whether or not their drinking water had tested positive for fecal contamination. Among households not purifying their water initially, this information led to an 11 percentage point increase in reported water purification as measured 8 weeks after information provision; about 42% of the study population purified (meaning either filtered, boiled, purchased bottled water, or, more rarely, chemically treated) their water at baseline. They also increased water purification expenditures by about $7. Households that initially purified their water but who received information that their water was probably not contaminated (based on tests of untreated water in their household) were not statistically more likely to change their purification behavior than the control group. This finding of an asymmetric response to testing is evidence for the idea that the channel through which information campaigns work is salience of some sort rather than Bayesian learning. Bayesian learners would respond to information that their water is safer or cleaner than they thought by reducing expenditure on purification. Luoto (2009) found through a randomized controlled trial in Kenya that sharing information on fecal contamination

Providing Clean Water: Evidence from Randomized Evaluations

with Kenyan households in a context in which treatment products were provided for free increases water treatment by 8–13 percentage points (or between 12% and 23% of baseline usage rates). The study also suggests that once information on the quality of source water is provided, providing additional information on the quality of water stored in the home has no further impact on takeup. When interpreting these results, it is useful to recall that, if people had unbiased expectations about water quality initially, information provision would lead some to revise beliefs about water quality downward and others to revise them upward, with ambiguous implications for water-treatment behavior. The fairly consistent finding that information provision increases takeup suggests that these results may reflect salience as much as Bayesian learning. Further evidence consistent with this is provided by Madajewicz et al. (2007) and Tarozzi et al. (2009), who studied as to how people respond to information about water quality in an area of Bangladesh where wells are frequently contaminated with arsenic. Madajewicz et al. evaluated the effectiveness of providing coarse information about well safety by providing a random sample of household information about whether their water source has arsenic concentrations above a threshold level. Households informed that their water exceeds this threshold are 37 percentage points more likely to switch sources than control households within 1 year. They increase their walking time 15-fold (about 4 min), on average, in response to the information. These responses to arsenic contamination information are further evidence for non-Bayesian decision making when combined with more recent research from Bangladesh following the introduction of a standardized labeling system for wells in which safe sources are labeled green and unsafe sources red. Tarozzi et al. (2009) performed an evaluation in this context in which all subjects receive the coarse information about water safety for all sources around them. A random sub-sample receives additional information about relative safety along a continuous scale. The relationship between arsenic and health is likely to be continuous. Thus, if households are Bayesian decision makers, continuous information should be more useful. Households far from any nearby uncontaminated well might switch to a well that is just above the cutoff level for being colored red, for example. Similarly, households using a well that is just below the cutoff might switch to a well that is much further below the cutoff. In practice, however, receiving continuous information does not substantially affect risk perceptions or the likelihood of switching sources. In fact, providing continuous information decreases the impact of the arsenic level on the probability of switching to a new source of drinking water. People are unable to use this additional information to improve their drinking water quality more than those people armed with only coarse information. This finding that coarse information may change behavior more than finer information poses a challenge to the idea that people can be modeled as Bayesian decision makers. These findings are instead consistent with the idea that information campaigns increase salience of water contamination and that it is this salience, which changes behavior rather than the precise information content itself. Similarly, the finding of an

73

asymmetric response to information, with those who find that their water is contaminated adopting safer behavior, and those with uncontaminated water not reducing their efforts to obtain clean water, is unexpected under a Bayesian model, but is consistent with the salience hypothesis.

1.05.4.2 Gain versus Loss Framing and Other Behavioral Marketing Given the evidence above that a simple Bayesian learning story is inadequate to explain behavior, we now turn to evidence on ideas from behavioral economics and psychology. Luoto (2009) provided households a variety of point-of-use watertreatment technologies for free in Kenya and then randomly assigned households to receive various promotional strategies to increase use of these products. This research generates a series of results that can inform marketing and distributional strategies. First, she examines whether emphasizing the gains from water treatment versus the losses from not treating water affected use. There are competing hypotheses in the literature for which framing should bring about the larger response. Prospect theory predicts that loss aversion will cause the lossframed message to realize a bigger effect on people’s choices and behavior (Tversky and Kahneman, 1981; Kahneman and Tversky, 1979). However, there is evidence that decisions regarding health behaviors respond more to gain-framed messages in some cases and more to loss framings in others (Rothman et al., 1999). The study compares a framing of safe water technologies as increasing health compared to one in which it is framed as both increasing health and avoiding disease. The latter approach increased usage by approximately four to six percentage points, a statistically significant difference. Luoto (2009) also tested whether a combination of commitment and a visual reminder to treat water changes behavior. A subset of the sample was assigned to make a commitment to treating their water to improving their family’s health, and also given a pictorial reminder to treat their water. This increased water treatment by five to eight percentage points, but was significant only in some specifications. A commitment to the interviewer had relatively large effects on households that showed evidence of high discount rates in responses to hypothetical questions about future payoffs.

1.05.4.3 Communal versus Individual Persuasion Kremer et al. (2009a) provided some evidence that a communal approach in which households are aware of the messages other community members receive is more effective than an individual approach in encouraging treatment of household drinking water with dilute chlorine disinfectant, although differences are limited to the case when households had to pay for the product. Their study tested three variants of a persuasion campaign in which promotional messages targeted to mother were delivered at either the household level or community level, or both. The treatment was cross-cut with providing subsidized (free) chlorine to households. The results confirm the importance of price as a key determinant of takeup. When chlorine was subsidized, community messages had no measurable impact on household

74

Providing Clean Water: Evidence from Randomized Evaluations

water treatment. The point estimate of the effect of messaging is actually negative, though statistically insignificant, and very small compared to the main effect (–0.02 as compared to 0.52). Messages can influence takeup, however, when positive prices are charged. At normal retail prices, treatment of household drinking water with chlorine increased by between three and five percentage points (as measured by testing household drinking water for chlorine) for the communitybased and combined scripts in the short run. There was no measurable impact of the household script alone, but community-based messaging, a much cheaper approach to marketing, had a small but positive effect. None of the promotion scripts had any significant effect on takeup at the medium-run follow-up 3–6 months after exposure. Considering the shortrun nature of the effects and the high cost of marketing during one-on-one conversations during household visits, or even through community-level meetings, such strategies do not appear to hold much promise as cost-effective means of promoting individually packaged retail chlorine takeup at scale. It is worth noting, however, that Kremer et al. (2009a) found little evidence for peer effects in takeup of chlorine packaged for household use. Using detailed data on conversation frequency and topics collected in the second and fourth survey rounds (of the first phase of the research), they found strong evidence that the distribution of free chlorine marketed as WaterGuard promoted conversations about the product as well as about drinking water more generally and, to a lesser degree, child health. In particular, conversations about WaterGuard were roughly 3 times more likely to occur if the respondent was a member of a treatment household and slightly more than twice as likely if the other household in a relationship pair was in the treatment group. Although the distribution of free WaterGuard prompted more conversations about the product, the evidence is consistent with the hypothesis of weak social network effects on actual use, with larger impacts on social desirability bias. They found statistically and economically significant effects of peer exposure on self-reported chlorination but pointed estimates that are much smaller and not generally statistically significant using positive chlorine tests in home drinking water.

1.05.4.4 Personal Contact Kremer et al. (2009a) also tested another approach to increasing demand – hiring local community members to promote chlorine use among their neighbors. This sort of personal contact has been previously identified as important to behavior change and adoption decisions (DellaVigna and Gentzkow, 2010; Manandhar et al., 2004). In this intervention, personal persuasion was accompanied by a price discount, however. Households were also given a coupon for one free bottle of dilute chlorine solution (equivalent to 1 month’s supply). The fraction of households with residual chlorine in their water was approximately 10 times as high in communities with a local promoter and a free sample of chlorine relative to comparison households in the short-run (3 weeks), at 40% roughly versus 4%, respectively. While takeup fell to a certain extent at the medium-run (3–6 months) as households used up their free bottle, communities with promoters were nonetheless able to

sustain adoption rates between 30 and 35 percentage points higher than the comparison group takeup rate of 8%. Personal contact was also successful in achieving high levels of takeup in the evaluation of the water-quantity intervention described in Section 1.05.2.1 (DeVoto et al., 2009).

1.05.5 Potentially Scalable Approaches to Improving Water Quality Section 1.05.2 documents the strong evidence that water treatment has the potential to improve health cost-effectively. Section 1.05.3 discusses that takeup of water quality interventions is extremely sensitive to price. Section 1.05.4 indicates that personal contact, salience, and convenience, and potentially having public information about water treatment can boost takeup. This section discusses potential low-cost, scalable models for water treatment based on the findings of Sections 1.05.2, 1.05.3, and 1.05.4. Kremer et al. (2009a) and DeVoto et al. (2009) developed and tested two alternative approaches to providing clean water, both involving free distribution. DeVoto et al. (2009) provided coupons for dilute chlorine solution to mothers who bring children to vaccination clinics sufficient to cover water supplies for the 12 months until children reach approximately age 2. Mothers are told how and where to redeem coupons and urged to treat water for their children during a vulnerable stage of their life. A second approach entailed switching to free delivery of chlorine by placing a container to dispense the product at water sources. This bulk supply dramatically reduced delivery costs relative to the retail approach that requires packaging chlorine in small bottles and makes free provision more realistic. Users can treat drinking water when they collect it. The required agitation and wait-time for chlorine-treated water are at least partially accomplished automatically during the walk home from the source. The source-based dilute chlorine disinfection approach to water treatment makes this act salient, convenient, and public, in addition to making it cheaper. The dispenser provides a daily visual reminder to households to treat their water at the moment when it is most salient – as water is collected – and maximizes the potential for learning, norm formation, and social network effects by making the dispenser public. Potential users can see others, who use the dispenser, and have the opportunity to ask questions and they will also know that others will see whether they use the dispenser. Takeup of chlorine provided through dispensers dramatically exceeded takeup of chlorine provided for in-home use. When communities were randomly assigned to treatment with a promoter and a community dispenser, takeup was about 40% in the short run (3 weeks) but had climbed to over 60% by the medium term (3–6 months), representing 37 and 53 percentage point gains respectively over the control group. In contrast to the takeup levels achieved with the dispensers, the clinic-based coupon distribution approach proved initially promising, but resulted in much lower coupon redemption over time. Over 40% of households, who were given coupons, redeemed them 8 months into the program in that sample, but this fell to 20% by 12 months. This suggests that the

Providing Clean Water: Evidence from Randomized Evaluations

success of the dispenser is not due only to the zero price, but also to the reduction in the psychic cost of remembering to treat water that is achieved by source-based treatment as well as other attributes, like the visual reminders. The chlorine dispenser is also extremely cost effective, with a cost per DALY saved of less than $20. The success of the chlorine dispensers at the proof-of-concept stage described here suggests that exploring how to scale up this approach to water treatment warrants further attention in its own right, as well. An important challenge for the future will be to determine how best to handle the supply side under free provision and, in particular, scaling supply chain management.

1.05.6 Methods and Theory: Contributions of Randomized Evaluations of Domestic Water The evaluations surveyed in this chapter have provided policy guidance on several questions related to health, technology adoption, and pricing regimes. The work has also made a number of methodological contributions that are of broader interest in resource economics. We review these contributions in this section.

1.05.6.1 Survey Effects A recent randomized evaluation of a water-quality intervention provides evidence that the act of surveying can affect behavior in ways that can interfere with estimates of treatment effects, a result with broader implications. Many studies have focused on measuring reported diarrhea (by mothers of young children) through household visits as a means of assessing the health impact of domestic water interventions. Kremer et al. (2009b) provided evidence that collection of self-reported diarrhea data through repeated interviews leads to health-protective behavior change in addition to respondent fatigue and social desirability bias, which are well-known concerns in survey administration. More frequent questioning exacerbates this survey effect. As part of a larger study of the impact of spring protection, households from a rural Kenyan population were randomly assigned to be interviewed about diarrhea either every 2 weeks or every 6 months. Kremer et al. (2009b) documented that frequent data collection, as is typically used in the epidemiology literature to measure diarrhea incidence, induces behavioral change in the form of higher levels of home water treatment, verified by tests for chlorine in water. The authors also find that frequent data collection leads to lower reports of child diarrhea by mothers relative to infrequent surveying. These effects are sufficiently large so as to change the conclusions about the effectiveness of the water-quality intervention being studied, that is, spring protection. The potential for survey effects implies that researchers relying on both self-reported or otherwise subjective data and objective data to measure outcomes should consider designing data collection strategies that minimize interaction with subjects. For example, outcome data could also be collected via administrative records maintained at clinics or schools. Purchases or collection of products from central locations can also be tracked without direct interaction with subjects.

75

In the particular case of the literature on water, sanitation, and hygiene, survey effect concerns imply that more research is needed that does not measure impacts via subjective reports of diarrhea. Researchers in this field should expand their datacollection strategies to emphasize other health outcomes that can be measured objectively and infrequently. This will likely require larger sample sizes to detect small treatment effects (e.g., on stunting, cognition, and ultimately, mortality) and longer study times, which funding will need to accommodate.

1.05.6.2 Valuation: Revealed Preference versus Contingent Valuation One common approach to understanding willingness to pay for environmental amenities is to use stated preference data from hypothetical situations to identify the price that households would be willing to pay. The survey-based approach for eliciting stated preference is a method known as contingent valuation (CV), the development of which is surveyed by Hanemann (1994). While data from such studies may be the most practical solution when estimating valuation of a nonrival good or one that is costly to offer in a real transaction, it is also subject to a number of pitfalls stemming from the fact that choices in hypothetical situations might not be the same as those that the respondent would make in the real world, facing real budget constraints and real benefits (Diamond and Hausman, 1994). In addition, it may be difficult for individuals to know beforehand how they will value a good and survey respondents may strategically misstate their willingness to pay (Whittington, 2002). Another approach is estimating valuation of goods not traded in markets to use discrete choice models to analyze nonexperimental survey data on households’ decisions. (There are a number of papers, including Mu et al. (1990) that estimate more general demand functions for water from various sources. We do not focus on these papers here, since they do not explicitly deal with water quality as a measurable attribute in the decision process.) While this has the advantage of evaluating real choices, and hence providing information on a part of the household preference function (McConnell and Rosado, 2000), a potential problem with this type of method is that there may be unobservable household characteristics correlated with households’ choices as well as supplier pricing decisions, which will lead to biased results. For example, suppliers may charge higher prices where demand is higher, resulting in a positive cross-sectional correlation in prices and adoption even if increasing prices would result in lower demand for each household. A third approach is to infer willingness to pay by randomizing prices or locations of new facilities, thus generating random variation in travel time. This enables analysis based on actual choices while also addressing omitted variable bias, addressing the main concerns with both CV data and results based on nonexperimental data. This method also has the potential to enable examination of the allocative role of prices in targeting populations of interest and the isolation of specific channels of causality for effects of prices on demand. The literature on domestic water that we review here has provided one of the first direct comparisons of revealed preference and stated preference valuations of an environmental

76

Providing Clean Water: Evidence from Randomized Evaluations

service in developing countries. Kremer et al. (unpublished manuscript) found that the stated preference approaches generate much higher valuation estimates than the revealed preference approach, by a factor of 3, with the contingent valuation survey approach exhibiting much greater dispersion, as well as considerable sensitivity to question framing, casting doubt on the reliability of stated preference methods. A revealed preference estimate of households’ valuation of cleaner water from source water-quality improvements can be assessed using a travel-cost approach that measures the number of trips made to the improved source relative to an unimproved source at a different distance, as long as source water-quality improvements are randomly assigned. Kremer et al. (unpublished manuscript) used such a method to estimate a model of demand for clean water from source quality improvements. As previously described, they exploited exogenous changes in the trade-off that households face when choosing between multiple water sources, some of which are close but contaminated and others of which are far but clean. They then contrasted this revealed preference estimate of willingness to pay for spring protection with two different stated preference methodologies: stated ranking of alternative water sources and contingent valuation. The divergence between the valuation estimates suggests that CV studies overestimate how much users value water-quality improvements.

1.05.6.3 Combining Randomized Evaluations with Structural Modeling Several recent papers combine data from randomized experiments with structural econometric methods in development economics (e.g., Todd and Wolpin, 2006). Kremer et al. (unpublished manuscript) combined experimental results with a structural model of water infrastructure investment to explore the implications of alternative property rights institutions on social welfare and assess the welfare impacts of alternative institutions governing water property rights. Using the valuation results discussed earlier as inputs into policy simulations, the authors can compare the welfare impacts of a number of potential scenarios under alternative sets of social norms regarding property rights. A hypothetical case of pure privatization, for example, in which landowners could restrict access to the spring and charge for water, results in relatively little investment in environmental protection (i.e., spring protection) since households’ willingness to pay for cleaner water is low, but leads to large static losses since landowners can extract consumer surplus by charging for even unprotected spring water even though the marginal cost of provision is zero. They concluded that, at low income levels, common property likely yields greater social welfare than private property, but that at higher income levels private property may yield higher social welfare.

1.05.7 Conclusion As noted in the introduction, the sole quantifiable environmental goal selected by the United Nations as part of the United Nations Millennium Development Goals is to reduce by half the proportion of people without sustainable access to

safe drinking water. Standard public finance theory suggests that local governments may ordinarily be best placed to allocate funds among competing local public goods, but it also suggests that international donors and national governments could reasonably intervene to support water projects due either to disease externalities across political jurisdictions, or to different distributional preferences than local decision makers. Both factors point to an emphasis on the impact of water on communicable diseases of children. While there is currently limited evidence on the health impact of increasing access to water without improving quality, there is strong evidence that improving water quality has health benefits for young children. Since households seem more willing to pay for access to increased quantities of water than for safer water, and since investments in water treatment are extremely cost-effective relative to other health expenditures, even expenditures such as vaccination, there seems a strong case for zero prices or even negative prices for water treatment. This chapter reviews evidence from randomized evaluations that can inform this policy debate on the quality versus quantity investment decision, and strategies to drive takeup of water treatment products. One promising approach is chlorine dispensers. Takeup of chlorination via communal chlorine dispensers (Kremer et al., 2009a), combined with a local promoter, is between 60% and 70% and the authors estimate that the long-run cost of supplying a community with bulk chlorine through a dispenser is only about 1/4 to 1/3 as much as with individually packaged bottles. This amounts to a cost of about 15 cents per person per year. The feasibility of this approach depends on the ability to solve the challenge of refill servicing and efforts to increase the density of dispensers (to drive refill costs down). Additional work to understand how to combine interventions and transition to greater levels of service as incomes rise remains an important area of policy-relevant work. The methodological lessons from the research on water-treatment takeup and valuation reviewed here can inform study design on the scale-up of alternative approaches to water-treatment and other experiments in resource economics, as well.

References Arnold B and Colford J (2007) Treating water with chlorine at point-of-use to improve water quality and reduce diarrhea in developing countries: A systematic review and meta-analysis. American Journal of Tropical Medicine and Hygiene 76(2): 354--364. Ashraf N, Berry J, and Shapiro J (in press) Can higher prices stimulate product use? Evidence from a field experiment in Zambia. American Economic Review. Banerjee A, Cole S, Duflo E, and Linden L (2007) Remedying education: Evidence from two randomized experiments in India. Quarterly Journal of Economics 122(3): 1235--1264. Banerjee A and Duflo E (2009) Experimental approach to development. Annual Review of Economics 1: 151--178. Bennear LS and Coglianese C (2005) Measuring progress: Program evaluation of environmental policies. Environment: Science and Policy for Sustainable Development 47(2): 22--39. Berry J, Fischer G, and Guiteras R (2008) Willingness to pay for clean water. Presented at Workshop on Scaling up Distribution of Water Treatment Technologies in Developing Countries, Harvard University, Cambridge, MA. Cardenas JC (2009) Experiments in environment and development. Annual Review of Resource Economics 1: 157--182.

Providing Clean Water: Evidence from Randomized Evaluations

Chattopadhyay R and Duflo E (2004) Women as policy makers: Evidence from a randomized policy experiment in India. Econometrica 72(5): 1409--1443. Clasen T, Brown J, Collin S, Suntura O, and Cairncross S (2004) Reducing diarrhea through the use of household-based ceramic water filters: A randomized, controlled trial in rural Bolivia. American Journal of Tropical Medicine and Hygiene 70(6): 651--657. Clasen T, Roberts I, Rabie T, Schmidt W, and Cairncross S (2006) Interventions to improve water quality for preventing diarrhoea. Cochrane Database of Systematic Reviews 2006, Issue 3. Art. No.: CD004794. DOI: 10.1002/14651858.CD004794. pub2. Conley T and Udry C (2001) Social learning through networks: The adoption of new agricultural technologies in Ghana. American Journal of Agricultural Economics 83(3): 668--673. Curtis V, Cairncross S, and Yonli R (2000) Domestic hygiene and diarrhoea – pinpointing the problem. Tropical Medicine and International Health 5(1): 22--32. Cutler D and Miller G (2005) The role of public health improvements in health advances: The 20th century United States. Demography 42(1): 1--22. DellaVigna S and Gentzkow M (2010) Persuasion: Empirical evidence. Annual Review of Economics 2 (doi:0.1146/annurev.economics.102308.124309). DeVoto F, Duflo E, Dupas P, Pariente W, and Pons V (2009) Happiness on tap: The demand for and impact of piped water in urban Morocco. Working Paper UCLA. Diamond P and Hausman J (1994) Contingent valuation: Is some number better than no number? Journal of Economic Perspectives 8(4): 45--64. Duflo E, Dupas P, and Kremer M (2007) Peer effects, pupil–teacher ratios, and teacher incentives: Evidence from a randomized evaluation in Kenya. Cambridge: Mimeo; Harvard University. Duflo E, Glennerster R, and Kremer M (2008a) Using randomization in development economics research: A toolkit. In: Shultz TP and Strauss J (eds.) Handbook of Development Economics, 4, pp. 1--2. Amsterdam: Elsevier. Duflo E and Kremer M (2008) Use of randomization in the evaluation of development effectiveness. Evaluating Development Effectiveness 7: 93--120. Duflo E, Kremer M, and Robinson J (2009) Nudging farmers to use fertilizer: Theory and experimental evidence from Kenya. Working Paper, 15131, NBER. Eckel C and Grossman P (1998) Are women less selfish than men? Evidence from dictator experiments. Economic Journal 108(448): 726--735. Esrey SA (1996) Water, waste, and well-being: A multicountry study. American Journal of Epidemiology 143(6): 608--623. Esrey SA, Potash JB, Roberts L, and Shiff C (1991) Effects of improved water supply and sanitation on ascariasis, diarrhoea, dracunculiasis, hookworm infection, schistosomiasis, and trachoma. Bulletin of the World Health Organization 69(5): 609--621. Fewtrell L, Kaufmann RB, Kay D, Enanoria W, Haller L, and Colford JM (2005) Water, sanitation, and hygiene interventions to reduce diarrhoea in less developed countries: A systematic review and meta-analysis. Lancet Infectious Disease 5: 42--52. Foster A and Rosenzweig M (2001) Imperfect commitment, altruism, and the family: Evidence from transfer behavior in low-income rural areas. Review of Economics and Statistics 83(3): 389--407. Galiani S, Gertler P, and Schargrodsky E (2005) Water for life: The impact of privatization of water services on child mortality. Journal of Political Economy 113(1): 83--119. Gamper-Rabindran S, Khan S, and Timmens C (2010) The impact of piped water provision on infant mortality in Brazil: A quantile panel-data approach. Journal of Development Economics 92(2): 188–200. Garrett V, Ogutu P, Mabonga P, et al. (2008) Diarrhoea prevention in a high-risk rural Kenyan population through point-of-use chlorination, safe water storage, sanitation, and rainwater harvesting. Epidemiology and Infection 136(11): 1463--1471. Greenstone M and Gayer T (2009) Quasi-experimental and experimental approaches to environmental economics. Journal of Environmental Economics and Management 57(1): 21--44. Han AM and Hlaing T (1989) Prevention of diarrhoea and dysentery by hand washing. Transactions of the Royal Society of Tropical Medicine and Hygiene 83(1): 2128--2131. Hanemann WM (1994) Valuing the environment through contingent valuation. Journal of Economic Perspectives 8: 19--43. Hoffman V (2009) Intrahousehold allocation of free and purchased mosquito nets. American Economic Review 99(2): 236--241. Holla A and Kremer M (2008) Pricing and access: Lessons from randomized evaluation in education and health. Cambridge: Mimeo; Harvard University. Imbens G and Wooldridge JM (2009) Recent developments in the econometrics of program evaluation. Journal of Economic Literature 47(1): 5--86.

77

Jalan J and Somanathan E (2008) The importance of being informed: Experimental evidence on demand for environmental quality. Journal of Development Economics 87: 14--28. Kahneman D and Tversky A (1979) Prospect theory: An analysis of decision under risk. Econometrica 47(2): 263--291. Khan MU (1982) Interruption of shigellosis by hand washing. Transactions of the Royal Society of Tropical Medicine and Hygiene 76(2): 164--168. Kremer M, Leino J, Miguel E, and Zwane A (2008) Managing rural water infrastructure in Kenya. Working Paper. Kremer M and Miguel E (2007) The illusion of sustainability. Quarterly Journal of Economics 112(3): 1007--1065. Kremer M, Miguel E, Mullainathan S, Null C, and Zwane A (2009a) Coupons, promoters, and dispensers: Impact evaluations to increase water treatment. Working Paper. Kremer M, Miguel E, Null C, Van Dusen E, and Zwane A (2009b) Measuring diarrhea: Quantifying Hawthorne effects in frequently collected data. Working Paper, University of California, Berkeley. Laibson D (1997) Golden eggs and hyperbolic discounting. Quarterly Journal of Economics 112(2): 443--477. Lancet (2007) Science at WHO and UNICEF: The corrosion of trust. Lancet 370: 1007 (editorial). Lipscomb M and Mobarak M (2008) Decentralization and water pollution spillovers: Evidence from the re-drawing of county boundaries in Brazil. Working Paper. molly.lipscomb.googlepages.com/Brazilwater113007.pdf (accessed March 2010). Luby S, Agboatwalla M, Painter J, Altaf A, Billhimer W, and Hoekstra H (2004) Effect of intensive hand washing promotion on childhood diarrhea in high-risk communities in Pakistan: A randomized control trial. JAMA 291(21): 2547--2554. Luby S, Mendoza C, Keswick B, Chiller T, and Hoekstra R (2008) Difficulties in bringing point-of-use water treatment to scale in rural Guatemala. American Journal of Tropical Medicine and Hygiene 78(3): 382--387. Luoto J (2009) Information and persuasion: Achieving safe water behavior in Kenya. Working Paper. Madajewicz M, Pfaff A, van Geen A, et al. (2007) Can information alone change behavior? Response to arsenic contamination of groundwater in Bangladesh. Journal of Development Economics 84: 731--754. Manandhar DD, Osrin B, Prasad N, et al. (2004) Effect of a participatory intervention with women’s groups on birth outcomes in Nepal: Cluster randomized control trial. Lancet 364: 970--979. Mansuri G and Rao V (2004) Community-based and -driven development: A critical review. World Bank Research Observer 19(1): 1--39. McConnell K and Rosado M (2000) Valuing discrete improvements in drinking water quality through revealed preferences. Water Resources Research 36(6): 1575--1582. Miguel E and Gugerty M (2005) Ethnic diversity, social sanctions, and public goods in Kenya. Journal of Public Economics 89(11–12): 2325--2368. Mu X, Whittington D, and Briscoe J (1990) Modeling village water demand behavior: A discrete choice approach. Water Resources Research 26(4): 521--529. Mullainathan S, Schwartzstein J, and Shleifer A (2006) Coarse thinking and persuasion. Working Paper, W12720, NBER. Munshi K (2004) Social learning in a heterogeneous population: Technology diffusion in the Indian Green Revolution. Journal of Development Economics 73: 185--215. Muralidharan K and Sundararaman V (2009). Teaching incentives in developing countries: Experimental evidence from India. Working Paper, 15323, NBER. Nowell C and Tinkler S (1994) The influence of gender on the provision of a public good. Journal of Economic Behavior and Organization 25(1): 25--36. O’Donoghue T and Rabin M (1999) Doing it now or later. American Economic Review 89(1): 103--124. Oster SM (1995) Strategic Management for Nonprofit Organizations: Theory and Cases. Oxford: Oxford University Press. Pattanayak SK and Pfaff A (2009) Behavior, environment, and health in developing countries: Evaluation and valuation. Annual Review of Resource Economics 1: 183--217. Ray I (2004) Water for all? Peri-urban and rural water delivery options: The case of India. International Conference of Engineers for a Sustainable World, Stanford Unversity. Rothman A, Martino S, Bedell B, Detweiler J, and Salovey P (1999) The systematic influence of gain- and loss-framed messages on interest in and use of different types of health behavior. Personality and Social Psychology Bulletin 25(11): 1355--1369. Sachs JD (2005) The End of Poverty: Economic Possibilities for Our Time. New York: Penguin.

1.06 Pricing Water and Sanitation Services D Whittington, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA & 2011 Elsevier B.V. All rights reserved.

1.06.1 1.06.2 1.06.3 1.06.4 1.06.5 1.06.5.1 1.06.5.1.1 1.06.5.1.2 1.06.5.1.3 1.06.5.1.4 1.06.5.1.5 1.06.5.2 1.06.5.3 1.06.6 1.06.7 1.06.7.1 1.06.7.2 1.06.8 References

Introduction The Costs of Providing W&S Services W&S Development Paths Objectives of Tariff Design Tariff Structures – the Alternatives Single-Part Tariffs Fixed charges Volumetric charges Uniform volumetric charge Block tariffs Increasing linear tariff Two-Part Tariffs Seasonal and Zonal Water Pricing Achieving Economic Efficiency and Recovering Capital Costs: Fundamentals of Dynamic Marginal Cost Pricing in the W&S Sector Subsidizing Capital Costs: Reaching the Poor Create a Well-Run System of Public Taps as a Safety Net for the Poor Preserve Options for the Poor Concluding Remarks

1.06.1 Introduction A tariff is an important management tool that can be used to assist with efforts to improve the delivery of water and sanitation (W&S) services. The pricing of W&S services is, however, controversial, and it is important to understand why there is so little consensus on W&S tariff issues. There are four main reasons. First, there is disagreement over the objectives of water pricing and tariff design. Water pricing decisions affect several different objectives or goals of policymakers, often in conflicting ways. This means that if one person is looking solely (or mostly) at the consequences of a particular water pricing policy (or tariff design) in terms of one objective, and another person is looking at the same water pricing policy in terms of its impact on another objective, they may reach quite different conclusions about the attractiveness of the pricing policy. Second, as people do not generally know what it costs to provide W&S services, it is difficult for them to judge what is a fair or appropriate price to pay. Third, there is disagreement over what would actually happen if different water tariffs were implemented. The empirical work is often lacking that would enable someone to know with reasonable confidence how changes in water prices would affect the quantity of water that different customers would use and whether or not price changes would affect customers’ decisions to connect (or stay connected) to the water distribution system (Nauges and Whittington, 2009). Fourth, although there is some competition in the water market, there is no market test for different water tariff structures. Many tariff structures are feasible and can partially accomplish some of the competing objectives of water pricing.

79 79 82 83 84 85 85 85 85 86 87 87 87 88 91 93 94 94 94

There are typically an insufficient number of providers of piped water services for customers to reject inappropriate tariff structures. Bad ideas thus do not get weeded out of either sector practice or policy discussions. Even in different private sector participation arrangements, water tariff structures are typically set by the regulatory agency, and the private sector operator has to treat them as given and manage the system as best he can (given this constraint). The purpose of this chapter is to provide the reader with a better understanding of the main issues involved in the design of W&S tariffs. Section 1.06.2 summarizes the costs of providing piped W&S services. Obviously, these costs vary widely depending on local circumstances, but the presentation of some estimates of different components of the costs of providing such services illustrates that they are not cheap. Section 1.06.3 discusses alternative development paths for moving from low levels of W&S service (or no service at all) to modern piped services, and shows the costs associated with various incremental changes. Section 1.06.4 presents the four main objectives of tariff design. Section 1.06.5 summarizes the main tariff options. Section 1.06.6 describes the basic ideas of dynamic marginal cost pricing in the W&S sector and illustrates how a two-part tariff can be used to achieve both economic efficiency and cost recovery objectives. Section 1.06.7 discusses how subsidies can be best used in the W&S sector to reach poor households. Section 1.06.8 offers some concluding remarks.

1.06.2 The Costs of Providing W&S Services A key feature of network W&S investments is that they are very capital intensive. The majority of costs are incurred early in the

79

80

Pricing Water and Sanitation Services

life of the project, and the subsequent stream of benefits (and revenues) occurs over many years. Baumann and Boland (1998) estimated that the ratio of annual investment in the US water industry to gross revenues was 0.43 in 1993. (They note, ‘‘no other major industry group in the United States even approaches this ratio of annual investment to revenue.’’) Someone (e.g., private investors and government) must take a long-term perspective and put large amounts of capital at risk. If future revenues are needed to pay for the high capital costs in the early years, investors need assurance to their rights to the revenue stream. In contrast, poor households in developing countries tend to have high rates of discount (Poulos and Whittington, 2000) and thus have short planning horizons. Such households cannot easily make the long-term commitments required to pay for network W&S services. They are also uncertain about the prospects for long-term economic growth and their ability to pay in the future. The challenge of W&S tariff design in developing countries must be understood within the context of this fundamental mismatch of perspectives between investors and consumers. In urban areas, there is widespread consensus that the long-term goal of W&S service providers in developing countries should be to offer 24-h potable water supply piped into people’s homes, to remove wastewater with a piped sewerage system, and to treat this wastewater to a standard sufficient to minimize the environmental effects of its discharge to surface water bodies. (Some people in the sector do question the wisdom of pursuing the goal of piped sewerage – see, e.g., Esrey and Andersson (2000).) Even in many rural communities, households aspire to this level of service. The treatment and delivery of water to households, and the removal and treatment of the wastewater generated, cost serious money. These costs must be paid by someone, households must either pay or receive subsidies (e.g., from richer households, industries, donors, and higher levels of government). Of course, costs vary depending on local circumstances, and estimates of what it will cost to provide a certain level of service may vary widely. Also, most investments are incremental in nature. Only rarely would a community incur the costs of complete (full service) piped W&S systems at a single point in time. Nevertheless, some rough calculations may prove useful for the discussion of water pricing and tariff design. The approach here is to present some illustrative average unit costs of providing an urban household with modern W&S services. First, representative unit costs per cubic meter for different components of W&S services are looked at. Second, some typical quantities of water that different representative households use in a month are provided. Third, representative unit costs are multiplied by typical monthly household water use to obtain estimates of the monthly economic costs of providing a typical household with improved, piped W&S services. The economic costs of providing a household with modern W&S services are the sum of seven principal components: 1. opportunity costs of diverting raw water from alternative uses to the household (or resource rents); 2. storage and transmission of untreated water to the urban area;

3. treatment of raw water to drinking water standards; 4. distribution of treated water within the urban area to the household; 5. collection of wastewater from the household (sewerage collection); 6. treatment of wastewater (sewage treatment); and 7. any remaining costs or damages imposed on others by the discharge of treated wastewater (negative externalities). Table 1 presents some illustrative average unit costs for each of these seven cost components, expressed in US$ per cubic meter. The unit costs of these different cost components could vary widely in different locations. For example, in a location with abundant freshwater supplies, the opportunity cost of diverting water from existing or future users to our illustrative household (item 1) and the damages imposed by the discharge of treated wastewater (item 7) may, in fact, be very low or even zero. However, in more and more places, these opportunity costs associated with water diversion and the externalities from wastewater discharge are beginning to loom large. Some cost components are subject to significant economies of scale, particularly storage and transmission (item 2), the treatment of raw water to drinking water standards (item 3), and the treatment of sewage (item 6). This means that the larger the quantity of water or wastewater treated, the lower the per-unit cost. On the other hand, some cost components are experiencing diseconomies of scale. As large cities go farther and farther away in search of additional freshwater supplies, and good reservoir sites become harder to find, the unit cost of storing and transporting raw water to a community increases. There are also trade-offs between different cost components: one can be reduced, but only at the expense of the other. For example, wastewater can receive only primary treatment, which is much cheaper than primary and secondary treatment, but then the negative externalities associated with wastewater discharge will increase. Table 1 Cost estimates: improved water and sanitation services (assuming 6% real discount rate, see Whittington et al. (2008) for details) No.

Cost component

US$/m3

1

Opportunity cost of raw water supply Storage and transmission to treatment plant Treatment to drinking water standards Distribution of water to households (including house connections) Collection of wastewater from home and conveyance to wastewater treatment plant Wastewater treatment Damages associated with discharge of treated wastewater

0.05

3

0.10

5

0.10

5

0.60

30

0.80

40

0.30 0.05

15 3

2.00

100

2 3 4

5

6 7 Total

% of total

Pricing Water and Sanitation Services

The cost estimates in Table 1 include both capital expenses, and operation and maintenance expenses. The opportunity costs of raw water supplies (item 1) are still quite low in most places, on the order of a few cents per cubic meter. Even in places where urban water supplies are diverted from irrigated agriculture, the unit costs will rarely be above US$0.25 per cubic meter. Desalinization and wastewater reclamation costs will set an upper limit on opportunity costs of raw water in the range of US$0.50–1.00 per cubic meter for cities near the ocean, but the opportunity costs of raw water are nowhere near this level in most places. Raw water storage and transmission and subsequent treatment (items 2 and 3) will typically cost US$0.20 per cubic meter. Within a city, the water distribution network and household connections to it (item 4) comprise a major cost component, in many cases on the order of US$0.60 per cubic meter. The collection and conveyance of sewage to a wastewater treatment plant (item 5) are even more expensive than the water distribution; this will cost about US$0.80 per cubic meter, 40% of the total cost. Secondary wastewater treatment (item 6) will cost about US$0.30 per cubic meter. Damages resulting from the discharge of treated wastewater are very sitespecific, but environmentalists correctly remind us that that they can be significant, even for discharges of wastewater receiving secondary treatment. Let us assume for purposes of illustration that these costs are of the same order of magnitude as the opportunity costs of raw water supplies (US$0.05). As shown, total economic costs are about US$2.00 per cubic meter in many locations. It is emphasized that costs shown here are not intended to represent an upper bound. For example, in small communities in the arid areas of the western United States’ costs of W&S services can easily be double or triple these amounts per cubic meter. Also, note that these cost estimates assume that financing is available at competitive international market rates, and that countries do not pay a high default or risk premium. Table 2 presents a reasonable lower-bound estimate of unit costs of piped W&S services. Here, the opportunity costs of raw water supplies and the damages from wastewater discharges are assumed to be zero. Only minimal storage is included, and the only intake treatment is simple chlorination. Costs for the water distribution network assume the use of polyvinyl chloride pipes and shallow excavation. Wastewater is collected with condominial sewers, and the only wastewater treatment is provided by simple lagoons. Given all these assumptions, one can manage to reduce unit costs of piped W&S services to about US$0.80 per cubic meter. How much water does a typical household in a developing country need? The quantity of water used by a household will be a function of the price charged, household income, and other factors. Currently, most households in developing countries are facing quite low prices for piped W&S services. One can look at water use data from households in industrialized countries to see how much water one might expect a household to use for a comfortable modern lifestyle. For households with an in-house piped water connection, in many locations residential indoor water use falls in the range of 110–220 l per capita per day. For a household of six members, this would amount to about 20–40 cubic meters per month (Table 3). At the current low prices prevailing in many

81

Table 2 Cost estimates: improved water and sanitation services for low-cost option for private water and sewer connections (assuming 6% real discount rate, see Whittington et al. (2008) for details) No.

Cost component

US$/m3

1

Opportunity cost of raw water supply (steal it) Storage and transmission to treatment plant (minimal storage) Treatment of to drinking water standards (simple chlorination) Distribution of water to households (PVC pipe) Collection of wastewater from home and conveyance to wastewater treatment plant (condominial sewers) Wastewater treatment (simple lagoon) Damages associated with discharge of treated wastewater (someone else’s problem)

0.00

2 3 4 5

6 7

Total

0.07 0.04 0.24 0.30

0.15 0.00

0.80

Table 3 Range of estimates of monthly water use (in-house, private connection) Per capita daily water use (l)

Persons per household

Days per month

Monthly household water use (m3)

55 110 220

6 6 6

30 30 30

10 20 40

cities in developing countries, such levels of household water use are not uncommon. Other things being equal, households living in hot, tropical climates use more water for drinking, bathing, and washing than households in temperate or cold climates. Assuming average unit costs of US$2.00 per cubic meter, the full economic costs of providing 20–40 cubic meters of water to a household (and then dealing with the wastewater) would be US$40.00–80.00 per month (Table 4), more than most households in industrialized countries pay for the same services and far beyond the means of most households in developing countries. One would expect poor households in developing countries with in-house water connections to respond negatively to higher W&S prices: they might curtail use to as little as 50–60 l per capita per day. For a household with six members, at 55 l per capita per day, total consumption would then amount to about 10 cubic meters per month. The full economic costs of this level of W&S service at this reduced quantity of water use (assuming our unit costs of US$2.00 per cubic meter remained unchanged) would then be US$20.00 per month per household. At entirely plausible levels of water use (110 l per capita per day), the total economic cost would be about US$40.00 per month for the same household. With the unit costs of the low-cost system depicted in Table 2, the full economic cost of providing 10 cubic meters per month would be US$8.00 per household per month. This estimate should be regarded as a

82

Pricing Water and Sanitation Services

Table 4 Range of estimates of the full economic cost of providing improved W&S services (in-house, private water connection; piped sewer) Monthly household water use (m3)

Average cost ¼ US$0.80/m3 (US$)

Average cost ¼ US$2.00/m3 (US$)

10 20 40

8 16 32

20 40 80

lower bound on the full economic costs of piped W&S services in most locations. In industrialized and developing countries alike, most people are unaware of the magnitude of the true economic costs of municipal W&S services. There are several reasons why these economic costs are so poorly understood. First, the capital costs are heavily subsidized by higher levels of government (and, in developing countries, by international donors), so that households with services do not see the true capital costs reflected in the fixed charges or volumetric prices they pay. Second, in many cities tariff structures are designed so that industrial water usage subsidizes residential usage; households thus do not even see the full operation and maintenance costs in the prices they pay. Third, as many water utilities run financial deficits (in effect running down the value of their capital stock), water users in aggregate do not even see the full costs of supply. Fourth, most cities do not pay for their raw water supplies: typically, the water is simply expropriated from any existing water sources (and their users) in outlying rural areas. Fifth, wastewater externalities are typically imposed on others (downstream) without compensation. Sixth, the subsidies provided to consumers of W&S services are not only huge, but also regressive. It is often not politically desirable for the majority of people to understand that middle- and upper-income households, who generally use more water, are thus actually receiving the most benefit from subsidies. Tariff designs may in fact be made overly complicated in order to offset this reality and appear to be helping poorer households (Komives et al., 2005). Most fundamentally, poor households are often not connected to the W&S network at all and hence cannot receive the subsidized services. Even if they do have connections, the poor use less water than richer households, thus receiving lower absolute amounts of subsidies. The estimates presented here are intended merely to suggest what W&S costs are like in many developing countries. A reasonable question to ask is whether costs differ much across countries in the developing world and between industrialized and developing countries. Labor costs are obviously lower in developing countries, but because W&S projects are capital intensive, the cost component has less of an impact on total costs than for other goods and services. There are, to our knowledge, no publicly available international indices of W&S project construction costs. To illustrate the magnitude of international cost differentials for some related goods and construction costs, Table 5 compares costs of rebar, cement, and industrial construction in 11 large cities in both

Table 5 Comparison of costs of rebar, cement, and industrial facility construction in 11 cities City

Rebar (US$/ton)

Cement (US$/ton)

Industrial Construction (US$/m2)

London Boston Los Angeles Shanghai Jakarta Bangkok Hanoi New Delhi Durban Nairobi Buenos Aires

981 1100 992 435 528 482 349 600 1028 NA 765

96 85 135 43 68 63 62 64 137 NA 82

850 915 699 592 269 301 409 247 516 291 NA

From Engineering News-Record (2004) 253. 24 (December 12), 32–37.

industrialized and developing countries. Costs are indeed lower in cities such as New Delhi and Hanoi than in London and Boston, and lower costs for inputs such as cement and steel will translate into lower costs for W&S projects. It is, of course, less expensive to provide intermediate levels of W&S services (e.g., public taps and communal sanitation facilities) than the costs in Table 2 would indicate. Monthly household costs for such services are, however, often quite considerable, roughly US$5.00–10.00 per month for much smaller quantities of water and much lower levels of sanitation services. These costs are often reported to be as low as US$1.00–2.00 per household per month, but such accounts often systematically underestimate key cost components and rarely reflect the real costs of financially sustainable systems. It is also important to appreciate that intermediate services impose additional costs on households in terms of extra time spent accessing the services and increased coping costs for the inconveniences of using off-site services (Bahl et al., 2004; Pattanayak et al., 2005).

1.06.3 W&S Development Paths The high capital costs of network water and sewer systems have important implications for water prices and tariff design. Decisions on how to price network W&S services in developing countries are typically made in a dynamic, changing environment. Pricing and tariff design decisions made today should not lock households into low-level equilibrium solutions that will constrain them from improving their W&S services as economic growth occurs. In-house piped W&S services are unaffordable today in many cities in developing countries, but as economic growth occurs, there is general agreement that this goal is both desirable and achievable. It is thus important to consider carefully how pricing and tariff-design decisions influence the evolution of W&S service provision and the ability of managers and planners to upgrade services when economic growth creates the resources to make this vision a reality. There are numerous strategies or development paths for moving from a situation where households have poor or no services to

Pricing Water and Sanitation Services

modern W&S services, and it is necessary to reflect explicitly on the pros and cons of each, and how pricing and tariff design decisions push service providers and households along a particular development path, or create hurdles that must be overcome to make progress. For purposes of illustration, Table 6 compares three levels of water services and four levels of household sanitation. Let us consider a household without either improved water or sanitation services (case 1). Within the parameters given in the table, such a household might progress from this status to full modern W&S services (case 12) along any of the four principal development paths. First, some water planners would advocate for a water-first development path (case 1 - case 2 - case 3 - case 6 - case 9 - case 12); here W&S service providers concentrate on first getting piped water services into the household; only after this stage is achieved would investments go to the installation of neighborhood sewers and then to wastewater treatment. Note that the household itself has important investments to make. On the water side, in-house plumbing is required to take full advantage of the piped water connection. Similarly, the household would typically be responsible for the installation of a private water-sealed toilet, without which the installation of neighborhood sewers would be of less value. Proponents of a water-first development path argue that people want water services first and do not recognize the need for removing wastewater from the household until water has been provided and wastewater removal has become a problem. Also, as described above, sewers and wastewater treatment are very expensive, so it is easier financially to provide the less-expensive services first. From a pricing perspective, a water-first strategy has important implications. Under this strategy, revenues from water sales should not be diverted to subsidize sewers or wastewater treatment, at least until the majority of the population has high-quality water services. Also, any available subsidies from

Table 6

83

higher levels of government should be used to push households toward in-house piped service. For example, subsidies might be used to reduce the upfront connection charge for a household water hookup. Public health professionals sometimes argue in favor of a second development path: not one without the other (case 1 - case 5 - case 9 - case 12). Proponents of this approach believe that there are important public health complementaries from providing improved W&S together, and that households should not be allowed to receive in-house piped water without hooking up to a sewer line. Engineers often point out that it is cheaper to install water and sewer lines at the same time, particularly in cities where this may entail tearing up streets, sidewalks, and other infrastructure. Such bundling of W&S services has important implications for tariff design. If households are required to have sewer services when they receive piped water services, then from a household’s point of view, W&S services cannot really be charged separately. (Also, note that W&S service providers cannot practically meter the amount of water that a household receives separately from the amount of wastewater that it discharges. Thus, even if a provider claims to calculate water and wastewater charges separately, and apply a separate volumetric charge to each flow, from the household’s perspective this is simply an accounting trick. The household effectively faces a single weighted volumetric rate for the combined service.) If the service provider attempts to recover the full costs of both services, and the household is willing to pay the cost of the water services but unwilling to pay for the sanitation services, the household will reject the entire bundle. Thus, when services are bundled and tariffs are designed to recover the costs of service, tariffs can easily become a barrier to the provision of full modern network services (case 12). A third development path might be termed ‘sanitation first’ (case 1 - case 4 - case 5 - case 8 - case 9 - case 12). The rationale here is that improved sanitation is a more important

Water and sanitation development paths Unimproved water source (e.g., pond and river)

Improved water source outside the home (e.g., hand-pump and public tap)

Improved water inside the home (private water connection or yard tap)

No improved sanitation

Case 1

Case 2 (US$5/month/household)

Case 3 (US$10/month/ household)a

On-site sanitation (e.g., VIP latrine and pour flush toilet)

Case 4 (US$5/month/household)

Case 5 (US$10/month/household)

Case 6 (US$15/month/household)

Water-sealed toilet þ neighborhood wastewater collection (e.g., small-bore or conventional sewers)

Case 7 (US$15/month/ household)b

Case 8 (US$20/month/household)

Case 9 (US$25/month/household)

Water-sealed toilet þ neighborhood wastewater collection þ wastewater treatment

Case 10 (US$25/month/ household)

Case 11 (US$30/month/ household)

Case 12 (US$35/month/ household)

a

Water costs are not cumulative because having a private connection does not require a public tap or handpump. Sanitation costs are cumulative, that is, level 3 includes the costs of in-house plumbing þ neighborhood wastewater collection.

b

84

Pricing Water and Sanitation Services

first step than improved water services in achieving the desired public health benefits. Thus, if resources are limited, public authorities should tackle sanitation problems before building piped water distribution networks. From a pricing perspective, if demand for improved sanitation services turns out to be low, this development path requires more initial subsidies, with revenues from water to follow when households eventually receive in-house water connections. A fourth set of development paths might be termed ‘demand-driven’ in that the paths are not selected by experts but rather by people themselves. There are numerous plausible development paths that households might choose (e.g., case 1 - case 3 - case 6 - case 9 - case 12; or case 1 - case 6 case 9 - case 12). If households’ preferences are allowed to shape the evolution of W&S services, prices and tariff design have an especially important role to play. Prices provide the signals about the real resource costs of the various steps from the status quo to full modern network services. If these signals are incorrect, households may take an unwanted or unnecessary detour on the road to case 12.

1.06.4 Objectives of Tariff Design Setting water (and sanitation) tariffs requires that one strikes a balance between four main objectives (see Boland (1993) and Whittington et al. (2002) for additional discussion of tariff objectives). Cost recovery. From the water supplier’s point of view, cost recovery is the main purpose of the tariff. (For example, the World Bank’s Operational Manual Statement No. 3.72 emphasizes the importance of the cost recovery objective and the financial autonomy of the borrower.) Cost recovery requires that, on aggregate, tariffs faced by consumers should produce revenue equal to the financial costs of supply. Moreover, the revenue stream should be relatively stable and not cause cash flow or financing difficulties for the utility. Economic efficiency. Economic efficiency requires that prices be set to ensure that customers face the avoidable costs of their decisions. In other words, prices should signal to consumers the financial, environmental, and other costs that their decisions to use water impose on the rest of the system and on the economy. In practice, this means that the volumetric charge should be set equal to the short-run marginal social cost of bringing one additional cubic meter of water into a city, delivering it to a particular customer, collecting and treating the wastewater, and discharging the treated wastewater into a receiving water body. In many cities, the cost of bringing in additional water is higher than the cost of supplying the water already on hand, as the cheapest sources tend to be developed first. The short-run marginal cost should include not only the financial cost of public works undertaken, but also the social cost of diverting water resources into public supply rather than using it for other purposes. An efficient tariff will create incentives that ensure, for a given water supply cost, that users obtain the largest possible aggregate economic benefits. Equity. The term ‘equity’ is often used to denote quite different things. Here, it is used to mean that the water tariff treats similar customers equally, and the customers in different

situations are not treated the same. This usually means that users pay monthly water bills that are proportionate to the costs they impose on the utility by their water use. Affordability. One objective of tariff design is to ensure that poor households are able to obtain adequate supplies of clean water. The terms ‘equity’, ‘fairness’, ‘poverty alleviation’, and ‘affordability’ are often used interchangeably to express this desire. It is preferred to treat affordability as an objective distinct from equity, fairness, and poverty alleviation, because a W&S tariff that is affordable may not be equitable or perceived as fair. Moreover, an affordable tariff may not pull poor households out of poverty. Many people feel that water services are a basic right and should be provided to people regardless of whether they can pay for the services. These considerations have led to recommendations that W&S tariffs should be kept low and that water should be provided free or at minimal cost, at least to the poor, through systems of subsidies. There are a number of trade-offs between these different objectives and the W&S tariffs used to calculate customers’ bills. For example, providing water free through private connections in order to achieve the objective of affordability conflicts with the objectives of cost recovery and efficient water use. Also, poor customers can sometimes be relatively expensive to serve (e.g., perhaps due to their outlying location), and hence it might not be regarded as equitable to charge them the same as, or less than, other customers. Additional objectives and considerations may be involved. For example, a tariff design should be easy to explain, understand, and implement. A tariff design should be acceptable both to the public and to the political leaders. This may require the tariff to conform to perceptions of fairness, often quite different from notions of equity. Water tariffs may be designed to discourage excessive uses of water, thus promoting water conservation, where excessive may be understood as a deviation from some notion of a fair amount. A successful W&S tariff design should not be controversial, nor should it become a focus of public criticism of the watersupply agency. Human beings are, however, acutely sensitive to situations perceived to be unfair, and fairness is often in the eye of the beholder. It can prove to be especially difficult to design a W&S tariff that is perceived to be fair when customers do not understand the true resource costs of providing modern W&S services. Consider the four cases in Table 7. If household members understand the real resource costs of supplying modern W&S services and believe the household should pay a share of these costs proportionate to its use of such services (case A), a W&S tariff that is perceived to be fair can be relatively easily designed. But if household members do not understand the real resource costs of supplying modern W&S services, it may prove to be difficult for them to believe that a tariff is fair even if they believe the household should pay a share of these costs proportionate to its use of such services (case B). For example, such a household may perceive a proposed W&S tariff to be price gouging even if it is not. On the other hand, household members may understand the real resource costs of supplying modern W&S services but not believe the household should pay a share of the costs proportionate to its use of such services (case C). This may be due to past injustices, a feeling that this household is more

Pricing Water and Sanitation Services Table 7 Households’ understanding of supply costs vs. agreement to pay a proportionate share of the costs of W&S services: four cases A household understands the real resource costs of supplying modern W& S services

A household does not understand the real resource costs of supplying modern W& S services

A household believes that it should pay a proportionate share of the costs of W& S services

Case A

Case B

A household believes that it should not pay a proportionate share of the costs of W& S services

Case C

Case D

deserving of help than others, or any number of reasons. Or household members may neither understand the real resource costs of supplying modern W&S services, nor believe the household should pay a share of the costs proportionate to its use of such services (case D). This is the most difficult situation for all stakeholders, and unfortunately it is quite common. In case C and especially in case D, the negotiation of W&S tariffs often becomes a political problem largely unrelated to the costs of service delivery.

1.06.5 Tariff Structures – the Alternatives A tariff structure is a set of procedural rules used to determine the conditions of service and the monthly bills for water users in various categories. (This section draws heavily on Whittington et al. (2002).) Table 8 presents a simple classification of the different types of water tariff structures. Two main types of tariff structures are used in the municipal water supply sector: a single-part tariff and a two-part tariff. With a singlepart tariff, a consumer’s monthly water bill is based on a single type of calculation. With a two-part tariff, a consumer’s water bill is based on the sum of two calculations. The single type of calculation used in a single-part tariff can be one of two types: a fixed charge or a water use (volumetric) charge; volumetric charges can be handled in several different ways. Figure 1 illustrates how the price of water to the consumer changes as the quantity of water used increases for some of these tariff structures. Figure 2 shows how the customer’s monthly water bill varies as the quantity of water used increases for selected tariff structures.

1.06.5.1 Single-Part Tariffs 1.06.5.1.1 Fixed charges In the absence of metering, fixed charges are the only possible tariff structure. With a fixed charge, the consumer’s monthly water bill is the same regardless of the volume used. In many countries, renters in multi-story apartment buildings have unmetered connections to their units and thus effectively pay

Table 8

85

Basic types of water tariff structures

Single-part tariffs A. Fixed charge: monthly water bill is independent of the volume consumed B. Water use charge a. uniform volumetric tariff b. block tariff: unit charge is constant over a specified range of water use and then shifts as use increases (i) increasing block (ii) decreasing block c. increasing linear tariff: unit charge increases linearly as water use increases Two-part tariffs Fixed charge þ water use charge

a fixed charge for water (perhaps incorporated into their rent). Fixed charges are still quite widely used in industrialized countries, such as Canada, Norway, and the United Kingdom (and until recently in New York City), where water has historically been abundant and hence metering is not widespread. The fixed charge itself can vary across households or consumer classes depending on characteristics of the consumer. For example, historically a common way to charge differential fixed charges was to set higher fixed charges on more valuable residential properties, sometimes on the assumption that people living in higher-value dwellings tend to use more water and/or have a greater ability to pay for the water they use. It was also common to assign businesses, a different fixed charge than households, on the assumption that firms use more water than households, and notions of fairness (e.g., that firms have a greater ability to pay for water than households). Another common approach is to charge different monthly fees depending on the diameter of the pipe used by the customer to connect to the distribution system: single-family domestic connections generally require a smaller bore than connections for larger concerns (e.g., businesses, hospitals, and apartments). From the perspective of economic efficiency, the problem with a fixed-charge system is that consumers have no incentive to economize on water use, as using more water will not increase their water bill. If the short-run marginal cost of supply is very low due to excess capacity in the system, this may not be a big problem. However, from a cost recovery perspective, a fixed-charge system creates a potentially large problem for the utility (or operator) if some households still lack individual connections: customers who do have a connection can supply water to other users (e.g., unconnected households and vendors) without incurring an increase in the household water bill. Moreover, because the fixed charge offers no incentive to economize on the use of water, a fixed charge that provided sufficient revenues at one point in time will become increasingly inadequate as the economy and incomes grow and water use increases. W&S service providers will be reluctant to expand coverage because more customers may mean more financial losses. Fixed-charge tariffs are thus especially prone to locking communities into low-level equilibrium traps of few customers, low revenues, and poor service (Whittington et al., 1990).

86

Pricing Water and Sanitation Services 1

Price ($/m3)

0.75

0.5

0.25

0 0

5

10

15 Quantity (m

Decreasing block

20

25

30

3)

Increasing block

Increasing linear

Uniform

Figure 1 Price of water vs. the quantity of water used for selected tariff structures.

30

Monthly bill ($)

25 20 15 10 5 0 0

5

10

15 Quantity

Decreasing block

20

25

30

(m3)

Increasing block

Increasing linear

Uniform

Figure 2 Monthly water bill vs. the quantity of water used for selected tariff structures.

1.06.5.1.2 Volumetric charges

1.06.5.1.3 Uniform volumetric charge

The second way to structure a single-part tariff is to base consumers’ water bills on the amount of water they use. In mathematical terms, the monthly water bill is thus a function of the quantity of water a consumer uses. The precise formula used for the calculation of the water bill can differ. There are three main options: (1) a uniform volumetric charge; (2) a block tariff where the unit charge is specified over a range of water use for a specific consumer, and then shifts as use increases; and (3) an increasing linear tariff whereby the unit charge increases linearly as water use increases. All volumetric charges require that the consumer has a metered connection and that this meter works reliably and is read on a periodic basis.

With a uniform volumetric charge, the household’s water bill is simply the quantity used (e.g., cubic meters) times the price per unit of water (e.g., US$ per cubic meter). This is the most common type of volumetric charge among water utilities in the United States, Australia, and a number of European countries and is also very common for industrial and commercial users throughout the world. A uniform volumetric charge has the advantage that it is easy for the consumer to understand, in part because this is how most other commodities are priced. From an economic efficiency point of view, it can be used to send a clear, unambiguous signal about the short-run marginal cost of using water.

Pricing Water and Sanitation Services 1.06.5.1.4 Block tariffs Block tariffs come in two main varieties: increasing and decreasing. They create a stepwise price structure as illustrated in Figure 1. With an increasing block tariff (IBT), consumers incur a low volumetric per-unit charge (price) up to a specified quantity (or block); for any additional water consumed, they pay a higher price up to the limit for a second block, even higher for the third, and so on. IBTs are widely used in arid areas such as Spain and parts of the Middle East, where water resources have historically been scarce. The use of IBTs is also widespread in many developing countries in Latin America and Asia. With a decreasing block tariff (DBT), on the other hand, consumers face a high volumetric charge up to the specified quantity in the first block, pay less per unit for additional water up to the limit for second block, then less still for the third, and so on. Thus, for both an IBT and a DBT structure, the water bill is calculated in the following manner: Let Q* ¼ amount of water sold to a specific consumer; Q1 ¼ maximum amount of water that can be sold in the first block at price P1; Q2 ¼ maximum amount of water that can be sold to a consumer in the second block at P2; Q3 ¼ maximum amount of water that can be sold to a consumer in the second block at P3. If Q*oQ1, then the consumer’s water bill ¼ (Q*) P1. If Q1oQ*oQ2, then the consumer’s water bill ¼ P1Q1 þ (Q* – Q1)P2. If Q1 þ Q2oQ*oQ3, then the consumer’s water bill ¼ P1Q1 þ P2Q2 þ (Q* – (Q1 þ Q2))P3. And so on for however many blocks there are in the tariff structure. The rationale commonly given for an IBT structure is that, in theory, it can achieve three objectives simultaneously. Proponents of an IBT argue that it promotes affordability by providing the poor with affordable access to a subsistence block of water (the lifeline rate). It can achieve efficiency by confronting consumers in the highest price block with the marginal cost of using water. It can raise sufficient revenues to recover costs. (Note: this argument assumes that the marginal cost of water is in fact higher than the first block price. But if a large expansion project has been recently completed, the short-run marginal cost of water may be very low.) The IBT structure has become so widely used in both industrialized and developing countries that many professionals working in the water sector assume that it must always be the most appropriate tariff structure. This is not the case. In practice, IBTs often fail to meet any of the three objectives mentioned above, in part because they tend to be poorly designed. An IBT may provide more expensive water to poorer households than to richer households, because in many cities the poor share connections, and in such cases the resulting higher volumetric use in turn results in higher prices for most of the water that those households consume (see, e.g., Whittington, 1992; Boland and Whittington, 2000; and Komives et al., 2005). Many IBTs also fail to achieve cost recovery and economic efficiency objectives, usually because the upper consumption blocks are not priced at sufficiently high levels and/or because the first subsidized consumption block is so large that almost all residential consumers never consume beyond that level. The DBT structure was designed to reflect the fact that when raw water supplies are abundant, large industrial

87

customers often impose lower average costs because they enable the utility to capture economies of scale in water-source development, transmission, and treatment. Also, large industrial users typically take their supplies from the larger trunk mains and thus do not require the expansion of neighborhood distribution networks. Although it is still used in some communities in the United States and Canada, the DBT has gradually fallen out of favor, in part because short-run marginal costs, properly defined, are now relatively high in some parts of the world, and there is thus increased interest in promoting water conservation by the largest customers. The DBT structure is also often politically unattractive because it results in high-volume users paying lower than average water prices.

1.06.5.1.5 Increasing linear tariff The increasing linear tariff structure is rarely used. It is of interest largely because it illustrates that there are many ways that water bills can be related to the quantity of water used. In this tariff structure, the price that a consumer pays per unit increases continuously (rather than in block increments) as the quantity of water used increases. (In other words, the water bill ¼ (Q*)P*, where Q* ¼ amount of water sold to a specific consumer; and P* ¼ (a1 þ a2)Q* and a1 and a2 are positive constants.) This tariff structure sends the consumer a powerful signal that increased water use is costly. Not only each additional unit of water used is sold at a higher price, but all the preceding units are sold at the last (high) price. A related but different tariff structure would require that only the last unit used would be sold at the highest price; other units would be sold at the price associated with the lower quantity. It is important to recognize, however, that an increasing linear tariff cannot send the proper economic signal to a consumer about the short-run marginal cost of additional water use. This is because the utility’s short-run marginal cost of providing water does not change appreciably as the water use of an individual household changes. An increasing linear tariff would thus be especially inappropriate if applied to large-volume industrial or commercial water users because it could drive the price they confront for increased water use far beyond the short-run marginal cost of supplying them additional water.

1.06.5.2 Two-Part Tariffs With a two-part tariff, the consumer’s water bill is based on the sum of two calculations: (1) a fixed charge and (2) a charge related to the amount of water used. There are many variations in the way these two components can be put together. The fixed charge can be either positive (a flat fee) or negative (a rebate). The water use charge can be based on any of the volumetric tariff structures described above (a uniform volumetric tariff, an increasing or decreasing block tariff, or an increasing linear tariff). In many cases, the fixed charge is kept uniform across customers and relatively low in value, and is used simply as a device for recovering the fixed administrative costs associated with meter reading and billing that are unrelated to the level of water consumption.

88

Pricing Water and Sanitation Services

1.06.5.3 Seasonal and Zonal Water Pricing In some circumstances the short-run marginal costs of supplying water to customers may vary by season. For example, a community may have relatively plentiful water supplies in the rainy season, but much more limited supplies in the dry season; water storage (reservoir capacity) will also be a factor. In such cases, it makes economic sense for water tariffs to reflect the varying circumstances. By charging higher rates in the dry season and lower rates in the wet season, water tariffs can be used to signal to customers that the water supply is not constant across the seasons, and that the costs of maintaining and distributing the water supply may vary as well. The higher dry season rate also serves as a reminder that each user’s consumption of water reduces the amount available for others. Chile is one of the few developing countries that currently uses seasonal water tariffs. Similarly, it may cost the water utility more to deliver water to outlying communities due to, for example, increased pumping costs to reach higher elevations or more distant settlements. A zonal water-pricing structure charges users who live in such areas more for their water because it costs the utility more to serve them. Zonal prices can be used as an economic signal to users that living in such areas involves substantially higher water supply costs and that such information should be factored into customers’ locational and water-use decisions. However, this practice is comparatively rare, in part because it requires the water supplier to collect detailed geographically referenced accounting information. This type of special tariff is only appropriate if the costs of serving the specially zoned areas are significantly higher than for the rest of the community. In fact, costs vary among all users, and a practical tariff always reflects averaged costs to some degree.

1.06.6 Achieving Economic Efficiency and Recovering Capital Costs: Fundamentals of Dynamic Marginal Cost Pricing in the W&S Sector The high costs of the capital investments necessary to build modern network W&S systems make the two-part tariffs described in the previous section especially attractive. They offer service providers a means simultaneously to achieve economic efficiency and cost recovery objectives and also to simplify the design of subsidies to aid poor households. Economic appraisal of W&S investments requires that stakeholders first determine the optimal price to charge for services, if the services are provided, and then determine whether the benefits are greater than the costs if the optimal price is charged. For large capital projects with no constraints on raw water supply, the volumetric charge (one component of a two-part tariff) may be very low in some circumstances because short-run marginal costs can be very low. The economic logic for setting price equal to the short-run marginal cost is straightforward (see, e.g., Layard and Walters, 1978, pp. 171–176). Consider a community with an inverse demand curve for W&S services p ¼ b1  b2x, where p ¼ price of the services, x is the quantity of W&S services that can

be supplied per time period, and b1 and b2 are positive coefficients. Let C equal the fixed costs per period of the W&S system, which is by definition assumed not a function of x. The investment is able to provide an amount of water Qc per period, where b1/b2 is less than Qc. Net benefits are maximized when the optimal quantity of W&S services x* is provided:

Total benefits  costs ¼

Z

x

ðb1  b2 xÞdx  C 0

¼ b1 x  12b2 x2 dðB  CÞ=dx ¼ b1  b2 x ¼ 0 x ¼ b1 =b2 Solving for the price that will achieve this optimal quantity, we see that in this simple example, in which there are no variable costs of water production and delivery and no opportunity costs of the raw water supply, the volumetric price should be set equal to zero (the short-run marginal cost):

p ¼ b1  b2 x p ¼ b1  b2 ðb1 =b2 Þ ¼ 0 If the price is set equal to zero to ensure customers receive the optimal quantity x*, the benefits of the project exceed the costs if

Z

Total benefits4 Costs b1 =b2

ðb1  b2 xÞdx4 C 0 1 2 2b1 =b2 4

C

Such a price will result in large financial deficits unless a fixed charge for capital recovery and other fixed costs is also imposed (the other component of a two-part tariff). The principles that a W&S service provider should follow to determine the volumetric and fixed-charge components of a two-part tariff in different circumstances have not been well understood in the water resources community. The key point is that short-run marginal costs change depending on the regional water resources situation, and both the volumetric and the fixed-charge components of the two-part tariff must change in response to changes in short-run marginal costs. A simple example can illustrate this point. Consider a community without a modern W&S network system that is thinking about undertaking a new project to develop such infrastructure along with arranging a new source of raw water supply. The various stakeholders consider the benefits and costs of such an infrastructure improvement and decide that the project is desirable (benefits exceed the costs). Assume that capital for this project is not available from a higher level of government or a donor agency; the city instead borrows the necessary funds from a bond market, promising to repay the loan from new revenues available from the sale of W&S services. The citizens of the community agree to allocate the responsibility for repaying this loan among all who use the W&S services. Upon the advice of the engineering firm

Pricing Water and Sanitation Services

responsible for designing the project, the community decides to build excess capacity into their W&S system in order to accommodate future population and economic growth. The engineers’ argument is that it is cheap to build this excess capacity now due to economies of scale in the various project components. Figure 3 presents the situation after this first project is built. The community now has the capacity to supply Qc. What volumetric price should the service provider charge? As the short-run cost of supplying additional water to the existing population is now low (because the capital costs have already been incurred), the volumetric charge should be low. The economic logic is that customers should not be discouraged from using more water if such use does not impose increased or significant costs on the W&S suppliers or neighbors. If a customer derives a benefit from using more water and this use does not hurt anyone else, why not permit the additional water use? However, the loan must still be repaid, so this consumer must pay a fair share of the capital costs. How a community determines a customer’s fair share is essentially a political matter. This decision will not affect the economically efficient outcome unless it significantly affects the number of customers who decide to connect to the W&S system. If large numbers of customers decide to disconnect from the system after a project is completed and the new tariff structure is imposed, in most cases this indicates a failure of the planning process. The voices of these customers were not likely to have been heard when the decision was made to build the project. A numerical example will help to clarify this argument. Assume that this W&S project has an average cost of US$0.75 per cubic meter and a short-run marginal cost of US$0.25 per cubic meter. Suppose that if the typical household were charged US$0.25 per cubic meter, it would use 20 cubic meters per month. In this case, the volumetric charge would yield revenue of US$5.00 per month (20 cubic meters  US$0.25 per cubic meter). However, when the loan repayment is considered, the utility actually needs US$15.00 per month from this household (US$0.75 per cubic meter  20 cubic meters). This implies that the fixed charge should be set equal to US$10.00 per month so that the utility can recover its average costs.

89

In this situation, the volumetric price sends a signal that there is more water available for households at this low shortrun marginal cost if households want to use it. The demand curve in period 1 (D1) intersects the short-run marginal cost curve at a point far below capacity Qc (Figure 3). Water is relatively abundant, and households, encouraged by a low volumetric charge, use plenty of water. They pay a significant fixed charge in order to repay the capital that they collectively agreed to borrow. In a well-governed community, households would have been made fully aware of the magnitude of the volumetric and fixed charges that would be necessary when they decided (voted) to undertake the new W&S project. Assume that this two-part tariff structure stays in place and that over time the population and economy of the community grow. As shown in Figure 4, the demand for W&S services shifts out and to the right. W&S services actually become more valuable to customers, but there is no need for the service provider to increase either component of the two-part tariff until point A in Figure 4 is reached, because the loan is being repaid and revenues are sufficient to pay the average costs of service. Customers thus enjoy an increasing consumer surplus on their W&S purchases. The citizens of this community are in effect reaping the benefits of their wise decision to invest in the new W&S project, and to include excess capacity into the project design. However, the water resources professionals can see that this excess capacity is being used up as growth continues, and the day will come when the community reaches the limits of its existing water situation (Qc). They make this known to the public. The community must then decide what to do before point A is reached, because it takes time to develop a new project. Essentially, it can either make do with the amount of water that it has (Qc) or build another water project. Suppose that there is another raw water source available to the community, but a project to develop this second source is more expensive than the one included in the first investment. Assume that this second project would result in a system-wide average cost of US$1.00 per cubic meter. Assume that the short-run marginal costs of the combined system (after this second project is built) would increase as well, to US$0.50 per cubic meter. Suppose that the community decides that the new, second project is too expensive (the benefits are less than the costs). The citizens vote against a bond referendum to raise money to

$0.50/m3 $0.50/m3

$0.25/m3

A

$0.25/m3 D1 D1

D2

Qc Qc Figure 3 First period: first project is completed, excess capacity.

Figure 4 Second period: demand grows as population and economic growth proceed; system capacity is reached at point A.

90

Pricing Water and Sanitation Services

undertake the new investment. Instead, they will try to make do with the water supply they already have. In this case the magnitude of the components of the two-part tariff must change, because the short-run marginal cost of using water changes. Now the volumetric charge must be used to ration the available water supplies, as shown in Figure 5. As population and economic growth proceed, the demand curve for water continues to shift up and to the right, but in this case the total quantity of water Qc available to the community is already being used by existing consumers (Figure 6). The shortrun marginal costs must now reflect the opportunity cost associated with taking water away from some customers: if one customer increases water use, another must decrease water use. The volumetric price of water thus keeps increasing to reflect the rising opportunity cost foregone (scarcity rent) and to ensure that the available water supply is accessible to users who need it most. Suppose that the community allows this process to go on, demand keeps growing, and the volumetric price needs to

$1.00/m3 B $0.50/m3 D3

$0.25/m3

A

Qc Figure 5 Third period: water from first source must be rationed.

$0.50/m3

D4

D5

D6

$0.25/m3

Qc Figure 6 Fourth period: demand continues to grow after completion of the second water project.

increase from US$0.25 to US$1.00 per cubic meter in order to ration the available supply. Assume that if the short-run marginal cost is US$1.00 per cubic meter and the service provider charges this price, average household consumption falls from 20 cubic meters per month to 12 cubic meters per month. Households economize on their use of water because the volumetric price has quadrupled. In effect, by cutting back on water use existing customers are leaving water available for new customers and new and expanded economic activities. This should not come as a surprise to existing consumers, because they themselves voted down the bond referendum that would have provided the finance for the second water project. Assuming that average costs do not change, the W&S service provider needs US$9.00 per month in revenue from the typical household customer (US$0.75 12 cubic meters). However, the volumetric charge yields US$12.00 in revenues (US$1.00 12 cubic meters). Most people would consider it unfair for the provider to reap windfall profits from the increase in the volumetric part of the tariff. The provider does not need the increased revenue to repay the loan or to pay its financial costs of operation. The purpose of the higher volumetric price is not financial, but rather to ration water use economically. The two-part tariff can be used to resolve this fairness problem associated with water rationing. The fixed charge should be reduced as the volumetric charge increases. Instead of a positive fixed charge of US$10.00, for example, a negative charge (rebate) of US$3.00 per month will result in a typical household water bill of US$9.00 per month, precisely the amount the provider needs to cover its costs. In this example, the fixed charge is negative, but this need not be the case. If the scarcity rent is small, the volumetric charge may not be large enough to recover the service provider’s costs, a positive fixed charge may still be needed. Now suppose that the rationing of the water available from this first project becomes an increasing burden on the citizens of the community, and they finally decide that it is worthwhile to build the second water project. This new project was projected to be more expensive than the first project, but nevertheless they vote to approve a bond referendum to finance it, because now they are paying a high volumetric price for water, US$1.00 per cubic meter, due to the high scarcity rent, and an increased supply will bring greater benefits than costs. Again, the community decides to build in excess reservoir capacity, to support further population and economic growth. After the second project is finished, what should the tariff be? The principles are the same as before: the volumetric component of the two-part tariff should be set equal to the short-run marginal cost, which has now risen from US$0.25 to US$0.50 per cubic meter. When the new water project opens, the citizens in this community are relieved that the constraint on their water use has been relaxed, and the volumetric price falls from US$1.00 to US$0.50 per cubic meter. Assume that in response to this decline in volumetric price, the typical household increases its water use to 16 cubic meters per month. Note that this volumetric price is less than in the previous period, when the price was being used to ration supplies, but more than in the first period, when the

Pricing Water and Sanitation Services

community was smaller and the first water project provided cheap and abundant supplies. If the typical household now uses 16 cubic meters and the volumetric charge is US$0.50 per cubic meter, the volumetric component of the two-part tariff yields US$8.00 per month. However, this is not enough for the W&S provider to recover its average costs, which now include loan payments on the second project. The total average cost of providing services is a weighted average of the first and second projects. Recall that this is assumed to be US$1.00 per cubic meter. In this case, the provider needs US$16.00 per month from the typical household (16 cubic meters  US$1.00 per cubic meter). The fixedcharge component of the two-part tariff must then be set at a positive US$8.00 per month. This example illustrates how a two-part tariff can be used to send the correct signal to customers about the economic value of water and at the same time address the financial needs of the W&S provider. The key point is that the volumetric charge should be continually adjusted to reflect the real short-run marginal cost of using water (including any opportunity costs associated with foregone uses), and the fixed-cost component should be adjusted to meet the financial needs of the utility. It is the community’s collective decision to agree (or not) to share the capital costs of the project that ensures that the benefits of the project exceed the costs and that the allocation of costs is considered fair by most parties. Note that regulatory authorities will have an important role to play in the establishment of an optimal two-part tariff. Particularly in times of water scarcity, when a high volumetric price and possibly a negative fixed charge is warranted, a regulatory body needs to ensure that the public understands the rationale for the pricing policy adopted. Unregulated private W&S service providers cannot be expected to reduce their fixed charge as the volumetric charge increases. The major objection to using a two-part tariff in this way is the possible instability in the volumetric price for services (in the example above, the volumetric price starts low, then quadruples, and then falls again). Some water-resource professionals and utility managers feel that changing volumetric prices will confuse customers and prevent them from engaging in careful long-range planning. From this perspective, price stability is a major objective of tariff design. Households and businesses are, however, able to deal with changing prices in the telecommunications and energy sectors, so there is reason to believe that these fears are exaggerated. Two-part tariffs are widely used in telecommunications pricing, although a negative rebate is not common because short-run marginal costs have continued to fall. (Actually, a form of negative rebate does occur with some mobile phone plans that provide expensive smart phones (e.g., Apple’s iPhone) to customers with long-term contracts at prices far below the cost that the mobile phone manufacturer sells the phone to the service provider.) Also, note that in some locations the period during which the volumetric price must be used to ration water use may be quite long. As cities need to go farther and farther afield in search of new supplies, managing water use with high prices may be increasingly attractive compared to incurring the rising capital costs of new projects. If the volumetric price of water cannot change in response to changing water resources’

91

circumstances, it will be increasingly difficult to develop rational W&S pricing policies. But what about poorer members of the community? How can they be provided with improved W&S services when such a two-part tariff is used?

1.06.7 Subsidizing Capital Costs: Reaching the Poor Any discussion of W&S subsidies should begin with the question ‘‘Why do many people (both those working in the water supply sector and others elsewhere) assume that it is a good idea to deliver subsidies to the poor by reducing the water bills of households with private connections?’’ What is it about a piped water distribution network that makes it a good candidate for the delivery of subsidies to the poor? It does not follow that because water itself is a basic need, a piped water distribution system provides an efficient and an effective way to deliver subsidies to the poor. After all, people also have basic needs for food, health services, and housing. The relevant question is not ‘‘How can piped water services be subsidized most effectively?’’ but ‘‘Which subsidy mechanisms reach the poor most efficiently and effectively?’’ It is also important to ask how households themselves view the importance of the good or service to be subsidized. There is strong evidence that households indeed want improved W&S services as their incomes increase; this correlation between W&S coverage and household income suggests that these services are normal goods. As economic growth occurs in developing countries, more and more people obtain improved infrastructure services. Progress is being made particularly in China and India. Figure 7 shows the percentage of households at different income levels that have four infrastructure services (piped water, sewer, electricity, and telephone); the data come from interviews with more than 55 000 households in 15 developing countries (Komives et al., 2003). What is noteworthy about these households is that at all income levels, more people have electricity than have piped water or sewer. Very few of the poorest households have piped water or sewer, yet almost a third of those households have electric service. As monthly household income increases from very low levels to US$300 per month, coverage of all of these infrastructure services increases rapidly; above US$300, coverage increases at a slower rate. The data in Figure 7 should be interpreted carefully. It could be that more households have electricity because W&S networks were not, but electricity was, available in their neighborhoods, or because electric service was less expensive than W&S service. But in fact, monthly household bills for electricity are almost always higher than for W&S service; thus, a comparatively lower cost of service does not explain the pattern seen here, where many households have obtained electricity even when they do not have piped water. Figure 8 shows the percentage of households with different infrastructure services at different income levels in Kathmandu, Nepal. All of these households had the option to connect to all three network infrastructure services: electricity, water, and sewer. The majority of the very poor chose electricity, but not water and sewer. At higher income levels the percentage of households with W&S services is also higher, but

92

Pricing Water and Sanitation Services

the percentage of households with electricity is always higher still. The important point to recognize from these examples is that although water itself is a necessity, this does not necessarily mean that people prefer piped water service to electric service. Indeed, because water is a necessity, households must already have some water source. The question is thus how much an improved source is worth to them. This will depend on many factors, but probably the most important is how poor the household’s existing water service actually is. W&S planners often present the need for improved services as a moral imperative or a basic human right, but given the

choice, many households in developing countries would appear to want electricity before an in-house piped water or sewer connection. In fact, it is unusual for a household in a developing country to have a piped water connection and not have electricity. Figure 9 shows how the prevalence of different infrastructure bundles changes as household income increases. Almost no one, at any income level, has only a piped water service. However, many people do have electricity and not water. Many households in fact have no infrastructure services at all, although that percentage declines rapidly as household income increases. These data suggest that although most households would certainly like improved W&S services,

% hhs with electricity % hhs with sewer connection

% hhs with in-house water tap % hhs with telephone

100

% hhs with service

80

60

40

20

0 200

0

400

600

800

1000

1200

Median monthly household income in 1998 (US$) Figure 7 Infrastructure coverage vs. household income. From Komives K, Whittington D, and Wu X (2003) Infrastructure coverage and the poor: A global perspective. In: Brook P and Irwin T (eds.) Infrastructure for Poor People: Public Policy for Private Provision, ch. 3, pp. 77–124. London: World Bank and Public–Private Infrastructure Advisory Facility.

Electricity

% hhs with service (among hhs with access to 3 services)

100 90

In-house tap

80 70

Sewer

60 50 40 30 20 10 0

28

45

62

74

Median monthly household income in 1998 (US$) Figure 8 Infrastructure choices vs. household income: Kathmandu, Nepal.

169

Pricing Water and Sanitation Services

93

% of hhs with bundle in home

50 40 30 20 10

e hr e

t ri el ec

d an In -h ou se

ta

p

an

ta

d

p

Al lt

ci

ile to

to d an

ty

In -h ou se

ci

ty

t

t ile

on ly ta t ri El ec

ci

ty

In -h ou se

tri El ec

p

on ly

on ly le t To i

No ne

0

Figure 9 Household infrastructure bundles vs. household income: Asia vs. rest of the world. From Whittington D and Komives K (2002) The challenge of demand assessment in pro-poor infrastructure projects. Presentation at the PPIAF/ADB Conference on Infrastructure: Providing Solutions for the Poor – The Asian Perspective. Manila, November.

this is by no means their most important development priority. Given the choice, in some locations households would probably prefer to have any available subsidies directed to other sectors (e.g., roads, power generation, and education). But suppose that a city’s public health professionals and other development experts decide that W&S services are merit goods that must be subsidized. How best can this be done? In his memoirs (Yew, 2000), the former prime minister of Singapore, Lee Kwan Yew insightfully summarizes his philosophy: subsidize investment and savings, not consumption. He succinctly states the advantage of the two-part tariff with respect to making W&S services affordable to poor households. Subsidizing consumption by selling water at low volumetric prices without an accompanying fixed charge is a never-ending distortion, a signal that continually sends customers the wrong message about how expensive the fixed costs of W&S services really are. But once the capital costs are sunk, low volumetric prices may be appropriate to let consumers know the short-run marginal cost consequences of their decisions. This logic means that any available subsidies for piped W&S services should be directed to (1) lowering connection charges and (2) reducing the recurrent fixed charge component of the monthly bill. Capital subsidies are, of course, not without problems. In theory, if the political process can ensure that only economically sound public investments are undertaken, capital subsidies can both assist poor households and foster economic growth. But capital subsidies for infrastructure investments in general, and for W&S investments in particular, require disciplined public sector decision making. Such discipline is extremely difficult when subsidies come from outside the community that is to benefit from the investment. In most circumstances, a community would be foolish to decline a capital grant for an infrastructure project with an associated stream of positive benefits. It is the high initial costs that are

typically the hurdles to W&S improvements, and if someone else volunteers to pay these costs, why not let them? In practice, it has proved almost impossible for national governments or donor agencies to conduct rigorous economic appraisals of W&S projects. Whenever it appears that a particular project might not pass a cost–benefit test, water professionals appeal to intangible benefits to argue that the investment will in fact pass the test. This is particularly the case in the evaluation of rural W&S investments in developing countries, where neither donors nor national agencies attempt serious project appraisal of W&S projects. It is not hard to provide justifications for subsidies of social overhead capital such as water, transportation, and power infrastructure; the problem is removing such subsidies when they are no longer needed. As Hirschman pointed out, ‘‘The trouble with investment in social overhead capital (e.g., water and sanitation investments) y is that it is impervious to investment criteria. y As a result, social overhead capital is largely a matter of faith in the development potential of a country or region. y Such a situation implies at least the possibility of wasteful mistakes’’ (Hirschman, 1958, p. 84, emphasis added.) This is precisely what we have witnessed in the W&S sector, where white elephants and poorly performing projects have been a standard feature of the sector landscape (Therkildsen, 1988). When higher levels of government (or donor agencies) pay the capital costs of W&S projects, numerous opportunities for rent-seeking and corruption arise (Lovei and Whittington, 1993; Olson, 2000; Davis, 2004). If subsidizing the water bills of households connected to piped networks is a bad idea, what policies can instead or additionally be put into place to protect poor households from the rise in piped water bills that will be required for effective improvements and reforms? There are in fact a number of regulations or policy initiatives that can be coupled with the tariff structure to protect poor customers. The most

94

Pricing Water and Sanitation Services

obvious is simply to identify poor households and give them cash assistance to pay their water bills. This is essentially the approach now used in Chile. But even without such means testing, two sets of appropriate pro-poor policies are available.

1.06.7.1 Create a Well-Run System of Public Taps as a Safety Net for the Poor In every locale, W&S providers and regulatory bodies planning to install or expand a piped W&S system need to look carefully at any existing system of public taps. (Here, the term ‘public taps’ refers to a system of fountains in public areas outside of people’s residences where anyone can go to collect water, perhaps for a per-bucket charge or a fixed monthly fee. These public taps do not necessarily need to be run by a public sector utility; they could be efficiently built and managed by a private operator.) In many places, public taps will become obsolete if and when piped services become available: where the majority of households have piped water connections, households without private connections will work out efficient ways of obtaining water from their neighbors at relatively low cost (Whittington et al., 1998). This solution depends on improving the piped distribution system so that connected households do not have to worry about running out of water if they give or sell water to their neighbors. (Public taps will become relatively high-cost sources of supply compared to purchasing from neighbors, because most unconnected households will have to walk farther to collect water from public taps than to obtain it from neighbors, and because the fixed costs of an attendant at the public tap will be large relative to revenues if only low volumes of water are sold.) Nevertheless, public taps may still have an important role to play because they may serve as a water source of last resort for the very poor. In some cases, it is even possible to provide water free from public taps without substantially reducing the revenues of the water utility. This can occur when the availability of free water from public taps does not reduce the number of households desiring private connections for their exclusive use, and when only a small number of households cannot afford private connections. (This is, in fact, the situation in many industrialized countries today. Water is often available free from public fountains, but the vast majority of households still demand private connections in their residences (see, World Bank Water Demand Research Team, 1990).) One source of potential revenue for financing a subsidized system of public taps is the excess revenues that are available if the volumetric price of water from private connections is higher than average costs.

1.06.7.2 Preserve Options for the Poor Poor households are hurt most when they have few options for self-help and when others have restricted their choices. In such cases, it is common to find poor households being exploited. This is as true in the W&S sector as elsewhere. One important way to protect poor households is to preserve their choices so that local mafia or other rent-seeking actors cannot exploit them. There are three main things that can and should be done.

1. Ensure that poor households (and others) can have a private water connection when they want it. Pro-poor policies should not trap poor households into always accepting a low level of off-site water service. If a poor household always has the option of choosing a private connection, when they can afford it, there are limits to the degree they can be exploited by rent seekers. 2. Legalize water vending and sale of household water to neighbors. Vendors and neighbors with private connections create options for poor households: they promote competition in local water markets, limit the reach of spatial monopolies, and drive down water prices. The poor will benefit most from these lower prices. The system of public taps described above also adds to the choices available to poor households, fosters competition, and thus protects the poor from exploitation. 3. Do not give private operators exclusive rights to provide water within a service area. Contracts with private operators should not contain exclusivity clauses. These limit competition and typically end up restricting poor households’ options. Small-scale providers can often lower the cost of providing piped water to poor households; they should be permitted to operate within the contract areas of larger private operators.

1.06.8 Concluding Remarks Two-part tariffs have an important role to play in enabling water utilities simultaneously to achieve economic efficiency and cost recovery objectives. If a large-capacity expansion project has recently been completed, the short-run marginal cost of raw water supply may be very low. Economic efficiency requires that water should be priced at short-run marginal cost. If a two-part tariff is used, however, the necessary revenues can be raised via a fixed charge, without distorting the price signal contained in the volumetric charge. However, in periods of water scarcity (e.g., just before the construction of a water supply augmentation project), the situation is reversed. In this case, pricing at short-run marginal cost implies that the volumetric charge must include the opportunity cost to the user who does not receive water due to scarcity. This scarcity rent causes the volumetric charge to be relatively high in order to ration the available water supply among competing users. Such high volumetric charges may produce revenues in excess of financial costs. This can be corrected by employing a negative fixed charge (rebate), while the volumetric charge remains high enough to send the correct signal to customers from an economic efficiency perspective. Such dynamic tariff design will require that W&S service providers, regulatory bodies, and public officials provide much more information to customers on the rationale behind sound pricing policies. As Hanemann (2005) has observed, it is extremely difficult for publicly owned W&S utilities to receive permission from political regulatory authorities for even modest rate increases, even though such increases are routinely granted to other service providers such as cable television. As water resources management becomes increasingly complicated, the public must become better informed about the challenges for tariff design posed by the high capital costs

Pricing Water and Sanitation Services

of W&S services, the long lives of the projects, and the tradeoffs between competing objectives. This degree of public understanding is unlikely to happen without increasing involvement and participation of stakeholders in the water resources planning and investment process in general and tariff design, and in rate setting in particular.

References Bahl RA, Sinha C, Poulos D, et al. (2004) Costs-of-illness of typhoid fever in Indian urban slum community: Implications for vaccination policy. Journal of Health, Population, and Nutrition 22(3): 304--310. Baumann D and Boland J. (1998) The case for managing urban water. In: Baumann D, Boland J, and Hanemann W. (eds.) Urban Water Demand Management and Planning, ch. 1, pp. 1–28. New York, NY: McGraw-Hill. Boland J (1993) Pricing urban water: Principles and compromises. Water Resources Update 92: 7--10. Boland J and Whittington D (2000) The political economy of increasing block water tariffs in developing countries. In: Dinar A (ed.) The Political Economy of Water Pricing Reforms, pp. 215--236. Oxford: Oxford University Press. Davis J (2004) Corruption in public services delivery: Experience from South Asia’s water and sanitation sector. World Development 32(1): 53--71. Davis J, Kang A, Vincent J, and Whittington D (2001) How important is improved water infrastructure to microenterprises? Evidence from Uganda. World Development 29(10): 1753--1767. Engineering News-Record (2004) 253. 24 (December 12), 32–37. Esrey SA and Andersson I (2000) Ecological sanitation – a missing link to sustainable urban development. Paper presented at the International Symposium ‘‘Urban Agriculture and Horticulture: The Linkage with Urban Planning.’’ Doma¨ne Dahlen, Berlin, Germany, 7–9 July. Hanemann WM (2005) The economic conception of water. In: Peter P, Rogers M, Llamas R, and Martinez-Cortina L (eds.) Water Crisis: Myth or Reality, pp. 61--91. London: Taylor and Francis. Hirschman A (1958) The Strategy of Economic Development. New Haven, CT: Yale University Press. Komives K, Whittington D, and Wu X (2003) Infrastructure coverage and the poor: A global perspective. In: Brook P and Irwin T (eds.) Infrastructure for Poor People: Public Policy for Private Provision, ch. 3, pp. 77–124. London: World Bank and Public–Private Infrastructure Advisory Facility. Komives KV, Halpern FJ, and Wodon Q (2005) Water, Electricity, and the Poor: Who Benefits from Utility Subsidies? Directions in Development. Washington, DC: World Bank. Layard PRG and Walters AA (1978) Microeconomic Theory. New York, NY: McGrawHill. Lovei L and Whittington D (1993) Rent-seeking in the water supply sector: A case study of Jakarta, Indonesia. Water Resources Research 29(7): 1965--1974. Middleton R, Saunders R, and Warford J (1978) The costs and benefits of water metering. Journal of the Institution of Water Engineers and Scientists 32: 11--122. Nauges C and Whittington D (2009) Estimation of water demand in developing countries: An overview. The World Bank Research Observer (forthcoming). Olson M (2000) Power and Prosperity: Outgrowing Communist And Capitalist Dictatorships. New York, NY: Basic Books.

95

Pattanayak S, Yang J, Whittington D, and Kumar B (2005) Coping with unreliable public water supplies: Averting expenditures by households in Kathmandu, Nepal. Water Resources Research 41(2): W02012. Poulos C and Whittington D (2000) Individuals’ rates of time preference in developing countries: Results of a multi-country study. Environmental Science and Technology 43(8): 1445--1455. Sara J, Gross A, and Berg C (1996) Rural Water Supply and Sanitation in Bolivia: From Pilot to National Program. UNDP–World Bank Water and Sanitation Program. Therkildsen O (1988) Watering White Elephants: Lessons from Donor-Funded Planning and Implementation of Rural Water Supplies in Tanzania. Uppsala: Scandinavian Institute of African Studies. United Nations Human Settlement Programme (UN-HABITAT) (2003) Water and Sanitation in the World’s Cities: Local Action for Global Goals. London: Earthscan. United Nations Millennium Project Task Force for Water and Sanitation (2004) What will it take? Water, sanitation, and the millennium development goals. Abridged Draft Final Report. November. New York, NY: SIWI. Van Wijk-Sijbesma C (1989) What Price Water? User Participation in Paying for Community-Based Water Supply. Occasional Paper Series, No. 10. The Hague: International Reference Centre for Community Water Supply and Sanitation. Warford J (1994) A marginal opportunity approach to municipal water pricing. EEPSEA Special Paper. Singapore. White G, Bradley D, and White A (1972) In: Drawers of Water: Domestic Water Use in East Africa. chs. 4 and 5. Chicago: University of Chicago Press. Whittington D (1992) Possible adverse effects of increasing block water tariff in developing countries. Economic Development and Cultural Change 41: 75--87. Whittington D (2003) Municipal water pricing and tariff design: A reform agenda for South Asia. Water Policy 5: 61--76. Whittington D (2006) Reflections on the goal of universal access in the water and sanitation sector: Lessons from Ghana, Senegal, and Nepal. In: Liberalisation and Universal Access to Basic Services: Telecommunications, Water, Sanitation, Financial Services, and Electricity, chap. 5, pp. 135–148. OECD, The World Bank. Paris: OECD Publishing. Whittington D, Boland J, and Foster V (2002) Water Tariffs and Subsidies in South Asia: Understanding the Basics. New Delhi: Water and Sanitation Program. Whittington D, Davis J, Komives K, et al. (2009a) How well is the demand-driven, community management model for rural water supply systems doing? Evidence from Bolivia, Peru, and Ghana. Water Policy 11(6): 696--718. Whittington D, Davis J, and McClelland E (1998) Implementing a demand-driven approach to community water supply planning: A case study of Lugazi, Uganda. Water International 23: 134--145. Whittington D, Hanemann WM, Sadoff C, and Jeuland M (2008) The challenge of improving water and sanitation services in less developed counties. Foundations and Trends in Microeconomics 4(6–7): 469–609. Whittington D and Komives K (2002) The challenge of demand assessment in propoor infrastructure projects. Presentation at the PPIAF/ADB Conference on Infrastructure: Providing Solutions for the Poor – The Asian Perspective. Manila, November. Whittington D, Okorafor A, Okore A, and McPhail A (1990) Strategy for cost recovery in the rural water sector: A case study of Nsukka district, Anambra state, Nigeria. Water Resources Research 26(9): 1899--1913. World Bank Water Demand Research Team (1990) The demand for water in rural areas: Determinants and policy implications. The World Bank Research Observer 8(1): 47--70. Yew L (2000) From Third World to First – the Singapore Story: 1965–2000. Singapore: Times Publishing Group.

1.07 Groundwater Management E Lopez-Gunn, Complutense University, Madrid, Spain MR Llamas, Complutense University, Madrid, Spain A Garrido, Polytechnic University of Madrid, Madrid, Spain D Sanz, University of Barcelona, Barcelona, Spain & 2011 Elsevier B.V. All rights reserved.

1.07.1 1.07.2 1.07.2.1 1.07.2.2 1.07.2.3 1.07.2.4 1.07.2.5 1.07.2.5.1 1.07.2.5.2 1.07.2.6 1.07.3 1.07.3.1 1.07.3.2 1.07.3.3 1.07.3.4 1.07.3.5 1.07.3.6 1.07.3.7 1.07.4 1.07.4.1 1.07.4.1.1 1.07.4.2 1.07.4.3 1.07.5 1.07.5.1 1.07.5.1.1 1.07.5.1.2 1.07.5.2 1.07.5.2.1 1.07.5.2.2 1.07.6 1.07.6.1 1.07.6.1.1 1.07.6.1.2 1.07.6.1.3 1.07.6.2 1.07.6.2.1 1.07.6.2.2 1.07.6.2.3 1.07.7 References

Introduction The Global Silent Revolution of Intensive Groundwater Use Introduction The Role of Groundwater in the Global Water Cycle Location of the Main Aquifers Groundwater Uses: Past and Present The Pros and Cons of the Intensive Use of Groundwater The complex meaning of sustainability in groundwater use The ethics of pumping nonrenewable groundwater (groundwater mining) The Social Sustainability of Groundwater Management The Economics of Groundwater Use Groundwater Costs of Abstraction and Groundwater Tariffs Productivity of Groundwater Use Perverse Subsidies in Water Policy Groundwater: From Open Access to Common Pool? Optimal Groundwater Pricing Departures from Optimality: Second-Best Solutions Internalizing the Value of Environmental Services Provided by Groundwater Regulatory Frameworks for Groundwater Multilevel Governance Diversity in Groundwater Regulatory Regimes The controversy over private, public, or community groundwater rights Implementation and Enforcement of Groundwater Legislation Multilevel Regulatory Frameworks Institutional Aspects of Groundwater Management Groundwater Institutions: Mapping Groundwater Institutional Design Boundary definition The role of groundwater-user associations An Institutional Audit of Groundwater Institutions Higher-level authorities: Supporting, legitimizing, and leading Transparency and participatory groundwater management The Complex Concept of Groundwater Sustainability and Future Management Issues Groundwater Management Externalities Degradation of groundwater quality Susceptibility to subsidence Interference with surface water and ecological impacts Groundwater: Future Risks and Opportunities for Management Groundwater and climate change Future management issues Groundwater: Issues of fit and political windows of opportunity Conclusion

1.07.1 Introduction The last half century has witnessed a spectacular development in groundwater use. Its use has increased from some 100 Mm3 to almost 1000 Mm3 in the period 1950–2000 (Shah, et al., 2007). In this chapter, water volumes are referred to in million cubic meter (106 or Mm3) and billion cubic meter (109, km3 or bcm).

97 98 98 99 99 99 99 100 100 103 103 104 104 105 105 106 106 108 110 110 111 112 113 114 114 114 116 117 117 117 118 119 119 120 120 121 121 123 123 124 124

The table of conversions is as follows:

• • • •

Mm3 ¼106 m3 km3 ¼109 m3 bcm ¼ 109 m3 hm3 ¼106 m3.

97

98

Groundwater Management

This groundwater development has been mainly used for irrigation in arid and semiarid regions. Nevertheless, the use of groundwater for rural and urban water supply is very important; in some countries, like in Italy, it represents 90% of all of the urban water supply. Although the accuracy of available water-use data is still rather illusory, it seems that today the economic value of groundwater irrigation is even greater than the corresponding value of surface-water irrigation. The main factors that have driven this spectacular development in groundwater use include (1) the availability of modern and relatively cheap rigs to drill water wells; (2) the invention and ease of use of the turbine pump that allows abstracting significant volumes of groundwater from deep water wells; and (3) the consolidation of hydrogeology as a reliable science and technology that has dispelled the mystery of groundwater (Figure 1). This extraordinary development of groundwater use has been described as a silent revolution (Forne´s et al., 2005; Llamas and Martı´nez-Santos, 2005) because it was due to the efforts of millions of farmers, with scarce planning and control by conventional water authorities. These public government bodies have been occupied for more than 50 centuries with surface-water systems, beginning in the hydraulic civilizations located in the valleys of large rivers, such as the Nile, the Ganges, and the Yellow River. Therefore, it is not surprising that most high-level water decision makers suffer from hydro-schizophrenia. This disease was described for the first time in 1973 by the American hydrologist Raymond Nace, as the mindset of those that completely separate surface water and groundwater, and usually forget the latter (Llamas, 2004).

From the dug-well to the deep borehole.

This silent revolution has produced great benefits to humankind because it has contributed significantly both to reduce malnourishment in poor countries and to provide drinking water to the rural and urban poor. Moreover, groundwater irrigation is generally a driver for positive social changes. However, the current and frequent situation of inadequate planning and control over groundwater development has also triggered problems, which are mainly related to negative ecological impacts on aquatic ecosystems and groundwater-quality degradation. These problems have been frequently exaggerated by many surface-water experts who have created the pervasive hydromyth of groundwater fragility in order to foster the traditional policy of surface-water infrastructure. In summary, our main message is that it is crucial that high-level water decision makers seriously consider the real role that groundwater is playing and can play in current and future water policy. This role is going to increase even more if the predictions by the International Panel for Climate Change (IPCC) of an increase in temperature and a decrease in precipitation in most arid and semiarid regions become true (Bates et al., 2008). Our emphasis in this chapter is on groundwater management and not on groundwater hydrology. This is, first, because one of the volumes in this treatise deals with hydrology; second, because in our view, the main current problem is not lack of knowledge about aquifer location, characteristics, and functioning, but rather about better ways to manage the aquifers as a common-pool resource. We refer the interested reader to the work of Chevalking et al. (2008) for some useful citations of websites pertaining to the various aspects of groundwater management.

From the water wheel to the pump.

Figure 1 The silent groundwater revolution: changes in global water management.

From the water-witches to hydrogeology.

Groundwater Management

Consequently, after this introduction, Section 1.07.2 is devoted to recall the value of groundwater as a strategic resource. The main specific characteristics of groundwater that require a different management style than surface water is emphasized. Data provided by the United Nations Economic Scientific and Cultural Organization (UNESCO)–International Groundwater Resources Assessment Centre (IGRAC), the International Association of Hydrogeologists and the book by Margat (2008) describe the location of the main aquifers and groundwater uses in most countries. Developing (or semi-developing or emerging) arid and semiarid regions, such as India or regions of East Asia, have experienced a spectacular development in groundwater irrigation during recent years (Shah, 2005; Shah et al., 2007). Such large regions may present a wide variety of conditions: from subsistence livelihoods to market economies and from large alluvial aquifer systems, which may sustain long-term groundwater development, to hard-rock aquifers, where small communities may rely on scarce resources and pumping may prove to be costly. Some arid regions are endowed with good aquifers: these may correspond to countries, such as Saudi Arabia or Libya, where groundwater mining is commonplace (see Section 1.07.2.5.2). Reliance on nonrenewable resources, however, does not seem to render these economies unsustainable. Contrary to the perception of some environmental organizations, a good number of authors and the UNESCO World Commission on the Ethics of Science and Technology (COMEST) consider the use of nonrenewable groundwater resources to be acceptable under certain circumstances (Selborne, 2001; Delli Priscoli et al., 2004; Llamas, 2004). In arid and semiarid regions of industrialized countries, for example, the USA (California, Texas) and Spain, intensive groundwater withdrawals for irrigation are a well-established practice. Development is essentially market driven, as the cost of obtaining groundwater generally amounts to a very small fraction of the crop value. Some authors argue that the depletion of groundwater levels results in an increase of pumping costs, and may ultimately yield these intensive uses economically unsustainable. However, empirical evidence in some areas seems to show the opposite. Farmers are not deterred from pumping despite depths in excess of 400 m (Garrido et al., 2006). This is because switching to highervalue, water-efficient crops may offset the increase of pumping costs, provided that groundwater quality does not worsen (Llamas and Martı´nez-Santos, 2005; Forne´s et al., 2005). It can also be explained by the so-called Gisser–Sanchez effect (see Koundouri (2004) for a detailed explanation), that is, ‘‘the nomanagement (competitive) dynamic solution of groundwater exploitation is almost identical (in terms of derived social welfare) to the efficient management (optimal control) solution’’ (p. 706). This is a management paradox because the serious depletion of aquifers is a major risk to many freshwater ecosystems; yet, the social benefits from managing groundwater extraction are numerically insignificant. This also has significant implication for water managers because it severely constrains the effectiveness of policy options, since implementing reduced extractions is not socially, economically, or politically costless.

99

1.07.2 The Global Silent Revolution of Intensive Groundwater Use 1.07.2.1 Introduction This section summarizes the main hydrologic characteristics of groundwater occurrence, availability, and past and present uses. A detailed description of these aspects can be found in Margat (2008), in the UNESCO–IGRAC website, and is also treated in one of the volumes in this treatise. Intensive use of groundwater is a recent phenomenon, less than half a century old in most places. This situation has occurred mainly in arid and semiarid countries, in some coastal zones, and close to a few mega-cities. This groundwater development has produced great socioeconomic benefits, mainly in developing countries. It has provided cheap drinking water that has helped improve public health. The new irrigated lands have contributed to eradication, or at least mitigation, of malnourishment among those living in poverty. Millions of modest farmers with scarce public or governmental planning, assessment, financing, and control have mainly carried out this intensive groundwater development. This intensive use has really been a kind of silent revolution. In most countries, the corresponding public water or irrigation agencies have been mainly devoted to designing, building, and operating large surface-water irrigation systems. The attitude of some water decision makers who strongly separate surface and groundwater projects, usually ignoring groundwater, was described as hydro-schizophrenia by the well-known American hydrologist, Raymond Nace, in the year 1973 (Llamas, 2004). This attitude has been commonplace in India, Mexico, Spain, and many other arid and semiarid regions worldwide. As a consequence, certain adverse effects have ensued in some places. For instance, in South Asia, the current situation concerning groundwater development has been frequently described as colossal anarchy (Shah et al., 2007). Most of the problems caused by this uncontrolled groundwater development could be avoided or mitigated if the corresponding government agencies had been more active in assessing and controlling groundwater use. On the other hand, surface-water officials have frequently exaggerated such problems. This has created a pervasive hydromyth on the fragility or weakness of groundwater as a reliable resource (Custodio, 2002; Lopez-Gunn and Llamas, 2008). Due to ignorance, vested interests, or, more frequently, because of the low credibility of the official warnings of watergoverning bodies about potential threats, most farmers are not reducing their intensive groundwater abstraction. On the other hand, there are practically no documented cases where intensive groundwater abstraction from medium- or largesized aquifers has caused serious social or economic problems similar to those caused by soil water logging and salinization, or by the people displaced or ousted by the construction of large dams.

1.07.2.2 The Role of Groundwater in the Global Water Cycle The inventory and the movement of water on planet Earth is well known and acceptably quantified, for at least a half century. According to Margat (2008), groundwater storage

100

Groundwater Management

globally, that is, the volume of groundwater stored in the geological formations defined as aquifers, is huge, about 107 km3. This is about 98–99% of all the liquid freshwater in the Earth, although it is only about 1% of the total volume on the hydrosphere (including oceans). The hydrological cycle indicates that water is continuously in motion, driven mainly by solar energy and gravity. This flow of water is very important. It is estimated that every year, precipitation is in the order of 100 000 km3, river flow about 40 000 km3, and groundwater recharge in the order of 10 000 km3. This means that the storage is 1000 times greater than the annual recharge. A large part of this groundwater recharge or runoff feeds the rivers and is included in river flow. However, these global numbers give only a preliminary idea and the specific situation in each region may be quite different. In Margat (2008) and in the UNESCO–IGRAC website, details can be found by country and continent. The main idea to keep in mind is that the volume of freshwater stored in the aquifers is usually huge in comparison with their yearly recharge (usually coming from precipitation) and with discharge (usually to rivers or wetlands). The annual recharge may change from practically zero in the most arid regions to more that 1000 mm yr1 in very humid areas.

1.07.2.3 Location of the Main Aquifers Most countries have made significant efforts to characterize the geological formations defined as aquifers, including their main parameters and functioning and relation with surfacewater bodies. These data are usually synthesized in different types of maps, known as hydro-geological maps. The UNESCO–IGRAC is a center that collects these data from all over the world and makes them available to the general public. The interested reader can find a good summary in the work by Margat (2008) of the situation in most countries (Figure 2) and a brief description of the main aquifer systems, with special emphasis on those that are transboundary (Figure 3). This means that they occupy areas in two or more countries. As an international transboundary resource, it is estimated that there are nearly 240 transboundary groundwater systems or aquifers (WHYMAP, Worldwide Hydrogeological Mapping and Assessment Programme; Lopez-Gunn, 2009). Margat (2008) classifies the aquifers into four groups according to their main geological characteristics: karstic, alluvial, hard rock, and volcanic. Significant attention is devoted to the aquifers in arid and semiarid regions because in these areas, the recharge of groundwater is small and its use may be relevant.

1.07.2.4 Groundwater Uses: Past and Present The use of groundwater coming from springs is as old as humanity. Margat (2008) mentions historical data of dug wells several millennia old. In Armenia, the first infiltration galleries to drain phreatic aquifers are recorded dating to 8 BC. This technology was soon extended to the whole Mediterranean region and to Asia. Hundreds of thousand kilometers of these galleries were constructed, some of them still in operation today. Nevertheless, the spectacular increase in the use of

groundwater has been mainly driven by the improvement in drilling technology and the invention of the turbine pump in the first-third of the twentieth century. Margat (2008) estimates that groundwater abstractions in 2004 were 800 km3 yr1; Shah et al. (2007), however, estimate that this amount is more than 1000 km3 yr1 and its use is on the increase. For the sake of comparison recent studies on the water footprint and virtual water trade (Aldaya et al., 2009), the total use of green and blue water is estimated as 7000 km3 yr1; of this, probably blue water accounts for about 3000 km3.

1.07.2.5 The Pros and Cons of the Intensive Use of Groundwater As mentioned in Section 1.07.1, a good number of authors emphasize the problems related to groundwater development. In this chapter, we intend to present an objective appraisal. Groundwater development produces great economic benefits due mainly to its general resilience to drought, and thus allowing supply to meet demand in a timely fashion. The fact that most groundwater development has been done by private persons, mostly modest farmers, with no, or a small public subsidy, is the best evidence. Some externalities of groundwater development have negative impacts. These externalities are described in Section 1.07.6.1. Nevertheless, in agreement with Llamas and Custodio (2003), in general, many of the negative externalities have been exaggerated and/ or could have been corrected or mitigated by good groundwater management.

1.07.2.5.1 The complex meaning of sustainability in groundwater use Whenever adverse effects of groundwater development begin to be felt, it is common to hear about ‘overexploitation,’ a term usually equated to pumping in excess of the recharge. While this practice is often dismissed as unsustainable, the concept of overexploitation is conceptually complex. This is the reason why a significant number of authors consider it simplistic and potentially misleading (Selborne, 2001; Delli Priscoli et al., 2004; Llamas, 2004). Probably, the most complete analysis is the one by Custodio (2002). As a consequence, more and more authors are changing to the expression intensive use of groundwater instead of using groundwater overexploitation. Intensive groundwater use denotes significant changes on natural aquifer dynamics (Llamas and Custodio, 2003). In contrast with aquifer overexploitation, intensive groundwater use does not convey a positive or negative connotation. It merely refers to a change in flow patterns, groundwater quality, or interrelations with surface-water bodies. It has been stated that the frequently encountered view – that the water policy of arid countries should be developed in relation to renewable water resources – is unrealistic and fallacious. Ethics of long-term water-resources development must be considered with ever-improving technology. It has been customary – as in the Spanish 1985 Water Law – to define overexploitation as the situation when groundwater withdrawal exceeds or is close to the natural recharge of an aquifer. The observation of a trend of continuous significant decline of the levels in water wells during several years is frequently considered as a clear indication of an unsustainable

Rio de Janeiro

Cairo

Medium groundwater recharge (15 − 150 mm/a) Low groundwater recharge ( 15 mm/a)

Medium groundwater recharge (15 − 150 mm/a)

Low groundwater recharge ( 15 mm/a)

2000

3000

4000

Dhaka

5000 km

Jakarta

Seoul

Shenyang

Sydney

Selected city

Continuous ice sheet

Large saltwater lake

Large freshwater lake

Major river

Surface water and Geography

© BGR Hannover / UNESCO Paris 2006

Tokyo Osaka

Manila

Hong Kong

Shanghai

Tianjin

Beijing

Madras Bangkok

Hyderabad

Calcutta

Delhi

Bangalore

Bombay

Lahore Karachi

Area with local and shallow aquifers

1000

Tehran

Figure 2 Groundwater resources of the world. From World-wide Hydrogeological Mapping and Assessment Programme (WHYMAP), special edition 2006.

High groundwater recharge (> 150 mm/a)

0

Istanbul

Moskva

Saint Petersburg

Kinshasa

Area with complex hydrogeological structure

Buenos Aires

Sao Paulo

Lagos

High groundwater recharge (> 150 mm/a)

Santiago

Lima

Bogota

New York

Paris

Major groundwater basin

Groundwater

Mexico City

Los Angeles

Chicago

London

Ruhr area

Figure 3 Map on transboundary aquifers of the world. From World-wide Hydrogeological Mapping and Assessment Programme (WHYMAP), special edition 2006.

Groundwater Resources of the World

Groundwater Management

situation. This is a simplistic approach that might be a long way from the real situation. It often corresponds to a transient state of the aquifer toward a new equilibrium (Custodio, 2002). Intensive groundwater use frequently depletes the water table. Depletions of the order of 0.5 m yr1 are frequent, although rates up to 5–10 m yr1 have been reported (Llamas and Custodio, 2003; Garrido et al., 2006). Farmers are seldom concerned with this issue, except in the case of shallow aquifers. The increase in pumping costs is usually a small problem in comparison with potential groundwater-quality degradation or equity issues such as the drying up of shallow wells or khanats (infiltration galleries), owned by the less-resourceful farmers and located in the area of influence of the deep wells (Wegerich, 2006). This may cause social-equity problems in regions where many farmers cannot afford to drill new wells, or the water authorities are not able to demand just compensation in terms of water or money to poor farmers. The opposite phenomenon (rise of the water table due to surface-water over-irrigation) is also a problem, for example, in Punjab, India, and in Pakistan, or in San Joaquin Valley in California. Raising the water table often results in significant social and economic troubles due to soil waterlogging and/or salinization. It is not easy to achieve a virtuous middle way. As Collin and Margat (1993) state: ‘‘we move rapidly from one extreme to the other, and the tempting solutions put forward by zealots calling for Malthusian under-exploitation of groundwater could prove just as damaging to the development of society as certain types of excessive pumping.’’ In a given aquifer, pumping rates for irrigation may prove to be sustainable from the hydrological viewpoint provided that storage and/or average recharge are large enough. However, water table drawdown may induce degradation of valuable groundwater-dependent ecosystems, such as wetlands, which may be considered unsustainable from the ecological point of view. Would a restraint from pumping be the most sustainable course of action? The answer to this question is difficult. If farmer livelihoods rely heavily on groundwater resources, a ruthless push toward wetland restoration may not be the most sensible solution to the problem. In that case, like in many real-life situations, the social and economic aspects of sustainability come into play, and may eventually offset environmental considerations. Llamas et al. (2007) provide a succinct overview of nine different aspects of groundwater sustainability: hydrological, ecological, economic, social, legal, institutional, inter- and intra-generational, and political. Throughout the text, a distinction is often made between developed and developing regions. This is because perceptions as to what is sustainable vary across geographical boundaries, and are often rooted in cultural, political aspects, and the socioeconomic situations. In this regard, the Hydrogeology Journal theme issue of March 2006 (Llamas et al., 2006) presents the socioeconomic analyses of a number of case studies from all over the world. Therefore, any study on economic sustainability of groundwater use should take into account the specific regional settings. In developing countries where easily accessible unconfined shallow aquifers exist, devices such as the treadle pump to access shallow water tables may constitute a catalyst

103

for irrigation development, while environmental concerns are generally subordinated to human development. This is the case in many small African villages.

1.07.2.5.2 The ethics of pumping nonrenewable groundwater (groundwater mining) Some arid regions have very small amounts of renewable water resources but huge amounts of fresh groundwater reserves, for example, the existing reserves under most of the Sahara desert. In such situations, groundwater mining may be a reasonable action if various conditions are met: (1) the amount of groundwater reserves can be estimated with acceptable accuracy; (2) the rate of reserve depletion can be guaranteed for a long period, for example, from 50 to 100 years; (3) the environmental impacts of such groundwater withdrawals are properly assessed and considered clearly less significant than the socioeconomic benefits from groundwater mining; and (4) solutions are envisaged for the time when the groundwater is fully depleted. Selborne (2001), former chairman of the COMEST, agrees with this approach. In Saudi Arabia, the main aquifers (within the first 300 m of depth) contain huge amounts – a minimum of 2000 km3 – of fresh fossil water that is 10 000–30 000 years old. It is considered that these fossil aquifers can supply useful water for a minimum period of 150 years. Current abstraction seems to be around 15–20 km3 yr1. During a couple of decades, the Saudi government had pumped several km3 yr1 of nonrenewable groundwater to grow low-cost crops (mainly cereals), which were heavily subsidized. The official aim of such an activity was to help transform nomadic groups into farmers. Now, the amount of groundwater abstraction has been dramatically reduced and the farmer nomads have become high-tech farmers growing cash crops. Another example is the situation of the Nubian sandstone aquifer located below the Western Desert of Egypt, where the fresh groundwater reserves are higher than 200 km3 and the maximum pumping projected is lower than 1 km3 yr1. Probably, similar situations exist in Libya and Algeria. Other examples of mining groundwater can be found in Llamas and Custodio (2003).

1.07.2.6 The Social Sustainability of Groundwater Management As previously stated, most aquifers present a large storage volume of groundwater in relation to their renewable resources (often two or three orders of magnitude higher). A practical consequence is that the potential problems do not usually become serious in the short term (within one or two generations). By then, the farmers may have experienced a positive social transition. Groundwater irrigation has proven to be an excellent catalyst for this social transition of farmers in arid and semiarid regions worldwide (Llamas and Martı´nez-Santos, 2005; Moench, 2003, 2007). Increased revenues result in, and allow for, a greater degree of social welfare. In addition, farmers are able to provide better education for their children, who may either move on to other economic sectors (generally more productive), or return to agriculture with a more productive outlook. Therefore, this transition means a reduction of global poverty (Lopez-Gunn and Llamas, 2008).

104

Groundwater Management

This social transition, triggered by groundwater together with the implementation of more efficient irrigation technologies, can often result in a sustainable use in the midterm. However, adequate groundwater management and governance remain as important challenges in areas of India, Spain, China, or the southern United States (Shah et al., 2007; Llamas and Custodio, 2003; Foster et al., 2004). Aquifers constitute an example of common-pool resources, and as in the majority of cases, all actors have direct access (legal or illegal) to groundwater. Therefore, aquifers should typically follow the widely articulated tragedy-of-the-commons pattern (Hardin, 1968). Nevertheless, after half a century of intensive groundwater use, the authors of this chapter do not know any cases of medium-sized or large good aquifers (those with a surface larger than 500 km2, and medium-tohigh transmissivity and storage-capacity values) where the tragic outcomes outlined by Hardin have taken place causing social or economic disturbances – at least not in the degree of magnitude of those caused by soil waterlogging and salinization (India, Pakistan, or California), or the serious social conflicts in relation to people displaced or ousted by the construction of large dams (Briscoe, 2005; Shah et al., 2007). The situation may be different in small or poor aquifers, where storage is not large enough to sustain development for over two or three generations. Although still uncommon, cases of small aquifers that have run out of groundwater have been some times reported verbally to the authors of this chapter. The reality is that even some poor aquifers, such as the Indian hard-rock aquifers, have played a key role in increasing food production. In India, groundwater-irrigated surface has increased by more than 40 million hectares (ha) during the last few decades (Shah et al., 2007). As a consequence, India, despite an almost 100% increase of its population in the last 50 years, has not only achieved food security in practice, but has also become an important grain exporter. However, uncontrolled aquifer development in arid and semiarid regions worldwide raises sustainability concerns, particularly whenever the natural rate of recharge is low.

It might be appropriate to point out the situation of some large aquifers that have undergone overdrafting or groundwater mining for many decades. In many such areas, pumping data are hardly reliable. Take for instance, California’s aquifers, where overdraft estimates range between 1.2 and 2.4 km3 yr1. Equally, the overdraft in California aquifers has not been adequately analyzed since the 1980s. It is perhaps the lack of willingness to monitor, rather than overdraft per se that may constitute the greatest intergenerational threat for groundwater resources.

1.07.3 The Economics of Groundwater Use It is estimated that more than two-thirds of available freshwater is groundwater, and it is currently the most extracted natural resource in the world. More than half the world’s freshwater, for uses like drinking, cooking, and hygiene, comes from groundwater; groundwater irrigates 20% of irrigated agriculture. Groundwater supplies 75–90% of drinking-water supply in European countries, and 95% of the US rural population public-water supply. Aquifers provide natural storage reservoirs with little evaporative loss at little or no cost. Equally, aquifers provide natural transmission of water from the various sources to the point of use. During periods of drought, groundwater provides reliable supplies, compared to surface water, by its use as supplementary irrigation water to surface-water supplies (Howe, 2002). Groundwater is an important economic resource for billions of people, in developed and developing countries. Ninety percent of urban supply in India, and 70% in Mexico are just examples of the socioeconomic importance of this key resource for humans. Figure 4 plots the share of agricultural groundwater use and total groundwater use in total use in 2002 in the Organization for Economic Cooperation and Development (OECD) countries. It shows the importance of both the agricultural sector with groundwater use (countries situated on left) and the percentage of groundwater use over total use (countries on the right).

Po

rtu g G al re ec e N S et pa he in rla n M ds U ni ex te d ico St at e O s C D Ko E re Tu a rk ey EU -1 Ja 5 pa Ire n D lan en d m a Fr rk an Sw ce ed Be en lg G ium er m an Sl ov Au y ak s t U ni Re ria te pu d ki blic ng do Ic m el a C ze Hu nd ch ng R ary ep ub lic

100 90 80 70 60 50 40 30 20 10 0

% share of agriculture use in the total groundwater use

% share of total groundwater use in total water use

Figure 4 Share of agricultural groundwater use in total groundwater use, and total groundwater use in total water use. From OECD (2008) Environmental performance of OECD Agriculture since 1990, Paris, France. Online at: www.oecd.org/agriculture/env/indicators

Groundwater Management

Rosegrant et al. (2002) have estimated that the sustainable yield of groundwater resources in the world would be approximately 861 km3, down from 925 km3 as evaluated in the year 1995, with half this amount abstracted in Asia, and 28% in developed countries. It is difficult to put a dollar value to the use of this resource, but considering a conservative figure of US$ 0.25 m3 (including capital, environmental, and resource costs), this generates an annual total value of US$ 231 billion as a preliminary estimation. The groundwater literature shows that in addition to the direct-use value, groundwater resources also have a significant stabilization value (Tsur and Parker, 1997) in cases where groundwater is conjunctively used with more unreliable surface waters. Estimates of the stabilization value of groundwater resources show that it can be as high as that of direct use. This is because groundwater provides reliability of supply and reduces the probability and severity of water shortages. In addition to households’ water services, groundwater resources are extensively used for food production. With few exceptions, irrigation is the result of uncoordinated efforts of small entrepreneurs and farmers all over the world. These groundwater users have sought to improve their livelihoods investing in private capital and small pumping equipment to improve farm productivity.

105

1.07.3.2 Productivity of Groundwater Use Despite the illusory accuracy of global irrigation data and the variability of the existing estimates, rough calculations yield the following conclusion: groundwater-based irrigation seems to be twice as efficient as surface-water irrigation in hydrological terms (m3 ha1), a ratio that increases to between 3 and 10 times from the social and economic points of view (US$ m3 and jobs m3). Regional-scale analyses carried out in Spain seem to confirm these figures (Herna´ndez-Mora et al., 2001) (see Table 1). Thus, it appears relevant and urgent to assess the comparative hydrological and socioeconomic efficiency of surface and groundwater irrigation at a global scale, carrying out similar studies in other regions of the world. Assessing the implications of this silent revolution should constitute a valuable contribution to the debate about global irrigation needs as perceived by many water experts. The required investment to assess the value and efficiency of groundwater irrigation versus surface-water irrigation can be afforded by most governments. Many high-value crops are watered with groundwater resources or by combining ground and surface water (Llamas and Martı´nez-Santos, 2005). For instance, in Table 1, Herna´ndez-Mora et al. (2001) show that, in Andalusia, irrigated agriculture using groundwater is economically over 5 times more productive and generates almost 3 times the employment than agriculture using surface water, per unit

1.07.3.1 Groundwater Costs of Abstraction and Groundwater Tariffs Groundwater unit volume costs increase with groundwater depth, as more energy is required for pumping and deeper wells might be needed. (In our experience, these costs usually range between US$ 0.02 and US$ 0.30 m3 depending on the country and the aquifer. However, according to Shah et al. (2007) the economic cost (value) of groundwater is about US$ 0.20–0.30 m3). It would be worthwhile to study this aspect worldwide in more detail since values appear very high in comparison to the general economic situation of Southeast Asia. One possible cause is the low technology used in the drilling of the wells and the performance of pumping devices. Groundwater irrigation cost per hectare also increases with time, albeit at a lower rate. This is because farmers begin to use a more efficient technology and switch (if soil and climate allow) to less-water-consuming crops: from maize or rice to grapes or olive trees, for instance. It is estimated that groundwater irrigation cost in Spain generally ranges between US$ 20 and US$ 1000 ha1 yr1. Despite the difficulties in setting tariffs for groundwater use, a number of countries have these in place. In many cases, tariffs are accompanied by quotas and licenses. In general, developed countries have in place a fixed fee plus a volumetric fee, but these levies are generally not adapted to recharge or movements in the water table. Essentially, this indicates that tariffs on groundwater use are environmental levies but not rationing instruments to manage aquifers. This is because, to ensure sustainable management, tariffs would need to be flexible enough to change according to scarcity costs and alluse externalities – and this – assuming that perfect monitoring and information are economically feasible.

Table 1 Comparing ground and surface-water irrigation productivity: Some irrigation economic indicators in Andalusia (Spain)

Irrigated area (ha) Average water consumption (m3 ha1) Total production (106 h) Production (h ha1) Employment generated (number jobs/100 ha) EU aid to income (% of production value) Gross water productivity (h m3) Total average water price to farmer (h m3)

Groundwater

Surface water

Total

244 190 (27%)

648 009 (73%)

893 009 (100%)

3900

5000

4700

2222

2268

4480

9100

3500

5100

23.2

12.6

15.4

5.6

20.8

13.4

2.35

0.70

1.08

7.2

3.3

3.9

From Herna´ndez-Mora N, Llamas MR, and Martı´nez-Cortina L (2001) Misconceptions in aquifer over-exploitation. Implications for water policy in southern Europe. In: Dosi C (ed.) Agricultural Use of Groundwater. Towards Integration between Agricultural Policy and Water Resources Management, pp. 107–125. Dordrecht: Kluwer.

106

Groundwater Management

volume of water used. This difference can be attributed to several causes: the greater control and supply guarantee that groundwater provides, which in turn allows farmers to introduce more efficient irrigation techniques and more profitable crops. Greater dynamism has normally characterized farmers who sought out their own sources of water and have had to bear the full costs of drilling, pumping, and distribution. Higher financial costs to farmers motivate them to look for more profitable crops that will allow them to maximize their return on investments. Surface and groundwater distinctions, therefore, should be taken into account in order to achieve an efficient allocation of water resources.

The hidden or open subsidies that have traditionally been a part of large hydraulic projects for surface-water irrigation are probably the main cause of the pervasive neglect of groundwater problems among water managers and decision makers. Surface water for irrigation is usually given at low or heavily subsidized costs to farmers and this often results in the wasteful use of a valuable resource. It is usual that water supply companies, farmer unions, etc., lobby the state for the construction of surface-water infrastructures that are primarily paid for through general revenues, instead of advocating a responsible use of groundwater resources. At times, this may lead to social conflicts – such as in the case of the Tagus–Segura transfer or the overruled Ebro transfer, both in Spain – between water-importing and waterexporting basins. Progressive application of the user pays or full-cost-recovery principle of the European Union (EU) Water Framework Directive (WFD) would probably make most of the large hydraulic projects economically unsound. As a result, a more comprehensive look at water planning and management would be necessary and, in turn, adequate attention to groundwater planning, control, and management would probably follow.

came earlier. In this slowness and imperceptibility lies a significant part of the difficulties of managing groundwater resources nowadays. There are a number of political and policy choices to offset the pervasiveness of open access, such as regulatory interventions, market instruments, or information and technical choices. Most often, the effectiveness of policy measures might mean the right mix of policy instruments from an existing portfolio where experience has already been gathered from groundwater management. For policymakers, there are a range of policy instruments for management (Table 2), with different strengths and weaknesses for groundwater management according to a range of criteria such as effectiveness, economic efficiency, technical efficiency, administrative feasibility, equity, and social or political acceptability (Hellegers and Van Ierland, 2003). An example of regulatory instruments for environmental protection is the case of the Edwards aquifer, in Texas (USA). The Edwards aquifer is a karstic aquifer, which means that the effects of pumping are quickly transmitted to large areas creating in effect an open-access resource since it operates under the rule of capture, that is, where under Texas groundwater law, landowners can pump without limit (Howe, 2002). Pumping for agricultural and urban purposes has represented 45–50% of the total discharge of the Edwards aquifer for the period from 1934 to 1999. Springs supported several species of fish and amphibians giving the Federal Fish and Wildlife Service the right to intervene to protect these species if the state failed to act. A lawsuit by the Sierra Club, an environmental nongovernmental organization (NGO), under the Endangered Species Act ended in a federal ruling, which meant that the Texas legislature set up the Edwards Aquifer authority in 1996 with extensive powers, including the issuing of permits to regulate groundwater withdrawals. For example, it required pumping limits to protect endangered species. A flow of 150 cubic feet per second must be maintained at the most sensitive springs.

1.07.3.4 Groundwater: From Open Access to Common Pool?

1.07.3.5 Optimal Groundwater Pricing

Economics deals with scarcity and the allocation of scarce goods. Groundwater resources until very recently, it could be argued, were not economic goods, because anyone interested in pumping water could do it ab libitum. Capital, energy, or time constraints were the only barriers for all potential users of groundwater resources. In this case, however, the economic problem, if any, was related to access to inputs, finance, or labor. Even in cases where the inputs required to pump groundwater are unlimited, deciding on how much water should be pumped may not be an economic problem. One of the most pressing institutional questions centers on inappropriate legal and administrative structures, which in effect means that groundwater rather than being an openpool resource becomes an open-access resource which can lead to excessive contemporary and inter-temporal externalities (Howe, 2002) (see Section 1.07.6). As in many other natural resources, the transition from an open access and unlimited resource to an exhaustible and rival one is gradual. It happens by the marginal adhesion of small groundwater users attracted by the benefits that are obtained by those who

An optimal groundwater price is a theoretical economic concept, and the solution to a dynamic and stochastic problem. It is dynamic because the optimal price varies with time, and is meant to ensure that pumping rates at any given moment maximize the discounted flow of benefits for an infinite time horizon. As Neher (1990) shows, there is an optimal tax that makes all users of a common pool internalize the (marginal) consequences of their pumping rate so that a socially optimal management use is achieved. Theoretical optimal rates can be obtained to account for irreversible effects – caused by excessive drawdown (Rubio and Fisher, 1997), for situations in which backstop technologies can become operative if pumping becomes too expensive (Gemma and Tsur, 2007), or when surface water and groundwater are conjunctively used (Pongkijvorasin and Roumasset, 2007). Pongkijvorasin and Roumasset’s (2007) main contribution is to combine two problems that have been addressed separately in the literature. One looks at the problem of managing large surface systems with conveyance losses (Roumasset and Chakravorty, cited by Pongkijvorasin and Roumasset, 2007),

1.07.3.3 Perverse Subsidies in Water Policy

Groundwater Management Table 2

107

Portfolio of policy instruments available to manage groundwater quantity and/or quality

Portfolio of groundwater-management tools

Example

Standards and regulations

High Plains Conservancy district No 1 outside Lubbock e.g., harim rule in Islamic law (Middle East) Indonesia EU Water Framework Directive For example, sprinkling bans to reduce low value agricultural groundwater abstraction in Holland; Yemen (ban on agricultural wells); and Egypt (ban on new wells) Netherlands, France, and parts of Germany have introduced a groundwater-abstraction tax Edwards aquifer Texas, (USA)a, Italy (Perugia province) Parts of Spain China

Limit pump capacity Well spacing Groundwater permits Safe-yield criteria Ban in most critical groundwater areas

Pumping tax, for example, reflect to the individual pumper the negative externalities Caps on abstraction Metering groundwater use Metering electricity use as groundwater use proxy Market-based instruments (economic incentives)

Administrative planning and education measures

Groundwater resource fee (money raised earmarked for aquifer management), cost of right (waterresource levy) Volumetric pricing of groundwater and/or electricity supply, sliding-scale pricing strategy Electricity subsidies Groundwater farms (i.e., purchase of land for associated groundwater rights) Issuing tradable permits (leasehold) with or without cap on total abstractions Water markets (freehold), for example, from rural to urban, from small to large farmers, from low-value to high-value crops, from irrigation to environmental flows Compensation program for third-party effects, for example, tax to area of origin Full cost recovery Groundwater banks Groundwater zoning, for example, aquifer-vulnerability maps to contamination Land-use planning

Education, for example, agricultural extension service and training Political education at senior level on the value of groundwater Public awareness campaigns, civil education, for example, value of groundwater Name-and-shame list Joint regulation and monitoring Participatory groundwater monitoring Institutional (incl. information and voluntary measures)

Incentives for private entrepreneurs for example, as franchisees for billing and collecting electricity dues Self-imposed correlative rights for example, % owned above the aquifer Voluntary agreements (self-regulation), devolved groundwater management Groundwater protection codes

Indonesia for industrial water users Holland has a levy under the 1983 Groundwater Act India and China For example, Mexico 1/3 of electricity costs Arizona Spain (Parque Nacional Tablas de Daimiel) Spain, for example, for environmental flows Spain, USA, Australia, Chile Mexico (including energy pricing coupled with water rights) Contemplated in Texas EU Water Framework Directive Trialled in Texasb United Kingdom, Holland, USA, and Canada Pakistan through informal committees For example, removal of invasive species and replacement with native vegetation in South Africa to help recharge shallow aquifers, India and South Africa studies on impact of forestry (native and plantations) on evaporation and recharge High Plains District Eastern La Mancha aquifer (Spain) e-Water India UNESCO For example, China water pollution map Eastern Mancha aquifer (Spain) India and China China

For example, Water Boards in Holland (interest, payment, authority) involved in groundwater-level management UK (Continued )

108 Table 2

Groundwater Management Continued

Portfolio of groundwater-management tools

Example

Technological

e.g. Spain and Mexico

Use of Geographic Information System (GIS) for monitoring Wastewater-reuse schemes Artificial aquifer recharge enhancement and/or storage Improved irrigation technology (for example, irrigation scheduling, micro-irrigation, land leveling) Changes in crop type, for example, higher-value crops if possible (i.e., no impact on livelihoods); ban on high water-consumptive crops

Central Valley (California), Rainwater harvesting (India); well recharge movement Israel, Spain, Australia, USA, and Mexico Areas in Spain Saudi Arabia

a

A system of marketable groundwater permits to be issued to all pumpers as a proportion to historical use subject to a total pumping cap. The cap was 450 mm3 yr1 which was accepted by all users, urban, agricultural and environmental. In 1998, a groundwater trust was set up to facilitate the trading of permits. In the period 1997–2001 there have been 403 trades with an average size of 235 000 m3. b For example, under the Irrigation Suspension program, water rights were purchased from 40 farmers representing 4000 ha to supply water to San Antonio. Irrigators were paid $ 98 to $ 1850 ha1 to stop irrigating, with water savings of 20 000 acre feet at a cost of $ 2–3 million paid for by cities, counties, and water companies. For additional ideas, please see Chevalking S, Knoop L, and Van Steenbergen F (2008) Ideas for Groundwater Management. Wageningen, The Netherlands: MetaMeta and IUCN.

and the other looks at the conjunctive use of surface and groundwater. The paper’s main accomplishment is to show that, with surface pricing as the only managing instrument, farmers switch from surface to groundwater and vice versa, when groundwater-scarcity rent diminishes, with the irrigation boundary contracting or growing depending on whether groundwater-scarcity rents increase or decrease. The institutional implications of these results are worrying, unless one can think of a much more restricting context in which pumping rights could be effectively enforced, and irrigation districts had nonmobile boundaries. Yet, Rijsberman (2004) quotes work by the International Water Management Institute (IWMI) which shows that groundwater use is largely beyond the possibilities of most water institutions in the developing world for a proper monitoring. Llamas et al. (2008) and Shah et al. (2007) present a similar general situation. The large literature on theoretical groundwater pricing yields somewhat impractical policy prescriptions (e.g., Molle and Berkoff (2006)) because of the number of externalities involved in many cases of intensively exploited aquifers (see Llamas and Custodio, 2003). Brown (2000) offers very convincing arguments to explain why optimal pricing has rarely been used to allocate renewable resources such as fisheries or groundwater.

1.07.3.6 Departures from Optimality: Second-Best Solutions In economics, second-best solutions are those that cannot achieve the results of the first-best optimal solution but are more applicable in practical terms. First-best solutions may be too information demanding or based on a perfect fine-tuning to the specific circumstances. In this section, we review some of the most commonly used second-best (or even third-best) solutions to properly manage groundwater resources. These policy approaches are always implemented to solve one or the other type of the sources of economic inefficiencies. The following are the policies from the less to the more sophisticated: user rights, pumping rights, pumping quotas, water tariffs, and water markets.

Issuing user rights is the simplest way to grant access to an aquifer. The authority may or may not control the pumping capacity and the type of equipment. In intensively exploited aquifers, granting user rights may not be sufficient to deter pumping and, in many situations, the outcome may not ensure that the exploitation is adequate. However, issuing user rights is a prerequisite to consider in any of the policy menus mentioned. In principle, user rights are increasingly required to obtain legal access to tap an aquifer in virtually all contexts in which water is scarce. The next instrument in the list is granting pumping rights, whereby users are entitled to pump fixed amounts. In principle, the ownership of pumping rights can be associated with private property if those rights cannot be encumbered by new users or forfeited by a public agency. Setting up pumping quotas enables a more flexible instrument, because these can be modulated to the aquifer’s recharge. In many cases, pumping rights are combined with water tariffs, but an interesting option would be to modulate tariffs to mimic the optimal price. Perhaps, the most advanced initiatives in the area of groundwater pricing can be found in some European countries, which have seized the opportunity of the EU WFD to implement the principle of full-cost-recovery pricing. Water markets can be found in multiple formats. In India, for instance, water markets occur as informal exchanges among small farmers (Saleth, 1996). In Australia, water markets are formally established and can facilitate groundwater trading or rights’ exchanges. In Spain, the basin authorities have offered farmers permanent buy-outs of water rights, or annual pumping quotas, following public offerings (see Box 1 for examples of water trading). A prerequisite of all these policies is to have control of (at least) the number of users, and ideally of the pumping yields of all users tapping an aquifer. However, even if control and surveillance can be guaranteed, it does not imply that the aquifer will be sustainably managed. Unfettered water rights can be as damaging to aquifer management as an aquifer that is not controlled. Groundwater management is usually based on rights or licensing, but enforcing compliance with these

Groundwater Management

109

Box 1 Water markets and voluntary exchange mechanisms to solve groundwater overdraft. From Garrido A and Llamas MR (2009) Water management in Spain: An example of changing paradigms. In: Dinar A and Albiac J (eds.) Policy and Strategic Behavior in Water Resource Management, pp. 125–146. London: Earthscan. Water banks or exchange centers, as these are called in Spain, received legal recognition in the 1999 Spanish Water Law reform. Not strictly a bank or agency, these centers are hosted, run, and located in the basin agencies themselves. It is widely believed that these centers are a much more efficient medium for promoting water exchanges, for a number of reasons, such as transparency, control, avoidance of third-party effects, and market activity and scope. Yet, the experience so far has been limited to the Ju´car, Segura, and Guadiana basins, since these water centers have been primarily used to tackle severe problems of groundwater intensive use. Since the enactment of the 1985 Water Law, which included special provisions to tackle the problem of overexploited aquifers, there have been at least four major initiatives to manage groundwater resources. In short, these were (1) the declaration of overexploited aquifers and the mandate to enforce regulations and implement management plans; (2) an EU agri-environmental program, only applicable to Aquifer 23 in the Guadiana Basin, with subsidies to farmers who curtail their water consumption; (3) the use of inter-basin transfers, both in the case of the southeast coastal areas and in the Upper Guadiana; and lastly, (4) the Special Upper Guadiana Plan (PEAG, Spanish acronym), and the creation of exchange centers in the Segura, Ju´car, and Guadiana basins. Clearly, the first option failed; the second one succeeded, but the financial cost was very high, and the third option failed because the second one was not sustainable. In the end, the PEAG was approved in 2007 with a total budget for 20 years of h5.5 billion (equivalent to the proposed Ebro transfer) and part of its subprograms are now operational, although under PEAG the basin would reduce to a meager 200 million m3. Underlying these initiatives, but undermining them too, was the recognition that tens of thousands of users in virtually all basins had no legal rights or concessions to the groundwater resources they had been tapping for years. Any effort to reduce total extractions in the over-drafted hydrogeological units had to be accompanied by the closure of the alegal or illegal uses. In 2005, it was clear to all managers, analysts, and users that something new had to be given a chance. The option to use buyouts of water rights, permanent or temporary, gave a rationale to the establishment of exchanges centers (centros de intercambio in Spanish). We review the different approaches taken in the Jucar and Guadiana. In the Jucar basin, the offer of public purchase (Oferta pu´blica de adquisicio´n de derechos, OPA) was targeted at farmers tapping groundwater resources near the Jucar’s headwaters. Its objective was to increase the water tables in Castille-La Mancha to ensure that the Ju´car flows to the Valencia region increase from historical lows. Farmers were given the option to lease out their rights for 1 year in return for a compensation ranging from 0.13 to 0.19 cents m3, the variation depending on the distance of the farmer’s location to associated wetlands or to the river alluvial plain. The OPA was launched in two rounds, the first with disappointing results in terms of farmers’ response, while the second had more success. The purchased waters served the unique purpose of increasing the flows, enabling more use downstream in Valencia. However, the OPA did not have any specific beneficiaries downstream, other than to increase flows. The OPAs of the Guadiana followed a completely different approach and were meant to address serious problems of overdraft in the Upper Guadiana. As stated before, the OPA formed part of the more ambitious program of aquifer recovery, the PEAG. The Guadiana’s OPA made offers to purchase permanent water rights to groundwater, paying farmers h6000–10 000 ha1 of irrigated land. Note that, since these farmers had seen their allotments reduced in preceding years, what the Guadiana basin was truly purchasing from the farmers was about 1500–2500 m3 ha1, effectively h2–4 m3. The Guadiana basin agency has the objective of purchasing the water rights of 50 000 ha1 of irrigated land, and is budgeting h500 million for the whole plan. A marked difference from the Jucar’s OPA is that the Guadiana exchange center will transfer part of these rights to other farmers (growing vegetables) and to the autonomous community of Castille-La Mancha. The Guadiana basin will grant less rights than it has purchased, allocating the difference to wetlands and to increasing the piezometric levels of the aquifers.

rights remains a major challenge for adequate groundwater management. For example, there is wide experience with water markets in the USA; according to Howe (2002), what water markets do best is to generate information on values for more rational, better-informed water allocation – for example, the sale or lease of water rights to off-site buyers such as cities. In Arizona, the government had acquired 200 000 ha of land by 1990 for the associated groundwater rights. These water farms or water ranches average about 12150 ha and are valued at US$15 million, and expected to supply 15 000 acre feet of groundwater per year for 100 years. In Colorado, Front Range water rights in 1990 sold for US$ 1000–4000 per acre foot (Colby, 1990 in Wagner, 2005). In systems where groundwater rights have been incorporated into the general water-rights system, groundwater rights can be bought and sold, transferred to other locations, or transformed into surface rights (e.g., tributary groundwater– groundwater that is intimately connected with surface water). In Texas, groundwater was purchased to secure water for urban centers such as Houston, San Antonio, and El Paso. For example, an old mining right from the Alcoa-Sandown mine was sold for US$ 688 per acre foot annually. The city of Amarillo also bought groundwater for US$ 679 ha1 for groundwater rights from 28 350 ha of lands, when the land itself sells for US$ 494 (Gillinland, 2004 in Wagner, 2005). The El Paso Water Utility purchased more than 19 000 ha of ranchland to pump 15 000 acre feet by 2010 (Texas Center for Policy Studies

in Wagner, 2005). Lucrative groundwater leases with at least four private water ranches on over 200 000 ha have been formed to sell or lease a significant amount of water to off-site users, principally cities. Another example is the Irrigation Suspension Program, where water rights were purchased from 40 farmers representing 4000 ha to supply water to San Antonio. Irrigators were paid US$ 98–1850 ha1 not to irrigate, with water savings of 20 000 acre feet at a cost of US$ 2–3 million paid for by cities, counties, and water companies (Wagner, 2005). Nevertheless, there is a range of issues that has to be considered in water marketing. Over-entitlement occurs when the sum of all legally defined rights are greater than 100% of the system’s potential. Overuse occurs when the quantity of water abstracted is greater than the system’s potential to supply. Sleeper or dozer licenses are legal entitlements that are not used at present but are legal (e.g., UK sleeper licenses). One major reason why some people might distrust markets is the fear that markets might fail to deal with issues of distributional equity, fairness, public concern, and community interest.

1.07.3.7 Internalizing the Value of Environmental Services Provided by Groundwater Water can be framed on current discussions on ecosystem functions from environmental services. Ecosystem functions refer to system properties and processes. Services represent the

110

Groundwater Management

benefits that society derives, directly or indirectly, from ecosystem functions. A summary of the authors’ evaluation of annual flows of water-related ecosystems at the world scale is presented. Humans avail many types of services from water-related ecosystems in addition to water supply. Note, for example, that 1 ha of wetlands can generate almost US$ 4200 yr1 in wastetreatment services (Costanza and de Groot, 1997). While this evaluation was certainly preliminary at the time it was produced, it conveys a clear idea about the costs and damages that water scarcity can provoke. The mere recognition of many of the identified services valuable for society has huge implications for drought-policy design and implementation. Chief among this is the fact that many of these services have public-nature features, which means that they are nonrival and nonexclusive goods. As scientists have learned to identify and value them, water policy must take into account and ensure that decisions are a compromise between both productive and nonproductive services (National Research Council, 2004). The Millennium Ecosystem Assessment undertaken in 2005 classified the goods and services provided by natural resources as provisioning, regulating, supporting, and cultural services. In this context, Bergkamp and Cross (2007) discuss the high-value ecosystems supported by groundwater. The total economic value (TEV) of groundwater resources is the sum of groundwater resources: use and non-use value, based on direct values, indirect values, option, and existence values. A number of methods have been developed by environmental and ecological economists to value these goods and services using a range of environmental techniques such as contingent valuation method (willingness to pay and willingness to accept), choice modeling, production-function approaches, surrogate markets, costs-based approaches, and stated preference. These methodologies are addressed in one of the chapters of this volume. This is a way of internalizing the value of groundwater services and also the economic value of groundwater externalities. This can help decision making to evaluate the costs of action and costs of inaction. It gives a clearer signal of whether it is better or not to opt for preventative policies when the full remediation costs of polluted groundwater are accounted for, that is, measure the potential benefits (or damages) of a range of effects. In 1996, the US Department of Defense invested in 75 pump-and-treat systems to remediate contaminated sites estimated at US$ 500 000 per site. Equally, a calculation on the externality costs of overabstraction in the Queretaro aquifer (Mexico) between 1970 and 1996 included: the increased pumping costs due to drop in piezometric levels estimated at US$ 6 million; the loss of water quality for public water supply at US$ 26 million; and damage to urban infrastructure due to land subsidence at US$ 26 million (6 million as private cost and 20 million as costs to taxpayers). Moreover, it is relevant to consider impacts on other policy sectors like the costs for public health due to the mobilization of naturally occurring arsenic through deep wells in Bangladesh, which is estimated to affect 30–35 million people (WHO, 2001). Opportunity costs are the foregone benefits that could have been generated if resource was allocated to the next-best use if water is not allocated to its highest use value and, in fact,

opportunity costs may be greater than the value generated by next-best use. In this case, the economy is subjected to inefficient and suboptimal groundwater allocation, although this may be justified in equity or sociopolitical terms. Therefore, it is equally important to properly include the positive services provided by groundwater. In terms of equity, Acharya and Barbier (2000) analyze losses to farmers from reduced groundwater recharge. On average, farmers could lose US$ 413 ha1 if the groundwater benefits are not accounted for, that is, evaluating the systemwide benefits associated with groundwater use. The increased recognition of the benefits associated with groundwater use is reflected, for example, in the growing aquifer-recharge movement. In Texas, aquifer recharge through open-space protection and cooperative groundwater allocation is a new paradigm in water management (Wagner, 2005), based on valuing the products and services that a functioning system provides. In India, meanwhile, the socalled decentralized recharge movement was a spontaneous response to groundwater depletion to help water tables rebound to predevelopment levels at the end of the monsoon season in pockets of intensive use. This is an example of contrasts between popular hydrogeology and formal hydrogeology; for example, scientists argue that hard-rock areas have too little storage and advocate recharge; meanwhile, the prolific growth of recharge structures is based on the value people attach to a check-dam even if their wells provide only 1000 m3 which – although small – is crucial for life-saving irrigation in times of delayed rain. Thus, rainwater harvesting can be used to recharge groundwater via recharge ponds, based on the main sources of recharge: rain, and infiltration from riverbeds and from the floodplain. Equally, in Australia and the USA, sand dams are used to make artificial shallow aquifers in streambeds to reduce evaporation of stored waters. The growth in the number of sand dams could substantially increase to compensate for potential climate variability and change making use of groundwaters’ buffering service (see Section 1.07.6.2.1). Failing to account for the opportunity costs of groundwater and surface/groundwater linkage often result in suboptimal outcomes. Groundwater recharge is one of the most important environmental functions brought about by wetlands. For example, households in rural Nigeria rely on groundwater for drinking and cooking, and in the arid north, particularly during drought, it often has the added advantage of its higher quality (Table 3).

1.07.4 Regulatory Frameworks for Groundwater Multilevel Governance One of the most obvious examples of the Cinderella status of groundwater in global water resources is reflected in the evolution of regulatory frameworks. Due to its silent (and relatively recent) rapid growth, groundwater traditionally had little or no regulation (i.e., as exemplified in the rule of capture in Texas), part of mining law, or of private-property rights (tied to land). This lack of prominence and the lack of concern over its management and state of preservation have historically been reflected in the law. Therefore, groundwater

Groundwater Management Table 3

111

Valuing groundwater goods and services

Total economic value of groundwater

Environmental goods and services

Definition

Example

Direct value

Provisioning service

Drinking water supply

Two billion people rely on groundwater directly for drinking water More than 50% of cities with population of more than 10 million rely on or make significant use of groundwater Forty percent of world’s food relies heavily on groundwater Land irrigated from aquifers has increased 113 times between 1990 and 1990. Aquifer supplies more than half the world’s irrigated land For example, manufacturing processes and geothermal and cooling systems Groundwater stores and releases water, sustains river flows, springs, and wetlands Through microbial degradation of organic compounds and potential human pathogens, microbiological and some chemical contaminants removed, retarded, or fragmented For example, by absorbing run off Primary buffer against climate variability and spatial variability of droughts Potential innovation as future use for anthropogenic carbon sequestration in the ground For example, groundwater recharge and discharge

For example, storage and retention Agriculture

Industrial Indirect value

Regulating services

Water regulation; Water purification and waste treatment Erosion and flood control Climate regulation

Option value and existence value

Supporting services

Cultural services

Necessary for the production of all other ecosystems services Non-material benefits people obtain from ecosystem services

Spiritual enrichment, cognitive development, religious value, and symbolism

From Bergkamp G and Cross K (2007) Groundwater and ecosystem services: Options for their sustainable use. In: Ragone S, de la Hera A, Hernandez-Mora N, Bergkamp G, and McKay J (eds.) Global Importance of Groundwater in the 21st Century: The International Symposium in Groundwater Sustainability, pp. 233–246. Alicante, Spain, 24–27 January 2006. Westerville, OH: National Groundwater Association Press.

was legally structured as one more facet of the right of ownership for a specific area of land. Starting out from that premise, the various laws gave shape to the depth of that right and regulated how it would fit in with the rights held by owners of adjacent pieces of land. All of the above emerged from an eminently private perspective imbued with the wealth of duties and rights conferred by ownership rights. Groundwater doctrine in Texas is based on the rule of capture, an English common-law approach based on absolute ownership, where landowners can pump without limit, as long as water is put to beneficial use, which allows unrestricted pumping by competing groundwater users as long as it is not wasted, and whereby property rights are not defined (Wagner, 2005). This is an example of one of the few natural resources in the USA not regulated by a central agency. The activities of 88 water districts in Texas are unusual because they have been based exclusively on a voluntary approach controlling wastage of water, recharge, enhancement, and water-conservation education rather than controlling abstractions. For nearly 100 years, the rule of capture has survived attempts to regulate groundwater use. Although government oversight and technical assistance are vital, a carefully crafted free-market system based on private rights to a communal resource becomes increasingly important. A bottom-up process created the State Water Plan of 2002, which incorporated regional water plans’

gradual increase in the security of mining rights from open access to other systems.

1.07.4.1 Diversity in Groundwater Regulatory Regimes Regimes with a civil-law tradition, inherited from the principles of Roman law, are clear exponents of this and not very far from them are those grouped under the parameters of common law. Under common-law regimes, the landowners were the right holders of groundwater, flowing under their properties, which could be harnessed (Embid Irujo, 2002). From the legal viewpoint, legislation on aquifers presents two main issues of concern: first, ownership and second, transferability or flexibility with ownership rights. The first one relates to whether groundwater resources should be public or private property. Ownership of groundwater resources shows a high level of diversity from completely private (e.g., in Texas USA), groundwater from the Ogallala aquifer is mainly private (Peck, 2007), to state-owned resources, such as in the case of Mexico, to plural legal systems, as in parts of Africa, and community based, such as in Bolivia. Legal provisions may confer ownership of groundwater directly on public authorities, as part of the public domain (Morocco, Italy, Spain, Zimbabwe, Israel, some US States, Jamaica, Mexico, Argentina, Australia, the Lebanon, Jordan, and Syria), give the authorities preferential rights in groundwater (South Africa and Uganda),

112

Groundwater Management

or make groundwater the common heritage of the nation and place conditions on its use (France). Particularly, in Muslim countries, the applicable legal regime is intertwined with principles related to religious beliefs which result in the classification of water as public in itself with cases of private appropriation being seen as exceptional (Caponera and Nanni, 2007). Whichever route is taken, the result is very similar: governmental authorities give themselves powers over groundwater with the aim and effect to implement public policies that they lay down. The progressive importance of groundwater in the definition of water resources in various countries has engendered the implementation of legal reforms that protect public intervention (Hodgson, 2006). When groundwater is public, the concept that is generally used by the rule makers is the permit, license, authorization, concession, or a similar instrument. This is the case of Israel, a number of states of the USA, Mexico, and many other countries. In other places, such as California, Chile, India, or Texas, groundwater is under private ownership. In all of these cases, a private party, individual, or community is granted the right to use a certain amount of groundwater. This right is subject to certain conditions relating to time or use. The right may or may not be granted according to whether it is consistent with the status of the resource and with the parameters of the planning regulations on water resources that must govern its contents. Those same premises must be used to determine the period for which the right is granted and the amount that may be extracted. It is also necessary for the law to outline the cases in which this right can be altered, restricted, or even eliminated as a result of damage to the aquifer, possible droughts, watersupply needs to the population, or similar events, or principles such as reasonable and beneficial use. A legal regime for groundwater should ideally consider some of these aspects:

• • • • • •

•

approval of compulsory legal norms for all groundwater users; determination of the legal rules and principles to be applied in the management of groundwater, including its relationship with surface water; legal parameters to define groundwater as a resource; institutional regime applicable to groundwater; specification and regime for uses of groundwater available for all citizens without being subject to specific control; determination of the rights on water (transitional rules in the event of amendments to the law; concession rules; registration processes; contents of the right – volume of water, term, conditions, and termination – transfer of rights; and dispute-resolution mechanisms); and rules on the protection of groundwater and measures to adopt if needed (control of the pollution of groundwater – rules on discharge and authorization; use restrictions; and prohibition on use).

1.07.4.1.1 The controversy over private, public, or community groundwater rights Evidence indicates that ownership per se (public, private, or common) does not guarantee or pre-empt sound management. The importance does not lie in what name is given to the legal title but rather in the contents given to that title. The

emphasis is placed on the aim to be achieved. Preferences on ownership are societal choices, which are subject to change and flux; the underpinning question is not ownership but whether management is according to some predefined a priori objectives, which in any case are themselves subject to constant negotiation and renegotiation as part of a normal political process. Some authors consider that the legal declaration of groundwater as a public domain is a conditio sine qua non to perform a sustainable or acceptable groundwater management. This assumption is far from evident. For many decades, groundwater has been a public domain in a good number of countries. Nevertheless, sustainable groundwater management continues to be a significant challenge in many of those countries. Highly centralized management of groundwater resources is not automatically the solution to promote solidarity in groundwater use as a common good because a key element is the internalization, by often thousands or hundreds of individual users, on the need for collective action. Groundwater management sometimes can successfully be devolved to stakeholders of the aquifer, in self-governance arrangements under the supervision of the corresponding water authority. Stakeholders’ participation has greater chances of success if it emerges bottom -up and is supported topdown. The practical application of a hybrid (public and private) system is exemplified by a few countries in the world where a range of systems coexist: one is the US, where states like Colorado, Arizona, Texas, and California exhibit a range of ownership rights to groundwater; and the other is Spain with a particularly interesting example of a mixed system. Wells drilled after 1 January 1986 require governmental permission, while those operational before 1986 remain private. Private groundwater may remain so for 50 years (provided the well owners reach an agreement with the government in exchange for administrative protection) or perpetually (if the owner wishes to preserve his/her rights under the 1879 Water Act). In any case, the Spanish situation is far more complex due to the lack of a reliable registry of groundwater rights. While the government is currently carrying out a series of remedial initiatives, these ignore a significant share of existing wells, and the registry or inventory is therefore incomplete. A key ingredient is the need for a strong political willingness to apply the laws. It seems clear that a reliable inventory of groundwater rights is desirable in order to ensure adequate management. The second issue refers to the way groundwater rights should be inventoried and to whether the possibility to trade with them should be allowed. This second aspect, usually equated with water markets and banks (discussed in the previous section) is perhaps subordinated to the first in terms of importance, even if significant informal markets already exist in some places (Mukherji, 2006). It cannot be ignored, however, that in other territories, the inseparable link between water, land, and private property has been maintained. The states of California and Texas or countries such as Chile and India are examples of this. These cases have maintained the private ownership of water and in many cases applied the doctrine of prior appropriation, although this is subject, depending on the territory and circumstances, to specific measures of administrative or court intervention, linked, for example, to the principle of reasonable use. It is also possible

Groundwater Management

113

Box 2 Groundwater-use rights in China. By Simon Howarth, based in Gansu, under the UK Department for International Development (DFID)-funded Water Resources Demand Management Assistance Project. The Shiyang River Basin in Gansu province in Northwestern China is an area of severe water shortage, where groundwater abstraction greatly exceeds recharge. A package of measures, including greater delivery of surface water and restrictions on the use of surface water, is being introduced. A key element in this process is the allocation of water rights to individual households. These are based on land allocations and household size (to take account of domestic use and livestock as well as agricultural demands). Both land and water are owned by the state in China, but user rights have been granted to individuals. Land rights have been granted since the 1980s and further reforms are in progress. Water rights have been formalized since 2007, when individual household water-rights certificates were issued (via village committees or WUAs). The rights are calculated to be sufficient for the locally recommended cropping pattern, and are being reduced each year (e.g., from 7200 to 6615 to 6435 m3 ha1 in one typical irrigation district) as recommendations are revised and farmer skills in water savings are developed. Awareness and capacity-building programs are being run simultaneously so that farmers can protect their livelihoods while coping with less water. Wells were developed by villages and are owned by them, but the amount of water that can be pumped from them is regulated by the state through a system of permits. Well permits are now being reissued to suit the new water rights, and electronic controls (IC cards) are being installed at all wells. These will limit the amount that can be pumped from the well to the annual total permitted for that well. These cards are held by the well operator who is responsible for ensuring that each household receives water in accordance with their individual rights. These systems are new, rely on both sophisticated technology and complex administrative systems, and have been introduced rapidly (in over 10 000 wells in 1 year). Not surprisingly, some teething difficulties have been encountered, but there is a very strong political will to solve these problems. Allocation of water rights is intended to enable trading of rights, although this does not yet happen on a formal basis in this area of China. It will be subject to certain restrictions – for example, the right will be salable at a maximum of three times the water-resources fee, which is small when compared to both the value of water and the cost of pumping. There is a large and growing requirement for water for industrial development, which is a more valuable use of water, but there is a competing requirement for food security – these competing demands cannot be managed purely by market measures but will require government control as well.

to outline mechanisms for exchanging water rights in the context of what has come to be called the water market (Chile, South Africa, Mexico, Spain, the UK, Australia, and the US). The use of water-market institutions for groundwater has its detractors who point to the risks involved. In our view, however, those risks do not necessarily warrant ruling out the option completely. Bringing flexibility to the allocation of resources and allowing them to be exchanged are not in themselves misguided concepts. The usefulness of water markets is usually associated with cases of multiple supply and demand sources, with transparent exchange mechanisms and the appropriate transportation networks that make them feasible (Melgarejo and Molina, 2005). The key lies in controlling their use and making them subject parameters of sustainability and protection (Box 2).

1.07.4.2 Implementation and Enforcement of Groundwater Legislation In the context of groundwater management, rules on the ground are crucial, for example, those related to time, well location and spacing, technology, or groundwater-abstraction quotas. In addition, another factor is the interaction between formal groundwater law and its operation on the ground. More attention is being paid increasingly to the implementability of regulation, since the problem with most groundwater legislation lies in its implementation and enforceability. For example, South Africa established an implementation team with the task of anticipating what the water domestic bill would require, with close interaction between the drafting and implementation teams to identify possible implementation problems before enactment. In the case of groundwater, it would be useful to develop implementation tools such as guidelines, procedures, information systems, user manuals, and organizational arrangements. Another option is to opt for framework laws, which specify general guidelines but leave implementation to detailed regulations as used in Uruguay.

Implementation requires time, and needs political support at the highest level since strong economic and political interests are usually affected by allocating or reallocating groundwater resources. As Gardun˜o (2003) states that implementable legislation is one that the government is able to administer and enforce and water users have the ability to comply with. Experience shows the education of stakeholders and widespread presence of groundwater-user associations is crucial for an adequate participatory bottom-up management approach. One of the main problems in groundwater governance is lack of enforcement in some cases of relatively sophisticated laws, such as in Spain. As stated earlier, institutions encompass not only rules in norm but also rules in use or institutional arrangements. In effect, the implementation and enforcement of groundwater laws have to be legitimized and supported by society (social norms). The involvement of groundwater users in groundwater-management regimes is a necessary (although not sufficient) condition for successful enforcement regimes. In traditional societies, social networks were denser and therefore transaction costs lower, whereas modern societies require complex institutional structures that constrain and regulate interactions among groundwater users. Groundwater users possess detailed local knowledge on water use, and these communities can apply for sanctions unavailable through formal institutions. For example, name and shame can resolve conflicts at the local level in a manner customized to local circumstances, which reduces transaction costs, which in turn are critical for economic performance. In fact, in a study undertaken in Spain, groundwater users had a clear perception of the kind of behavior that should be penalized and how sometimes sanctions devised by farmers do not mirror sanctions designed by higher-level authorities (Lopez-Gunn, 2003). This can be rooted in different perceptions of equity and fairness. For example, farmers in an aquifer in Spain would prefer to be sanctioned in the following irrigation season with water as a penalty, in lieu for the same amount of water that farmers abstract over their quota in the previous

114

Groundwater Management

irrigation season, instead of the current (formal) sanctions of a monetary penalty. Thus, groundwater users can reduce the transaction costs of enforcement and devise adequate sanctions. However, it should not be forgotten that authorities in most cases ultimately hold this legal responsibility to protect public goods. Higher-level authorities will often have to be imaginative with monitoring and sanctioning regimes. Many Asian administrations lack the capacity to perform complex tasks, for example, urban groundwater management with joint monitoring of industrial groundwater abstraction and wastewater discharges. A feasible alternative tried in Indonesia was to select a random sample and thoroughly monitor these users. In cases of noncompliance, the weight of the law should be applied and widely publicized in the media; as capacity grows, the sample could be enlarged. Limited administrative capacity is a key constraint to groundwater management and revenue-raising fees can be re-invested toward capacity programs.

1.07.4.3 Multilevel Regulatory Frameworks The section above described and discussed briefly some of the main challenges for national groundwater law. It is increasingly recognized, however, that national groundwater law is only part of the regulatory framework. Other levels (both conceptual and in terms of scale) have to be taken into account: first, international conventions currently being negotiated, for example, for transboundary aquifers or the rise in the human right to water; second, a pragmatic approach on the advantages and limitations of legislation and litigation; and third, a consideration of the legal principles that have to underpin legal norms and an evolution in our understanding of how laws will be drafted in the twenty-first century. First, in relation to international conventions, there are two conventions that are applicable to groundwater: the first relates to transboundary aquifers and the second refers to the International Convention on Human Rights (1948) and its new impetus to recognize a human right to water (or HRW). Until only a few years ago, international law did not pay too much attention to groundwater. This state of affairs has changed, aided by the Convention on the Law of the NonNavigational Uses of International Watercourses (1997) (Eckstein, 2004). This Convention, yet to be ratified, only partially covered transboundary groundwater, that is, those connected to rivers, and thus left many aquifers uncovered. As a result of this situation, in 2008, the International Law Commission delivered to the United Nations General Assembly, draft articles for the law on transboundary aquifers. After reaffirming the protective and environmental approach to the use of groundwater, they ratified the application of the principle of fair use (1997) and of sensible damage. They also outlined measures for the following: first, cooperation between states; second, the regular exchange of data and information; third, the promotion of bilateral and regional agreements; and fourth, measures for the protection and preservation of ecosystems, and the prevention, reduction, and control of pollution. Along these lines, they provided that where appropriate, a shared management mechanism will be established.

Claims for the right to water to become a fundamental right, and thus protected, are increasing. This is probably highly applicable to groundwater since in many countries public water supply (to which the HRW is addressed) is supplied largely by groundwater. This is the case, for example, in Africa, the continent lagging most behind in the Millennium Development Goals. In relation to lack of access to water and sanitation by 2015, Africa is the continent most off target where groundwater is the daily source of drinking water for more than 75% of the population The first reference in this respect is to be found in articles 11 and 12 of the International Covenant on Economic, Social, and Cultural Rights of 19 December 1966. While not expressly mentioning the right to water, its wording has led the United Nations Committee on Economic, Social, and Cultural Rights (2002) to define the HRW as one which entitles everyone to sufficient, safe, acceptable, physically accessible, and affordable water for personal and domestic uses, and even links this right to the International Bill of Human Rights (1948). This reference to the right to water has been kept in recent documents such as the Plan of Implementation of the World Summit on Sustainable Development (2002), the Charter of Water of the Senegal River (Mali et al., 2002), or the Third World Water Forum Ministerial Declaration (2003). Second, there are advantages and disadvantages to a pure regulatory approach. Recognized limitations include symptoms such as the existence of rigid overly bureaucratic administrative procedures, the large number of authorities involved in taking decisions on groundwater, scarcity of technical and human resources to enforce compliance with legislative requirements, the often-absent citizens’ participation in decision-making processes on deliberation and decision making, or the confrontation of interests among the various government departments. These are clear examples of what we could call organization sickness. Legal proceedings that are prolonged, costly, hard to enforce, or construed poorly with practical needs of water management make it problematic for courts to be able to solve groundwater conflicts. Crucial and fundamental advantages to regulatory processes remain, such as its role as leverage and recourse for aggrieved third parties in court. This is why it is crucial in the case of groundwater to facilitate legal literacy or legal empowerment improving the capacity of communities to know and use the law – training in techniques such as interest-based negotiation, mediation, and facilitation. Third, the step from regulation by rulemaking process to negotiated rulemaking can never replace the public decisionmaking process, with the participation of all interested parties, or generate inequality. A series of legal principles have to be embodied in formal groundwater regulation, leaving more freedom or flexibility in terms of the implementation and enforcement (Table 4).

1.07.5 Institutional Aspects of Groundwater Management The problem of groundwater over-use has often been portrayed by the tragedy of the commons, that is, Hardin’s seminal essay in 1968 (Hardin, 1968) which describes how the

Groundwater Management Table 4

115

Legal principles applicable to groundwater legislation and its implementation

Legal principles

Rationale and justification

How

Effectiveness and efficacy

Efficacy of water management must be sought, by implementing or furthering measures

Cooperation

Participation and subsidiarity

Cooperation between authorities as fundamental (Declaration on Groundwater in the Mediterranean, 2006)a Aarhus Convention (1998)b

Sustainability and precautionary

Rio and Johannesburg Summitsc

Common responsibility

Commission on Sustainable Development, United Nations Economic and Social Council (2008)d

Adapting organization and competent authorities to conform to the natural characteristics of the resource – normally identified as a drainage basin Fostering the participation of users and interested third parties which has already been identified as a mechanism to secure acceptance and implementation of the agreed measures. Encouraging planning related to the allocation of resources or water-quality protection or restriction measures, rules on improvements and irrigation transformations, guidelines on recharge and aquifer protection. Cooperation, either through procedures or by agreeing to specific conventions, must enable more effective, allow the views of each of the players with responsibilities in the area to be known, avoid subsequent defects in the implementation of agreements and, in short, allow views to be joined to find the best solution. Environmental governance that is transparent, legitimate, and efficient. Public authorities, as the necessary guardians of correct application of the legal framework, may confer an especially important role on user associations directly involved in the management of, for example, groundwater resources. There has already been a certain amount of international experience in this area in countries such as Argentina, Colombia, Spain, the US, Indonesia, Mexico, Nepal, the Philippines, Sri Lanka, or Tunisia. Besides, it acquires greater importance in relation to groundwater as it is a way of surmounting the management difficulties caused by having multiple users. The implementation of sustainable development must pervade decisions on territorial and urban planning and the performance of specific projects, the approval of new protection rules, to end, cease or modify granted rights to groundwater and, especially, the economic development and growth initiatives in every country. Groundwater is a common good and therefore the responsibility for its protection and correct management belongs to everyone.

User and polluter pays principle

Solidarity

Levels of solidarity in groundwater management

To determine the obligation to repair and replace the resource base to their original state. In addition, it will be absolutely necessary to establish the strict liability regime in these cases, notwithstanding any potential exceptions linked to the state of technology or the grant of approvals.  Intergenerational solidarity. Future generations must be considered when adopting initiatives.  International solidarity. Not all countries have the same difficulties. Ranging from the actual exchange of water to the transmission of technology and knowledge. An example is the Johannesburg Declaration on Sustainable Development of 2002.  Regional solidarity. The areas within a state must seek points of consensus and foster instruments of cooperation in the rational and sustainable use of groundwater. It will undoubtedly be fundamental for this task to be able to plan and study the circumstances of each specific case, but it is important to take as reference the need to share and join forces in searching for the balance sought by all.

a

Ma´laga-Marrakech Declaration on Groundwater in the Mediterranean, 2006. (This Declaration is the result from two international congresses organized in 2006, AQUAinMED’06 – Ma´laga – and GIRE3D – Marrakech). b The United Nations Economic Commission for Europe Convention on access to information, public participation in decision making, and access to justice in environmental matters, Aarhus (Denmark), on 25 June 1998. c Johannesburg Declaration on Sustainable Development (World Summit on Sustainable Development, United Nations, 2002). d Commission on Sustainable Development, Report on the sixteenth session May 2007 and 2008 (Economic and Social Council, United Nations). Author: D. Sanz.

rational actions of individual actors, in our case groundwater users, lead to the demise of all, that is, aquifer over-use. This is because groundwater, which is a classic example of a common-pool resource, is defined by two characteristics: the

resource is largely rival and nonexcludable. These commonpool resources exist at different scales from transboundary to regional or small local aquifers. The works of Ostrom (1990) and other institutionalists have demonstrated that this case

116

Groundwater Management

underestimated the capacity of the users to self-regulate their actions, that is, to develop rules in norm and rules in use to prevent aquifer overuse. The groundwater silent revolution described earlier in the chapter has however outpaced the capacity to develop institutions suitable for good groundwater governance in terms of resilience, while maintaining a level of flexibility and adaptability to cope with a high degree of change.

1.07.5.1 Groundwater Institutions: Mapping Groundwater Institutional Design A number of conditions have been well documented in the literature for the successful management of common-pool resources. These factors are summarized in Box 3 (Schlager and Lopez-Gunn, 2006) in relation to groundwater.

1.07.5.1.1 Boundary definition The first tenet of institutional theory refers to boundaries. This refers on the one hand to natural boundaries and on the other to institutional (property right) boundaries. In the first case, the definition of natural aquifer boundaries for management purposes has the added complication that groundwater aquifers do not necessarily coincide with surface-water systems. In addition, groundwater suffers from the same problem that surface water had traditionally experienced, lack of overlap between administrative and natural boundaries (i.e., problem of fit), that is, the boundaries of for example, regional administration do not coincide with river-basin boundaries, with the added twist that surface basins and aquifers often do not coincide, which further increases the complexity. An interesting example is currently pursued under the EU WFD, which has adopted a twin-track approach of managing water according to river basins while simultaneously mapping groundwater bodies, while setting the objective to achieve

good status for all water bodies in the EU (surface as well as groundwater) by 2015. According to Howe (2002), assigning well-defined groundwater property rights, for example, through pumping permits (discussed earlier) enhances the value of water, which creates incentives to use water more effectively or to transfer rights and/or use to third parties who are willing to pay for pumping rights. Therefore, the most complex challenge for water laws is the administration of water rights, that is, the granting of licenses, concessions, permits, and other legal deeds for the abstraction of groundwater, and for the discharge of waste water directly or indirectly into the aquifer. Groundwater, in particular, offers additional problems because of the following: first, the potentially large and often heterogeneous number of users, and second, the boom in use which has often overwhelmed administrations. In Spain, 20 years after the 1985 Water Law, the registration of groundwater rights was required (1988), but the administration has still not finished the process; or in places such as Mexico an ecological price has been effectively been paid for the process of registering 330 000 water rights by 2003, by over-allocating groundwater resources. The new 2002 Water Law in China established the need to obtain groundwater permits. Yet, the issuing of water permits in China by counties is proceeding very slowly, and there is lack of consistency between authorized abstractions via permits and groundwater-resource availability (Foster et al., 2004). Furthermore, growing experience in the process of assigning groundwater property rights has shown that it is crucial to take context into account when assigning groundwater rights, for example, in South Africa, where plural legislative frameworks (formal and customary) coexist. These plural, often dual legal systems have important implications for the registration of groundwater property rights, due to overlapping legal orders. The diversity and flexibility of customary laws, principles, and practices may be intentionally or

Box 3 Ostrom’s institutional design principles applied to groundwater institutions. Reproduced by Lopez-Gunn E from Ostrom E (1990) Governing the Commons: The Evolution of Self-Governing Irrigation Systems. Cambridge: Cambridge University Press; Schlager E and Lopez-Gunn E (2006) Collective systems for water management: Is the tragedy of the commons a myth? In: Rogers P, Llamas MR, and Martı´nez-Cortina L (eds.) Water Crisis: Myth or Reality?, pp. 43–60. London: Taylor and Francis; and Cleaver F and Franks T (2005) How Institutions Elude Design: River Basin Management and Sustainable Livelihoods. BCID research paper 12, ICID Conference, London. *

* *

* *

*

*

*

Clearly defined boundaries. Both the boundaries of the aquifer and the individuals or households with groundwater rights from the aquifer are clearly defined. This principle refers both to the physical boundary of the aquifer and a clear identification of groundwater rights (legal boundary on groundwater). Collective choice agreements. A clearly defined groundwater-user group or community should be involved in groundwater management. Appropriation rules. Operational rules in relation to time, location, technology, or groundwater-abstraction units should include the groundwater users affected by these rules and should be included in decision-making processes to modify these appropriation rules. Monitoring. Monitors who actively audit physical conditions and behavior are accountable to groundwater users and/or are groundwater users themselves. Graduated sanctions. Sanctions are devised for noncompliance with collective rules (operational rules). Groundwater users who violate operational rules are likely to receive graduated sanctions by other groundwater users, by officials accountable to these groundwater users, or by both. These sanctions have to be applied consistently, impersonally, and rapidly. Conflict-resolution mechanisms. Groundwater users and officials have access to low-cost local arenas to resolve conflict among users or between users and officials. Conflict-resolution mechanisms should be clear, accessible, and quick. Legitimacy. The legitimacy of groundwater users to organize and set up their own institutional arrangements is not challenged by external government authorities. Nested enterprises. Local groundwater institutions are nested within other levels of decision making, in multiple layers, which facilitate governance (in terms of consistent operational rules, monitoring, and enforcement and conflict resolution).

Groundwater Management

unintentionally replaced by new water laws, and uniform rigid principles and requirements. Legal frameworks empower if these recognize rights of existing water-user communities, and enable legal recourse if rights are harmed. Plural groundwater property-rights systems ideally have to be based on the principles of good governance: transparency, accountability, and the rule of law. There is also a risk in idealizing customary groundwater rights, which might not necessarily, for example, be gender neutral.

1.07.5.1.2 The role of groundwater-user associations Decentralization of groundwater-resource management is coherent with the creation of collective institutions like groundwater-user associations that can be directly involved in groundwater management. There is a range, nevertheless, on the degree of management devolution to groundwater-user groups, for example, market co-production, co-management, or regulated autonomy. In the late 1990s, decentralization was a consequence of rolling back the state, and transferring management directly to users – participatory-irrigation management (PIM; or irrigation-management transfer) since the dominant use of groundwater globally is agriculture. This is part of the wider trend in PIM (Merrey et al., 2007). In the case of groundwater, PIM has interesting twists and turns because at least two types of groundwater-irrigation systems can be identified: first, the case of collective wells which are managed as very small surface-water systems, and second, and most common, individual farmers exploiting their well for productive agriculture and/or livelihoods. The creation of wateruser groups and PIM would be similar to surface water in the first case, and would face similar limitations as those recently put forward for surface water, that is to say, that this is no panacea and it is suitable in some cases but not necessarily in all cases. These water-users associations (WUAs) are much smaller than WUAs for surface irrigation: this makes it simpler to organize them but it is also less important for them to be formal organizations. Informal groups are generally adequate for managing irrigation from individual wells, even when managed by groups of up to 50 farmers, such as in China. The second case is a true case of collective action because individual users have to be persuaded externally (top-down) or realize (internally) that the benefits of self-organization are higher than the costs, and that free riding on the collective action of others is now penalized either through formal sanctions or through informal, social norms. The objective in this case is regulation of the aquifer (i.e., the source of water) rather than equitable management of the distribution of water from the source. The case of PIM in groundwater is fascinating because there is evidence from groundwater-user associations that have been both created top-down and others which have emerged bottom-up, spontaneously. This is the case of Spain, where Comunidades de Aguas Subterraneas – which is in effect, part of the water authority and instigated by the administration, and which manages groundwater as part of the public domain – coexists with Comunidades de Usuarios de Aguas Privadas, where private groundwater-user groups have been created through user initiative. More research is needed on delivery of management outcomes; what is already evident is that the scale of

117

these groundwater-user groups is large, managing aquifers which can cover areas from 7000 km2 to 300 km2, and where success is mixed in terms of sustainable aquifer management. In China, groundwater management has gradually become more decentralized, and bottom up, with increased stakeholder participation at all levels, and closer interaction between users at the local level and the responsible authority, the Water Resource Bureau, normally designated at county level for rural areas and district level for urban municipalities, while some groundwater-user associations have also been established (Foster et al., 2004). The few examples of groundwater-user associations that have become effective resource managers have two things in common: they have successfully articulated common goals and objectives, and they have established mutually accepted rules regarding resource access and use, in order to guarantee the long-term availability of groundwater to users. For example, in Mexico, in the early 1990s, due to intensive groundwater use in the central and northern part, many groups started to emerge concerned with the problem of intensive groundwater use and negative externalities: for example, the spontaneous creation of the Grupo del Agua in the Comarca Lagunera (1991) and the Grupo del Agua of Santo Domingo valley a year later (1992). Other groups appeared in other areas. Initially, there was lack of clarity on the regulatory structure of these groups and their financing, which meant there was little support from the federal level. Initially, the Mexican Federal Government did not legitimize these spontaneous water-user groups until the mid-1990s, when these groups reorganized themselves as Comites Tecnicos de Aguas Subterraneas or COTAS, starting in the Queretaro valley, and then spreading to other aquifers in the central and northern part of Mexico. In the state of Guanajuato, local authorities encouraged the formation of COTAS in all aquifers in the state, supporting them financially (Escolero and Martinez, 2007). Meanwhile in the USA, local landowner associations in Texas have been experimenting with the feasibility of selfmonitoring and regulation under local groundwater districts, which would set pumping limits and well placement based on hydrologic models, to deliver public goods such as open-space protection and aquifer recharge through cooperative landowners associations (Wagner, 2005). In India there is evidence of the spontaneous creation of WUAs, through what Shah (2005) calls swayambhoo (self-creating), involving entrepreneurial efforts, which are normally present since most groundwater users are by definition smallscale entrepreneurs. It is estimated that over a quarter of Indian irrigated areas operate through this kind of spontaneous creation of informal water markets (Shah, 2005). The challenge is when swayambhoo institutions have to be scaled up, whether motivation can shift to longer term, and collective self-interest, then, can also start to internalize externalities (Box 4).

1.07.5.2 An Institutional Audit of Groundwater Institutions Added research and experience have however highlighted new dimensions to the institutional framework analyzed above (Cleaver and Franks, 2005; Cleaver and Franks, 2008). The socalled post-institutionalist turn has added some caveats and

118

Groundwater Management

Box 4 WUAs for groundwater management in China. By Simon Howarth (UK DFID-funded Water Resources Demand Management Assistance Project. Water-users associations (WUAs) are being set up in each village in the Shiyang River Basin, an arid internal river basin in Gansu Province of Northwestern China. They are being promoted by the government to assist in the management of groundwater, but with a primary focus on achieving water savings. Existing tube-well management arrangements, by local production groups (subdivisions of villages) supported by water-management stations (WMSs) (government) at township level, are believed to be effective and equitable, but insufficiently focused on reducing total water use – with the result that groundwater levels are dropping at 50– 100 cm yr1, making agriculture unsustainable. These WUAs thus have different objectives to WUAs set up elsewhere in the world, which are required to improve management of large surface canal flows and hence ensure greater equity of water distribution. Each village (or WUA) typically includes 20–50 tube-wells (each serving 5–20 ha, farmed by 10–50 households) which are managed by production groups (water-user groups). The tasks of the village-level WUA include assistance to the WMS in many of the new groundwater-management procedures, such as issuance of household water-rights certificates, enforcement of permits, and collection of fees – all of which are aimed at reducing the amount of water that farmers use. These are onerous requirements and thus the WUA are repaid part of the water-resource fees collected in order to cover a small salary for directors and vicedirectors and some administrative costs – in recognition of the role that WUAs play in water-resources management. This formal process of paying staff from part of the newly introduced water-resources fee is important for ensuring that the WUAs are effective and sustainable. In addition to these responsibilities for assisting the government, WUAs also have a small role in water management which includes improving maintenance; reducing conflicts; planning, implementing, and monitoring water distribution; monitoring groundwater levels; and ensuring effective communication between WMS, WUA, and farmers. Much of this work is done by well-established informal means by production groups, but the WUA coordinates between production groups and provides services at a higher level – such as employing a maintenance technician who is available to all groups, linking groups to governmentsponsored training programs, and assisting in contracts with crop-grower associations, seed suppliers, and markets. These WUAs are intended to be independent, autonomous, democratic, village-level organizations, but for practical reasons, the staff are often largely drawn from existing village committees (these are elected, but all candidates are required to be vetted and approved in advance). On paper, it is a strong system, but it is newly established and not yet fully effective. Many questions remain unconfirmed, including sustainability of financial arrangements, ability to deliver a positive service to farmers, and the willingness of farmers to accept the restrictions on water use. Further work on WUAs will require a combination of administrative measures at provincial, municipal, and county levels, and capacity building among WUAs. This capacity building will in turn require awareness-raising at the various levels of government, where there is typically greater faith in top-down controls (such as IC cards) or infrastructural improvements (canal lining) than there is in local institutional methods for water savings. Nevertheless, early indications are that the strong commitment to water savings by the government will ensure that WUAs will be effective, but that their role and responsibilities will be modified and simplified as they are implemented.

new dimensions to a strict application of Ostrom’s institutionalist framework. The main criticisms are that it is does not provide a causal analysis for the processes underlying these design principles. In particular, the areas that are increasingly perceived as fundamental to sound groundwater governance are: first, the key role of social capital and higher-level authorities; and second, the relevant role of political leadership and acknowledging the politics and vested interests of groundwater use, which are played out in the prioritization of groundwater use among competing users; and third, the potential problem of corruption as a symptom of a malfunctioning groundwater systems and the antidote of transparency and participatory groundwater management.

1.07.5.2.1 Higher-level authorities: Supporting, legitimizing, and leading The relevance of higher-level authorities comes to the fore as an essential supporting element for effective institutions and the development of organizational capacity since both authority structures and social norms (e.g., collective action by users) have to support and underpin the functioning of design principles. Higher-level authorities are key as facilitators for local groundwater management and for the vertical integration between the different spatial scales (farm level, aquifer, and regions, national, and international scales). For example, it appears that cross-scale linkages exist in China where there is a provision for transboundary issues across provinces, and also indirect leadership, as professional guidance from higher-level

authorities without any hierarchal subordination. County government can issue groundwater regulations within its boundaries, in agreement with provincial and national legislation (Foster et al., 2004). Higher-level authorities are increasingly perceived as a necessary condition to support local institutional arrangements. One of the most important roles for higher-level authorities is either to provide leadership or to facilitate leadership. In many cases, leadership is actually in the form of legitimizing or supporting local leaders. These local leaders in turn can drastically reduce the transaction costs of institutional change. In India, for example, the common aspect of all successful tank institutions was a leader or a leadership compact, which could sway the community and thus drastically reduce the transaction costs of ‘‘enforcing institutional arrangement that would either not work in their absence nor survive them’’ (Shah, 2005, p. 17).

1.07.5.2.2 Transparency and participatory groundwater management This chapter starts from the tenet that there is no global physical groundwater crisis; rather, there is a crisis of groundwater governance. Governance in this context is defined as the interplay of actors (public, private, and civic) to promote societal goals and the production of collective goods. One of the key basic assumptions of effective functional groundwater management is transparency and participation by all groundwater users in the decision-making process in

Groundwater Management

line with the Aarhus Convention on Public Participation (UNECE, 1998). The problem of corruption often damages those most vulnerable, weakening the rule of law, and fostering social norms that systematically prioritize private gain over social well-being. Corruption according to Transparency International is about breaking socially established expectations of appropriate behavior (Stalgren, 2006). As stated earlier, groundwater has some inherent characteristics that should make it less prone to corruption. In the case of groundwater, the timescale and size of investment is normally smaller than in the case of surface-water projects. Evidence of corruption in the case of groundwater tends to refer to drilling concessions, bribing meter readers, distorted site selection for boreholes, for example, for those with more political or economic influence; bribery to obtain drilling permits or to cover up excessive abstraction, to obtain preferential treatment for services or repairs, and also to falsify meter readings (Transparency International, 2008). Advances are constantly made to facilitate transparency, accountability, and decentralization in groundwater management and use, for example, in the use of technology (Calera et al., 1999). Three measures are considered crucial in the case of effective groundwater management. First, reduce the complexity in regulation, licensing, and control; this is to prevent a weak and ineffective legal system which can encourage a clientelistic patronage system. Second, facilitate and incentivize so-called participatory monitoring. As stated early, transparency, monitoring, and sanctioning are part of healthy groundwater institutional arrangements. Robust groundwater institutions can benefit from advances in participatory geographical information systems, that is, use of technology jointly by groundwater users and regulators to increase transparency in water use and allocation. A good example is currently being implemented in the Mancha region in Spain, where satellite information is being used directly by farmers through an irrigation advisory service, which integrates realtime data to help farmers improve water use by different crops, while optimizing production (Calera et al., 1999). Third, encourage transparent access to data on groundwater use, licensing, and subsidies. This can be strengthened by partial decentralization to water users, to involve them in decision making, which would decrease the transaction costs of obtaining good-quality information while increasing the level of information available (Box 5). In summary, good, symmetric information equally accessible to both users and the regulator is crucial to facilitate cooperation among aquifer stakeholders. This information

119

ideally has to be externally audited and contrasted, to allow for advocacy and disclosure of illicit behavior. Often, easy access to good and reliable data on abstractions, water quality, and aquifer water levels is a prerequisite to succeed in groundwater management. Current information technology can help information to be made easily and economically available to an unlimited number of users. Nevertheless, in a good number of countries, it will be necessary to change the traditional attitude of water agencies of not facilitating easy access to water data to the general public. This partly comes from a shift in mentality that strengthens accountability of public authorities downward to users and civil society.

1.07.6 The Complex Concept of Groundwater Sustainability and Future Management Issues An economic, efficient use of an aquifer would imply maximizing the present value of the resource in the case of the Ogallala, in Texas, where abstractions are much greater than natural recharge. It was discussed earlier in Section 1.07.2.5 that this exemplifies the complexities in defining what is meant by sustainable groundwater management. The exhaustible nature of the resource would raise the issue of appropriate long-term economic and demographic development of the region. The availability of open-access inexhaustible resources such as groundwater often invites gold-rush patterns of excessive fast exploitation and maladapted patterns of infrastructure and social development, that is, so-called boom towns (Howe, 2002). Economic efficiency in a renewable aquifer may imply drawing down the aquifer during droughts and allowing its recharge in periods of good surface flow. However, ecological dimension of sustainability used to equate recharge equal to abstractions is what some authors consider the renewable yield. Nevertheless, the EU WFD introduces a more complex concept, the achievement of good ecological health of aquatic ecosystems, which depends on the available yield. This new concept, not fully applied yet, may imply significantly lower amount of groundwater allowed abstraction than the renewable yield. Literature, such as Moench (2003) and Shah et al. (2007), and examples in Indonesia, China, USA (Ogallala), Spain, and Mexico, seem to indicate that de facto development (e.g., in terms of agricultural productivity) is prioritized over longerterm ecological groundwater sustainability. In Indonesia, regulation of groundwater is perceived to be at cross-purposes with industrial growth. In China, pure economic growth is

Box 5 Participatory groundwater monitoring in China. By Simon Howarth, based on Gansu under the UK DFID-funded Water Resources Demand Management Assistance Project. Simple monitoring by villages of the volumes of water abstracted and of the groundwater level is valuable for developing an understanding of groundwater. This has three classes of benefit: promoting awareness of groundwater, which is commonly less well understood than surface water; this can profitably be incorporated into school curricula so that children become aware of water issues: * * *

Enabling WUAs to manage their resource better, and understand why restrictions are being introduced. Providing data to supplement formal data-collection programs by government hydrology bureaus. Involving communities increases ownership of the concepts and reduces asymmetry of information. The information can easily be published on village notice boards.

120

Groundwater Management

seen as a central part of development policy, in the policymakers’ mindset toward pure economic accumulation. In Mexico, continued overdraft in Hermosillo is driven by problems of reconciling economic efficiency and ecological sustainability (e.g., problem of saltwater intrusion), and where huge economic returns derived from groundwater in terms of income create an incentive to search for additional water resources through surface-water transfer (Escolero and Martinez, 2007). This is similar to the case in Southeast of Spain, in the Murcia and Almeria regions, where the intensive use of aquifers for highly productive agriculture started to drive national water policy to transfer water to this area, because of the political difficulty of controlling this intensive but highly profitable intensive groundwater use (Llamas and Martı´nezSantos, 2005; Llamas and Martı´nez-Cortina, 2009). The strength of negative externalities depends partly on aquifer characteristics (e.g., transmissivity storage), spacing of wells, and connection to surface water. These questions over preferred criteria as against competing uses go to the heart of the meaning of groundwater sustainability and to what extent this is feasible or, indeed, it is perceived as feasible or possible due to the institutional path dependencies of choices made in the past.

1.07.6.1 Groundwater Management Externalities In groundwater, externalities are the rule rather than the exception. The real issue is not the elimination of externalities (usually physically impossible) but rather, whether the impacts on third parties are excessive according to certain criteria. The relevant policy question is whether these externalities are considered excessive and for which criteria they are used or prioritized: economic efficiency, groundwater sustainability, or Table 5

social equity. In the EU, according to the WFD, the goal is to restore the ecosystems to a good ecological status unless the cost of this recovery is economically or socially very difficult. In this case, member states have to report in detail to the European Commission on the extenuating circumstance to ask for derogations. This is a process currently underway, but it appears that countries are anticipating the difficulty of complying with the WFD by 2015. Here, we summarize five indicators of typical problems of intensively used aquifers, but it is important to mention that these are sometimes used inadequately (Table 5).

1.07.6.1.1 Degradation of groundwater quality Groundwater abstraction can cause, directly or indirectly, changes in groundwater quality. The intrusion into a freshwater aquifer of low-quality surface water or groundwater, because of the change in the hydraulic gradient due to groundwater abstraction, is a frequent cause of quality degradation. This degradation of groundwater quality may not be related to excessive abstraction of groundwater in relation to average natural recharge. Other causes may be responsible, such as return flows from surface-water irrigation, leakage from urban sewers, infiltration ponds for wastewater, septic tanks, urban solid-waste landfills, abandoned wells, mine tailings, and many other activities not related to groundwater development (Custodio, 2002). For instance, the groundwater-quality degradation in many Central and Northern European countries is related to intensive rainfed agriculture. Saline intrusion may be an important concern for the development of aquifers adjacent to saline water bodies. This is a typical problem in many coastal regions of semiarid or arid areas. Moreover, in this case, the relevance of saline-water

Typology of groundwater externalities

Type of externality

Externality

Explanation

Environmental

Affected ecosystems

Socioeconomic

Pumping costs externality

Damage to ecosystems or surface-water features dependent on discharge from aquifers; spring-flow reduction Increase in pumping costs due to drop in aquifer levels, these costs can be fixed or marginal costs that one user imposes on another when pumping lower from a water level, external costs can be reduced by selecting a better well location Due to salinization or marine intrusion Water quality varies with depth (normally more saline with more depth) and also location specific, for example, aquifer located in coastal areas and islands (e.g., saltwater intrusion), the spread of low-quality water within an aquifer Water pollution due to intensive agriculture, for example, with nitrates and/or pesticides Decrease in pore-water pressure, related to amount of groundwater withdrawn Aquifer compaction with possible resultant reduction in aquifer storage capacity For example, when small farmers cannot adjust to the drop in aquifer levels Increased opportunity cost due to an increase in scarcity value due to intensive use Reduction to buffer value provided by groundwater against drought Diminished economic activity in the area, reduced water availability for other water-right holders, and reduced land-use options for future inhabitants

Potential loss of agricultural land Groundwater quality externalities

Land subsidence Aquifer compaction

Option and intertemporal externalities

Social externalities Increased scarcity value Buffer value of a groundwater stock Intertemporal externalities

Groundwater Management

intrusion depends not only on the amount of the abstraction in relation to the natural groundwater recharge, but also on well-field location and design, and the geometry and hydrogeological parameters of the pumped aquifer. In most cases, the existing problems are due to uncontrolled and unplanned groundwater development and not to excessive pumping. It appears, for example, that the last half-century seawater intrusion has been well controlled in the coastal plains of Orange County (California) and Israel.

1.07.6.1.2 Susceptibility to subsidence When an aquifer is pumped, the water-pore pressure decreases and the aquifer solid matrix undergoes a greater mechanical stress. This greater stress may produce compaction of the existing fine-grained sediments (aquitards) if the stress due to the decrease in water-pore pressure is greater than the so-called preconsolidation stress. This situation has occurred in some aquifers formed by young sediments, such as those in Mexico City, Venice, and others. In Bangkok (Thailand), parts of the city were sinking at a rate of 10 cm yr1, with an increased risk of flooding and damage to roads and buildings. Caves and other types of empty spaces may exist under the water table in karstic aquifers. When the water table is naturally depleted, the mechanical stability of the roof of such empty spaces may be lost and the roof of the cave collapses. This is a natural process that gives rise to the classical dolines and poljes in karstic landscapes. When the water table depletion or oscillation increases due to groundwater abstraction, the frequency of karstic collapses can also increase. There are a number of well-known examples of land subsidence due to intensive groundwater use. In both cases, the amount of subsidence or the probability of collapses is related to the decrease in pore-water pressure, which is related to the amount of groundwater withdrawal. Nevertheless, the influence of other geotechnical factors may be more relevant than the amount of water abstracted in relation to the renewable groundwater resources of the aquifer. In Tianjin city (China), excessive abstraction of deep groundwater caused a land subsidence of up to 3.0 m (Foster et al., 2004). In Texas, Galveston and adjacent counties have experienced subsidence due to the long-term drop in aquifer levels (Wagner, 2005). Meanwhile, land surfaces in parts of central Arizona have fallen by 20 m in the last 20 years (Howe, 2002).

1.07.6.1.3 Interference with surface water and ecological impacts There are potential conflicts due to groundwater pumping and its interaction with surface waters and riparian habitat. For example, if there is no source of capture, a well will continue to withdraw water from storage until, either the aquifer is depleted, or the drawdown exceeds the well depth. The groundwater pumped from an aquifer is derived from a decrease in storage in the aquifer, a reduction in previous discharge from the aquifer, an increase in the recharge, or a combination of these changes. Capture may be defined as the increase in recharge plus the decrease in discharge. Examples of capture are: (1) an increase in groundwater recharge from losing streams (or increased infiltration); (2) a decrease in groundwater discharge to gaining streams (or interception of

121

baseflow); and (3) the reduction in the component of evapotranspiration that is derived from the saturated zone. If we restrict the groundwater pumping to the capture, there will no longer be a decrease in storage. Restricting the groundwater withdrawals to what may be captured is a definition of safe yield or sustainable yield for the aquifer. Nevertheless, the definition of safe yield is controversial for some authors. Groundwater mining, that is, when water is abstracted mainly from storage (as discussed in Sections 1.07.2.5 and 1.07.2.6) may also be considered a safe yield, which can be valid under certain conditions. This is because sustainability has to consider different aspects, including economic, ecological, and social, and the real-life difficulties of implementing the concept of sustainability. Most river systems have a hydrology that is simple in concept but complex in detail. Some anthropogenic activities may have a significant impact on the catchment hydrologic cycle. For instance, the intensive use of groundwater for irrigation in the Upper Guadiana basin (Spain) has resulted in serious water-table depletion (B30–40 m). The most alarming consequences of the water-level drop were changes in the groundwater flow patterns and in the form, function, and quality of many wetlands. Areas that had received the natural discharge from the aquifer became natural recharge zones (Herna´ndez-Mora et al., 2003). This has produced a spectacular decrease in total evapotranspiration from the water table and wetlands, evaluated between 100 and 200 Mm3 yr1 (Martı´nez-Cortina, 2001). From the point of view of the water budget, there is an important increase (almost 50%) of the annual renewable resources, understood as the water that can be abstracted from the aquifer maintaining the water level as in the previous year, and calculated as the difference between aquifer recharge from precipitation and losses from evapotranspiration. This artificial depletion of the water table can also change dramatically aquifer–streams relationship, as in the previous example. Gaining rivers fed by aquifers may become dry except during storms or humid periods when they may become losing rivers, an important source of recharge to the aquifer. Nevertheless, this new water budget may present legal problems if the downstream water users have previous water rights (Llamas and Martı´nez-Cortina, 2009). The ecological impacts, mainly caused by water-table depletion as it has been showed in the Upper Guadiana basin case, are becoming an important new constraint in groundwater development in some countries, especially in the 27 countries of the EU because of the requirements of the WFD. A famous case is the Tablas de Daimiel National Park, a Ramsar site, whose main source of water used to be aquifer discharge before intensive irrigation made the area a recharge area rather than a natural outflow for the aquifer (Figures 5 and 6). Decreasing or drying up of springs and wetlands, low flow of streams, disappearance of riparian vegetation because of decreased soil moisture, alteration of natural hydraulic river regimes, and changes in microclimates because of the decrease in evapotranspiration, can all be used as indicators of ecological impact. Reliable data on the ecological consequences of these changes are not always available, and the social perception of such impacts varies in response to the cultural and economic situation of each region. The lack of adequate

122

Groundwater Management

Figure 5 The Tablas de Daimiel National Park, October 2008. Photo courtesy: Pedro Zorrilla).

Figure 6 The Tablas de Daimiel National Park, October 2008. (Photo courtesy: Pedro Zorrilla).

scientific data to evaluate the impacts of groundwater abstraction on the hydrologic regime of surface water bodies makes the design of adequate restoration plans difficult. For instance, wetland-restoration programs often ignore the need to simulate the natural hydrologic regime of the wetlands, that is, restore not only its form but also its hydrological functions (Bergkamp and Cross, 2007). Similar problems result in trying to restore minimum low flows to rivers and streams. Oftentimes, minimum stream flows are determined as a percentage of average flows, without emulating natural seasonal

and year-to-year fluctuations to which native organisms are adapted (Llamas and Garrido, 2007; Garrido and Llamas, 2009).

1.07.6.2 Groundwater: Future Risks and Opportunities for Management 1.07.6.2.1 Groundwater and climate change The latest report of the IPCC by Bates et al. (2008) only marginally addresses the role groundwater can play in the

Groundwater Management

adaptation or mitigation of the potential negative effects of climate change. The report says ‘‘There is a need to improve understanding and modeling of climate changes related to the hydrological cycle at scales relevant to decision making. Information about the water related impacts of climate change is inadequate – especially with respect to water quality, aquatic ecosystems and groundwater (added emphasis) – including their socio-economic dimensions’’ (p. 4). A detailed analysis of this topic is outside of the scope of this chapter. However, it seems that the role of groundwater development will increase significantly if the pessimistic predictions of the IPCC reports for the arid and semiarid areas become true. These pessimistic forecasts are mainly related to the increase in evaporation rates, due to the increase in temperature, and to the higher drought frequency. As it is shown in this chapter, the evaporation of groundwater from the aquifers is usually irrelevant and one important property of most groundwater reservoirs or aquifers is their resilience to dry spells. The UK Groundwater Forum, for example, studied potential scenarios for groundwater as a result of climate change, and some of these scenarios pointed to a long-term decline in aquifer storage, increased frequency, and severity of groundwater-related floods, mobilization of pollutants due to seasonally high water tables, and saline intrusion due to sealevel rise (Bergkamp and Cross, 2007). However, these predictions have to be contrasted with others across the world, with different climatic regimes. The Edwards aquifer is one of the largest freshwater aquifers in the USA with a total area of 15 640 km2, and a primary source of water (agricultural and municipal) for southern Texas (Loaiciga, 2003). It has been identified as one of the areas most vulnerable to complex, nonlinear climate feedbacks, and where potentially aquifer-exploitation strategies must be adapted to climate variability. In fact, when climate and groundwater-use changes are considered together, the role of groundwater use over climate prevails, that is, changes in groundwater use due to population growth and changes in land use or economic preferences may cause more profound aquifer impacts than those associated with global warming. For example, climate change in San Marcos springs could increase spring flow relative to the base condition by 17%, while the groundwater use alone in the year 2050 can reduce springwater flow by 22%, that is, groundwater use dominates over climate change. Therefore, the primary threat to the Edwards aquifer comes from the rise in groundwater use associated with predicted growth not from climate change. The latter in fact would increase spring flow in the study area (Loaiciga, 2003). There is also increased understanding that the vegetation response to climate change could either increase or decrease recharge. Climate change in fact could increase aquifer recharge according to recent simulation models, although highly dependent on geological settings (Green et al., 2007). In study areas characterized by sandy top soils and large interconnected aquifers, groundwater levels rose significantly. In Australia, simulations of twice the existing CO2 led to significant changes in the rate of groundwater recharge in Mediterranean and subtropical climates. Water recharged from 34% slower to 119% faster in the Mediterranean climate and from 74% to 500% faster in the subtropical climate.

123

Opportunity for decrease exists but the general trend is toward increase in recharge (Green et al., 2007). In the context of groundwater and climate change, it is important to note the spatial and timescales of aquifer and climate systems. First, aquifers often operate in scales of much less than 106 km2 and in a great majority of cases, groundwater basin encloses areas of 104 km2, while global climate models (GCMs) operate on 200 km  200 km (4.104 km2) and regional climate models (RCMs) with resolutions of the order of 20 km  20 km (Loaiciga, 2003); therefore, the scales do not necessarily match up. Second, the nature of medium- and long-term climate predictions and the contrast between floodimpact studies with temporal scale from minutes to days, and drought impact studies precipitation and temperature temporal scale from days to years depending on inter- and intraseasonal variation. These uncertainties in both space and time make predictions, in terms of climate change and variability, difficult; what is clear is that in this context, aquifers have the natural capacity to act as climate regulators, that is, buffering capacity for drought and floods.

1.07.6.2.2 Future management issues There are a number of future management issues that become apparent and whose importance is increasing. One of the main issues is the joint use of surface and groundwater and the linkages between water quality and water quantity. There are good examples across the world of successful joint management of surface and groundwater that play to the strengths and weaknesses of both. For example, the case of Israel, and the case of the cities such as Barcelona (Spain) and Phoenix (Arizona), which rely on groundwater supplies as a strategic resource in times of drought. However, there are also examples on lack of joint management or in fact disjointed management, that is, when poor groundwater management leads to surface-water transfers to compensate. This is the case, for example, of lobbies pushing for surface-water transfers or the authorities pushed to find additional water supplies due to groundwater-quality problems. For example, both in London and in the coastal metropolitan area in Barcelona, there is a problem with aquifer rebound, and these polluted groundwater resources signify that it is easier to invest on large desalination plants to augment supply, since the costs of cleaning polluted groundwater are prohibitive. Meanwhile, in both Spain and Mexico, pressure builds in areas where aquifers are intensively used to bring surface-water supplies. The city of Hermosillo (Mexico) is heavily dependent on groundwater for its level of productivity and the residents of the area have been lobbying the Mexican government for a large water transfer that would bring water from the State of Sinaloa 485 km away (Plan Hidra´ulico del Noroeste). As groundwater storage continued to fall, the rising shadow (marginal price) value for groundwater rose. This shadow value can indicate what would be the optimum timing for the water transfer to occur, which would be when the values of the transferred water are lower (US$ 0.0222 m3) than the shadow groundwater value price (US$ 0.0224 m3). This highlights that the ideal time to build the project would be in the 29th year. This indicates that if the water project was built at the current costs of groundwater abstraction, it would

124

Groundwater Management

have been built prematurely (Howe, 2002). This, however, is only if one applies economic-efficiency criteria. If other criteria (such as environmental sustainability) were used, this would call for a much quicker solution (e.g., due to the external costs of saltwater intrusion). In Spain, lack of groundwater management in the southern Mediterranean coastal belt, triggered in large, the conflict to divert water from the Ebro river in the north to help compensate for rapidly depleting aquifers (Llamas and Martı´nez Santos, 2005). Water agencies tend to build projects far in advance of their justifiable need on pure economic terms (Howe, 2002), and often fail to capitalize on the synergy between effective joint surface and groundwater use, which by definition implies the management of both. Restrictions on groundwater allocations are a direct loss attributed to decision makers and thus unpopular, whereas a loss of income due to over-abstraction is a probabilistic loss. Therefore, it is politically rational for decision makers to prefer users to continue pumping than to take the (unpopular) decision to cut allocations and instead opt for the politically more popular water transfers. There are very few systems of explicit conjunctive management. Until the 1940s, one main reason was the lack of understanding of hydrogeological knowledge and therefore the poorly developed model of surface and groundwater interaction. There are recent examples of regulatory innovations to deal with groundwater due to a much clearer understanding and a capacity to play to the strengths of surface and groundwater joint use – for example, in Colorado, where tributatory groundwater rights have been incorporated into the prior appropriation (priority) doctrine of water law, which in theory could preclude conjunctive management since groundwater tends to be junior rights compared to surface senior water rights (Howe, 2002).

1.07.6.2.3 Groundwater: Issues of fit and political windows of opportunity There is increasing evidence of the institutional diversity of groundwater user groups, from landowners associations, local landowner cooperatives, natural resource cooperatives, and water districts in Texas, to tube-well cooperatives in India, to agricultural transformation societies, and both public and private user communities in Spain. There is also some evidence that spikes in groundwater scarcity can trigger organization of these groundwater-user groups and in fact provide a window of opportunity for collective action, for example, drought can act as a motivator for self-organization or increased competition for groundwater resources among users. Drought intensifies conflicts and yet stimulates short-term and long-term efforts to modify rules and procedures for regulating rights. Scarcity value encourages the spontaneous creation of cooperative groups as groundwater becomes scarcer and groundwater economic value raises the cooperative model of groundwater pumping. Increasingly, it is appreciated that issues such as droughts rather than perceived as short-term crises are in fact also an opportunity for institutional change and adaptation. These are windows of opportunity for political action and social change since most stakeholders are receptive to the need for effective responses – that is, they provide opportunities for institutional innovation.

Strategically, choices can be made on the value of groundwater resources and prioritization of use, for example, for domestic water supply while at the same time providing incentives for economic transition. In addition, there is increased recognition on the importance of context and that there are no ideal aquifer-management regimes; rather governance arrangement. In particular groundwater basins are highly individualized, there is no single best-practice model for groundwater management; rather there are context-specific, multiple management scenarios which have to be negotiated. Institutional solutions that are viable in a particular context have to be framed within the inherent limitations of scientific knowledge, and centered on core objectives rather than specific groundwater parameters (Moench, 2007). That is, focus on livelihoods and environmental values rather than sustainable yield, which by itself is increasingly a contested concept (Llamas and Garrido, 2007). Responses have to be suitable for specific socioecological context rather than politically correct integrated management, which sometimes can be too rigid or overtly focused on technical ideals. Management in groundwater has to be pragmatic because timescales in the case of groundwater vary, on the one hand between the resource itself, which has an inbuilt lag, and requires a long-term perspective, and groundwater users themselves who often operate on a much shorter timeframe, normally driven by economic development. These two (long and short-term timescales) somehow have to be synchronized. The stabilization of the North China plain aquifer will be a long-term process when one considers that in 1988 the Hai river basin exceeded recharge by some 8800 Mm3 yr1, and an average recharge deficit of 40–90 mm yr1 (Foster et al., 2004). As discussed earlier in the chapter, at times it might be rational to use an aquifer intensively due to the associated socioeconomic and generational changes, which in time might decouple livelihoods from groundwater dependence, for example, through education or a more diversified economy. Issues of spatial scale and fit as raised earlier are relevant since often human institutions do not match groundwater boundaries. In the case of groundwater, in many cases, it is the individual, micro-level which drives groundwater use as the aggregate demand of thousands of groundwater users. To this, one has to add the high levels of uncertainty, which are an integral part of groundwater management, possibly magnified due to climate variability and change. For groundwater institutions, these limitations are inherent rather than situational (Moench, 2007).

1.07.7 Conclusion Suggestions or recommendations to achieve sustainable and ethical groundwater management have been presented in many conferences. The Alicante Declaration (Ragone and Llamas, 2006) is one of the recent ones. We include here what we consider relevant aspects: 1. There is no doubt that agriculture is the main blue and green water consumer. The virtual water-trade analysis

Groundwater Management

2.

3.

4.

5.

6.

7.

seems to show that in many countries the motto ‘more crops and jobs per drop’ is changing to ‘more cash and nature per drop’. Usually hydrological and economic productivity of groundwater irrigation is significantly greater than the corresponding productivities of surface-water irrigation. Detailed analyses on this aspect in different climates and countries should be done to confirm or reject these preliminary data. The spectacular increase in groundwater use that has occurred in the last five or six decades can be classified as a silent revolution because it has taken place with scarce planning and control by governmental agencies. This is the main cause of some observed negative impacts, which could be avoided or mitigated with adequate groundwater management. It is extremely difficult to provide a general guide to good groundwater management, as complying with all the different dimensions may not be possible in most cases. Emphasis on one or another is likely to depend on economic, social, cultural, and political constraints. Groundwater management requires a higher degree of user involvement than surface-water developments. Experience shows that sustainable aquifer use cannot be solely achieved by means of top-down control-and-command measures. User participation requires a degree of hydrogeological education which is still absent in most places. Steps should be taken to make the peculiarities of groundwater resources known to all, from politicians and water decision makers to direct users and the general public. This should begin at the school level. Appropriate groundwater management requires a significant degree of trust among stakeholders. This implies that groundwater data should be transparent and widely available (e.g., via the Internet). In addition, the system should be able to punish those who act against the general interest.

References Acharya G and Barbier EB (2000) Valuing groundwater recharge through the agricultural production in the Hadejia-Nguru wetlands in northern Nigeria. Agricultural Economics 22: 247--259. Aldaya MM, Llamas MR, Varela-Oretga C, Novo P, and Rodriguez- R (2009) Challenging the conventional paradigm of water scarcity through the water footprint: The Spanish example. In: Hoekstra A (ed.) Global Water Governance. London: Earthscan. Bates BC, Kundzewicz ZW, Wu S, and Palutikof JP (2008) Climate Change and Water, Technical Paper of the Intergovernmental Panel on Climate Change, 210pp. IPCC Secretariat, Geneva. Bergkamp G and Cross K (2007) Groundwater and ecosystem services: Options for their sustainable use. In: Ragone S, de la Hera A, Hernandez-Mora N, Bergkamp G, and McKay J (eds.) Global Importance of Groundwater in the 21st Century: The International Symposium in Groundwater Sustainability, pp. 233–246. Alicante, Spain, 24–27 January 2006. Westerville, OH: National Groundwater Association Press. Briscoe J (2005) India’s Water Economy: Bracing for a Turbulent Future. Washington, DC: World Bank. Brown GM (2000) Renewable natural resource management and use without markets. Journal of Economic Literature 38(4): 876--915.

125

Calera A, Medrano J, Vela A, and Castan˜o S (1999) GIS tools applied to the sustainable management of water resources. Application to the aquifer system 0829. Agricultural Water Management 40: 207--220. Caponera D and Nanni M (2007) Principles of Water Law and Administration, National and International, 2nd edn. London: Taylor and Francis. Chevalking S, Knoop L, and Van Steenbergen F (2008) Ideas for Groundwater Management. Wageningen, The Netherlands: MetaMeta and IUCN. Cleaver F and Franks T (2005) How Institutions Elude Design: River Basin Management and Sustainable Livelihoods. BCID research paper 12. http:// www.bradford.ac.uk/acad/bcid/research/papers/ResearchPaper12CleaverFranks.pdf (accessed March 2010). Cleaver F and Franks T (2008) Distilling or diluting? Negotiating the water researchpolicy interface. Water Alternatives 1(1): 157--177. Collin JJ and Margat J (1993) Overexploitation of water resources: Overreaction or an economic reality? Hydroplus 36: 26--37. Costanza R and de Groot R (1997) The value of the world’s ecosystem services and natural capital. Nature 387: 253--260. Custodio E (2002) Aquifer overexploitation: What does it mean? Hydrogeology Journal 10(2): 254--277. Delli Priscoli J, Dooge J, and Llamas MR (2004) Water and Ethics: Overview. Essay 1, 31pp. Paris: UNESCO. Eckstein G (2004) Protecting a hidden treasure: The U.N. International law commission and the international law of transboundary ground water resources. Sustainable Development Law and Policy Winter): 5--11. Embid Irujo A (ed.) (2002) El Derecho de Aguas en Iberoame´rica y Espan˜a: cambio y modernizacio´n en el inicio del tercer milenio, vols. I and II, Madrid: Civitas. Escolero O and Martinez S (2007) The Mexican experience with groundwater management. In: Ragone S, de la Hera A, Hernandez-Mora N, Bergkamp G, and McKay J (eds.) Global Importance of Groundwater in the 21st Century: The International Symposium in Groundwater Sustainability, pp. 97–104, 233–246. Alicante, Spain, 24–27 January 2006. Westerville, OH: National Groundwater Association Press. Forne´s JM, De la Hera A, and Llamas MR (2005) The silent revolution in groundwater intensive use and its influence in Spain. Water Policy 7(3): 253--268. Foster S, Gardun˜o H, Evans R, Olson D, Zhang W, and Han Z (2004) Quaternary aquifer of the north China Plain – assessing and achieving groundwater resource sustainability. Hydrogeology Journal 12: 81--93. Gardun˜o H (2003) Administracio´n de derechos de agua. Experiencias, asuntos relevantes y lineamientos. FAO Legislative Study, Paper 81. Garrido A and Llamas MR (2009) Water management in Spain: An example of changing paradigms. In: Dinar A and Albiac J (eds.) Policy and Strategic Behavior in Water Resource Management, pp. 125--146. London: Earthscan. Garrido A, Martı´nez-Santos P, and Llamas MR (2006) Groundwater irrigation and its implications for water policy in semiarid countries: The Spanish experience. Hydrogeology Journal 14: 340--349. Gemma M and Tsur Y (2007) The stabilization value of groundwater and conjunctive water management under uncertainty. Review of Agricultural Economics 29(3): 540--548. Green T, Taniguchi M, and Kooi H (2007) Potential impacts of climate change and human activity on subsurface water resources. Vadoze zone Journal 6(3): 531--532. Hardin G (1968) The tragedy of the commons. Science 162(3859): 1243--1248. Hellegers P and Van Ierland E (2003) Policy instruments for groundwater management in the Netherlands. Environmental and Resource Economics 26(1): 163--172. Herna´ndez-Mora N, Llamas MR, and Martı´nez-Cortina L (2001) Misconceptions in aquifer over-exploitation. Implications for water policy in southern Europe. In: Dosi C (ed.) Agricultural Use of Groundwater. Towards Integration between Agricultural Policy and Water Resources Management, pp. 107--125. Dordrecht: Kluwer. Herna´ndez-Mora N, Martı´nez-Cortina L, and Forne´s J (2003) Intensive groundwater use in Spain. In: Llamas R and Custodio E (eds.) Intensive Use of Groundwater: Challenges and Opportunities, pp. 387--414. Lisse, The Netherlands: Balkema, Swets and Zeitlinger. Hodgson S (2006) Modern Water Rights. Theory and Practice. FAO Legislative Study, Paper 92. Howe C (2002) Policy issues and institutional impediments in the management of groundwater: Lessons from case studies. Environment and Development Economics 7: 625--641. Koundouri P (2004) Current issues in the economics of groundwater resource management. Journal of Economic Surveys 18(5): 703--740. Llamas MR (2004) Use of Groundwater. Series on Water and Ethics, Essay 7, 33pp (ISBN 92-9220-022-4). Paris: UNESCO. Llamas MR and Custodio E (eds.) (2003) Intensive Use of Groundwater: Challenges and Opportunities, p. 478pp. Dordrecht: Balkema.

126

Groundwater Management

Llamas MR and Garrido A (2007) Lessons from intensive groundwater use in Spain: Economic and social benefits and conflicts. In: Giordano M and Villholth KG (eds.) The Agricultural Groundwater Revolution: Opportunities and Threat to Development, pp. 266--295. Wallingford: CABI. Llamas MR and Martı´nez-Cortina L (2009) Specific aspects of groundwater use in water ethics. In: Llamas MR, Martı´nez-Cortina L, and Mukherji A. (eds.), pp.187– 204. Leiden: CRC Press/Balkema. (ISBN 978-0-415-47303-3). Llamas MR, Martı´nez-Santos P, and de la Hera A (2007) Dimensions of sustainability in regard to groundwater resources development: An overview. In: Ragone S, de la Hera A, Hernandez-Mora N, Bergkamp G, and McKay J (eds.) Global Importance of Groundwater in the 21st Century: The International Symposium in Groundwater Sustainability. Alicante, Spain, 24–27 January 2006. Westerville, OH: National Groundwater Association Press. Llamas MR, Martı´nez-Santos P, and de la Hera A (2008) Hydropolitics and hydroeconomics of shared groundwater resources: Experience in arid and semiarid regions. Paper presented in the conference of the NATO Advanced Study Workshop. Varna, Bulgaria, 1–12 October 2006. In: Darnault C (ed.) Overexploitation and Contamination of Shared Groundwater Resources, pp. 415– 431. Dordrecht: Springer. Llamas MR, Shah T, and Mukherji A (2006) Guest Editors’ preface. Hydrogeology Journal 14(3): 269--274. Loaiciga H (2003) Climate change and groundwater. Annals of the Association of American Geographers 93(1): 30--41. Lopez-Gunn E (2003) The role of collective action in water governance, a comparative study of groundwater user association in La Mancha aquifers, Spain. Hydrogeology Journal 28(3): 367--378. Lopez-Gunn E (2009) Governing shared groundwater: The controversy over private regulation. Geographical Journal, 175(1): 39–51. doi: 10.1111/j.14754959.2008.00313.x. Lopez-Gunn E and Llamas MR (2008) Re-thinking water scarcity: Can science and technology solve the global water crisis? Natural Resources Forum 32: 228--238. Margat J (2008) Les Eaux Souterraines dans le Monde. 178pp. Orleans: BRGM e´ditions. Martı´nez-Cortina L (2001) Estimacio´n de la recarga en grandes cuencas sedimentarias mediante modelos nume´ricos de flujo subterra´neo. Aplicacio´n a la cuenca alta del Guadiana (Recharge Estimation in Large Sedimentary Basins Using Groundwater Flow Models. The Case of the Upper Guadiana Basin), 418pp. PhD Thesis, University of Cantabria, Spain. Melgarejo J and Molina A (eds.) (2005) Los mercados del agua. Ana´lisis jurı´dicos y econo´micos de los contratos de cesio´n y bancos de agua. Madrid: Civitas. Merrey D, Meinzen-Dick R, Mollinga P, and Karar M (2007) Policy and institutional reform: The art of the possible. In: Molden D (ed.) Water for Food, Water for Life: Comprehensive Assessment of Water Management in Agriculture. London: Earthscan. Moench M (2003) Groundwater and poverty: Exploring the connections. In: Llamas MR and Custodio E (eds.) Intensive Use of Groundwater: Challenges and Opportunities, pp. 441--456. Lisse, The Netherlands: Swets and Zeitlinger. Moench M (2007) When the wells run dry but livelihoods continue: Adaptive responses to groundwater depletion and strategies for mitigating the associated impacts. In: Giordano M and Villholth KG (eds.) The Agricultural Groundwater Revolution: Opportunities and Threats for Development, pp. 173--194. Wallingford: CABI. Molle F and Berkoff J (2006) Cities versus agriculture: Revisiting intersectoral water transfers, potential gains, and conflicts. Comprehensive Assessment of Water Management in Agriculture Research Report 10. Colombo: International Water Management Institute. Mukherji A (2006) Is intensive use of groundwater a solution to world’s water crisis? In: Rogers PP, Llamas MR, and Martı´nez-Cortina L (eds.) Water Crisis: Myth or Reality? Marcelino Botin Water Forum 2004, pp. 181–193. London, UK: Balkema/ Taylor and Francis Group. National Research Council (2004) Valuing Ecosystem Services. Washington, DC: National Academies Press. Neher PA (1990) Natural Resource Economics: Conservation and Exploitation. New York: Cambridge University Press. Ostrom E (1990) Governing the Commons: The Evolution of Self-Governing Irrigation Systems. Cambridge: Cambridge University Press. Peck JC (2007) Groundwater management in the high plains aquifer in the USA: Legal problems and innovations. In: Giordano M and Villholth KG (eds.) The Agricultural Groundwater Revolution: Opportunities and Threats for Development, pp. 296--319. Wallingford: CABI. Pongkijvorasin SP and Roumasset J (2007) Optimal conjunctive use of surface and groundwater with recharge and return flows: Dynamic and spatial pattern. Review of Agricultural Economics 29(3): 531--539.

Ragone SE and Llamas MR (2006) The alicante declaration: Steps along the pathway to a sustainable future. Ground Water Reader’s Forum 44(4): 500--503. Rijsberman F (2004) Sanitation and access to water. In: Lomborg B (ed.) Global Crises, Global Solutions, pp. 498--527. New York: Cambridge University Press. Rosegrant M, Cai X, and Cline S (2002) World Water and Food to 2025. Washington, DC: IFPRI. Rubio SJ and Fisher AC (1997) Adjusting to climate change: Implications of increased variability and asymmetric adjustment costs for investment in water reserves. Journal of Environmental Economics and Management 34: 207--227. Saleth RM (1996) Water Institutions in India: Economics, Law and Policy. New Delhi: Commonwealth. Selborne J (2001) The ethics of freshwater use: A survey. Report of the Commission on the Ethics of Science and Technology (COMEST), 62pp. Paris, France: UNESCO. Schlager E and Lopez-Gunn E (2006) Collective systems for water management: Is the tragedy of the commons a myth? In: Rogers P, Llamas MR, and Martı´nez-Cortina L (eds.) Water Crisis: Myth or Reality? pp. 43--60. London: Taylor and Francis. Shah T (2005) The new institutional economics of India’s water policy. In: International Workshop on African Water Laws: Plural Legislative Frameworks for Rural Water Management in Africa0 . Johannesburg, South Africa, 26–28 January 2005. Shah T, Burke J, Villholth K et al. (2007) Groundwater: A global assessment of scale and significance. In: Molden D (ed.) Water for Food, Water for Life: Comprehensive Assessment of Water Management in Agriculture, pp. 395–423. London: Earthscan. Stalgren P (2006) Corruption in the Water Sector: Causes, Consequences and Potential Reforms, Swedish Water House Policy Brief No. 4. SIWI. Transparency International (2008) Global Corruption Report 2008: Corruption in the Water Sector. Cambridge: Cambridge University Press. Tsur Y and Parker D (1997) Decentralization and Coordination of Water Resource Management. Boston: Kluwer. UNECE (1998) Convention on access to information, public participation in decisionmaking and access to justice in environmental matters. Aarhus, Denmark, 25 June 1998. http://www.unitar.org/egp/sites/default/files/aarhus_convention.pdf (accessed March 2010). Wagner M (2005) Wildlife and Water: Collective Action and Social Capital of Selected Landowner Associations in Texas, 150pp. PhD Thesis, Texas A&M University. Wegerich K (2006) Groundwater institutions and management problems in the developing world. In: Tellam JH, Rivett MO, and Israfilov RG (eds.) Urban Groundwater Management and Sustainability. Dordrecht: Springer. WHO (World Health Organization) (2001) Arsenic in Groundwater – Factsheet. http:// www.who.int/mediacentre/factsheets/fs210/en/print.html. (accessed March 2010).

Relevant Websites http://www.acaciawater.com Acacia Water. http://www.aueas.org Asociacio´n Espan˜ola de Usuarios de Aguas Subterra´neas. http://www.whymap.org BGR: UNESCO; World-Wide Hydrogeological Mapping and Assessment Program (WHYMAP). http://www.connectedwater.gov.au Connected Water: Managing the Linkages between Surface Water and Ground Water. http://www.ecolex.org ECOLEX: The Gateway to Environmental Law. http://www.empowers.info EMPOWERS Thematic Group: Advancing participation and dialogue in local water governance in the MENA Region. http://www.epa.gov EPA: United States Environmental Protection Agency; Groundwater and Drinking Water. http://www.ewater.eu eWater Portal. http://www.water.nstl.gov.cn GWRTAC: Ground-Water Remediation Technologies Analysis Center. http://www.waterandfood.org Groundwater Governance in Asia: Water is Divine. http://www.sg-guarani.org Guarani Aquifer System.

Groundwater Management

http://www.indiawaterportal.org India WaterPortal. http://www.igrac.nl International Groundwater Resources Assessment Centre. http://www.inpim.org International Network on Participatory Irrigation Management. http://www.waterlaw.org International Water Law Project: Addressing the future of water law and policy in the 21st century. http://www.isarm.net isarm Internationally Shared Aquifer Resources Management; Transboundary Aquifers. http://aguas.igme.es No se encuentra la pa´gina. http://www.groundwatermanagement.org Participatory Groundwater Management. http://www.ploppy.net Ploppy (educational material for children (in Spanish) on groundwater).

http://www.projectwet.org ProjectWET; Worldwide Water Education. http://www.groundwater.org The Groundwater Foundation. http://www.worldbank.org The World Bank; GW-MATE. http://www.iah.org The WorldWide Groundwater Organisation: International Association of Hydrogeologists. http://www.water-ed.org Water Education Foundation. http://www.wfdvisual.com WFDVisual: Water Framework Directive Visualisation Package. http://www.groundwateruk.org UK Groundwater Forum: Raising Awareness of Groundwater.

127

1.08 Managing Agricultural Water J Ramirez-Vallejo, Universidad de los Andes, Bogota, Columbia & 2011 Elsevier B.V. All rights reserved.

1.08.1 1.08.1.1 1.08.1.2 1.08.1.3 1.08.1.3.1 1.08.1.3.2 1.08.1.3.3 1.08.1.3.4 1.08.1.3.5 1.08.1.3.6 1.08.1.3.7 1.08.1.3.8 1.08.1.4 1.08.1.4.1 1.08.1.4.2 1.08.1.4.3 1.08.1.4.4 1.08.1.4.5 1.08.2 1.08.2.1 1.08.2.2 1.08.2.2.1 1.08.2.2.2 1.08.2.2.3 1.08.2.2.4 1.08.2.3 1.08.3 1.08.3.1 1.08.3.2 1.08.3.3 1.08.4 1.08.4.1 1.08.4.2 1.08.4.3 1.08.4.3.1 1.08.4.3.2 1.08.4.4 1.08.4.4.1 1.08.5 1.08.6 1.08.7 1.08.8 1.08.9 References

Introduction and Overview Trends in Water Management for Agriculture Water Scarcity: Is It a Demand or a Supply Problem? Challenges Facing Agricultural Water Management The policy and institutional challenge The economic and financial challenge The problem of declining investment The challenge of technology and water resources to supply growing demand Poor performance of public managed irrigation systems The neglect of environmental impacts of agricultural water management Neglect of water management for rainfed agriculture The poverty and rural incomes challenge Potential/Promise of Science and Technology Advances Hydro-climatic forecast prediction Drought-tolerant crops Remote-sensing technology on the estimate of crop ET Cost-effective irrigation systems Improving irrigation water management Water Productivity in Agriculture Economic Value of Water for Agriculture Example of Estimates of the Economic Value of Water: The Case of Mexican Agriculture Indirect method Residual method Water markets Math programming method Agricultural Trade Protection Water Management and Competitiveness Framework Farmers, Water, and the Process of Economic Development Water Resource Management and the Regional Economic Strategy Water Resource Management, Institutions, and Implementation Integrated Water Resource Management Participatory Irrigation Management Lessons Learned from Participatory Management The Philippines Mexico Institutions and Water Governance Water management principles Water Management and the Environment Water for Agriculture and Poverty Reduction Water Management of Rainfed Agriculture Policy Actions for the Future Summary

1.08.1 Introduction and Overview Managing water for agriculture is a topic that covers broad dimensions in the development of countries, such as management of water for rainfed agriculture, irrigated agriculture, water-use recycling, conservation of water and land, and watershed management. It is a collection of activities that

129 129 130 132 132 132 132 132 132 133 133 133 133 133 133 134 134 134 134 135 135 135 136 136 136 136 137 137 139 139 140 140 140 141 141 142 142 143 144 145 146 147 148 149

lies between four areas of economic development: rural development, agriculture, water supply, and environmental management. Water management has achieved an impressive record during the past century but there is a larger challenge ahead. Irrigation is responsible for approximately 75% of water demand in developing countries and low-cost waters are already

129

130

Managing Agricultural Water

a scarce resource. Only 17% of all cropland is irrigated, but provides 30–40% of the world’s food production, and over 60% of the world’s irrigated area is in Asia, mostly devoted to the production of rice (IUCN and WBCSD, 2008). Water needs to be more productive as an input to generate a more dynamic development of rural areas, to supply the food for the increasing demand in the world, and to leave enough water for other key uses such as domestic and industrial uses. At the same time, water use in agriculture needs to be friendly with the environment and society in general. A new context for water management in agriculture has been generated by the changing national and global trends. The high rates of population growth, the dynamic economies of countries such as China, India, and East Europe that bring each year millions of people out of poverty with increasing levels of disposable income to spend on higher value food, and the stronger desire of the world to live in an environmentally safe planet are just some of the trends that delineate the context in which water will need to be managed for agriculture. Water connects all ecosystems across the landscape and the competition for its use is becoming increasingly intense over time. In many locations, water supply is limited compared to its demand for various uses, and given that agriculture is by far the largest consumer of water, the future effective supply will depend on the productivity of water in agriculture. This water productivity is a function of the management, innovation, and governance of the water resource systems. More efficient irrigation systems, more precision agriculture, and the upgrading of rainfed systems and wastewater management will generate a better social and economic water use when agricultural production is increased. The above scenario suggests the importance of dealing with water management in agriculture as a way to optimize its use for society. This chapter first presents the context and main trends that will influence water management in agriculture. It presents the main challenges facing agricultural water management, and its impact on poverty reduction. Subsequently, two related sections present the link between water management for agriculture and competitiveness as a framework to explore water productivity in agriculture, with emphasis on the value of water for irrigation. Topics such as water management trends, water management in rainfed agriculture, and the impact of water management on the environment are also covered. Finally, governance and institutional aspects of water management are delineated.

1.08.1.1 Trends in Water Management for Agriculture According to the United Nations (2008), it is estimated that by year 2050 the world population can reach 8.9 billion, up by 47% of today’s population, and will place a significant pressure to world agriculture to satisfy food demand. Global agriculture must double in the next 30 years to sustain the population, and given the constraint on land expansion, current agriculture will need to become more efficient to meet this demand. Given this reality, the obvious question is how the world is going to use the available water to produce enough food to feed the world population in the future without compromising the environment and with a minimum

external cost for society? Also, given that it is expected that the urban population will exceed the rural population by a significant margin, how can water be more productive to deal with a shrinking rural population that needs to feed increasing urban population as well as themselves? Although the share of world’s workers who are employed in agriculture has decreased from 42% in 1996 to 36% in 2006 (International Labour Organization, 2007), this sector will remain to be an important source of occupation for many people, especially those from developing countries. Water will remain with an important role of increasing the well-being of many rural workers and their families, directly through consumption of the fruits of their labor and by increasing their income. Consumers will define the function of water management in agriculture. As income increases so does the demand for food, not only in terms of quantity but also in the composition of the traditional food intake. It has been shown that wealthier consumers demand more animal protein, such as meat and milk products, something that requires more land to be used for agriculture and livestock production. As an example, in the recent evolution of diet in China, meat consumption has more than doubled in the last 20 years and it is projected to double again in 2030 (Centre for World Food Studies, 2008). Another observed trend is the concern of the consumer about food safety where traceability becomes important and water becomes an attribute of agricultural products when consumers take the final purchasing decision. In addition, people in developed countries will demand more fruits and vegetables, a trend that will induce the production of fewer calories per hectare, reducing field crops such as cereals and producing more vegetables and fruits that require more water. Food production to satisfy a person’s daily dietary takes about 3 m3 of water, a little more than 1 l calorie1 (IUCN and WBCSD, 2008). Another trend that will transform the composition of food intake is the change in consumer taste and expectations when people place more emphasis on doing their best with the limited time at their disposal. People are eating healthier convenient food such as fresh-cut snacks, a market that has grown from 8.8 billion in 2003 to 10.5 billion by 2004, according to the International Fresh-Cut Produce Association. These trends are becoming stronger over time. Water plays then an active role in the virtuous cycle of development. Water is needed to generate high-value products and income to rural people that will demand more agricultural products. In addition, more sophisticated products will create an incentive for innovation processes. The pressure to increase water productivity in agricultural ecosystems will induce water to be used in a more friendly manner with numerous species, such as microorganisms in the soil and pests and predators that are part of the agricultural biodiversity needed to have plant stability and sustainable crop production. Water is then a necessary input to implement management systems such as integrated crop management and conservation agriculture. Water pollution is increasing and the sources of water are being affected by the augmentation in water consumption mainly for agricultural production. Irrigation water has

Managing Agricultural Water

generated important increases in income in many parts of the world with a clear impact on poverty but with negative effects on the environment. The problems of soil erosion, salinization, intrusion of seawater, and pollution, in general, are limiting the effective water availability for agriculture and other uses. Most of these processes are real consequences of badly managed water resources in settings with institutions that fail or are slow to change processes. Land and water productivity had raised agronomic yields from 1.4 tons per hectare to 2.7 tons per hectare over the past four decades. However, according to the IWMI (2007), the number of malnourished people is above 850 million (IWMI, International Water Management Institute), and the average daily per capita food supply in some regions of the world, such as South Asia and sub-Saharan Africa, is rising too slowly to generate a significant impact on poverty alleviation. These levels are far below from the ones observed in industrialized countries. In addition, it is also estimated that a third of what it is produced in agricultural products is lost before it is consumed. Another trend that is set to play an important role in future water management in agriculture is the world balance of energy. It is estimated that alternative fuels will provide 5% of the United States’ energy by 2020, up from 1% today, and something similar will be experienced by other developed and developing countries. Bioenergy, currently made largely from sugar cane and from corn, will increase the demand for water and generate water-associated contamination. More certification standards and enabling regulatory and policy frameworks will most likely be set to obtain sustainable practice of this economic activity. Moreover, competition for land use to provide food and fiber will affect food prices and impact lowincome consumers. There is still some debate regarding the extent to which climate change will impact agricultural productivity at the global level (World Bank, 2008). Climate change is affecting precipitation patterns and temperatures, and it is estimated that the areas of the world that are poorer will be the most adversely affected in its impact on water supplies. Groundwater levels are declining in many areas of the world, such as North Africa, North China, India, and Mexico, because of overexploitation, and there is not enough water to meet all demands of the different uses. Some estimates show that if nothing is done, climate change by the 2080s would have an impact in agricultural production capacity by about 16% if carbon fertilization is omitted and by about 3% if it is included. Some parts of the world will be more affected than others, particularly India and a number of countries in Africa will be impacted. Finally, the reconfiguration of global value chains that redefine the role of location on generating advantages to companies is a trend that will modify the geography of agricultural output and production of goods and services that are used as inputs in agriculture.

1.08.1.2 Water Scarcity: Is It a Demand or a Supply Problem? When confronting the future, the relevant question is whether the relative water scarcity is a demand or a supply challenge. Some authors are optimists and some others are pessimists. The optimists, for example, Allan (2001), the World Bank

131

(2003), IFPRI (1995, 1997), Dyson (1996), Rosegrant et al. (1995), Rosegrant and Cai (2002), and Brichieri-Colombi (2004, Who Speaks for the River; PhD Thesis, University of London; unpublished) think there is sufficient freshwater and soil water in the world to meet current and future water needs if we accept the demographic predictions. On the other hand, we find pessimists such as Postel (1999) and Postel and Richter (2003) who argue that there are not enough water supplies to meet the estimated demand, given the population and income projections. For them, the prospect looks darker when considering the water-scarce Middle East and the heavily populated South Asia and China. For some, the difficulty can be described as a technology and economic problem to make water available where it is needed for distinct uses. In other words, there is plenty of water in the world – the issue is how to induce the right costefficient technology to obtain the desired quality of water available to transport it to where it is needed. Theoretically, the existing pressure on freshwater use will generate the right incentive to innovate new ways of obtaining water, such as (1) the production of freshwater by desalination of brackish or saltwater (mostly for domestic purposes), and (2) the reuse of urban or industrial wastewaters (with or without treatment), which increases the overall efficiency of use of water (extracted from primary sources), mostly in agriculture, but increasingly in the industrial and domestic sectors. This category also includes agricultural drainage water. However, these technologies are still high-cost options for most developing countries, a fact that leads to reframe the water scarcity problem as a demand problem – whether economic resources are available at the country, region, company, and farm levels to pay for the cost of these technologies. The two major cost items are treatment and transportation to users. Most likely, in a scenario with increasing water demand accompanied by scarcity of local water supply, the induced innovation hypothesis will take place (Hayami and Ruttan, 1985), in which technical and institutional changes will be induced through the responses of farmers, agribusiness entrepreneurs, scientists, and public administrators to water endowments and to changes in the supply and demand of water and other inputs. Based on this, the state of relative endowments and the accumulation of the primary resources of land, labor, and water, would be critical elements in determining the pattern of technical change that will occur in water resources and agriculture in the future. Technical change embodied in new and more productive inputs may be induced primarily to save labor, to save land, or to save water. However, this works in the right direction if the agricultural inputs are priced according to their opportunity cost (Ramirez-Vallejo and Rogers, 2004). If subsidies exist, which is the case in the majority of developing and developed countries, water will remain undervalued by society and future changes in technology will not reflect the reality of water scarcity. A trade-related concept has been proposed to solve the water-supply–water-demand correspondence. The trade-in embodied water or virtual water concept comes from the idea that water should be treated as a production factor and is equivalent to the volume of water needed to produce a commodity or service. The virtual water argument, on the other hand, shows that the importation of agricultural products

132

Managing Agricultural Water

that require significant amounts of water represents the importation of water into a water-scarce country. According to Allan (2004), ‘‘trade in commodities has already achieved a pivotal position in enabling the comparative advantages and disadvantages of regional water endowments to be balanced.’’ The concept of trade-in embodied water can be used to study the linkages that exist between trade reform and water resources development. The trade-in embodied factor or the factor content of trade approach was first employed by Leontief (1953) in his well-known test of the Heckscher– Ohlin (H–O) theorem and later formalized theoretically by Travis (1964), Melvin (1968), Deardorff (1982), and many others. However, there are only few studies in the literature that discuss trade-in embodied environmental factors. This approach has been extended to analyze the linkage between trade, food security, and water resources by using the virtual water concept and argument. The concept originates from the idea that water should be treated as a production factor and virtual water is the volume of water needed to produce a commodity or a service. In other words, countries sometimes do not have a surplus amount of water to produce water-intensive agricultural commodities such as rice, cotton, and tropical and subtropical fruits such as bananas and oranges, and use trade to compensate for this shortage. Food trade then becomes an instrument to augment water supplies on the scale needed to meet the domestic food demand. This concept has become popular since Allan (1996) used it first. An example of this concept is that it takes approximately 1500 m3 of water to grow 1 ton of grain. Equivalent figures can be estimated for virtually any traded agricultural product; however, these figures can only be used as an approximate number given that the real demand and supply of water to an individual crop depend on a large number of variables (i.e., the geographic location of the region where the crop is grown, the irrigation system used, the management of the irrigation system, soil type, climatic and socioeconomic conditions, etc.). The concept of virtual water then becomes a useful indicator of future water resources use. The virtual water argument can be also seen as an application of the well-known H–O theorem. The assumptions of this model are that factors can move without cost among industries within a country, but are completely immobile internationally, which might be the case of water as a production factor. As such, production functions for all countries (products) exhibit constant return to scale and each country has the same productive technology for each good, and consumers have the same utility functions. Under these conditions, the theorem states that a country exports products that use intensively a relatively abundant input. However, the level of protection to agricultural products via tariffs and duties and nontariff instruments by all countries, developed and developing, has distorted the virtual water movement worldwide. Ramirez-Vallejo and Rogers (2004) showed that, for example, using International food Policy Research Institute (IFPRI)’s IMPACT model results, a scenario of full liberalization of agriculture compared to a baseline scenario would have a significant net effect of virtual water flows from the relocation of meat and cereals trade. When the net effect of the meat and cereals markets are added together,

the two major contributors to the increase in virtual water trade would be the United States, which would increase its annual virtual water exports in about 86 km3, and Latin America would have a similar increase of 89 km3. These become the two water surplus regions in the world. The major changes in virtual water imports would occur in Asia in general (South Asia, Southeast Asia, and East Asia) with an increase of 112 km3, sub-Saharan Africa with an increase of almost 40 km3, and the former Soviet Union with an increase in water imports of 22 km3, mostly because of an increase in meat imports. West Asia and North Africa together, on the other hand, would decrease the level of virtual water imports to about 7 km3, but would remain as an important net importer of virtual water of about 176 km3. Ramirez-Vallejo and Rogers (2009) showed that the concept of virtual water is useful to educate public officials and society in general that water in some parts of the world is a scarce resource and that agriculture uses the great majority of water resources available on earth. The argument also has an implicit lesson underscoring the importance of running efficiently irrigation districts so that water could be allocated to other uses, including ones benefiting the environment.

1.08.1.3 Challenges Facing Agricultural Water Management Under the described trend scenarios, agricultural water management faces many challenges, especially as a result of its strong links with many global and economic issues. Agricultural management is not a unique goal by itself but it is part of a process to manage one of the most important inputs to income generation. According to the World Bank (2006), some of the challenges facing agricultural water management include the policy and institutional challenge; the economic and financial challenge; the problem of declining investment; the challenge of technology and water resources to supply growing demand; the poverty and rural incomes challenge; and the environmental dimension and sustainability imperative.

1.08.1.3.1 The policy and institutional challenge The most difficult challenge has been how to conciliate agricultural and macroeconomic and social policy. Countries want low-cost food products for the population and at the same time to improve farmer’s income levels. Low-cost products for local consumers imply low trade barriers and higher competition from international competitors, and lower support levels to agriculture for a food security policy. Generating incentives for agricultural development through water management for agriculture has its direct trade-offs with specific macroeconomic goals. The incentives for agricultural water management need to be integrated with agricultural policy and new institutional capacity needs to be built for a better water allocation and priority setting at the basin, regional, and national levels. New management skills and a better understanding of the political economy of reforms need to be developed.

1.08.1.3.2 The economic and financial challenge Use of water in agriculture generates the lowest value-add compared to other uses such as municipal, domestic, and industrial uses. When water is assigned to the higher value use, agriculture is usually ranked lower and other benefits different

Managing Agricultural Water

from direct economic and financial need to be added to justify allocation. Water scarcity increases and competition creates the incentive to improve the returns on water. The economic challenge is how to build an incentive system to encourage efficient water use and profitable agriculture of high value. A broader economic challenge deals with the generation of an incentive framework for all types of investment and to promote environmentally responsible use. On the financial side, the challenge is to achieve cost recovery in traditional agricultural systems that are characterized by high subsidies, selfsufficiency goals, and low-productivity agriculture and to create the investment environment under the existing distorted incentive frameworks.

1.08.1.3.3 The problem of declining investment The global irrigated area doubled from 1960 to 2000, developing faster at the beginning of the period and slowing down in later years (Cleaver and Gonzalez, 2003; Winpenny, 2003). The world experienced a decrease in the construction of dams for the development of surface irrigation, which has been compensated for by the growth of groundwater irrigation. According to the World Bank (2005), governments are investing less in agriculture worldwide; public investment in agriculture has dropped, and investment in irrigation, drainage, and other agricultural water management projects has also been declining worldwide. The situation for groundwater irrigation is delicate. Most of the past investment has been predominantly private and done in an unregulated environment ending up in overexploiting groundwater resources. Another area that demands significant resources for investment is drainage. Unfortunately, it is difficult to provide the right investment incentives for drainage infrastructure because usually the benefits are underestimated and the cost recovery is a difficult task.

1.08.1.3.4 The challenge of technology and water resources to supply growing demand Irrigated agriculture supplies close to 40% of the world’s food and occupies only 17% of the cultivated land. However, for the future, the Food and Agriculture Organization (FAO) estimates that by 2030, food production needs to grow at 1.4% a year, and about half of this growth would have to be generated from irrigated agriculture. Therefore, the main challenge has to do with the capacity of the agricultural sectors to meet this additional demand. Technological change has slowed down from the significant advances of the Green Revolution of the 1960s, and the water base is overexploited in many ecosystems, particularly in developing countries where institutions and regulations failed to control water exploitation. However, technology is available, but often is not disseminated and adopted. Cost incentives or profitable market opportunities generate the incentive for farmers to invest in technology. A study by the World Bank found that there is ‘‘more technology available than we know what to do with’’ (World Bank, 2005). Many innovations are available to improve productivity or conserve water but have not been adopted as it was expected, such as drip technology, for example, that has been adopted on less than 1% of irrigated lands worldwide with affordable costs.

133

1.08.1.3.5 Poor performance of public managed irrigation systems Publicly managed irrigation systems account for half the irrigated area in developing countries and their performance has been generally below their technical and economic potential. Water service is poor, unreliable, and untimely. Among the causes are the decline of agricultural prices and subsidies to basic crops that have hampered diversification to high-value crops. However, perhaps the major cause comes from poor institutions with weak organization structures, bureaucratic with inefficient top-down approaches to service. Deficient management has led to lower collection rates of operation and insufficient maintenance fees, which have led to higher needs for rehabilitation intervention, which creates a never-ending pervasive cycle. Scarce farmers’ participation in the management processes has been recognized as one of the main reasons for the lack of accountability and efficiency in the public institutions.

1.08.1.3.6 The neglect of environmental impacts of agricultural water management As a result of agriculture being by far the largest user of land and water resources, there is a long list of environmental costs and risks of using water for agriculture, including land degradation, salinization, and erosion; loss of environmental water flows; pollution; destruction of natural habitats and livelihoods through drainage of wetlands and through land expansion and deforestation; and waterborne disease. Drainage was neglected in the rapid expansion of irrigation and irrigated land has become waterlogged and salinized due to the rise of the water tables and accumulation of salts, becoming a constraint to productivity. Land degradation caused by agricultural water management practices and by lack of drainage is affecting some of the world’s most fertile basins and dams, and irrigation infrastructure modifies flows, affecting their seasonality and frequency of floods. In addition, irrigated agriculture is a source of pollution, as a result of the technology package promoted with the green revolution that included the important use of chemical inputs. However, the main challenge of the management of water in agriculture is to simultaneously conserve biodiversity and diminish any external impact of water use to increase production, to secure enough production to meet the increasing demand for food, and, finally, to improve the prosperity of the rural people all around the world. Management of water in agriculture needs not only further increase in the productivity of existing farmland, but also to do it simultaneously with a support of biodiversity and ecosystem services and efficiently manage the natural resources. It needs to foster healthy populations and help them realize their development potentials and to increase prosperity in general, by generating income and improving the livelihoods of the rural communities and increase the value of the agricultural products.

1.08.1.3.7 Neglect of water management for rainfed agriculture Rainfed areas are home to most of the poor habitants of the rural areas and account for 60% of the current agricultural output. These areas are characterized by having less use of

134

Managing Agricultural Water

available technology. The technological options for improved water management in rainfed agriculture are few, mainly because the incentives have not generated the relevant research results around the world to offer higher yielding varieties for water stress and environmentally sensitive conditions. As a result, improvements in water management have been limited in these areas.

1.08.1.3.8 The poverty and rural incomes challenge Agricultural growth is central to poverty reduction. Seventy percent of the world’s poor live in rural areas, and most of them are dependent on agriculture. Typically, the rural poor live on marginal lands or on drylands, with little or no access to controlled water sources. Their technological options for improved water management are limited, and they face high risks from rainfall variations. The poor are exceptionally vulnerable to drought, floods, effluent discharge, aquifer depletion, waterlogging, salinization, and water quality deterioration. Thus, the key agricultural water challenges for the poor are food security, risk mitigation, and income growth. One of the Millennium Development Goals (MDGs) – eradicate extreme poverty and hunger – can be achieved only if agriculture grows and can provide access to food for the poorest and most vulnerable. Improved management of available water thus has a critical role to play in poverty reduction and food security. Other MDGs such as gender equality, child nutrition, and market access also depend directly or indirectly on pro-poor agricultural growth and related management of scarce water.

1.08.1.4 Potential/Promise of Science and Technology Advances At the same time, there are challenges, some of which might be more optimistic concerning agricultural water management because of the opportunities provided by advances in science and technology. Some of them are the hydroclimatic forecast and prediction, which could lead to water saving and profit gain, remote-sensing technology on the estimate of crop evapotranspiration (ET), drought-tolerant or salt-tolerant crops, improvement of cost-effective irrigation systems, and best management practices, mainly for water quality. Some of these are explained below.

1.08.1.4.2 Drought-tolerant crops Drought conditions are a well-known restriction to crop production. Research in public and private institutions is producing promising results in terms of developing varieties that demand less water while maintaining a higher level of agronomic yield. Local plant biologists are working hard to develop new crop strains that will produce abundant food even when water is scarce. Canola, in Canada, for example, is highly sensitive to water stress, which is one of the main factors that limits crop yield (Wanna et al., 2009). Development of drought-tolerant canola was then considered an important and urgent mandate for the canola industry. A breakthrough in this area should increase yield and permit expansion of canola growth regions. However, to be an effective solution, the new strain must produce normal yields under nondrought conditions and produce greater yields than conventional varieties under stress conditions. Research efforts in order to create drought-tolerant crop plants have been intensive, but the results remain incremental. Monsanto and other agricultural companies have drought-tolerant soybeans, cotton, and other crops in the pipeline. Universities are also engaged in intensive research aimed at growing more food with less water. ‘‘A more rapid growth in rainfed yield and production could compensate for reduced investments in irrigation or reduced groundwater pumping to eliminate groundwater overdraft, but that achieving the required improvements in rainfed production would be a significant challenge’’ (Rosegrant et al., 2001).

1.08.1.4.3 Remote-sensing technology on the estimate of crop ET Spatiotemporal information on actual ET helps users to better understand evaporative depletion and to establish links between land use, water allocation, and water use. Satellite-based measurements, used in association with energy-balance models, can provide the spatial distribution of ET for these linkages and optimize the use of water for irrigation (Bastiaanssen et al., 2005). Remote-sensing and hydrological models are applied to irrigation projects to estimate the water balance to support water use and productivity analyses. A case study in the Yakima River basin (Washington State) demonstrates how ET from remote sensing can be used for evaluating water-conservation projects.

1.08.1.4.4 Cost-effective irrigation systems 1.08.1.4.1 Hydro-climatic forecast prediction Expert system techniques have been used in irrigated agriculture to optimize the use of water in districts and to maximize profits by increasing agronomic yields. In the Havana Lowlands region, Illinois, USA, for example, researchers incorporated different types of weather forecasts into irrigation scheduling for corn production. The results showed that if farmers just use the real-time soil moisture information and the empirical rules set, profits might increase by 16%; and over the five testing years, it was found that the proper use of the 7-day forecast could save irrigation water and increase crop yield in dry years. The study found that farmers at the study site sometimes applied more water than necessary (Bastiaanssen et al., 2005).

Water is also saved through innovation in irrigation systems. Cost-effective irrigation systems allow for a better allocation of water resources. One way to achieve this is through the reengineering of the irrigation systems, which consists of designing the most cost-effective answer to the redefined water service within the scheme. This is a solution that demands consideration of the spatial distribution of the effective demand for the water service and the spatial distribution of the physical infrastructure characteristics. Some of the people in rural areas have developed their own water system with cost-reducing adaptations. One of these is the development of a solar system, which requires low operation cost compared to an electrical system to drive the motors or engines that finally converts the electrical energy to

Managing Agricultural Water

mechanical energy, so as to pump the water to the required destination.

1.08.1.4.5 Improving irrigation water management Best management practices are also a way to save water in irrigation systems. Improved performance in irrigation water management can usually be achieved through rehabilitation, process improvement, which consists of intervening in the process without changing the rules of the water management, and modernization. The introduction of modern techniques is a process improvement, and modernization, which is a more complex intervention implying fundamental changes in the rules governing water resource management. It may include interventions in the physical infrastructure as well as in its management.

1.08.2 Water Productivity in Agriculture Water scarcity can be seen under a multidimensional framework of physical, economic, managerial, institutional, and political water scarcity (Molle and Mollinga, 2003): physical water scarcity in which water availability is limited by natural availability; economic water scarcity when human and financial resources constrain availability of water; managerial water scarcity where availability is constrained by management limitations; institutional water scarcity where water availability is constrained by institutional shortcomings; and political water scarcity where political forces bar people from accessing available water resources. These types of scarcity can occur concomitantly, increasing both the severity and the impacts of water scarcity. Molden et al. (2003) estimated that, by 2020, approximately 75% of the world’s population will live in areas experiencing physical or economic water scarcity, precisely where most of the poor and food-insecure people live. Water productivity is a concept that has meaning and use when water is a scarce resource. Higher water productivity becomes part of the answer to the challenges described above. However, productivity is a concept used frequently by economists, engineers, and even biologists, with some degree of confusion. The concept of water productivity relates to the desire to have a higher level of output, economic value-add, or just value for the aggregate production per unit of water used. In other words, it is the way of achieving more using scarce water resources. In order to better understand the concept of water productivity, it is necessary to review the definition of efficiency. Irrigation efficiency is defined as the ratio of water consumed to water supplied. Irrigation systems in the developing world typically function at a level of irrigation efficiency from 30% to 40% (Seckler et al., 2003), showing the big difference that exists in terms of quantity between the point of water diversion in the irrigation system and the water available in the root zone of the plant. Agriculture depletes the water resources through evapotranspiration, and this signifies less water for other uses in the system. Climate plays an important role in the total amount of water that is evaporated and transpired by plants.

135

Although there is not a single definition, water productivity is then the ratio of crop output to water either diverted or consumed, being expressed in either physical or monetary terms, or some combination of the two (Barker and Molden, 2003). Another related concept used by economists is the economic efficiency, which takes into account value of output, opportunity costs of inputs and externalities, and it is achieved when water is allocated and used so the net value or net returns are maximized. It is a criterion that describes the conditions that must be satisfied to achieve the largest possible net resource (Wichelns, 1999). It is a concern for economists that, in general, the true value of water is not reflected by the prices or charges for irrigation water, in which the allocation does not follow socially optimal scenarios. Management of water in agriculture should be targeted to allocate water to the highest level of water productivity as a social optimum. In theory, the allocation of water among competing uses involves an economic optimization exercise. If we were to consider only private returns to the agricultural production activity, some alternatives to improve water productivity would be (Wallace and Batchelor, 1997):

• • • •

agronomic improvements (e.g., improved crop husbandry, cropping strategies, and crop varieties); technical improvements (e.g., improved and lower-cost technologies for extracting groundwater); managerial improvements (e.g., improvements in farmlevel resource management or system operation and maintenance (O&M)); and institutional improvements (e.g., introduction of water pricing and improvement in water rights).

The first two categories are innovations or new technologies that lower cost or increase the value of output per cubic meter of water. The third category refers to an increase in technical efficiency, and the fourth relates to allocative efficiencies encouraged by the creation of market incentives (Barker and Molden, 2003). The economic value of water – economic water productivity – can be increased by different actions: (1) increasing the agricultural yield per unit of water used; (2) switching from low value to high-value crops; (3) relocating water from low to higher value water uses; (4) lowering the cost of inputs for the same water used; (5) increasing health benefits and the value of ecological services of agriculture; (6) deceasing the social, health, and environmental costs; (7) obtaining multiple benefits per unit of water; and (8) achieving more livelihood support per unit of water, such as, more jobs, nutrition, and income. However, economic theory shows that if a new practice offers net benefits and does not have any negative effects on third parties off the farm (termed externalities by economists), then the adoption of this practice is advantageous for the society as a whole, not just for the farmer. Unfortunately, there are many externalities in the agricultural sectors of developing and developed countries that make it more complicated to identify the optimal practices that should be adopted by private agents (farmers).

136

Managing Agricultural Water

Policy-induced incentives are necessary to achieve an optimal water allocation. In order to assess the optimality of the allocation of water as a scarce resource, it is necessary to value all agricultural inputs and outputs at social prices that generally diverge from market prices. The difference between these two prices is a result of distortions, subsidies, and transfers among different actors of the economy.

1.08.2.1 Economic Value of Water for Agriculture The value of water is the maximum a user is willing to pay for the resource. In other words, water has value when users are willing to pay a price instead of not having it. Water has economic value when its supply is constrained compared to the amount that is demanded, and when water as an input becomes a constraint to the economic and social development process. This concept has become important particularly after water was declared as an economic good in Dublin, 1992. Water needs to be evaluated for the benefit of humanity, and to ignore its value would lead to an overexploitation of the resource and an incorrect allocation among alternative uses. Rational decisions that support water development, its allocation, and final use, demand a measurement effort to value water for alternative uses. When defining investment priorities in a country or a region, the level and the geographical variability of the economic value of water become a key element in the decision process. It is used as a selection criterion when compared to what users are actually paying for the resource (cost recovery and operation and maintenance fees). It also becomes a good indicator of the future demand of the resource for irrigation. There exists a set of methodologies to estimate the value of water for agriculture: direct methods such as the water markets and the contingent evaluation approach. There are also indirect methods such as the estimation of demand equations and its integration, and a series of methodologies that are derived from considering water as an intermediate good. Among these are the residual value methodology, the alternative cost, and the Hedonic price methodologies. Finally, when primary information does not exist to estimate the value of water, a methodology called meta-analysis is employed to extrapolate economic values of water from other watersheds or regions to the region of interest.

1.08.2.2 Example of Estimates of the Economic Value of Water: The Case of Mexican Agriculture Mexico is an interesting country case to observe the application of alternative methodologies to estimate the value of water for irrigation. Ramirez-Vallejo and Rogers (2004) used an indirect methodology proposed by Moore (1999) to estimate the value of water for irrigation, and also applied a residual method in various irrigation districts in Mexico. Others estimated direct values from water markets and the use of optimization exercises of the producer’s behavior at the farm level. Kloezen and Garce´s-Restrepo (1998) documented rent transactions of water concessions, and Florencio-Cruz et al. (2002) applied math programming to estimate values of irrigation water. As seen below, the estimates of the economic value of water for irrigation vary significantly depending on

the methodology employed, a situation that complicates the application of water value as a criterion for allocating water.

1.08.2.2.1 Indirect method The indirect methodology employed by Ramirez-Vallejo and Rogers (2004) was justified by the fact that the water prices are divorced from the production costs in the long run. This is something that has been criticized in the past with the argument of efficiency (Ciriacy-Wantrup, 1954; Bain et al., 1996). For the case of Mexico, a revenue function was defined that related the multiproduct to fix inputs, estimated as

Rðp; x; zÞ ¼ maxfpyA Yðx; zÞ; p4 0g

ð1Þ

where p is the vector of product prices and y is the vector of product production for each product, x is the amount of water, z is the amount of the mix input, and Y(x,z) is the set of production probabilities. Chambers and Just (1989) developed the properties of the revenue function, which was estimated econometrically, and the shadow prices were derived using a quadratic normalized form suggested by Lau (1978) to estimate the willingness to pay. Applying this methodology for 14 irrigation districts using data from 1993 to 2001 resulted in the following shadow prices for irrigation water (see Table 1).

1.08.2.2.2 Residual method Using an alternative residual method, Ramirez-Vallejo and Rogers (2004) computed the value of the production factors different from water and then subtracted this value from total sale income from cultivated products in the irrigation districts. This difference was assigned to water as its economic value. Young (1996) clarified that this imputed value is valid if two Table 1

Shadow prices for irrigation water ($ m3)

Irrigation district

Average shadow price 1993–2001

Average shadow price 1997–2001

001 Pabellon, Ags. 005 Delicias, Chih. 010 Culiacan Y Humaya, SIN. 011 Alto Rio Lerma, GTO. 014 Rio Colorado, B.C. 017 Region Lagunera, DGO. Y C. 023 San Juan Del Rio, QRO. 024 Cienega De Chapala, MICH. 038 Rio Mayo, SON. 041 Rio Yaqui, SON. 075 Rio Fuerte, SIN. 076 Valle Del Carrizo, SIN. 92A R.Panuco, Tamps. ‘‘ANIMAS’’ 92B R.Panuco Pujal Coy, S.L.P.

1.160 0.844 1.455

1.568 0.929 1.557

0.630 1.649 1.572

0.888 1.132 2.050

1.856 1.691

2.000 1.578

0.810 0.695 0.190 1.341 1.679

0.677 0.907 0.987 1.737 2.637

1.073

1.329

From Ramirez-Vallejo J and Rogers P (2004). Virtual water flows and trade liberalization. Journal of Water Science and Technology 49(7): 25–32.

Managing Agricultural Water

conditions are satisfied: first, the inputs and products are in competitive markets and are not regulated, that is, the price is equal to the value marginal product. Second, the production function should behave in such a way that an increment in each of the inputs generates an equal relative increment in the product. The residual value of water for irrigation was estimated by

Residual value of water ¼ PQ 

X

Wi Ni

where P is a vector of prices of products in the district and Q is a vector of the amount produced. Wi is the quantity of input i used and Ni its price. Besides the traditional inputs (land, capital, and labor), other inputs were considered such as the administration and the expected utility of the economic activity. The shadow price was then the residual value divided by the amount of water used. The results of applying this approach to some irrigation systems in Mexico are shown in the Table 2.

1.08.2.2.3 Water markets When water markets exist, prices applied in the resource transaction become good proxy of the economic value of water. Furthermore, if the market allows observing the resource–demand curve, the value would be determined by the area under the curve. There exist some water markets in irrigation districts in Mexico where some transactions involve the water concessions and others the right to use water for specific time periods. Table 2

137

Kloezen and Garce´s-Restrepo (1998) documented the rent transactions of the water concessions for the irrigation district of Alto Rio Lerma (Table 3). The observed market prices were below the economic values estimated using alternative methodologies, a situation that could be explained by the distortions of rental markets of concessions in the irrigation districts as a result of direct intervention of user associations; social and political motivations were present in the exchange process of water rights, which did not allow water markets to function properly. Other registered transaction values of concessions for the irrigation district of Lagunera varied from $0.05 to $0.1 m3. Finally, the government, in its Aquifer Program, rationalized the use of water by purchasing some of the users’ concessions. In 2004, the government, using market mechanisms, purchased back concessions at an average price of $2.5 m3, equivalent to an annual value of $0.25 m3 after applying the perpetuity formula with a discount rate of 10%.

1.08.2.2.4 Math programming method Some authors have implemented the math programming method to estimate the economic value of water, optimizing the net income subject to various constraints, among them, the budget constraint, and the water availability. The execution of these exercises for different levels of water available allows estimating the water demand equation, which, when integrated, gives the economic value of water, or willingness to pay for the resource. Florencio-Cruz et al. (2002) executed this exercise to the Alto Rio Lerma irrigation district and a significant dispersion of the estimate was found (Table 4), depending on the season considered.

Shadow price using the residual method

1.08.2.3 Agricultural Trade Protection Typology

Irrigation district

(Shadow price Pw)

Rehabilitation New New New New New

Yaqui Yaqui Angostura Carrizo Aguascalientes Queretaro

0.159 0.242 0.879 0.394 0.151 0.212

From Ramirez-Vallejo J and Rogers P (2004). Virtual water flows and trade liberalization. Journal of Water Science and Technology 49(7): 25–32.

Table 3

Agriculture is one of the most important sectors in many developing countries in terms of social and political stability, and thus it is heavily circumscribed by global trade protection, much of it originating from Organization for Economic Cooperation and Development (OECD) countries. Agricultural policies in OECD countries cost consumers and taxpayers over $280 billion every year (Anderson et al., 2006). The value of total agricultural support in OECD countries was more than 5 times higher than total spending on overseas development

Rental price in the irrigation district of Alto Rio Lerma

Season

Seller

Buyer

Volume (m3  1000)

Bought as a % of total used

Price ($m 3)

Summer 1995

Acambaro Acambaro Acambaro Acambaro Acambaro Acambaro Valle Valle Jaral

Cortazar Salvatierra Huanimaro Salvatierra Abasolo Huanimaro Salamanca Abasolo Salamanca

10 000 10 000 2000 8000 3000 2000 3500 1000 450

25 21 23 36 19 86 18 6 2

0.004 0.009 0.009 0.020 0.034 0.034 0.035 0.034 0.035

Summer 1996 Summer 1997

From Kloezen W and Garce´s-Restrepo C (1998). Assessing Irrigation Performance with Comparative Indicators: The Case of the Alto Rio Lerma Irrigation District Mexico. Colombo, Sri Lanka: IWMI.

138

Managing Agricultural Water

Table 4 Value of irrigated water for Alto Rio Lerma District using math programming method Groundwater February April Surface water January May

0.7–1.8 $ m3 1.6–2.4 $ m3

•

•

3

0.5–1.2 $ m 1.3–1.9 $ m3

From Florencio-Cruz V, Valdivia-Alcala´ R, and Scott CA (2002) Productividad del agua en el Distrito de Riego 011, Alto Rio Lerma. Agrociencia 36: 483–493.

• assistance and twice the value of agricultural exports from developing countries (OECD, 2001). Reductions in distortions to international trade of agricultural products will most likely increase real income and stimulate change in the composition and location of production and consumption of water, as a primary production factor. The impact of agricultural trade on water resource development will differ between countries depending on how protected and distorted their agricultural sectors are as well as their water resources endowment, its use, and their policy and institutional framework. That is, where trade liberalization leads to a decrease in agricultural production, there is likely to be a reduction of water use and its associated environmental impact, but where it increases production, water will be allocated to agriculture from other actual or potential uses (Ramirez-Vallejo and Rogers 2009). Therefore, a change in trade barriers and agricultural subsidies in general could impact differently water allocation, and, therefore, water policy adoption. In theory, to achieve a rational decision to allocate water within alternative uses, it is important to measure the value of water in these uses. When social and market prices are equivalent and no distortions are present, the market as a clearing mechanism yields the optimum allocation of water for a society as a whole. When this is not the case, the externalities should be taken into consideration for the analysis. However, in practice, there is often little time, money, knowledge, or will to conduct serious economic analyses of benefits and costs to consider in the water allocation disjunctives. To summarize

•

•

• •

The concept of water productivity relates to the desire have a higher level of output, economic value added, or just value for the aggregate production, per unit of water used. In other words, it is the way of achieving more using less scarce water resources. Water productivity is then the ratio of crop output to water either diverted or consumed, being expressed in either physical or monetary terms, or some combination of the two. Management of water in agriculture should be targeted to allocate water to the highest level of water productivity as a social optimum. Some alternatives to improve water productivity are: (1) agronomic improvements; (2) technical improvements; (3) managerial improvements; and (4) institutional improvements.

The value of water is the maximum a user is willing to pay for the resource. Water needs to be evaluated for the benefit of humanity, and to ignore its value would lead to an overexploitation of the resource and an incorrect allocation among alternative uses. There exists a set of methodologies to estimate the value of water for agriculture: direct methods such as the water markets, and the contingent evaluation approach; indirect methods such as the estimation of demand equations and its integration; and a series of methodologies that are derived from considering water as an intermediate good. The impact of agricultural trade on water resource development will differ between countries depending on how protected and distorted their agricultural sectors are as well as their water resources endowment, its use, and their policy and institutional framework.

1.08.3 Water Management and Competitiveness 1.08.3.1 Framework Water management is intrinsically linked to the competitiveness upgrading process. Under the current global scenario, the goal of many countries and regions is to increase the level of competitiveness to achieve a higher level of prosperity. Regions and countries are competing to offer a more competitive environment to allow agricultural firms and farmers to be more productive. The opportunity that farmers and firms have of obtaining the right amount of water at the right time to optimize the value-generation process becomes an important characteristic of the business environment. However, the link between water management and competitiveness is not well understood and most countries assign higher priority to other types of policies. Developing countries focus most on reforms to adjust their macroeconomic, political, legal, and social conditions that are necessary conditions, but far from being sufficient to increase prosperity. Wealth is created by the productivity of all production factors that a nation or region can utilize, and heavily depends on the microeconomic conditions, understood as the elements that are located outside the production unit (i.e., farms) and the conditions that take place within the boundaries of companies and farms. Unless these microeconomic capabilities improve, the growth process of prosperity will be truncated (Porter, 1998). Competitiveness is the value that firms can generate per unit of input used, that is, capital, land, labor, and so on. This real value is created by the private sector, by the farmers themselves, when transforming these inputs into products and services that consumers are willing to pay for. A farm is more productive when it generates the same amount of output but the production is sold in the market for a higher price. In addition, a farm is more productive if it generates more physical output per unit of input. True competitiveness is then synonymous with productivity, and the higher the level of productivity of the agricultural units, the higher the wages they can support, and the more important the influence on prosperity. In the agricultural context, wealth is created at the microeconomic level when farmers and firms produce valuable

Managing Agricultural Water

goods and services using water more efficiently. The wealth creation process is a result of their production activity and not a direct result of government or other societal institutions’ intervention. Water improves farm productivity in various ways. First, it creates a direct impact on the possibility of technology adoption by farmers and agricultural firms. Without water, for example, the choice of agricultural varieties would be limited. Water also allows the use of other agricultural inputs such as fertilizers, and causes a clear impact on production stability that helps to generate a higher output value. Moreover, paradoxically, water scarcity also generates the right incentive for innovation, a phenomena that explains new technology advances available for rainfed agriculture. Under a water management for agriculture framework, the foundations of productivity rest on two interrelated areas: (1) the sophistication and capabilities with which farms and agricultural companies compete, which are highly dependent on the availability of water resources and (2) the quality of the business environment in which they operate. More productive farm and firm operating practices require highly skilled people, technology and availability of water resources, better information, more efficient government processes, improved infrastructure, better suppliers, more advanced research institutions, and more intense competition, among other things. The competitiveness of a farmer and an agricultural company, then, depends on both, their internal capabilities and the characteristics of their location, which are influenced directly by water management. Water is not only a regular input of the agricultural production and transformation process, but also a facilitator of the strategic upgradation of farmers and agricultural firms. Agricultural firms and farmers must upgrade their modes of competing and capabilities in order to generate economic

development. They need to shift from competing on costs and inherited endowments to create competitive advantages from efficient and distinctive products and internal processes. Competition on cost structure or price is usually the traditional way of competition within agricultural sectors in most countries, and a characteristic of earlier stages of development. Moving to more sophisticated ways of competing depends on parallel changes in the business environment. The business environment can be understood in terms of four interrelated areas: the quality of factor (input) conditions, the context for firm strategy and rivalry, the quality of local demand conditions, and the presence of the related and supporting industries. Because of their graphical representation (see Figure 1), the four areas have collectively become referred to as the diamond (Porter, 1998). Water management impacts directly the quality of the business environment for agricultural development. First, water improves the access to high-quality agricultural inputs, such as infrastructure, fertilizers, human resources, and capital. Second, water user associations usually generate rules and incentives that affect investment and productivity, and foster or inhibit a vigorous local competition. Third, advanced water management practices induce the sophistication of local customers and needs for example, strict quality, safety, and environmental standards. Fourth, improving the availability and reliability of water supply facilitates the existence of suppliers and supporting industries of agricultural development. Improving competitiveness is a special challenge because no single strategy or action of an individual institution can improve significantly the productivity of a region. Many things matter for competitiveness. The quality of schools and roads, the financial markets penetration, the consumer sophistication, the supplier networks, the rules for competition, the quality of the private and public demand, and many more

Context for firm strategy and rivalry

Agricultural water management

139

• Water user associations could generate rules and incentives that encourage investment and productivity, and foster vigorous local competition

Factor (input) Conditions

Demand conditions

• Water improves the access to highquality agricultural inputs

• Advanced water management

Related and supporting industries • Improving the availability and reliability of water supply facilitates the existence of suppliers and supporting industries of agricultural development

Figure 1 Water management and quality of the business environment for agricultural development.

practices induce the sophistication of local customers and needs –e.g., strict quality, safety, and environmental standards

140

Managing Agricultural Water

attributes act as determinants of competitiveness of the regions where agricultural products and services are produced. Water resource management is one of them and needs to be articulated with reforms in other areas to upgrade competitiveness.

1.08.3.2 Farmers, Water, and the Process of Economic Development Improving competitiveness is a collaborative process in which private sector, academia, government, institutions, and labor unions contribute in designing and executing a competitiveness strategy. Farmers, entrepreneurs, and managers of agricultural companies play a crucial role in improving competitiveness and defining regional policy. Governments play an important role in upgrading competitiveness because it affects all its determinants. Government agencies affect regulatory standards, invest in public goods, set the rules and incentives, coordinate policies, and have an effect on the quality of its purchases of goods and services. Universities and schools impact the role of water in improving competitiveness through knowledge and human resource skill formation to improve the technology. The educational system needs to be connected with farmers and government to adjust research agendas and curriculum on water resources. Farmers and business people depend on the business environment and also contribute to shaping it for their own benefit. This explains the need of implementing communication mechanisms between leaders from the private sector, government, and academia to coordinate and collaborate on public policy and specific actions to improve water management. Engaging the private sector in water resource development is key to sustaining progress in the long term. Under this scenario, changes in national and local governments would less likely impact the continuity of policies and institutions that deal with water management. To achieve an effective collaboration, it is necessary to build an organizational structure that connects all actors and fosters collective activities. Organizations for water management at the basin level should be open institutions with the participation of all stakeholders.

1.08.3.3 Water Resource Management and the Regional Economic Strategy Water contributes with economic development through backward linkages that make the agricultural input supply sectors more dynamic, and via forward linkages strengthening agro-processing industries, trade, and other activities downstream in the value chain. Some of these activities are made feasible by the available technology for water agriculture. In addition, higher agronomic yields lower unit costs and impact on cheaper food available for the populace. Water also allows the cultivation of export crops that require a more regular supply and higher technology, and has an impact on off-farm employment. In a global scenario, countries and regions compete to offer the best location based on productivity to facilitate the value generating process that finally translates in prosperity. Globalization is putting pressure on countries and regions to

improve the determinants of competitiveness and they are adopting best practices in many aspects such as water management, water technology, agricultural technology, farm operation, infrastructure development, and human resource development, among other determinants of competitiveness. This competition leads to work on ways how to make a region distinctive from other competitor regions, to differentiate and achieve a distinct role in the country and in the world. Water then becomes an interesting resource to achieve regional differentiation, allowing competition and incentivizing innovation. Water management works as a channel to build a unique mix of strengths in the business environment, to differentiate and to improve the productivity potential of farms and agricultural businesses.

1.08.4 Water Resource Management, Institutions, and Implementation 1.08.4.1 Integrated Water Resource Management During the past couple of decades, agricultural water management has been approached within the integrated water resource management (IWRM) framework defined as a systematic process for sustainable development, allocation, and monitoring of water resource use in the context of social, economic, and environmental objectives (Cap-Net, 2009). This approach recognizes that water is used in various economic sectors, it has many uses, and as such it needs to be managed with an integrated approach. It recognizes that water management needs a multidisciplinary approach to deal with water sources and demands at the basin level. It also recognizes that water is a resource that needs to be sustained over time, and it is an instrument to achieve social, economic, and environmental goals at the basin level. Recently, this framework has been broadened to incorporate participatory decision making of all stakeholders. The IWRM is considered a paradigm shift that departs from traditional approaches in three ways: first, its multiple goals and objectives are cross-cutting so that this new approach departs from the traditional sectoral approach; second, the spatial focus is the river basin instead of single water courses; and third, there is a departure from narrow professional and political boundaries and perspectives and is broadened to incorporate participatory decision making of all stakeholders. The IWRM framework deals holistically with all water, all interests, all stakeholders, all levels and all relevant disciplines, and sustainable in all senses (Jaspers, 2001). In this framework, water is considered an economic good as well as a social good, and the market is given an important role for water pricing, and to deal with water scarcity, allocation efficiency, and environmental protection. Since the Dublin principles in 1992, there has been some progress in understanding the IWRM framework and its importance. However, there are many obstacles in its implementation process as a result of different sectoral interests, cultural values, and political constraints. Negotiation situates at the center of conciliating many thematic and special interests and beliefs at the time of allocating water resources. In addition, significant institutional and legal reforms are needed to reorient the traditional water management approach to a

Managing Agricultural Water

more integrated comprehensive management. Water development professionals still have difficulties in working together to come to a unique set of principles of IWRM.

1.08.4.2 Participatory Irrigation Management One of the most applied modalities in agricultural water management has been the participatory approach. Participatory irrigation management (PIM) refers to the involvement of irrigation users in all aspects and all levels of irrigation management. Users participate in all stages of the project life cycle; they begin with the participation on the design of the new (rehabilitation) project, and continue during its construction phase, including financing. Finally, and perhaps the most decisive stage, there is the participation of the users with the operation and maintenance of the system. It has been shown that PIM in real practice helps to solve some of the most prevalent problems with irrigation management, such as the inadequate water availability at the lowest outlets; poor condition/maintenance of the system; lack of measuring devices and control structures; inadequate allocation for O&M; inequitable distribution of water; and lack of incentives for saving water and poor drainage (Merrey, 2007). Since the beginning of the 1990s, this new perspective of irrigation management has been adopted in many developed and developing countries and has helped understand why some policies work better than others, and why some irrigation systems perform better producing higher impacts in terms of productivity and prosperity for the rural inhabitants. In many countries, PIM encouraged the creation of water user associations (WUAs) that took the responsibility of management of the irrigation systems once they were constructed. Taking management responsibilities demand from organizations the capacities to represent the users and institutions endowed with an adequate governance structure to take difficult decisions and solve the multiple obstacles, particularly those experienced during periods of water scarcity and conflict resolution. Before a WUA was formed, governments usually set a package of incentives for both users and irrigation agencies to make transfer programs sustainable. Irrigation agencies in many countries built processes of organization creation and strengthening, and delegated management of irrigation systems. One of the main objectives of the new management is to make the irrigation district financially sustainable, and to limit the amount of subsidies needed from the government. To finance the O&M activities, O&M fee schemes are usually implemented. The share in O&M contribution distributed between the government and users varies significantly among districts and countries; they are particularly biased toward a higher government contribution in those systems that were built with limited user participation. Sometimes the O&M fees are financed directly or indirectly through land or other type of agricultural taxes. To estimate the level of the fees, the portfolio of services of these organizations is identified, and costs are allocated. Once this information is available, usually the government and users enter into a negotiation process in which the responsibilities are assigned between the government and the WUA.

141

User participation in defining service levels is crucial for achieving financial sustainability. The higher the user involvement in defining the set of activities that is needed, the more dynamic the future demand for those services will be and the higher the willingness to pay for the corresponding fee. In Mexico, for example, following the transfer of management of irrigation districts to WUAs, irrigation plans were prepared by WUA-hired managers, taking into account cropping plans, conveyance losses, and equitable distribution. These were then negotiated with the national irrigation agency to determine the final allocation, generally based on an arranged demand pattern. Operation and maintenance costs at the level of the secondary and below were met by user fees managed by the WUAs. In addition, these associations contribute with part of O&M costs (World Bank, 2009). In Mali, joint committees were established for O&M in every region. The committees had 5–10 representatives of producers and 5–10 representatives of the agency. These committees decided on types of services, costs including procurement matters, and water service fees. They also made decisions on the use of 50% of the user fees collected for O&M. In Chile, the national federation of WUAs was consulted in the design of an irrigation project. The federation and local WUAs played an active role in project preparation especially in discussions of service options and costs. Subsequently, the project incorporated the condition of WUA-approval for investment proposals and other project components. The fees usually have a component that covers some fixed costs and another variable component as a function of the water consumed or the irrigated area. Linking the fee to the irrigated area makes it easy to estimate it and to understand it. It is recommended that the easier to understand the fee structure, the simpler its administration. On the other hand, volumetric charges that are a function of the discharge generate the right incentive for water saving but are more difficult to implement because of the necessary water-measuring devices and control mechanisms that are costly and operationally complex. Agencies often favor a combination of the two methods. In Mexico, for instance, the agency charges the WUAs volumetrically at the turnout of the secondary canal, and the WUAs base their water charges to individual members on area irrigated and type of crop. In addition to unit area and volume, WUAs have also resorted to other bases such as charges for the entire season, which clearly favor the highvolume user. In some other irrigation districts, a property tax is employed as a charge based on the increase in the land values due to water availability. The higher the land valuation, the higher is the tax level as a consequence of available water supply. However, the problem observed with this system is that it does not generate the water-saving incentives that result with other mechanisms such as the volumetric approach, and usually these taxes are collected by municipalities, which generate discontinuity between the source and use of the resources. Another system, less frequently found, is the in-kind contribution system, mainly through labor, materials or both, applied directly to O&M activities. In Vietnam, the provincial and national governments finance schemes down to 150 ha for new irrigation development. Below this level, farmers must build the channels with the government providing support for

142

Managing Agricultural Water

survey and design, as well as contributing with materials, in some instances. After schemes are completed and taken over by farmers, each member must provide up to 20 person days per year toward maintenance of the tertiary as well as secondary systems (World Bank, 2009). But perhaps the most difficult hurdle to overcome is in replacing the assets of the irrigation districts. As any system, the infrastructure that is used for storing and delivering water to users depreciates. Buildings, canals, motors, pumps, vehicles, and heavy equipment need to be replaced once they complete their life cycle. If the government and/or the WUA do not implement a saving mechanism to cover these replacement costs, the sustainability of the system was compromised. Poor management leads to short service life with higher acceleration rates that then demand an increase in fees to cover future expenses. Provisions made for collecting these asset replacement funds by the WUA are usually not accepted by farmers who adopt a free-rider behavior and wait for government agencies to intervene when major depreciations do not let the system function properly. In parts of Vietnam, the estimate of costs of services included provision for depreciation of assets. This approach demanded knowing the location and conditions of all assets within the system, and to implement a training and preparation program on the formulation of an O&M program. Linking fees to services in the irrigation districts is important, particularly when an agency collects O&M fees and these funds are then allocated later to other government agencies, or when the budget of the O&M of the irrigation district is done independently of the collection effort. In these cases, there needs to be an agreement between the WUA and the government agency on mutual rights and responsibilities. The cost of collection by WUAs is lower relative to government agents and sanctions for nonpayment by individuals are easier when enforced by WUAs. The Philippines experience has shown that fee collection is better in systems where WUAs are organized and where they have a role and incentive for collection. In the pilot projects in Maharashtra, India, WUAs were allowed to retain a proportion of collections as a bonus (World Bank, 2009).

1.08.4.3 Lessons Learned from Participatory Management Since the 1990s, there has been an important effort on transferring the management of irrigation districts to WUAs. This new trend in water management has increased net irrigated area in large public irrigation projects. Important lessons have accumulated over time on various management schemes. Some lessons from the Philippines and Mexico and documented by the World Bank are presented below that exemplify the trend toward participation (these examples are taken from the Electronic Learning Guidebook for Participatory Irrigation Management).

1.08.4.3.1 The Philippines Just as the state’s involvement – or micro-management – was reaching its peak in the 1980s, there were countervailing forces appearing. In the Philippines, the process can be traced to the mid-1970s when President Marcos ordered the National Irrigation Administration to move toward self-financing. The

agency responded by withdrawing from the small communal systems which had once been self-managed, but had grown dependent on the government. This weaning process was accomplished through intensive grass roots organizing and capacity building both among farmers and within the agency itself. Water user associations were formed to take over operations and maintenance, and to contribute to capital costs of improvements. Beginning in 1980, this organizing approach was applied to the state-run systems that had no prior history of self-management. As with the management transfer in communal systems, the goal here was cost savings to the agency, both through direct recovery of water fees and the replacement of some low-level agency operational functions by association volunteers. Until recently, this modest level of joint management was the dream of irrigation policy reformers, and the Philippines served as the model. The paradigm was one of joint management where farmers would become management partners with the agency, and decisions would be made jointly. However, the relationship is asymmetrical; the state controls the technical expertise and subsidizes maintenance and improvements even in the canals operated by farmers.

1.08.4.3.2 Mexico Independently from the trend toward joint management at the lower ends of the system, a model of irrigation management transfer was evolving in Latin America, in response to structural adjustment pressures. This model constitutes the qualitatively different paradigm where the users dominate, and the state facilitates. In the mid-1980s when Mexico was in the throes of a debt crisis, the government was bankrupt. The large irrigation districts under Federal control suffered as maintenance was deferred and the productivity of unpaid, demoralized engineering staff declined. Out of necessity, the government reorganized the state irrigation agency to create the National Water Commission (or CNA in its Spanish acronym), with a mandate to turn over the management of the irrigation districts to associations of users created specially for this purpose. In 1990, Mexico transferred the first irrigation district to the users. By 1995, more than two-thirds of the country’s 3.2 million hectare network – divided into 80 irrigation districts – had been transferred to 316 irrigation associations. The transfer program was initially in the most productive irrigation districts, which were best organized and with the most commercially oriented farmers. The most important criterion for selecting districts was the potential of the user organization to become financially self-sufficient, with users paying the fees to cover the costs of operations, maintenance, and administration. What could the government offer the farmers as an incentive to accept higher costs for their irrigation? In fact, there was a carrot as well as a stick. The carrot was management autonomy. The farmers would be free to set their own rules for when to clean the canals, and how to distribute the water. The farmers would hire their own technical staff – engineers and accountants – to run their system. The canal would be theirs on a 20-year concession, which is in practice a transfer of ownership.

Managing Agricultural Water

However, there was also a stick. If farmers refused to take over management, the government could offer no assurance that the canal network could be kept in repair. The government in effect threatened to default on its conventional understanding with farmers regarding levels of subsidy in the irrigation sector because it no longer had the financial means to do so. The government, however, also promised and provided technical, organizational, and legal assistance in realizing the transfer. Many farmers, and particularly the commercially oriented ones, could not accept the risk that the irrigation infrastructure might collapse. They preferred to take over the management, and with a few exceptions, they have not looked back. They are paying much more for their water without the government subsidy, but the reliability and responsiveness of their new management structure is well worth the price. For them it is a win situation, and for the government as well. What are farmer’s comparative advantages when they are managing for themselves? They have direct incentives to manage irrigation water in a productive and sustainable manner; they offer an on-the-ground presence that even the most dedicated off-site agency staff cannot equal, and they have an intimate knowledge about their fellow irrigators. The state’s comparative advantage is in the depth of financial and technical resources and the regulatory and administrative capacity for managing water supplies to competing interests.

exercise of economic, political, and administrative authority to manage a country’s affairs at all levels, which comprises the mechanisms, processes, and institutions through which citizens and groups articulate their interests, exercise their legal rights, meet their obligations, and mediate their differences (United Nations Development Programme, 2001). Rogers (2002) argued that governance is not restricted to the perspective of government as the main decision-making political entity, and governance covers the manner in which allocative and regulatory politics are exercised in the management of resources (natural, economic, and social) and broadly embraces the formal and informal institutions by which authority is exercised. According to Rogers (2002), there are different models of water governance:

•

•

• 1.08.4.4 Institutions and Water Governance Before presenting the argument of the need of institutional and organizational reform, some key concepts need to be defined. According to North (1990), institutions are the rules of the game in society. They are social arrangements that shape and regulate human behavior and have some degree of permanency and purpose transcending individual human lives and intentions. In agricultural water management, these types of institutions are user associations, rules of water allocation, market mechanisms, and property rights. Organizations refer to the group of people with shared goals and formally defined roles, such as water associations, government irrigation agencies, water companies, nongovernmental organizations, and regulation bodies. A policy is ‘‘a set of interrelated decisions taken by a political actor or group concerning the selection of goals and the means of achieving them within a specified situation where these decisions should, in principle, be within the power of those actors to achieve’’ (Howlett and Ramesh, 1995; Jenkins, 1978). On the other hand, governance is the way authority is organized and executed in society, and often includes the normative notion of the necessity of good control. The Global Water Partnership defines water governance as ‘‘the range of political, social, economic, and administrative systems that are in place to develop and manage water resources, and the delivery of water services, at different level of society’’ (Rogers and Hall, 2003). Governance is therefore a broad term that includes institutions, organizations, and policies. The World Bank broadens the definition to include the process by which those in authority are selected, monitored, and replaced, and the effectiveness of government in implementing sound policies (Jayal, 1997). The United Nations defines it as the

143

•

•

The bureaucratic politics and process model. This model is based on political–bureaucratic bargaining in a federal system. Its focus is typically on the executive branch, with the elected legislature hardly in the picture. The congressional behavior model. A second federal model concentrates on the elected congress, with the view that to understand congressional behavior is to understand that congressmen are single-minded seekers of reelection. It follows from this that congressmen’s goals are to improve the welfare of their constituents in the shortest possible time frame. The interest group model. In some cases, legislators see only a few dominant interests involved with water policy. These interest groups often have overlapping concerns and overlapping memberships. It is also valuable to carry out an interest group analysis of the feasibility of pursuing specific governance goals. They examined the likelihood that the various interest groups would be powerful enough to influence the investment and management decisions in their direction. Principal–agent theory. Principal–agent models have been employed in many different academic fields, including economics, in order to explain relationships among actors in which the consumer is the principal and the producer is the agent; and in various political science subfields, in which members of the legislatures are the agents of their constituents, or bureaucrats are agents of the executive, or the governments of Third World countries are the agents of international lending institutions, and so on. Regime theory and public choice. The use of the concentration or diffusion of costs and benefits of public choices to predict what decision-making system will prevail.

It has been found that water governance for agriculture is important for the management of irrigation systems and rainfed agriculture involving different stakeholders to achieve more effective water allocation. Good governance in irrigation systems matters for achieving high levels of economic, social, and environmental performance. There are some necessary conditions for good governance: inclusiveness, accountability, participation, transparency, predictability, and responsiveness. A failure on these conditions leads to poor governance and difficulty to deal with system problems and presents a risk in terms of sustainability. Better governance will lead to higher levels of prosperity of the users

144

Managing Agricultural Water

of the agricultural systems and becomes an essential input of country and regional development. There is also a strong relationship between the level of governance of water resource systems and the income per capita, lower infant mortality, and higher literacy (Kaufmann, 2005). New institutions and governance are needed to implement water management if the goals are to achieve food security, economic growth, and poverty reduction under favorable environmental conditions. This demands institutions and organizations responsible for good agricultural water management and policymaking. There is a need to reform not only the organizations at the national and international level that deal with water management, such as the irrigation agencies and multilateral institutions, but also local and informal institutions that act at the basin level that need to be redesigned and in some cases built from scratch. The current institutional arrangement has failed to meet the challenge to respond to new technologies and rules to deal with the challenging goals of developing agriculture to reduce poverty, and generate prosperity with equitable growth under minimum environmental impacts. In the water sector, as with any sector in the economy, some institutional arrangements have shown little promise so far. On the one hand, agricultural markets in developed and developing countries present high distortions that generate strong market failures that need government intervention. On the other hand, the government itself frequently fails to make the right adjustments to align market outcomes. In the irrigation sector, arrangements for authorization, payment, and accountability impact on service provision (Huppert et al., 2001). The failure of public organizations to offer socially and environmentally optimal allocations of water resources has induced the creation of new institutional schemes. The decades of the 1990s and the 2000s have experienced an upsurge of private companies and market mechanisms not only in the construction of water systems for agriculture but also for their operation and maintenance. Some of these new organizations present a heavy involvement of the private sector. Privatization of O&M has been a component of many irrigation programs, particularly where the operation of these systems required specialized skills that individual farmers did not have. Two major reforms have been put in place in many countries to generate economic incentives for water allocation: water pricing and tradable water rights. Water pricing mechanisms are used to create incentives for water conservation but require volumetric measurement devices that are expensive and difficult to control. In addition, the strong legacy of government subsidies and public intervention in water management has generated famers’ resistance to pay for irrigation service. In addition, investment cost recovery demands such high fees that unless there exist financial markets it is difficult to obtain payments from farmers. The second privatization mechanism that has been used frequently in many countries during the past two decades is the tradable water rights. The idea behind the water rights market is that the user who has a higher value use of the resource would be willing to pay more for the right to use the water. In other words, the market allows water to move from lower to higher value uses. However, these markets are

complicated to build since specific topologies of irrigation systems to transfer water from buyer to seller are required. They also need institutional arrangements to control, facilitate, and enforce transactions, and protect against negative impacts on third parties when water is transferred (Easter et al., 1998; Rosegrant and Binswanger, 1994).

1.08.4.4.1 Water management principles In 1992, at the International Conference on Water and the Environment, convened in Dublin, Ireland, four main principles of water were adopted that have since then shaped water management. The first principle is that water is a finite and vulnerable resource, essential to sustain life, development, and the environment. This principle recognizes that water is critical to sustaining life, and it is a finite resource because of the economic and technical complications of transferring water from one place to another. This principle also suggests a holistic approach to deal with water management in which all dimensions, social, economic, and environmental are taken into account. The second principle establishes that water development and management should be based on a participatory approach, involving users, planners, and policymakers at all levels. It suggests democratization in the decision-making process when allocating water among uses, and the importance of the input of multiple stakeholders in the design, construction, and operation and maintenance of systems such as irrigation districts. The third principle is that women play a central part in the provision, management, and safeguarding of water. In many parts of the world, women are the ones who collect and safeguard water for domestic and agricultural use. However, women do not have as important a role as men in water management, and their involvement has been shown to be important for achieving higher levels of efficiency and sustainability. The fourth principle, perhaps the most controversial at that time, and the one that has created a major impact in water management, is that water has an economic value in all its competing uses and should be recognized as an economic good as well as a social good. This principle highlights the need of assigning value to water not only as a social good but also as an economic good if water is to be allocated efficiently and if water systems need to be sustainable in all dimensions. This principle has generated debate for its implicit recommendation of charging for water.

1.08.5 Water Management and the Environment Almost by definition, water management impacts the environment; changes in agriculture cause modifications in land cover and watercourses, and degrade ecosystems and impact their services for human development. In order to increase productivity, land and water are manipulated in different ways with consequences on the environment. Some of these manipulations include the following (Falkenmark et al., 2007). (1) Shifting the distribution of plants and animal, when the native vegetation is cleared and is replaced by crops and wild animals are replaced with livestock. (2) Coping with climate

Managing Agricultural Water

variability to secure water for crops. As water is a key material for photosynthesis, crop productivity depends intimately on securing water to ensure growth. (3) Maintaining soil fertility. The conventional way to secure enough air in the root zone is by drainage and ditching through plowing to ensure that rain water can infiltrate, a process that leads to erosion and the removal of fertile soil by strong winds and heavy rain. (4) Coping with crop nutrient needs. The nutrient supply of agricultural soils is often replenished through the application of manure or chemical fertilizers. (5) Maintaining landscape– scale interactions. When natural ecosystems are converted to agricultural systems, some ecological processes (such as species mobility and subsurface water flows) that connect parts of the landscape can be interrupted. This can have implications for agricultural systems as it can affect pest cycles, pollination, nutrient cycling, and water logging and salinization (Lansing, 1991; Cumming and Spiesman, 2006; Anderies, 2006). The Millennium Ecosystem Assessment, an international assessment by more than 1300 scientists of the state of the world’s ecosystems and their capacity to support human wellbeing, identified agricultural expansion and management as major drivers of ecosystem loss and degradation and the consequent decline in many ecosystem services and human well-being (MEA, 2009). This study showed that by year 2000, a quarter of the global land cover had been converted for cultivation, with cropland covering more than 50% of the land area in many river basins in Europe and India and more than 30% in the Americas, Europe, and Asia. It also showed that the development of water infrastructure and the regulation of rivers for many purposes, including agricultural production, fragmented rivers and generated impoundment of large amounts of water (Revenga et al., 2000; Vo¨ro¨smarty et al., 2005). Failure of agricultural water management increases the effect of natural and human-induced disasters, such as droughts and famine, on poor people. The rural poor, who are highly vulnerable, have a high dependency on the ecosystems where they farm or live and adequate water management diminishes the negative impact of low precipitation events (Silvius et al., 2000; WRI et al., 2005; Zwarts et al., 2006). Issues related to gender in management are also important in managing water for agriculture, as well as recognizing the importance of extreme events for the evolution of ecosystems. Water use for irrigation for agricultural production alters the excess supply of water for other uses, such as surface and groundwater for aquatic use downstream. The spatial water distribution changes significantly after an irrigation system is built and reduces the availability of water for downstream uses, impacting ecosystems. Irrigation infrastructure, dams, and canals alter the waterscape and impact the ecosystem, and irrigation water alters also the hydrologic cycle at the basin level. Part of the transfer water from streams or aquifers moisturizes soil and infiltrates, and the rest either is used by the plants or is evaporated. Unfortunately, increasing agronomic yields require not only water use but also the application and increased use of agrochemicals. Irrigation indirectly contributes to the application of chemicals and creates a negative impact on water quality with externalities downstream where water is also needed. The challenge is then how to increase agricultural

145

production without compromising the sustainability of ecosystems needed by society in general. General recommendations are being contemplated to deal with the negative impact of water use in agriculture (Falkenmark et al., 2007). The implementation of an integrated approach to manage land and water resources and ecosystems that acknowledge the multifunctionality of agroecosystems in supporting food production and ecosystem resilience are recommended. To accomplish this, a better understanding of the services that are generated by agroecosystems and the value of biodiversity is required. Unfortunately, zero-environmental-cost agricultural water management is almost unobtainable. This obligates to understand the value of maintaining biodiversity and the interaction of ecosystems and their future water requirement to sustain ecosystem health and biodiversity and to evaluate how these benefits compare to short-term agricultural development. Given that new irrigation systems are converting land from its original ecosystems and regulating rivers are creating an impact on spatial water distribution, it is recommended to improve existing agricultural systems to obtain an increase in production as opposed to the expansion of agriculture. There is significant room for improvement in terms of technologies and management practices with techniques more environmentally safe to lower the impacts from the additional agricultural production needed to satisfy increasing food demand. There is also a need to reduce or reverse the ecosystem degradation through rehabilitation and restoration of agricultural systems. Mitigation of negative impacts on ecosystem services is achieved through an integrated management of land, water, and other dimensions of ecosystems at the watershed level. This approach requires a participatory process in which users can understand the benefits and costs of various development options, and to foster stakeholder discussion about the tradeoffs. Tools are available to assist this process, from economic evaluation to environmental assessment techniques, and should be complemented with more sociologic and human interaction methodologies. New tools need to be developed to monitor the evolution of ecosystems and to generate the required feedback to take the appropriate corrective decisions. Unfortunately, the agricultural development process and its impact on the environment are full of uncertainties that need to be dealt with when evaluating trade-offs and taking decisions. Using existing and developing new techniques to deal with uncertainty within the planning process is necessary. Tools such as scenario planning could improve the assessment and learning of the development decisions and their potential impacts on the environment. Increasing food demand requirements and minimizing its negative impact on the environment require building new institutions and organizations that could better address the different development options and their impacts on the ecological functions. To generate the best setting for decision making, it is necessary to educate all stakeholders about the ecological services of ecosystems, and the social and economic benefits of alternative policy options. A higher level of abstraction is needed so that interested people from different disciplines can internalize the relevant information for decision making.

146

Managing Agricultural Water

1.08.6 Water for Agriculture and Poverty Reduction High-income countries have shown important increases in agricultural productivity and lower-income countries have benefited less from improved varieties from the Green Revolution, which has been a major contributor to growth in staple crop production in the developing world. Production growth in sub-Saharan Africa was based almost exclusively on the extension of cultivated area, and in Asia, irrigation was part of the explanation since most of the innovations in agricultural varieties demanded supplemented water via irrigation. In developing countries, the population will remain predominantly rural until 2020 when the size of the rural population declines due to slower population growth and rapid urbanization (World Bank, 2008). Poverty is more prevalent in the rural areas and the World Bank (2008) estimates that gross domestic product (GDP) growth generated in agriculture is, on average, 4 times more effective in alleviating poverty and benefiting half of the poorest population than growth generated outside agriculture. This effect declines as the countries have higher incomes. There is a clear relationship between water management and poverty reduction. Case studies from research and international institutions show that water management is a good investment that not only contributes to poverty reduction, but also is generally cost efficient, and has the potential to generate wealth. As a result of this linkage, the global community has united to fight poverty through actions that bring different interests and organizations together in effective partnerships around the MDG agenda. Following the conceptual framework developed by the Poverty–Environment Partnership (2002), water management contributes to poverty reduction in four dimensions:

•

•

Enhanced livelihood security. Water is considered an input to livelihood activities and a determinant of health and productivity of ecosystems. Water in agriculture positively impacts the ability of poor people to use their agricultural land and capabilities to make a living in conditions of greater security and sustainability. Ensuring continuity in water flows and minimum levels of water quality is essential for maintaining the integrity of ecosystems. Making sure that adequate and reliable water supplies are available for agricultural activities (including livestock, aquaculture, horticulture, and other types of production) is key to poverty reduction throughout the developing world. Therefore, the adequate design of water irrigation districts makes water available for activities that impact directly the livelihood of rural habitants. Reduced health risks. Water-borne and water-related vectorborne diseases, such as diarrhea and malaria, are the main killers in many parts of the developing world, and, in particular, affect children and other vulnerable groups. Some of these diseases have their origin in irrigation systems and adversely affect the most vulnerable population, women, and children, generating disabilities, poor nutritional conditions, and, eventually, death. Well-designed irrigation systems, with the right hydraulic infrastructure and with an adequate administration and operation, together with educational campaigns, is the most effective

strategy to improve health in the rural areas. It has been found that, in many cases, the economic benefit from these ameliorating activities is higher than the benefit derived from an increase in the regional economic value-added in many parts of the developing world. Integrating water management in agriculture with health systems development is one of the most effective strategies to reduce poverty through health improvement of the rural population. Every day diarrheal diseases cause nearly 5000 deaths, mostly among children under 5 years of age (United Nations, 2006). The World Health Organization estimates that the number of deaths from infectious diarrheas amounts to 1.8 million for all age groups, with a heavy toll among children under 5–1.6 million deaths. Malaria kills about 1 million people in the world every year, mainly in Africa, and about 80% of these cases are children under 5 years of age. It is estimated too that 160 million people are infected with schistosomes and 133 million suffer from high-intensity intestinal infections. These diseases have a direct impact on the welfare of the population and a long-term effect on future generations.

•

•

Reduced vulnerability. Water in agriculture is often too much, too little, or too contaminated. It is rare to observe equilibrium of natural water supply with the right amount of water potentially demanded by the agricultural systems. This variability of water supply generates environmental, economic, and social threats as a result of sudden impact shocks and long-term trends. Droughts, floods, and highintensity rainfall have the potential to destroy livelihoods and increase poverty in the rural areas. In addition, water management in agricultural systems should compensate for the negative impacts of climate change, pollution, and solid degradation. Well-designed and managed agricultural systems reduce the resilience of the poor and help stabilize the income variability of rural landholders, and participants of the agricultural value chain. Pro-poor economic growth. Agricultural water management has the potential to generate the economic growth needed to reduce poverty in the rural sector and, in particular, to create income-generating opportunities for the poor. Water supply is a necessary input in productive activities and allows technologies adoption. Agricultural water management opens the opportunity for entrepreneurs to participate in ventures upstream and downstream of the agricultural value chain.

The potential economic dynamism generated by the induced economic activity by an improved water management will generate returns and benefits to the local economy with many multiplier effects. National and regional economies benefit with significant improvement in the management of agricultural systems, especially if these activities are complemented with measures to make input markets work, additional diversification opportunities, and investment in related infrastructure. Water management improvement should be accompanied by institutional strengthening and knowledge-generation initiatives to increase its impact on poverty reduction. Based on the previous framework, there are evident opportunities to reduce poverty through water management in

Managing Agricultural Water

agriculture. The UN Task Force on the MDG found four core areas on which it is necessary to work to eliminate the constraints (United Nations, 2006). First, it is necessary to work on policy, legal and regulatory reform, including issues of rights of access to water. Attention should be given to work on policies that target needs and opportunities of the poor for improved access to water for agriculture. Second, it is necessary to plan and select technology choices that are consistent with poverty reduction targets, assessing the possible impacts on the most vulnerable people and the water resources. Third, it is necessary to develop financial mechanisms such as investment incentives and cost-recovery mechanisms to have credit and financial management systems and to create a regulatory regime and climate where private investment is encouraged. Finally, institutional reform and coordination among government agencies are needed to deal with agricultural water management to support the investment and involvement from the private sector and other regional actors, and improve the management of existing systems.

1.08.7 Water Management of Rainfed Agriculture Water productivity in rainfed agriculture is low despite being the traditional practice on 80% of the world’s agricultural area, which presents an opportunity for investment to improve productivity and boost agricultural yields. In many communities in developing countries, rainfed agriculture remains as the primary source of food, particularly grain. Historically, many developing countries that practiced rainfed agriculture had chosen to increase production through brute force by simply expanding their agricultural area. Despite yields having increased by a significant amount in these countries, they still remain far from the yields found in developed countries such as the United States and Europe, which serves to demonstrate the potential for yield growth in developing areas (Rockstro¨m, 2007). On average, these developing countries have only reached approximately 30% of their achievable yields, while some countries such as Yemen and Pakistan only reaching approximately 10% of their achievable yield. Furthermore, historical evidence has demonstrated a ‘‘growing yield gap between farmers’ practices and farming systems that benefit from management advances’’ (Wani et al., 2003), which serves to show that neglecting water management practices not only prevents farms from achieving higher yields, but in fact it diminishes their yield as well. Moreover, it has been shown that improving rainfed agriculture has led to decreased poverty rates (Irz and Roe, 2000), which further emphasizes the need for investment in rainfed agriculture to heighten productivity. The key problem facing farmers that has hindered yield growth cannot simply be generalized as the amount of rainfall: depending on where in the rainfall zone spectrum an area lies, the core limiting factor varies. In arid regions, the amount of rainfall does in fact prove to be the largest problem for rainfed agriculture. However, in other regions, such as the semiarid and dry subhumid zones where the minimum crop water need is already met, the problem of yield growth becomes focused on high variability of rainfall. Challenging previous beliefs that the amount of rainfall was the main

147

limiting factor for higher yields, this sporadic availability of water has helped to frame rainfed agriculture by demonstrating that basic water management strategies can double yields, on average. This variability in rainfall, characterized by seldom and high-intensity rainfall as well as frequent dry spells and high-intensity droughts, creates an insufficient water supply to meet water demand, otherwise known as water stress. Water stress has two primary divisions: one is the dichotomy between dry spells, which are 2–4-week periods during the rainy season that temporarily halts production during critical stages of growth, and droughts, which are a less common (once a decade) lack of rainfall that entirely halts crop growth, while the second division is one between man-made variability (agricultural) and natural variability (meteorological). Further exacerbating the problem of water stress is the variability in amount of rain that reaches the crops’ roots through soil moisture, which is generally only 70–80% of total rainfall, and sometimes can be as low as 40% in areas of poorly managed land. These problems incited by poor water management techniques are referred to as agricultural dry spells and droughts, and provide additional evidence that investment in water management can drastically improve yields. When rainfall is not efficiently used for plant absorption due to faulty management practices such as poor soil fertility or land degradation, the shortage of food is blamed on an agricultural drought. For example, in semiarid regions, nonplant growth rainfall, which includes drainage, nonproductive evaporation, and runoff, account for up to 85% of rainfall. Transpiration, which directly attributes to plant growth, only accounts for up to 30% of the rainfall. In certain areas with severely degraded land, only 5% of rainfall is used productively for plant growth. Rainfed agriculture yields are often limited by deficient soil fertility that comes as a result of nutrient depletion, loss of organic matter, and other forms of soil degradation. Studies in India have shown that this phenomenon is oftentimes human induced, where practices such as subsistence farming have depleted the soil of many of the necessary plant growth nutrients. Fortunately, problems concerning soil fertility can be easily rectified through investment in water management, such as injecting micronutrients into the deficient soil, which has been shown to substantially increase crop yields and rainwater productivity in India. As a result of this increased production, economic returns also increased by 50–75%. These studies further emphasize the capabilities and the need for investment in rainfed systems. The solution to the low productivity of rainfed systems is much easier said than done. A multitude of challenges stand in the way of simply investing more in rainfed agriculture and increased yields, the most fundamental of which is the adaptation and implementation of new agricultural innovations to every farm worldwide. Oftentimes, what drive innovation to be realized and adopted are social and ecological crises. However, simply waiting until the advent of the next crisis is not a viable option. Moreover, most farming households are small and marginalized and do not necessarily have access to the latest agricultural innovation. Rainfed areas also tend not to have the necessary infrastructure to implement

148

Managing Agricultural Water

innovations because throughout history, only high-potential irrigated areas have received large innovations, leaving these smaller areas in the dust. Other roadblocks include ‘‘limited information of the options available, social and economic constraints to adoption, lack of enabling environments and backup services, poor market linkages, and weak infrastructure’’ (Rockstro¨m, 2007). Developing avenues for better water management practices, such as through governance, policy, institutions, practices, and technologies, requires more attention than what has currently been devoted to it. Current strategies aim to make more efficient use of rainfall by increasing plant water availability and plant water uptake capacity through external and in-site water harvesting systems, soil and water conservation, evaporation management, and integrated soil, crop, and water management. While there are efforts to reduce water loss by increasing canopy cover, most strategies focus on securing more water produced by rainfall. In order for these strategies to be successfully implemented, they require a larger focus on small-scale water harvesting. Since small households make up a large proportion of farmers, unless these techniques are financially and structurally feasible, they simply cannot be used. The structure of rainwater harvesting is divided into three parts: ‘‘a watershed area used to produce runoff, a storage facility (soil profile, surface reservoirs, or groundwater aquifers), and a target area for beneficial use of the water (agriculture, domestic, or industry)’’ (Siegert, 1994). Another angle of attack that accounts for rainfall variability includes maintaining yields throughout dry spells and droughts. One important strategy lacking prevalence is the technique of supplemental irrigation, which collects runoff from external rainwater harvesting systems and transfers it to rainfed cropland. This technique, when compared with strictly rainfed systems, has shown substantially higher crop yields, with increases in yield ranging from approximately 30% to 400% (Oweis and Taimeh, 1996). Arguably, the most valuable quality of supplemental irrigation is its ability to supply water to a region when it is undergoing a dry spell, thus diminishing the frequency of interruptions to crop cycles. Studies have shown that 50–200 mm of supplemental irrigation is enough to combat the yield-reducing effects of dry spells (Oweis and Taimeh, 1996). Furthermore, the cost of supplemental irrigation systems is low, even for family-sized farms and small communities, in that the cost of a reduced yield due to dry spells is effectively eliminated. The technique of supplemental irrigation has already been adopted and implemented by a multitude of commercial farms in countries such as Australia, South Africa, and India; however, the benefits of such a system must be realized through simultaneous practice of other water-management techniques in order to maximize crop yield.

1.08.8 Policy Actions for the Future IWMI (2007) has completed an assessment of water management for agriculture and recommended eight policy areas, recognizing first that different actions are required for different situations, that actions for the sub-Saharan Africa will be

needed in terms of infrastructure, and where infrastructure has already being developed, as in Asia, the improvement of water availability relied on increases in productivity, relocating supplies, and rehabilitating systems. The eight policy actions are discussed below:

•

•

•

•

•

•

•

•

Change the way we think about water and agriculture. To achieve food security, reduce poverty, and conserve the ecosystems, it is necessary to think differently about water. A broad focus should be adopted, with reformed institutions and viewing rain as the ultimate source of water that should be managed. Agriculture should be seen as a system integrated with other uses, providing services and integrating with other ecosystems. Fight poverty by improving access to agricultural water and its use. To effectively fight poverty, it is necessary to secure water access to smallholder farmers, using pro-poor technologies, and investing in roads and markets. In addition, systems that have multiple uses, such as aquaculture and agroforestry, are an ideal solution to alleviate poverty. Good governance is fundamental to any poverty-reduction strategy. Good governance means creating a fair legal, policy, and regulatory framework in which the rights of people to access water resources for agriculture are secured. It deals also with improving the effectiveness, accountability, and transparency of government agencies, ensuring the participation of the poor in decision making and enhancing the role of civil society guarantying basic security and political freedom (United Nations, 2006). Manage agriculture to enhance ecosystem services. Agricultural management could enhance other ecosystems to promote services beyond the production of traditional agricultural products. Water and land use will probably be intensified in the future but should be articulated with other services. Increase the productivity of water. More food per cubic meter of water should be the goal to increase effective water supply and reduce effective water demand. It is estimated that 35% increase in water productivity could reduce additional crop water consumption up to 80%. The poor can benefit from water productivity gains as well as largeholder farmers by introducing higher-value products. Upgrade rainfed systems – a little water can go a long way. The way to upgrade rainfed systems is by conserving soil moisture and providing supplemental irrigation where feasible. This action has a great potential to increase the income of an important number of farmers, especially in water-deficit regions such as sub-Saharan Africa and parts of Asia. Mixed crop and livestock systems have a good potential for improving the productivity of these systems. Adapt yesterday’s irrigation to tomorrow’s needs. Emphasis should be given to adapt current irrigation systems to future demands. Rehabilitation, modernization, and management techniques should be used to increase water productivity and its impact on poverty alleviation. Reform the reform process – targeting state institutions. Government institutions should be reformed if new management approaches are to be adopted and implemented. A collaborative process of all actors, private sector, and other institutions, is important to ignite a renovated process to satisfy local needs.

Managing Agricultural Water

•

Deal with trade-offs and make difficult choices. There should be good information to allocate water among conflicting uses. Informed multi-stakeholders are necessary to achieve effective negotiations and to make decisions about the use and allocation of water. Reconciling competing demands on water requires transparent sharing of information.

1.08.9 Summary Agricultural water management deals with the administration of a key input to agricultural production, and up to now it has been responsible for the majority of water use in the world, has improved nutrition, alleviated poverty, and increased production in 2.5 times the level in four decades. Agricultural water management has had the simultaneous challenge of meeting future food needs, making farms profitable and reducing poverty. New technologies are needed to intensify land and water use and to increase agronomic yields to produce the additional food to feed the world population and minimizing the impact on the environment. However, most importantly, agricultural water management needs to provide efficient solutions to improve the income levels of the farmers and rural people. These water resource challenges require better allocation of the resource using the social and economic value of the resource as distribution criteria. Agriculture has induced the degradation of many ecosystems, including those that are essential for food production. Aquifer contamination and depletion, drainage of wetlands, surface water contamination, land degradation and erosion, and aquatic ecosystems conflict are just some examples of these impacts. Therefore, water in agriculture should be managed by taking into account the array of services of the ecosystem, which are crucial for society. Adequate planning needs to take into account the trade-off between water for food production and ecosystem services, and address the social impact of poor rural people who often suffer the consequences of environmental negative externalities. Agricultural water management has the potential to become the best road to improve agricultural practices through technology adoption and diminish the environmental impact of the use of water resources. Water managers for agriculture will face many challenges while trying to accomplish multiple objectives. Sacrifices will be needed and so the appropriate institutional arrangement is to deal with these trade-offs. That is why an integrated approach to land use, water allocation, and ecosystem sustainability will be essential. In addition, understanding the tradeoffs between the conflicting goals will require an increase in the body of available knowledge to have a more accurate approximation as to the effect of irrigation to the environment and the effect of climate change on the future water supply for agriculture. The trend of irrigation investment has decreased by half during the past two decades compared to the growth experienced in the 1960s and 1970s. Constraints are larger as easy sites are already been exploited, groundwater sources are unsustainable in many places and the world is now more concerned with environmental and social issues that set a higher

149

standard in agricultural water management, a situation that generates the incentive for new technology development. At the same time, nation and regional development processes have generated new uses for water that are competing with agricultural use. One of the great setbacks of agricultural water management is the severe failure of many institutional arrangements at the national and local basin levels to deal with water management. New rules and incentive mechanisms need to be put in place in order to face the future challenge to reflect our understanding of the interaction of water resources with ecosystems and human activity. Irrigation systems involve people with different interests, and therefore good governance is needed as a framework (institutional and administrative) to allow cooperation, coordination, and, most importantly, conflict resolution. Finally, water management for agriculture could become an interesting tool to achieve regional differentiation, allowing competition and incentivizing innovation. Water management could help building the strengths needed to create a differentiated business environment and therefore creating the conditions to achieving higher levels of regional and national prosperity through increments in productivity.

References Anderies JM (2006) Minimal models and agroecological policy at the regional scale: An application to salinity problems in southeastern Australia. Regional Environmental Change 5(1): 1--17. Allan JA (2001) The Middle East Water Question: Hydro-Politics and the Global Economy. London: I. B. Tauris. Allan JA (2004) Virtual water: Part of an invisible synergy that ameliorates water scarcity. Paper presented at the Water Crisis Workshop. Santander, Spain. Anderson K, Martin W, and Valenzuela E (2006) The Relative Importance of Global Agricultural Subsidies and Market Access. Washington, DC: World Bank. Bain J, Caves R, and Margolis J (1996) Northern California’s Water Industry. Baltimore: The Johns Hopkins Press. Barker R., Molden DJ (2003) Comprehensive Assessment of Water Management in Agriculture Series No. 1. pp. 37–51. Wallingford and Cambridge, MA: CABI Publishing. Bastiaanssen W, Noordman EJM, Pelgrum H, Davids G, Thoreson BP, and Allen RG (2005) SEBAL model with remotely sensed data to improve water-resources management under actual field conditions. Journal of Irrigation and Drainage Engineering 131(1): 85--93. Centre for World Food Studies (2008) Brief: China’s rapidly growing meat demand: A domestic or an international challenge? Amsterdam. Chamber SR and Just R (1989) Estimating multioutput technologies. American Journal of Agricultural Economics 71(Nov.): 980--995. Ciriacy-Wantrup S (1954) Cost allocation in relation to western water policies. Journal of Farm Economics 36(Feb.): 108--129. Cleaver K and Gonzalez F (2003) Challenges for Financing Irrigation and Drainage. Washington,DC: Agriculture and Rural Development Department, World Bank. Cumming GS and Spiesman BJ (2006) Regional problems need integrated solutions: Pest management and conservation biology in agroecosystems. Biological Conservation 131: 533--543. Deardorff AV (1982) The general validity of the Heckscher–Ohlin theorem. American Economic Review 72(4): 683--694. Dyson T (1996) Population and Food: Global Trends and Future Prospects. London: Routledge. Easter KW, Rosegrant MW, and Dinar A (1998) Markets for Water: Potential and Performance. Boston, MA: Kluwer. Falkenmark M, Finlayson CM, and Gordon LJ (2007) Agriculture, water, and ecosystems: Avoiding the costs of going too far. In: David M (ed.) Water for Food, Water for Life, ch. 6. London: Earthscan and International Water Management Colombo: Institute.

150

Managing Agricultural Water

Florencio-Cruz V, Valdivia-Alcala´ R, and Scott CA (2002) Productividad del agua en el Distrito de Riego 011, Alto Rio Lerma. Agrociencia 36: 483--493. Hayami Y and Ruttan V (1985) Agricultural Development: An International Perspective, Revised and extended edition. Baltimore: John Hopkins University Press. Howlett M and Ramesh M (1995) Studying Public Policy: Policy Cycles and Policy Subsystems. Oxford: Oxford University Press. Huppert W, Svendsen M, and Vermillion DL (2001) Governing Maintenance Provision in Irrigation: A Guide to Institutionally Viable Maintenance Strategies. Wiesbaden, Germany: Universum Verlagsanstalt. IFPRI (1995) Global food security. IFPRI Report, October 1995. Washington, DC: International Food Policy Research Institute. IFPRI (1997) China will remain a grain importer. IFPRI Report, February 1997. Washington, DC: IFPRI. International Labour Organization (2007) The Key Indicators of the Labour Market, 5th edn. Geneva: International Labour Office. Irz X and Roe T (2000) Can the world feed itself? Some insights from growth theory. Agrekon 39(3): 513--528. IWMI (International Water Management Institute) (2007) Water for Food, Water for Life. A Comprehensive Assessment of Water Management in Agriculture. London: Earthscan, and Colombo: International Water Management Institute. IUCN and WBCSD (World Business Council for Sustainable Development) (2008) Agricultural ecosystems: Facts and trends. International Food Policy Research Institute (IFPRI) Annual Report. IUCN and WBCSD. Jaspers F (2001) Institutions for Integrated Water Resources Management Training Manual. Delft: UNESCO-IHE. Jayal NG (1997) The governance agenda. Making democratic development dispensible. Economic and Political Weekly 32(8): 407--412. Jenkins WI (1978) Policy Analysis: A Political and Organizational Perspective. London: Martin Robertson. Kaufmann D (2005) Debunking myths on worldwide governance and corruption: The challenge of empirics and implications. World Bank. Institute David B. Goodman Lecture, Bennett Lecture Hall, University of Toronto. Kloezen W and Garce´s-Restrepo C (1998) Assessing Irrigation Performance with Comparative Indicators: The Case of the Alto Rio Lerma Irrigation District Mexico. Colombo, Sri Lanka: IWMI. Lansing JS (1991) Priests and Programmers: Technologies of Power in the Engineered Landscape of Bali. Princeton, NJ: Princeton University Press. Lau L (1978) Applications of profit functions. In: Fuss M and McFadden D (eds.) Production Economics: A Dual Approach of Theory and Applications, 1, Amsterdam: North-Holland. Leontief WW (1953) Domestic production and foreign trade: The American capital reposition re-examined. In: Caves RE and Johnson HG (eds.) Readings in International Economics. London: Allen and Unwin. MEA (Millennium Ecosystem Assessment) (2009) Ecosystems and Human Well-Being: A Framework for Assessment. Washington, DC: Island Press. Melvin JR (1968) Production and trade with two factors and three goods. American Economic Review 58: 1248--1268. Merrey DJ (2007) Policy and institutional reform: The art of the possible. In: Molden D (ed.) Water for Food, Water for Life. A Comprehensive Assessment of Water Management in Agriculture, ch. 5. London: Earthscan, and Colombo: International Water Management Institute. Molden D, Murray-Rust H, Sakthivadivel R, and Makin I (2003) A water productivity framework for understanding and action. In: Kijne JW, Barker R, and Molden D (eds.) Water Productivity in Agriculture: Limits and Opportunities for Improvement, pp. 1--18. Wallingford: CABI Publishing. Molle F and Mollinga P (2003) Water poverty indicators: Conceptual problems and policy issues. Water Policy 5: 529--544. Moore M (1999) Estimating irrigators ability to pay for reclamation water. Land Economics 75(Nov.): 562--578. North D (1990) Institutions, Institutional Change, and Economic Performance. Cambridge: Cambridge University Press. OECD Organization for Economic Co-Operation and Development (2001) Agricultural Policies in Countries: Monitoring and Evaluation. Paris: OECD. Oweis T and Taimeh A (1996) Evaluation of a small basin water-harvesting system in the arid region of Jordan. Water Resources Management 10(1): 21--34. Porter M (1998) On Competition. Boston: Harvard Business School Publishing. Postel S (1999) Pillar of Sand: Can the Irrigation Miracle Last? New York, London: W.W. Norton. Postel S and Richter B (2003) Rivers for Life: Managing Water for People and Nature. St. Louis: Washington University Island Press. Poverty–Environment Partnership (2002) Linking Poverty Reduction and Environmental Management: Policy Challenges and Opportunities. Washington, DC: The International Bank for Reconstruction and Development/The World Bank.

Ramirez-Vallejo J and Rogers P (2004) Virtual water flows and trade liberalization. Journal of Water Science and Technology 49(7): 25--32. Ramirez-Vallejo J and Rogers P (2009) Failure of the virtual water argument: Possible explanations using the case study of Mexico and NAFTA. In: Biswas AK and Cline SA (eds.) Global Change: Implications for Water and Food Security. Washington: IFPRI. Revenga C, Brunner J, Henniger N, Kassem K, and Payner R (2000) Pilot Analysis of Global Ecosystems, Freshwater Systems. Washington, DC: World Resources Institute. Rockstro¨m J (2007) Water in rainfed agriculture. In: Molden D (ed.) Water for Food, Water for Life. A Comprehensive Assessment of Water Management in Agriculture, ch. 8. London: Earthscan, and Colombo: International Water Management Institute. Rogers P (2002) Water Governance in Latin America and the Caribbean. Washington, DC: Inter-American Development Bank, Sustainable Development Department, Environment Division. Rogers P and Hall A (2003) Effective Water Governance, GWP Technical Committee Background Paper 7. Stockholm: Global Water Partnership. Rosegrant M, Cai X, Cline S, and Nakagawa N (2001) The role of rainfed agriculture in the future of global food production (invited background research paper). In: International Freshwater Conference. Bonn, Germany. Rosegrant MW, Agcaoli M, and Perez ND (1995) Global Food Projections. Canada: Renouf Publishing. Rosegrant MW and Binswanger H (1994) Markets in tradable water rights: Potential for efficiency gains in developing country water resource allocation. World Development 22(11): 1--11. Rosegrant MW and Cai X (2002) Global water demand and supply projections: Part 2. Results and prospects to 2025. Water International 27(2): 170--182. Rosegrant MW and Cai X (2004) Implications of water development for food security. In: Lawford R, Fort D, Hartmann H, and Eden S (eds) Water Science, Policy, and Management,. Water Resources Monograph Series, vol. 16, 422pp. Washington, DC: American Geophysical Union. Seckler D, Molden D, and Sakthivadivel R (2003) The concept of efficiency in waterresources management and policy. In: Kijne JW, Barker R, and Molden DJ (eds.) Water Productivity in Agriculture: Limits and Opportunities for Improvement, Comprehensive Assessment of Water Management in Agriculture Series No. 1, pp. 37–51. Wallingford and Cambridge, MA: CABI Publishing. Siegert K (1994) Introduction to water harvesting: Some basic principles for planning, design and monitoring. In: Water Harvesting for Improved Agricultural Production, Proceedings of the FAO Expert Consultation, Water Report 3. Cairo, 21–25 November 1993. Rome: Food and Agriculture Organization. Silvius MJ, Oneka M, and Verhagen A (2000) Wetlands: Lifeline for people at the edge. Physical Chemistry of the Earth B 25(7–8): 645--652. Travis WP (1964) The Theory of Trade and Production. Cambridge: Harvard University Press. United Nations (2006) Millennium Development Goals Report. New York: United Nations. United Nations (2008) Trends in Sustainable Development. Agriculture, Rural Development, Land, Desertification and Drought. New York: United Nations United Nations Development Programme (2001) Partnerships to Fight Poverty. Annual Report. New York. United Nations Development Programme (2001) Partnerships to Fight Poverty. Annual Report. New York. Vo¨ro¨smarty CJ, Le´veˆque C, and Revenga C (2005) Fresh water. In: Hassan R, Scholes R, and Ash N (eds.) Ecosystems and Human Well-Being: Current State and Trends – Findings of the Condition and Trends Working Group. Washington, DC: Island Press. Wallace JS and Batchelor CH (1997) Managing water resources for crop production. Philosophical Transactions of the Royal Society London B 352: 937--947. Wana J, Griffithsa R, Yinga J, McCourtb P, and Huanga Y (2009) Development of drought-tolerant canola (Brassica napus L.) through genetic modulation of ABAmediated stomatal responses. Crop Science Society of America 49: 1539--1554. Wani SP, Pathak P, Sreedevi TK, Singh HP, and Singh P (2003) Efficient management of rainwater for increased crop productivity and groundwater recharge in Asia. In: Kijne JW, Barker R, and Molden DJ (eds.) Water Productivity in Agriculture: Limits and Opportunities for Improvement, Comprehensive Assessment of Water Management in Agriculture Series No. 1, pp. 199–215. Wallingford and Cambridge, MA: CABI Publishing. Wichelns D (1999) Economic efficiency and irrigation water policy with an example from Egypt. Water Resources Development 15(4): 543--560. Winpenny J (2003) Report of the World Panel on Financing Water Infrastructure. World Water Council. World Bank (2003) Managing Water as an Economic Good: Rules for Reformers, Water Resources Sector Strategy. Washington, DC: World Bank.

Managing Agricultural Water World Bank (2005) Shaping the Future of Water for Agriculture: A Sourcebook for Investment in Agricultural Water Management. Washington, DC: World Bank. 10pp. World Bank (2006) Reengaging in Agricultural Water Management. Challenges and Options. Washington, DC: World Bank. World Bank (2008) World Development Report 2008. Agriculture for Development. Washington, DC: World Bank. World Bank (2009) Electronic Learning Guidebook for Participatory Irrigation Management, http://www.worldbank.org/wbi/pimelg/charg.htm (accessed March 2010). WRI (World Resources Institute), United Nations Development Programme, United Nation Environment Programme, and World Bank (2005) World Resources 2005: The Wealth of the Poor – Managing Ecosystems to Fight Poverty. Washington, DC: World Resources Institute.

151

Young R (1996) Measuring Economic Benefits for Water Investments and Policies, World Bank Technical Paper No. 338. Washington, DC: World Bank. Zwarts L, Van Beukering P, Kone B, Wymenga E, and Taylor D (2006) The economic and ecological effects of water management choices in the upper Niger river: Development of decision support methods. International Journal of Water Resources Development 22(1): 135--156.

Relevant Website http://www.worldbank.org/ World Bank: Electronic Learning Guidebook for Participatory Irrigation Management.

1.09 Implementation of Ambiguous Water-Quality Policies DH Moreau, University of North Carolina, Chapel Hill, NC, USA & 2011 Elsevier B.V. All rights reserved.

1.09.1 1.09.2 1.09.2.1 1.09.2.2 1.09.3 1.09.3.1 1.09.3.2 1.09.4 References

Nonpoint Sources and the CWA Intrastate Cases Neuse River Nutrient Management Strategy Jordan Lake Stormwater Rules Interstate Nonpoint Management Mississippi River Basin and Hypoxia in the Gulf of Mexico Chesapeake Bay Program Summary and Conclusions

Ambiguities in water policies may create significant barriers to implementation and lead to unpredictable outcomes, especially in the United States’ federal system where the national government has made some parts of water policy unambiguous but left other parts to states with ambiguous mandates. The Clean Water Act (CWA) is a case in point. The goal of the act is to protect and enhance the quality of all waters of the United States to levels that are sufficient to support their state-designated uses, at a minimum to make them fishable and swimmable. Policies set forth in the Act establish several levels of technology-based effluent limits for point sources, publicly owned wastewater treatment plants, and a large number of categories of industrial dischargers. Enforceable permits issued to those sources contain effluent limits, monitoring protocols, and reporting requirements. The permit program also covers other sources such as urban stormwater runoff and concentrated animal feeding operations (CAFOs). There is little ambiguity if the least stringent effluent limits are sufficient to satisfy water-quality standards, but several ambiguities arise when minimal requirements covering these sources are not sufficient to satisfy water-quality standards. These ambiguities are especially important when sources of degradation are dominated by agriculture and urban stormwater runoff. In these cases, states are obligated to establish maximum allowable loads for those segments of water bodies that have been designated as being degraded. Then, states are obligated to allocate a portion of the allowable load to point sources, another portion to nonpoint sources, and reserve a third portion as a factor of safety. If more stringent effluent limits on point sources are insufficient to upgrade water quality to satisfy standards, then states are required to submit plans to control nonpoint sources to the maximum extent practicable (MEP). However, most agricultural sources are excluded from federal permits. Management of those sources is largely by voluntary participation in a variety of incentive programs. There are at least two unresolved issues. First, what is the intent of the Act? Are states expected to satisfy waterquality standards when available technologies are not sufficient to accomplish the task? Second, how do states control nonpoint sources to the MEP when those sources are largely exempt from regulation?

153 154 154 156 157 157 159 160 160

In this chapter, several nonpoint source management strategies are examined with respect to how ambiguities in the CWA either were addressed during implementation or have created as-yet insurmountable barriers to implementation. Particular attention is focused on policies for control of nonpoint sources of nutrients leading to hypoxia in receiving water and the management of urban stormwater. Cases include water-quality management for the Neuse River Basin, implementation of urban stormwater management in Jordan Lake watershed, North Carolina, control of nutrients in the Mississippi River Basin, and control of nutrients entering the Chesapeake Bay.

1.09.1 Nonpoint Sources and the CWA None of these issues were adequately addressed when fundamental changes to the federal water pollution control policy were made in 1972. That shortcoming is readily understood in the context of information about sources of pollution at the time. The Federal Water Quality Administration (FWQA) undertook the first national assessment of water quality in 1969. That assessment was largely a compilation of professional judgments by state water-quality officials. Results indicated that 33% of all stream mileage in the United States was polluted to some degree. One estimate in that assessment was that industrial sources accounted for 24% of degraded streams, municipal sources 22%, and agriculture 11% (USEPA, 1971). With that presumed factual basis, it is not surprising that Congress believed that strong regulatory limits on municipal and industrial sources would lead to substantial elimination of polluted waterways. Section 303(d)(1)(C) of Amendments to the Federal Water Pollution Control Act of 1972 (referred to as the CWA beginning in 1977) required establishment of total maximum daily loads (TMDLs) for all stream segments in which the application of technology-based effluent limits on municipal and industrial sources would be insufficient to implement the applicable water-quality standards. TMDLs had to account for seasonal variations in stream properties and a margin of safety to cover any lack of knowledge between effluent limitations and water quality. Section 302 stated that where technology-based standards were

153

154

Implementation of Ambiguous Water-Quality Policies

insufficient to satisfy water-quality standards, more stringent effluent limitations for point sources should be established at levels that can be reasonably expected to bring about compliance. USEPA introduced regulations to implement that portion of the act, using the following language (USEPA, 1975):

• • •

For each water quality segment, a total allocation for point sources of pollutants and a gross allotment for nonpoint sources of pollutants. A specific allowance for growth shall be included in the allocation for point sources and the gross allotment for nonpoint sources. The total of the allocation for point sources and the gross allotment for nonpoint sources shall not exceed the total maximum daily load.

The clear implication of the language in the statute and the regulations is that nonpoint sources could be addressed simply by reserving a portion of the TMDL for that purpose. The language did not address at least two possibilities: (1) that no economically reasonable available treatment technology for point sources would be sufficient to meet standards and (2) that loads from nonpoint sources on a stream segment would exceed the TMDL for that segment. Efforts to implement the policy quickly came up against realities of these two possibilities. The only option available to a state in this situation was to place a low priority on setting a TMDL for such a water body, thereby scheduling its development as far into the future as possible. When TMDL regulations were revised as of July 1987 (40 CFR 130.2), there was little change in the basic language. The loading capacity, defined as the maximum load of a pollutant on a stream segment that still satisfies related water-quality standards, is to be divided between a load allocation for nonpoint sources and background levels and a waste load allocation for point sources. These regulations went on to say that if best management practices (BMPs) to reduce nonpoint sources make more stringent load allocations for those sources practicable, then waste load allocations for point sources can be made less stringent. The regulations again failed to address the case where nonpoint source loads dominate the system and BMPs are insufficient to reduce actual loads below the TMDL. The latest version of TMDL regulations (40 CFR 130.7) states that TMDLs will be established ‘‘yat levels necessary to attain and maintain the applicable narrative and numerical water quality standardsy’’ taking into account all sources that are contributing to nonattainment of the standard. The directive to establish TMDLs does not carry with it any additional authority to control nonpoint sources. Amendments to the CWA in 1987, specifically the addition of Section 319, required states to take a number of additional steps to control nonpoint sources, including:

• •

preparation of an assessment report that identified navigable waters for which existing BMPs could not be expected to attain or maintain water-quality standards; identify BMPs and other measures to reduce to the MEP loads from each category of nonpoint sources; and

•

identify state and local programs for implementing BMPs and other measures.

States were then required to submit a management program ‘‘yto reduce pollutant loadingsy’’ from nonpoint sources ‘‘yto the maximum extent practicable.’’ Nowhere did the amendment state how much reduction was necessary. Section 319 did not require national technology standards or guidelines for nonpoint sources comparable to those for point sources in Section 302. Nor did it require any permitting, monitoring, inspection, and enforcement actions comparable to those for point sources as required under Section 402. State and USEPA officials were left to negotiate a mutually agreeable set of nonpoint source control measures, limited by whatever authorities and financial resources states and other federal agencies had to implement. Similar language was added in 1987 to Section 402(p) to address stormwater runoff from industrial activities and municipal separate storm sewer systems (MS4s). Although stormwater discharge was made subject to discharge permits, the operative provision for municipal permits is that dischargers are required to reduce pollutant loads to the MEP. That provision leaves open the question of what is the MEP. In particular, do the stormwater regulations apply to existing development as well as to new development? Portions of the stormwater provisions are rather straightforward, but, in addition to MET, other provisions are problematic. USEPA defined MS4s very broadly to include not only the conventional elements of urban stormwater management systems, but also roads with drainage systems, ditches, and man-made channels in urbanized areas that are owned by a state or any type of local government. At least two significant ambiguities have arisen from this definition and related requirements. First, urbanized areas do not necessarily coincide with boundaries of political jurisdictions to which the necessary regulatory authority has been delegated. Second, how would these provisions apply to large-scale real estate developments that convert rural areas to urbanized areas? In 1997 when EPA issued new policies for the TMDL process, it acknowledged that implementation of the program was moving at an unacceptably slow pace (USEPA, 1997). At the time, EPA regulations required each state to submit its list of impaired waters every 2 years and identify which of those would be scheduled for TMDL development over the following 2 years. The problem was that there was no schedule for all impaired segments. The revised policy urged states to develop schedules for all segments over a period of 8–13 years. Even then, implementation of nonpoint source controls was limited to the requirements in Section 319. When it issued proposed numerical nutrient criteria for the State of Florida in January 2010, EPA acknowledged that it could take many years before affected waters could be brought into compliance. Among many other provisions, the proposal asked for comments on Restoration Water Quality Standards that would allow achievement of water-quality standards in several phases over a period of up to 20 years so long as adequate progress was achieved at each stage (USEPA, 2010). This approach would allow considerable time to reach the final numerical criteria, but writing criteria does not grant additional authority to satisfy them.

Implementation of Ambiguous Water-Quality Policies

1.09.2 Intrastate Cases Two intrastate cases from North Carolina illustrate the use and limits of state authority to address problems of excessive nutrient loadings to lakes and estuaries, including urban stormwater runoff.

1.09.2.1 Neuse River Nutrient Management Strategy North Carolina’s Division of Environmental Management (DEM), the state’s water-quality management agency at the time, initiated a third-generation basinwide water-quality planning process in 1991 to coincide with 5-year renewals of the National Pollutant Discharge Elimination System (NPDES) discharge permits. The first of 17 basin plans was for the Neuse River Basin (NC Division of Environmental Management, 1993) shown in Figure 1. A supplemental classification of Nutrient Sensitive Waters had been assigned to the basin in 1983 to the Neuse River, and with that designation, all significant point sources had been required to meet a phosphorus effluent limit of 2.0 mg l 1 over the period of 1988–93. The 1993 report estimated that only 21% of the phosphorus and only 12% of nitrogen were coming from point sources. Agriculture accounted for approximately two-thirds of the balance. In compliance with USEPA policy, the report identified a wide range of state and federal programs to address nonpoint sources. It described 10 programs for agriculture, including the NC Agricultural Cost Share Program and provisions of the 1985 and 1990 Farm Bills. Four urban stormwater programs were discussed; runoff from construction was covered by the state sediment control act; and regulations on mining activities were discussed. Other programs for hydrologic modification, on-site disposal of wastewater, concentrated animal feedlots, solid waste, forestry, wetlands, and groundwater were also included in the mix.

The 1993 basinwide management plan, based on the CWA approach to nonpoint sources, failed to achieve its objective in late summer 1995 when massive fish kills occurred in the Neuse River estuary. Immediately thereafter, the North Carolina Environmental Management Commission (EMC) directed DEM to prepare a new plan to address deficiencies in the 1993 basin plan. The governor weighed in on the matter in the following March (Hunt, 1996). Draft administrative rules to implement the plan were sent to public hearing in May 1996 under the title of Neuse River Nutrient Sensitive Waters Strategy or the NSW Strategy. The state legislature provided legislative authority for a reduction of nitrogen by 30% from a 1991–95 baseline for both point and nonpoint sources (Session Law 1995 Chap. 572, ratified June 1966). This target was recommended by a special legislative committee that included knowledgeable academics and staff of DEM. Final rules were approved by the EMC in December 1997 (15A NCAC 2B.0232-0242 (Title 15, Chapter 2, Subchapter B, Sections of the North Carolina Administrative Code)), and specific state statutory authority was cited as the basis for these regulations. Several provisions of those rules established new initiatives to fill gaps in the basinwide management plan. For the first time, general agricultural operations were subject to waterquality regulations. Two options were made available to agricultural operators to satisfy the 30% reduction for nitrogen. They could act either individually to implement specified BMPs or collectively in a local (county) plan where some form of its trading among participants would be possible. A new program for certifying fertilizer applicators was established, covering all operations that applied fertilizer to 50 acres or more each year. Stormwater regulations were required for a list of municipalities and counties in the basin (they would later be covered by Phase II of USEPA regulations). Protection of existing riparian buffers was implemented in Section 233. A table of uses that are either exempt, allowable, allowable with mitigation, or prohibited applies to a

Durham

Wilson

Raleigh

Greenville

Goldsboro Kinston

Neuse River 0

Miles 10 20 30

Figure 1 Neuse River Basin, North Carolina.

40

155

New Bern

Neuse River Estuary

156

Implementation of Ambiguous Water-Quality Policies

50-ft-wide buffers directly adjacent to intermittent and perennial streams, lakes, ponds, and estuaries in the basin. A trading program for point sources was established in Section 02B.0235, building on experiences with the Tar RiverPamlico Sound trading program established in 1991. General powers delegated to the EMC by the state legislature were cited as the authority on which these administrative rules were adopted. EPA was in the process of formulating its policies for trading under the CWA. Unlike the Clean Air Act where national trading programs had been explicitly authorized to address acid rain, the CWA contained no mention of trading. EPA issued a policy statement on effluent trading in watersheds in January 1996, followed 4 months later by a draft framework for implementing the policy (USEPA, 1996). Basic principles for trading under the CWA were laid out in Chapter 2 of that document, but final rules were not formally proposed until proposed in the Federal Register until May 2002. Final trading rules were adopted in January 2003 (USEPA, 2003). Because many of the provisions in the NSW Strategy were new, they faced considerable obstacles and a few revisions before becoming effective. Considerable care was taken to involve stakeholders as the package of rules was being formulated, but public interest in finding a fix was very high. Not all stakeholders were enthusiastic about the rules adopted by the EMC. Under North Carolina’s Administrative Procedures Act, all administrative rules are subject to review by the Rules Review Commission (RRC), and no rule can become effective until after the next session of the legislature except in special cases where temporary rules are permitted. With very strong political support to get the rules in place, the legislature allowed all but the initial riparian buffer rule to become effective. This rule was slightly modified to meet objections by one group of stakeholders.

While the rules in Section 235 and 238 established a regulatory program on agriculture for the first time, they were limited to the installation of BMP and an annual reporting of what BMPs were in place and how much acreage was subject to BMPs. There is very limited instream monitoring to determine effectiveness of the rules, and inspections of operations subject to the rule are limited to sparse visits by agricultural agencies.

1.09.2.2 Jordan Lake Stormwater Rules A similar strategy was developed for the protection of B. Everett Jordan Lake, a critical source of public water supply in the Research Triangle of North Carolina. An important difference between the Neuse River and Jordan Lake nutrient loads is the relative importance of urban stormwater from existing development, a difference that was addressed in formulating a management plan for the lake. In the wake of actions in 1996 to address nutrient problems in the Neuse River estuary, the state legislature passed the Clean Water Responsibility Act (CWRA; also referred to as House Bill 515) in August 1997 to protect inland lakes that are designed as nutrient sensitive by the EMC. CWRA established default limits in the form of maximum concentration values for phosphorus and nitrogen in effluents from dischargers above such lakes. It also provided that alternate mass load limits could replace default values if they were based on a calibrated nutrient response model. Impacts estimated by mass loading would have to show compliance with the waterquality standard of 40 mg l 1 of chlorophyll-a. Primary targets of the CWRA were Jordan Lake in the Cape Fear Basin, shown in Figure 2, and Falls Lake in the Neuse Basin, both US Army Corps of Engineers multipurpose

Kernersville

Durham Burlington Greensboro Chapel Hill Jordan

Miles 0

5

10

15

20

Figure 2 Jordan Lake Drainage Area, Cape Fear Basin, North Carolina.

Lake

Implementation of Ambiguous Water-Quality Policies

reservoirs with a combined safe yield of about 165 million gallons per day for public water supplies in the area. Before Jordan Lake was found to be in violation of the water-quality standard in 2002, local governments in the watershed had already initiated development of a nutrient response model. A series of nutrient delivery models for point sources and nonpoint sources were developed over the period of 2001–03. A model of lake quality was delayed by errors in data collection, and a final report was not delivered until February 2005. Twenty-one stakeholder meetings were held in the process of developing the model (North Carolina Division of Water Quality, 2006). An initial draft of an NSW and TMDL strategy was submitted by the NC Division of Water Quality (DWQ) to the EMC in April of 2005, and draft rules were adopted by the Commission in November 2005. A set of technical issues were hashed out in meetings with stakeholders in 2006, before the proposed regulations were sent to formal public hearings in March 2007. The EMC approved a revised set of rules in May 2008, and rules that were passed were approved by the RRC in November 2008 in time to be considered by the 2009 session of the state legislature. Jordan Lake rules are very similar in structure to the Neuse Rive NSW Strategy with one very important exception. All prior regulations in NC directed toward control of pollutants in urban stormwater had been limited to the management of runoff from new development. However, in the process of making load allocations for Jordan Lake, DWQ could not find a feasible and equitable distribution of loads among sources without requiring reductions from existing urban development. An existing development rule was included as part of the management program. Several major urban areas in the watershed, including Burlington, Chapel Hill, Durham, and Greensboro would be affected by the proposed rule. The EMC proposed an adaptive approach to reducing loads from existing development. Municipalities and counties were given

157

3 years and 6 months from the effective date of the rule to conduct feasibility studies and prepare implementation plans. Following approval by DWQ, they would have another 4 years and 6 months to initiate implementation (North Carolina Division of Water Quality, 2008). Affected municipalities and counties sought relief from the existing development rule in the 2009 session of the legislature. The legislature responded by disapproving the EMC’s existing development rule, but included much of the content of the rule in the same statute (Session Law 2009-216). The primary change was the schedule for implementation. That action concluded a 10-year process for revising protection strategies adopted in the Cape Fear Basinwide Water Quality Plan approved by the NC EMC in October 1996.

1.09.3 Interstate Nonpoint Management In the Neuse and Cape Fear River Basins, both general and explicit state authorities over nonpoint sources were used to fill gaps in the CWA in efforts to bring waters in those basins into compliance with water-quality standards. Interstate problems are much more intractable, especially in the absence of a strong federal role.

1.09.3.1 Mississippi River Basin and Hypoxia in the Gulf of Mexico An initiative to address the growing problem of hypoxia in the Northern Gulf of Mexico by reducing nutrient inputs from the Mississippi River Basin is a case in point. The areal extent of the hypoxic zone, located along the Louisiana coast as shown in Figure 3, is highly variable from 1 year to the next, but the trend is upward, and over the period 1996–2000 the average was over 14 000 square kilometers (about 5500 square miles).

Upper Mississippi

Missouri

Ohio

Arkansas Red Tennessee Lower Mississippi Atchafalaya

Hypoxic Zone Gulf of Mexico

Figure 3 Mississippi–Atchafalaya River Basin and Hypoxic Zone in the Gulf of Mexico.

158

Implementation of Ambiguous Water-Quality Policies

Concerns about expansion of the zone led to the formation of the Mississippi River/Gulf of Mexico Watershed Nutrient Task Force (MR/GMWNTF) in 1997 comprised of a federal interagency working group and representatives of states and tribes. In November 1998, Congress passed the Harmful Algal Bloom and Hypoxia Research and Control Act of 1998 (HABHRCA, Title VI of P.L. 105-383), gave statutory authority to the task force, and directed it to conduct a scientific assessment of the problem and develop an action plan to control it. The Task Force was administered by the Committee on Environment and Natural Resources of the National Science and Technology Council. The first Action Plan produced in 2001 established the following goal (MR/GMWNTF, 2001): By the year 2015, subject to the availability of additional resources, reduce the 5-year running average areal extent of the Gulf of Mexico hypoxic zone to less than 5000 square kilometers through implementation of specific, practical, and cost-effective voluntary actions by all states, tribes, and all categories of sources and removals within the Mississippi/Atchafalaya River Basin to reduce the annual discharge of nitrogen into the Gulf.

An extensive body of scientific material was produced by the Task Force. Estimates of nutrient and sediment loads, their sources, and amounts delivered to the Gulf have been developed by a group of researchers at the United States Geological Survey (Alexander et al., 2008). Nitrogen and phosphorus delivered to the Gulf arrive from numerous tributary basins as shown in Table 1. Table 1 Percentage distribution of nutrients delivered to the Gulf of Mexico by Source Basin Basin

Upper Mississippi Missouri Central Mississippi Ohio and Tennessee White Arkansas Red Lower Mississippi and Atchafalaya Total

• •

•

Phosphorus

6.5 12.8 25.0 36.1 1.6 4.5 1.0 12.6

5.4 9.3 21.3 35.8 2.6 5.0 1.9 18.7

100.0

100.0

establishment of several subbasin committees; development of an integrated monitoring, modeling, and research strategy; additional monitoring of the hypoxic zone; increased assistance to agricultural producers through US Department of Agriculture programs to restore wetlands, add stream buffers, and install other best management practices; and completion of a major reassessment of the science that supports action items.

Other action items were not implemented, however, substantially impeding reductions in nutrient loads. Among the impediments cited in the report are:

• • •

failure to develop an integrated federal budget to support voluntary nutrient reduction; not only a failure to expand a long-term monitoring program, but also discontinuance of some stations; and the slow pace of development of sub-basin strategies.

Lack of progress on reducing nutrients may be attributable to several factors. First, although there may be strong desires among Gulf states and at the federal level to reduce hypoxia in the Gulf, upstream states lack a compelling interest to impose regulations on nonpoint source generators. Second, competing priorities in the Gulf states, particularly recovery from Hurricanes Katrina and Rita, could have delayed funding of the program. Third, the CWA is ambiguous on authority of

Percentage distribution of nutrients delivered to the Gulf of Mexico by type of source

Source

Nitrogen Mississippi

Urban and population-related sources Atmospheric deposition Crops Pasture/rangeland Forest Shrub and barren lands Total

• •

Percent of total Nitrogen

Table 2

Activities from which loads delivered to the Gulf are generated are shown in Table 2. Point sources account for less than 10% of the nitrogen load; they account for only 10–12% of phosphorus loads. Crops and pasture/rangelands account for approximately 70% of nitrogen and about 80% of phosphorus. The system is clearly dominated by nonpoint sources. Numerous interstate cooperative efforts among both governmental and nongovernmental agencies and organizations have been organized to address a range of water-resource issues along the Mississippi River, including water quality, navigation, and public water supply. Some of the more significant organizations are covered by a report of the National Research Council (NRC, 2008). MR/GMWNTF, with its 2001 Action Plan and its 2008 update (MR/GMWNTF, 2008) developed under authority of HABHRCA of 1998 and 2004, is the most effective action taken to date. The 2008 Action Plan reported progress on action items in the 2001 Plan, including:

Phosphorus Atchafalaya

Mississippi

Atchafalaya

9.1 16.2 65.6 5.0 4.1 0.2

8.9 18.0 62.2 5.6 5.2 0.2

12.3

10.6

43.1 37 7.5 0.4

40.0 39.4 9.2 0.97

100.0

100.0

100.0

100.0

Implementation of Ambiguous Water-Quality Policies

USEPA to take or even threaten to take regulatory action to achieve the goal. HABHRCA did not grant any additional authority to implement the 2001 Action Plan, and, as cited above, the Task Force conditioned the goal of reducing the hypoxic zone on the availability of additional resources and voluntary actions by all states. Several provisions of the CWA explicitly address interstate cooperation on nonpoint sources. These include:

• • •

Section 103(a) – Interstate Cooperation and Uniform Laws Section 319 – Nonpoint Source Programs Section 320 – National Estuarine Program.

Section 103(a) is a very brief admonition to USEPA to ‘‘yencourage cooperative activities by the Statesy’’ to manage pollution, promote uniform state laws, and encourage states to form interstate compacts when necessary to address cross-border effects. Specific authority is granted in Section 301(b) for two or more states to form compacts and agreements to control pollution. Section 319(g), part of the 1987 amendments, gives a state that does not meet water-quality standards due to nonpoint sources in other states the right to petition USEPA to convene a management conference. If USEPA finds that available information is sufficient to support the petition, tributary states are to be notified and a management conference is to be convened. USEPA is also given authority to initiate a management conference without a petition. The purpose of a conference is to develop an agreement among the participating states to improve water quality, but it is quite unclear as to what happens if the states do not reach agreement. The language is: ‘‘To the extent that the States reach agreement through such conference, the management programs of the Statesywill be revised to reflect such agreement.’’ Section 319(g)is silent on what happens in the absence of an agreement among the states. The National Estuary Program, authorized under Section 320 of the CWA, similarly directs USEPA to convene management conferences for the protection of designated national

estuaries in which water-quality standards are not being met. Estuarine management conferences are intended to assess water-quality trends, identify causes of pollution, establish relationships between pollution loads and water quality, develop comprehensive management plans, and identify federal financial assistance. The Barataria–Terrebonne Estuarine Complex along the Louisiana coast is the primary system that would be covered by this program. Section 303 of CWA – Water-Quality Standards and Implementation Plans does not mention other states in its requirements for listing of impaired waters and setting of TMDLs. Nonetheless, the Committee on the Mississippi River and the CWA (NRC, 2008) argued that USEPA has interpreted the CWA as imposing obligations on each state to protect downstream water quality in other states when setting TMDLs. A fundamental barrier to implementation of these authorities is that neither CWA nor HABHRCA grants USEPA or other federal agency any authority to set enforceable limits on nonpoint sources at levels sufficient to attain and maintain water-quality standards. In all cases, implementation depends on the will of tributary states to adopt effective nonpoint source programs, and USEPA is obligated to accept a state management plan if the state satisfies the criterion to identify BMPs and other measures to reduce to the pollution loads to the MEP.

1.09.3.2 Chesapeake Bay Program Urban growth, more intensive agricultural operations, and other factors in the watershed resulted in significant deterioration of water quality and related ecosystems in the Chesapeake Bay. Among other undesirable outcomes was widespread hypoxia due to excess nutrient loads. As shown in Figure 4, areas that drain to the Chesapeake Bay cover portions of several states. In 1983, 1987, and again in 2000, several states in the Chesapeake Bay watershed, the District of Columbia, the Chesapeake Bay Commission, and the US EPA entered into agreements to form the Chesapeake Bay

New York

Pennsylvania Basin boundary NJ

MD

West Virginia

Figure 4 Chesapeake Bay Drainage Area.

159

DE Virginia

160

Implementation of Ambiguous Water-Quality Policies

Table 3

Distribution of nutrient loads delivered to Chesapeake Bay by source

Source of nitrogen

Percent of total load

Source of phosphorus

Percent of total load

Septic Municipal and industrial wastewater Manure (agriculture) Chemical fertilizer (agriculture) Chemical fertilizer (nonagriculture) Atmospheric

4.3 18.9 17.8 15.4 10.1 33.5

Manure (agriculture) Chemical fertilizer (agriculture) Municipal and industrial wastewater Natural (wildlife and forest) Other fertilizer

27 18 22 3 30

Source: Chesapeake Bay Program, www.chesapeakebay.net/tribtools.htm#allocations.

Table 4 Direct funding provided by the federal agencies, states, and District of Columbia, fiscal years 1995 through 2004, in millions of constant 2004 dollars Federal agencies Department EPA Department Department Department

of Defense of Agriculture of the Interior of Commerce

Total – federal agencies

States 355.4 253.7 230.4 77.4 55.5

Maryland Virginia District of Columbia Pennsylvania

1862.4 752.6 41.8 28.1

972.4

Total – all states

2684.8

Source: United States Government Accountability Office, 2005.

Program (CBP) with the intent of restoring the Bay’s water quality and ecosystem. Maryland, Pennsylvania, and Virginia account for 88% of the nitrogen load and 87% of phosphorus. Distributions of sources of nitrogen and phosphorus that are delivered to the Bay are given in Table 3. As part of the 1987 agreement, the signatories committed to reduce 40% of nutrient loads that were controllable. Tributary-specific nutrient reduction strategies were adopted in 1992. The 2000 agreement asserted that where the 1992 goals had not been achieved, additional steps would be taken (CBP, 2000). Direct federal and state expenditures over the decade 1995–2004 are shown in Table 4. Federal agencies spent an average of nearly $100 million (in constant 2004 dollars) a year, and the three states and District of Columbia spent about $270 million a year (United States Government Accountability Office, 2005). Current estimates of distributions of nitrogen and phosphorus loads entering the Bay indicate that runoff from agriculture and urban areas account for large percentages of total loads. Nearly 60% of nitrogen and 80% of phosphorus loads are attributed to those sources (CBP, 2009). More recently, President Obama, citing lack of adequate progress toward restoration, issued Executive Order (EO) 13508 in May 2009, directing federal agencies to take a more active program to protect and restore water-quality and related ecosystems. Among the key challenges listed in the EQ are efforts to: (1) strengthen permit conditions for CAFOs and urban stormwater runoff and (2) enhance federal and state initiatives for conservation practices on agricultural operations (Federal Leadership Committee, 2009). USEPA also issued a notice of intent in September 2009 to establish a Bay-wide

TMDL for nutrients and sediments (USEPA, 2009). Preliminary targets for nitrogen and phosphorus loads have been established pursuant to the EO (Early, 2009). Maryland, Pennsylvania, and Virginia were allocated 88% of the nitrogen target and 89% of the phosphorus target. Even with these initiatives coming on authority of the President, it is far from clear as to how much can be achieved given ambiguities in the CWA. CAFO reductions may be more predictable because of existing permit authority, but controls on urban runoff will still be limited by the MEP criterion. All of the agricultural runoff programs cited in the EO are voluntary. Regulatory authority over urban stormwater and nonCAFO agriculture will be based on authorities of the states and the District of Columbia. In this case, the three states that are big contributors have a compelling interest and a history of strong political support for those policies, but it remains questionable as to how far that authority extends and how much political support there will be as reductions from these sources approach the limits of MEP.

1.09.4 Summary and Conclusions The several cases described in this chapter point to a conclusion that, in the absence of a compelling state interest to improving water quality, ambiguities in the CWA are likely to present strong impediments for achieving the goals of the Act. The goal of the Act is protect and enhance water quality sufficient to support designated uses of all stream segments, the lowest acceptable use being propagation of fish and aquatic ecosystems. Few ambiguities exist when it comes to managing point sources; effluent limits must be set to satisfy waterquality standards, and state boundaries do not present an insurmountable barrier. TMDLs set by one state must consider effects on waters in downstream states. However, CWA nonpoint sources require only that states employ management practices that reduce loads to the maximum extent practical, a judgment call that is left to the states. A similar requirement applies to stormwater runoff from urbanized areas. In the Neuse River, Jordan Lake, and Chesapeake cases discussed in this chapter, compelling local and regional interests have led to invocation of state authority to shore up gaps in federal legislation. It may be argued that Congress intended just that, but, in these cases, states have followed very closely the language of the CWA. States have made judgments as to what BMPs satisfy the criterion of MEP without effective

Implementation of Ambiguous Water-Quality Policies

monitoring, inspection, and enforcement procedures to insure that water-quality standards will be attained and maintained. While it may not be realistic to expect compliance with waterquality standards over short time horizons given existing land uses and agricultural and urbanization practices, ambiguities in present federal and state policies about nonpoint sources leave the public poorly informed as to what is possible and the extent to which management programs are being effective.

References Alexander RB, Smith RA, Schwarz GE, Boyer EW, Nolan JV, and Brakebill JW (2008) Differences in phosphorus and nitrogen delivery to The Gulf of Mexico from the Mississippi River Basin. Environmental Science and Technology 42(3): 822--830. CBP (Chesapeake Bay Program) (2000) ‘‘Chesapeake 2000’’, the 2000 Partnership Agreement. http://www.chesapeakebay.net/content/publications/cbp_12081.PDF (accessed April 2010). CBP (Chesapeake Bay Program) (2009) Bay Barometer: A Health and Restoration Assessment of the Chesapeake Bay and Watershed in 2008. http:// www.chesapeakebay.net/news_baybarometer08.aspx?menuitem=34917 (accessed April 2010). Early WC (2009) Acting Regional Administrator, Region III, Letter to L. Preston Bryant, Jr., Virginia Secretary of Natural Resources, 3 November 2009. http://www.epa.gov/ region3/chesapeake/bay_letter_1209.pdf (accessed April 2010). Federal Leadership Committee (2009) ‘‘Executive Summary for Draft Reports Addressing Key Challenges to Chesapeake Bay Protection and Restoration’’. Reports prepared in response to Section 202 of Executive Order 13508, 9 September 2009. Hunt JB (1996) Letter to David H. Moreau, Chairman, North Carolina Environmental Management Commission, 12 March. MR/GMWNTF (Mississippi River/Gulf of Mexico Watershed Nutrient Task Force) (2001) Action Plan for Reducing, Mitigating, and Controlling Hypoxia in the Northern Gulf of Mexico, Washington, DC. http://www.epa.gov/msbasin/pdf/ actionplan2001.pdf (accessed April 2010). MR/GMWNTF (Mississippi River/Gulf of Mexico Watershed Nutrient Task Force) (2008) Gulf Hypoxia Action Plan 2008. Washington, DC: Office of Wetlands, Oceans, and Watersheds, United States Environmental Protection Agency.

161

NC Division of Environmental Management (1993) Neuse River Basinwide Water Quality Management Plan, Raleigh, NC. North Carolina Division of Water Quality (2006) Jordan water supply nutrient strategy and rules. Report to the NC Environmental Management Commission, 11 January 2006. North Carolina Division of Water Quality (2008) ‘‘Jordan Lake nutrient strategy’’. http:// portal.ncdenr.org/web/wq/ps/nps/jordanlake (accessed April 2010). NRC (National Research Council) (2008) Mississippi River Water Quality and the Clean Water Act. Washington, DC: National Academy Press. United States Government Accountability Office (2005) Chesapeake Bay program: Improved strategies are needed to better assess, report, and manage restoration progress. Report to Congress. Washington, DC: United States Government Accountability Office. USEPA (United States Environmental Protection Agency) (1971) The Cost of Clean Water, Volume I, Annual Report (4th) to Congress in Compliance with Section 26(a) of the Federal Water Pollution Control Act, Senate Document 92-23. Washington, DC: United States Government Printing Office. USEPA (United States Environmental Protection Agency) (1975) Preparation of water management plans. Federal Register 40(230): 55345--55346 (28 November 1975) USEPA (United States Environmental Protection Agency) (1996) Draft Framework for Watershed Based Trading. EPA Report No. 800 R 96 001. Washington, DC: USEPA (30 May 1996). USEPA (United States Environmental Protection Agency) (1997) ‘‘New Policies for Establishing and Implementing Total Maximum Daily Loads (TMDLs),’’ Memorandum from Robert Perciasepe, Assistant Administrator to Regional Administrators Regional Water Division Directors, 8 August 1997. http:// www.epa.gov/OWOW/tmdl/ratepace.html (accessed April 2010). USEPA (United States Environmental Protection Agency) (2003) Water Quality Trading Policy. Washington, DC: Office of Water. (13 January). USEPA (United States Environmental Protection Agency) (2009) Clean Water Act Section 303(d): Preliminary Notice of Total Maximum Daily Load (TMDL) Development for the Chesapeake Bay. Federal Register 74(179): 47792--47794 (17 September 2009). USEPA (United States Environmental Protection Agency) (2010) Water Quality Standards for the State of Florida’s Lakes and Flowing Waters: Proposed Rule. Federal Register 75(16): 4291 (26 January 2010).

1.10 Predicting Future Demands for Water B Dziegielewski and DD Baumann, Southern Illinois University Carbondale, Carbondale, IL, USA & 2011 Elsevier B.V. All rights reserved.

1.10.1 1.10.1.1 1.10.1.2 1.10.2 1.10.2.1 1.10.2.2 1.10.2.2.1 1.10.2.2.2 1.10.2.2.3 1.10.2.2.4 1.10.3 1.10.3.1 1.10.3.1.1 1.10.3.1.2 1.10.3.2 1.10.3.2.1 1.10.3.2.2 1.10.3.2.3 1.10.3.3 1.10.3.3.1 1.10.3.3.2 1.10.4 1.10.4.1 1.10.4.2 1.10.4.2.1 1.10.4.2.2 1.10.4.2.3 1.10.4.3 1.10.4.3.1 1.10.4.3.2 1.10.4.4 1.10.4.4.1 1.10.4.4.2 1.10.4.5 1.10.4.5.1 1.10.4.5.2 1.10.5 1.10.5.1 1.10.5.2 1.10.5.3 1.10.5.4 1.10.6 References

Water Supply and Demand Changing Objectives of Water-Supply Development Emergence of Water Conservation Water-Use Data Definitions and Measurement of Water Use Accessibility of Data on Water Use Public-supply sector data Industrial and commercial sector data Power generation sector data Irrigation sector data Water-Demand Relationships Theoretical Models of Water Demand Derived demand of producers Final demand of consumers Empirical Models of Water Use Configuration of data sets Functional forms and model parameters Elasticities with respect to major determinants of water use Other Water-Use Relationships Cooling water requirements Supplemental irrigation water requirements Demand Forecasting Techniques Forecasting Principles and Criteria Forecasting Models and Procedures Time trend forecasting Water requirement forecasts Demand forecasts Dealing with Forecast Uncertainty Model-dependent prediction intervals Dealing with forecast assumptions error Forecasts with Conservation End-use accounting system Baseline and restricted forecasts Forecasting Software: The IWR-MAIN Program Model structure and procedures IWR-MAIN conservation forecasts Example of a Regional Multisector Forecast Water-Use Relationships Effects of Key Forecast Assumptions Effects of Future Climate Forecast Summary Conclusion

1.10.1 Water Supply and Demand Water is an essential natural resource, which plays a vital role as input into many economic activities adding to the quality of human life and supporting the health of ecological systems. Water is also a commodity which has an economic value in all its competing uses as has been recognized by the main international conferences on water (e.g., ICWE, 1992;

163 163 163 164 164 165 165 165 166 166 166 166 167 168 169 169 170 171 172 172 173 175 175 175 175 176 177 177 178 178 178 179 179 179 179 180 181 182 182 183 185 186 187

UNCED, 1992). Despite this, the importance of water supply may not be widely appreciated by the general public because only some water resources and water uses are easily noticed while others are not. Surface water resources such as rivers and lakes are highly visible and well recognized for their cultural and amenity values as well as for their important functions in outdoor recreation and transportation. Less recognized by the general public is the portion of water

163

164

Predicting Future Demands for Water

resources – groundwater – which in the United States (the lower 48 states) actually represents three-fourths of all freshwater storage (Dziegielewski and Kiefer, 2006). Similarly, some human uses of water are easily noticed while others are not. Use of water for hydropower production or for irrigation can be easily seen and appreciated, while large water flows to urban and industrial users are usually hidden in underground pipes. For example, the network of public water supply pipes in the US carries an average flow of 67 000 cubic feet per second (cfs) – this is equivalent to the average discharge of a large river. The flow of water for thermoelectric cooling is even greater; in the US in the year 2000, it reached 302 000 cfs – more than the average annual discharge of the Ohio River. Given the overall importance of water, it is understandable that the long-term adequacy of water supply is a major national concern in many countries. However, in order to assess the future adequacy of supplies, it is necessary to determine the amount of water that is used currently and the amount that will be demanded in the future. These demands have to be compared with the future availability of water in existing and potential sources of supply. To do this, water-supply planners need appropriate tools for quantifying future water demands and assessing the effects of future climate and other factors on both water demand and water availability. Credible long-term forecasts of water demand are essential to all type of planning. Without such forecasts water planners cannot achieve an efficient allocation of water supplies among competing uses or ensure their long-term sustainability.

1.10.1.1 Changing Objectives of Water-Supply Development The traditional approach to water-supply planning has evolved during the past century, as cities, industries, and irrigation districts have expanded their water-supply infrastructure (Blake, 1956; Grigg, 1986). In urban areas, municipal water-supply agencies considered their responsibility to be on-demand delivery of sufficient quantities of drinking quality water while maintaining adequate pressure for consumption and fire protection. Even very high expenditures on the development of waterworks could be justified because adequate community water supply was considered an essential service that ensured public health and safety, economic activity, and a general community well-being. Also, accurate long-term forecasts of water demand were not very important because rapid urban growth made it possible to add large increments of water-supply capacity without the risk that it would remain unused for long periods of time. The longstanding practice of augmenting water supply by securing reliable sources, protecting water quality at the source, and developing adequate facilities for transmission, treatment, and distribution was based on several advantages of this strategy. These were summarized by Platt (1993). Specifically, somewhat distant hinterland sources provided a cheap and abundant supply of high-quality water without the need for treatment. Cities in Northeast and Northwest of the US relied on such sources prior to the adoption of water chlorination. In addition, storage reservoirs at relatively high elevations permitted gravity-flow transmission of water without the need for pumping. Also, urban governments could acquire

water rights and watershed lands for source protection at a minimum cost, often under condemnation powers, while the hinterland regions offered little political resistance to water exports. Finally, large professionally managed sources of supply offered significant economies of scale. The advantages of the hinterland sources provided little incentive for a careful matching of water supplies with water demands.

1.10.1.2 Emergence of Water Conservation Since the 1970s, the viability of the traditional supply augmentation approaches has begun to decline gradually because of several obstacles (Platt, 1993). The hinterland regions have begun to offer some political and legal resistance to water exports. Environmental legislation and interests have introduced significant barriers to a continuing expansion of offstream uses of water for urban and agricultural purposes. Major droughts over the last 40 years have also contributed to the increased competition for water supplies between urban and agricultural interests. Other factors such as the depletion and contamination of groundwater sources, difficulties in financing major construction programs, especially those sponsored by the federal government, and increasing costs of water treatment for regulated contaminants have made the traditional options of supply augmentation less viable. As a result, the range of options has expanded to include both some unconventional supply-side alternatives and the opportunities for modifying the growth in water demand. The introduction of demand management alternatives represents an important change in water-supply planning. In the early 1980s, the growing attractiveness of long-term demand management measures began to catch the attention of urban water-supply agencies (Boland et al., 1982; Dziegielewski and Baumann, 1992). Demand reduction programs allowed some agencies to balance future supply and demand at a cost that is below the economic, social, and environmental cost of new supply development and thus result in net benefits to society (Dziegielewski et al., 1983). Baumann et al. (1984) developed a practical definition of long-term water conservation as ‘‘yany beneficial reduction in water use or in water losses.’’ The authors pointed out that a water management practice constitutes conservation if it conserves a given supply of water through reduction in water use (or losses) and if it results in a net increase in social welfare where the resources used have a lesser value than those saved (p. 431). In other words, the beneficial effects of the reduction in water use (or loss) must be considered greater than the adverse effects associated with the commitment of other resources to the conservation effort. This definition provided an important guidance for long-term conservation; however, it could not be easily applied to short-term conservation measures which are usually aimed at curtailing water demand during a drought. Temporary restrictions on water use are usually undertaken in order to prevent adverse impacts of severe shortages in the future if the drought continues and their outcomes cannot be easily analyzed through benefit–cost analysis. Other marked enhancements in theory and practical knowhow for planning and evaluation of demand management alternatives have also been achieved (Dziegielewski et al., 1993). These were followed by the development of computer

Predicting Future Demands for Water

software programs for disaggregate water-use forecasting, for the analysis of demand reduction alternatives, for optimization of long-term water management plans, and for the monitoring of water demands over time (Dziegielewski and Boland, 1989; Dziegielewski, 1993; Baumann et al., 1998). This chapter provides a review of key water-demand concepts and presents analytical methods for quantifying demands by different user groups, forecasting future demands and analyzing demand management alternatives. The chapter begins with the discussion of the measurement and analysis of water use and related concepts.

1.10.2 Water-Use Data Our knowledge of water demands necessarily derives from the measurement and estimation of water use. In practice, it is impossible to know precisely all water uses – there are many different types of water users and specific purposes of use and only some uses are metered. Instead, various estimation methods are usually employed to determine the quantity of water use. Because water use depends on many factors, the analysis of water demands requires data on those factors. This chapter describes the structure of water demand and identifies the factors affecting water use as well as methods for analyzing the available water-use data.

1.10.2.1 Definitions and Measurement of Water Use From the hydrologic perspective, water use is a part of the water budget. At the most general level, water use can be defined as all water flows that are a result of human intervention within the hydrologic cycle. Accordingly, all water uses can be divided into in-stream and off-stream uses. In-stream use represents water that is used, but not withdrawn, from a natural water source for purposes such as hydroelectric power generation, navigation, water-quality improvement, fish propagation, and recreation. Off-stream use represents water withdrawn or diverted from a groundwater or surface water source for public water supply, industry, irrigation, livestock, thermoelectric power generation, and other uses (Hutson et al., 2004). The term ‘water withdrawal’ is used to designate the amount of water that is taken out from natural water sources such as lakes, rivers, or groundwater aquifers. The difference between the amount of water withdrawn and water returned to the source (also referred to as discharge) is usually taken to represent consumptive use. This is the ‘‘part of water withdrawn that is evaporated, transpired, incorporated into products or crops, consumed by humans or livestock, or otherwise removed from the immediate water environment’’ (Hutson et al., 2004). The part of amount withdrawn and returned back to the source is called nonconsumptive use. The quantity of water consumed is utilized in calculating regional annual and monthly water budgets, and represents a measure of the volume of water that is not available for repeated use. While a major portion of water withdrawn for purposes such as public water supply, power generation, and industrial use represents nonconsumptive use, these withdrawals can have significant impacts on water resources and other uses of

165

water. For example, water withdrawn from an aquifer and then returned into a surface water body may have a positive impact on streamflow or lake water levels, but a negative impact on the source of groundwater. Similarly, water withdrawn from a river for public water supply must be continuously available at the intake but not for withdrawal for other uses upstream or immediately downstream from the intake. A more restrictive definition of water use refers to water that is actually used at a specific site or for a specific purpose. Individual residential or commercial buildings, industrial facilities, and other locations can obtain water from their own sources of supply or through connections to a public or private distribution system. Individual users of water within a defined geographical area can be classified into different categories and their combined use can be summed up into broader categories, or user sectors. For example, the United States Geological Survey compiles data on water withdrawals for individual counties for eight categories of users: public supply, domestic, industrial, commercial, mining, power generation, livestock, and irrigation. Some of the categories are further subdivided based on the purpose of use. For example, irrigation use is subdivided into cropland and golf course irrigation. Similarly, public-supply use can be subdivided into residential, commercial, industrial, and public categories with each category further subdivided into two or more subcategories (e.g., residential single-family and residential multifamily). Measurement of water use can take place at the point of withdrawals or at the point of water use. Also, some measurements could be taken at the point of water treatment or along water transmission routes. Direct measurements of water volume being transmitted over a given period of time are made by meters which register the volume of flow (such as displacement meters) or by measuring and recording instantaneous flow (such as in Venturi meters). The measurements of water use are reported as water volume per unit of time. The volumetric units include cubic meters, cubic feet, gallons and liters, and their decimal multiples. In some cases, composite volumetric units (e.g., acre feet) or units of water depth (e.g., inches or centimeters of rainfall) may be used. The time periods used may include a second, minute, hour, day, month, and year. Because the annual and monthly volumes of water use generally involve large numbers, the numerical data on water use are often reported as the average daily quantities used. Two popular units are thousand cubic meters per day (1000 m3 d1) and million gallons per day (mgd). Also, in order to make the estimates of water use easy to comprehend and to make meaningful comparisons of water use for various purposes (and various users), the annual or daily quantities can be divided by some measures of size for each purpose of use. The result is an average rate of water use such as gallons per capita per day (gpcd), gallons per employee per day (ged), or other unit-use coefficients. Finally, it is important to note that the reported quantities of water use can be in the form of direct measurements obtained from water meters or they may be estimates. Estimates of water use that are derived from the measurements of water levels in storages or from pumping logs are generally more accurate than those derived from related data on the volume

166

Predicting Future Demands for Water

of water-using activity. For example, the estimates of water use for hydroelectric power generation may be obtained by multiplying the amount of generated power by a water-use coefficient. In analyzing water demands, it is important to recognize the sources and nature of the data on water use. For example, when deriving statistical water-demand relationships, it makes little sense to use data which are estimates based on the volume of water-using activity or other correlates of water use. Only actual measurements of water volumes, which are withdrawn or used over time, can accurately capture the temporal and spatial variability of water demand and provide a basis for deriving econometric or other models of water use.

1.10.2.2 Accessibility of Data on Water Use The availability of data on water use depends on user sector. The best data are available for public-supply sector (also referred to as municipal and industrial use). Data for other sectors, which typically rely on self-supplied sources of water, are usually less precise and their accuracy depends on whether the withdrawals of water are regulated and are required to be metered or if only an annual reporting of the estimated quantities is required. The discussion of the available data for the different sectors is given in the following.

1.10.2.2.1 Public-supply sector data Water use in the public-supply sector can be characterized with respect to: (1) the demands of different types of customer classes (e.g., single-family residences, hotels, food-processing plants, etc.); (2) the purposes for which water is used (e.g., end-uses such as sanitary needs, lawn watering, or cooling); and (3) the seasonal variations in water use. The breakdown of total urban water use into customer groups, specific end-uses, and seasons can serve as a basis for modeling water demands and for conducting impact evaluations of water conservation programs. Generally, there are three types of water measurement records that are maintained by public water-supply utilities: (1) water production records (i.e., amounts of water pumped into the distribution system); (2) water billing records (i.e., records which detail each customer’s account activity); and (3) water sales records (i.e., summaries of total water sales or water sales by customer groups). Water production records show the amount of water pumped from treatment plants and are typically generated daily or hourly. Production data can be used for analyzing: (1) unaccounted water use (comparing production with water sales data); (2) impacts of water-use restrictions on total water demand; (3) relationships between total water demand and weather conditions; and (4) peak water use during different time intervals (e.g., peak day, peak hour, day of week, etc.). Table 1 (Dziegielewski and Chowdhury, 2008) shows an example of water production records for a sample of cities and water-supply systems in the Greater Chicago Area in Illinois. The last column shows the calculated rate of usage in gpcd. Water billing records represent the individual customer account data which are generally maintained by retail watersupply agencies. The individual record usually includes: (1) name and address of account holder, (2) type of account

Table 1 Examples of water production and purchase data by public water supply systems in selected communities in Northeastern Illinois in 2005 Community or system name

Aurora Bedford Park Belvidere Central Lake Co. JAWA Chicago Crystal Lake DeKalb DuPage Water Com. Elgin Evanston Glencoe Hammond WSS Highland Park Joliet Kankakee Aqua Illinois Lake County PWD Lake Forest Morris North Chicago Northbrook Northwest Sub. M. JAWA Oak Lawn Oswego Waukegan Wilmette Winnetka Total study area

Reported production in mgd

Estimated population served

Per capita production in gpcd

18.1 25.4 3.7 21.2 729.6 5.4 4.4 90.6 15.5 45.7 1.9 18.4 11.8 16.5 12.9 3.0 4.8 1.6 4.7 6.1 35.9

1 70 000 130 415 23 500 197 446 3 960 041 40 440 40 000 728 427 142 572 354 258 8 600 133 035 59 580 130 830 67 000 29 536 21 477 13 282 19 127 36 975 309 084

106.5 194.5 155.7 107.4 184.2 134.3 109.0 124.4 109.0 129.1 217.4 138.0 197.5 125.9 192.4 101.9 221.2 123.5 245.2 164.4 116.2

36.6 2.4 9.7 12.9 3.8 1142.3

316 389 23 000 101 919 90 391 17 600 7 164 924

115.6 102.6 94.8 142.3 217.6 159.4

mgd, million gallons per day; gpcd, gallons per capita per day. From Dziegielewski B and Chowdhury FJ (2008) Regional water demand scenarios for Northeastern Illinois: 2005–2050. Project Completion Report. Prepared for the Chicago Metropolitan Agency for Planning, Chicago, IL, USA, 15 June 2008.

(single family, commercial, industrial, institutional), (3) meter size, (4) meter readings and the dates of meter readings, (5) water use between meter readings, and (6) billing information (charges incurred, dates paid, etc.). The customer billing system is usually computerized and individual customer accounts can be sorted by customer type, geographical area (e.g., pressure zone), and other characteristics. Finally, water sales records are summaries of the individual water billing records. The sales data are aggregated by the billing cycle (i.e., monthly, bimonthly, semiannually, or annually) and by customer type. They show how much water is being sold to different types of customers but they do not show for what specific purposes the water is being used.

1.10.2.2.2 Industrial and commercial sector data Data on self-supplied commercial and industrial use within a given geographical area are available in areas where annual or monthly reporting of water is practiced. The industrial subsector includes water used for ‘‘industrial purposes such as fabrication, processing, washing, and cooling, and includes

Predicting Future Demands for Water Table 2 Estimates of self-supplied industrial and commercial water demand in 11 counties in Northeastern Illinois

Table 3

Water use in a sample of large power plants in Illinois

Plant name County

Boone Cook DeKalb DuPage Grundy Kane Kankakee Kendall Lake McHenry Will Total/ave.

Self-supplied withdrawal in 2005 (mgd)

0.57 123.73 2.54 0.96 6.99 4.34 5.09 0.78 13.88 6.58 24.97 190.43

Employment in self-supplied establishments

1 200 22 364 4 025 11 024 656 6 329 157 5 229 19 495 8 515 13 727 92 721

Unit selfsupplied withdrawals per employee (gped) 475.0 5 532.6 631.1 87.1 10 655.5 685.7 32 420.4 149.2 712.0 772.8 1 819.0 2 053.8

mgd, million gallons per day; gped, gallons per employee per day. From Dziegielewski B and Chowdhury FJ (2008) Regional water demand scenarios for Northeastern Illinois: 2005–2050. Project Completion Report. Prepared for the Chicago Metropolitan Agency for Planning, Chicago, IL, USA, 15 June 2008.

such industries as steel, chemical and allied products, paper and allied products, mining, and petroleum refining,’’ and the commercial subsector includes water used for ‘‘motels, hotels, restaurants, office buildings, other commercial facilities, and institutions’’ (Avery, 1999). For a given geographical area such as a county or a hydrologic basin, industrial and commercial water withdrawals will depend on the number, type, and size of water users. Table 2 shows an example of self-supplied commercial and industrial withdrawal data for 11 counties in Northeastern Illinois.

1.10.2.2.3 Power generation sector data In the US, water withdrawn by power plants is classified as thermoelectric generation water use. It represents the water applied in the production of heat-generated electric power. The main use of water at power plants is for cooling. Nearly 90% of electricity in the United States is produced with thermally driven, water-cooled generation systems which require large amounts of cooling water (Dziegielewski and Bik, 2006). The three major types of thermoelectric plants include: conventional steam, nuclear steam, and internal combustion plants. In conventional steam and nuclear steam power plants, water is used primarily for cooling and condensing steam after it leaves the turbine. In this type of generation, the use of cooling water is essential because the collapse of steam volume in the condenser creates a vacuum, which affects the rotation of the turbine. Because the level of the vacuum depends on the removal of waste heat by cooling water, the cooling system is an integral part of the power generation process. Precise estimates of thermoelectric water use are difficult to obtain. The only consistent source of thermoelectric water-use data is the annual survey of power plants by the US Energy

167

2005 Water withdrawals (mgd)

Once-through flow plants Crawford Plant 503.3 Fisk Street Plant 222.2 415.6 Dresden Nuclear Planta Waukegan Plant 758.6 Joliet 29 Plant 942.6 Joliet 9 Plant 415.3 Will County/ 917.9 Romeoville Plant Clinton Plant 810.4 Dallman Plant 328.1 Lakeside Plant 43.2 All once-through 5357.2 plants

2005 Gross generation (MWh yr1)

2005 Rate of withdrawals (gal. kWh1)

3 201 844 1 603 949 14 031 125

57.4 50.6 10.8

4 909 907 5 767 994 1 922 330 5 658 996

56.4 59.6 78.9 59.2

9 014 690 2 328 492 229 855 48 669 182

32.8 51.4 68.6 40.2

Closed-loop makeup water plants Vermilion Plant Powerton Plant Braidwood Nuclear Plant Kendall Co. Gen. Facility All closed-loop plants

2.8 25.9 49.8

702 950 10 120 133 20 390 274

1.43 0.93 0.89

2.5

1 367 008

0.67

80.9

32 580 365

0.91

a

Dresden plant uses a combination of once-through and pond recirculation system. From Dziegielewski B and Chowdhury FJ (2008) Regional water demand scenarios for Northeastern Illinois: 2005–2050. Project Completion Report. Prepared for the Chicago Metropolitan Agency for Planning, Chicago, IL, USA, 15 June 2008 and WHPA and Dziegielewski (2008).

Information Agency (EIA). The resultant EIA-767 database consists of a series of data tables that present data on different aspects of the power plant operation (Table 3). The EIA-767 data tables are the main data sources for the analysis of thermoelectric water withdrawals (intake) and consumptive use. However, since 2006 the EIA discontinued collection of operational data from power plants; therefore, the more current data on water withdrawals have to be obtained directly from individual plants.

1.10.2.2.4 Irrigation sector data The irrigation sector includes self-supplied withdrawals of water for irrigation of cropland, turfgrass-sod farms, and golf courses. In the existing inventories of water use, the designation of irrigation water withdrawals includes ‘‘all water artificially applied to farm and horticultural crops as well as selfsupplied water withdrawal to irrigate public and private golf courses’’ (Solley et al., 1998). Irrigation water use is rarely measured and the reported data on water withdrawals are based on the inventory of the total acreage of irrigated area. The data on irrigated land are collected and reported by the US Department of Agriculture (USDA, 2009).

168

Predicting Future Demands for Water

1.10.3 Water-Demand Relationships

water. That is, in general, for k inputs,

1.10.3.1 Theoretical Models of Water Demand Guidance for empirical modeling of water demand can be derived from economic theory. From an economic perspective, water is considered to be a commodity (or an economic good) and it can be conceived as a final good to consume or as an input to the production of some other good or service. Demand for water therefore can be a final demand if the user is a consumer or a derived demand if the user is a producer whose demand is driven by the demand for other goods produced through the use of water. Theoretically, industrial and most of the commercial and public uses can be viewed as derived demands by producers, while the residential component of water demand can be viewed as a final demand by consumers (Hanemann, 1998). Different economic theories apply to these two economic categories of demand. The derived demand is described by the economic theory of production. The final demand can be described by the economic theory of consumer demand. Good examples of a complete mathematical treatment of the application of these two theories to water demand are provided by Hanemann (1998) and Renzetti (2002).

1.10.3.1.1 Derived demand of producers According to Hanemann (1998) the derived demand for water by a firm can be represented by four different types of relationships. These include both the long-run and short-run versions of conditional and unconditional demand functions. The fixed-output conditional demand function reflects optimization of the amounts of inputs (which include water) to produce a given level of output. The variable-output or unconditional demand function determines how the firm should select the level of output together with the corresponding inputs. The other two functions introduce the distinction between the short-run and long-run input demand functions. In a short run some inputs are assumed fixed, whereas in a long run all inputs are variable. The long-term conditional demand functions can be obtained by combining the production function (which represents production technology) with the firm’s behavior (i.e., cost minimization or profit maximization). The derivation of conditional or unconditional demand functions would start with an explicit formula for the production function and then solve the optimization problem. Two production functions which were frequently used in the past include the Cobb–Douglas and the constant elasticity of substitution (CES) formulas. A newer approach uses the duality relationship between a production function and the cost or profit functions. Hanemann (1998) used the duality approach to show the two demand functions – conditional and unconditional demand. The conditional function can be represented by the derivatives of the associated cost function or profit function with respect to the price of the input of interest. Accordingly, the conditional water-demand function can be written as a derivative of the cost function with respect to the price of

xk ¼ gk ðo1 ; y; on ; yÞ ¼

q Cðo1 ; y; oN ; yÞ ; k ¼ 1; y; N q ok ð1Þ

and the unconditional demand function is equal to minus the derivative of the profit function with respect to the price of water:

xk ¼ hk ðo1 ; y; oN ; pÞ ¼ 

q pðo1 ; y; oN ; pÞ ; k ¼ 1; y; N q ok ð2Þ

where C denotes the firm’s total cost of production; ok the price of the kth input xk, with k ¼ 1,y, N; y the volume of output produced per unit of time; p the profit; and p the price of output. There are many possible formulas for cost or profit functions that can be used with this approach including production functions which are less restrictive than the Cobb– Douglas or CES functions with respect to their implications for complementarity or substitution among inputs. As an example, Hanemann (1998) provides an illustration of the application of the derivatives in Equations (1) and (2) using the translog cost function:

lnC ¼ lnb0 þ by lny þ þ

X

X

ln ok þ 1=2dyy ðlnyÞ2

k

dky ln ok lny þ

k

1X dik lnoi lnok 2 ik

ð3Þ

where b‘s and d‘s are coefficients to be estimated with dik ¼ dki. The application of Equation (1) in combination with Equation (3) leads to a relatively simple share equation which represents the share of total cost devoted to each input:

ok gk ðo1 ; :::; oN ; yÞ C ¼ bk þ by ln y þ dky ln y þ dkk ln ok X þ oki ln oi

sk ðo1 ; :::; oN ; yÞ 

ð4Þ

ia k

The theoretical relationships given above provide some guidance for the development of empirical water-demand equations for industrial and commercial sectors. The main points are: (1) derived demand for water depends on the price of water (own elasticity) as well as the prices of all other inputs (cross-price elasticity), and (2) demand also depends on either the quantity or prices of the outputs. These theoretical relationships should be incorporated into econometric models of industrial water demand. However, there are some practical difficulties which limit a strict adherence to the economic theory of production. One limitation is the level of data aggregation. Water-use data are often aggregated for an entire industrial sector while the production and optimization relationships need to be estimated separately for specific industries based on the firm-level data. Other difficulties relate to (1) specification of the price of water, (2)

Predicting Future Demands for Water

availability of prices for inputs other than water, and (3) specification of the mathematical form of the demand function. Examples of empirical relationships which were obtained in studies of industrial demands can be found in the work by Renzetti (1992) and Dupond and Renzetti (1998).

1.10.3.1.2 Final demand of consumers Economic theory of consumer demand provides a basis for deriving relationships for final demand for water. The theory is based on the concept of the consumer’s utility function and the optimizing behavior of a consumer faced with a limited budget. Utility functions can take many forms; examples include the Cobb–Douglas function and its variant, the Stone– Geary function (Deaton and Muellbauer, 1999). For example, Hanemann (1998) derived a water-demand function based on the Stone–Geary function for the case of two goods (i.e., water and all other goods) of the form

x1 ¼ ð1  a1 Þg1 þ a1

y p2  a1 g2 p1 p1

ð5Þ

where x1 is the quantity of water consumed, y the consumer’s budget, g1 and g2 the minimum consumption levels (which specify the assumption that the consumer derives utility from a commodity only if the consumption exceeds gi), p1 and p2 the prices of the two goods, and a1 the exponent of x1 in the original utility function. The own price, cross-price, and income elasticities can be obtained by taking derivatives of this demand function. Generally, single-demand equations used in empirical studies of residential water demand (even if they include price of water, income, and price of other goods) are formally inconsistent with economic theory because their functional forms (i.e., model specification) cannot be derived from the maximization of the utility function (Hanemann, 1998). The correct specification of a linear demand function with these three variables would be

x1 ¼ a1  b

p1 y þg p2 p2

ð6Þ

Equation (6) indicates that demand for water depends on the relative prices and relative income and not on their absolute values which are typically used in the empirical equations which are found in the literature. Another issue in modeling final demand is related to the choice of explanatory variables to be included in the demand function along with the economic variables of price and income. Variables such as those describing the climatic conditions, demographic characteristics, or physical settings need to be included in the equation because they also affect the consumer’s utility. Different values of these variables would cause different levels of utility from the same level of consumption. Hanemann (1998) suggests that other variables should be introduced into demand functions by making one or more of the coefficients in Equation (6) a function of those variables. In summary, the theoretical relationships described above provide some guidance for the development of empirical

169

water-demand equations for residential sectors. The main points are as follows: 1. Final demand for water depends on the price of water (own elasticity) as well as the prices of all other goods (crossprice elasticity) and the level of income. 2. Demand also depends on factors other than price and income and these factors have to be incorporated into the demand function in conformity with economic theory. However, a strict adherence of water-demand studies to economic theory is rarely achieved. Equations (1)–(6) represent demand functions by individual producers or consumers for specific uses of water. Because the empirical data sets are usually aggregated over individual users and often combine different uses of water, Hanemann (1998) observes that empirical studies of existing data make some leap from a theory that applies to individual agents to more aggregate data. For example, a number of empirical studies addressed aggregate water use in the residential sector. Some notable examples include Howe and Linaweaver (1967), Foster and Beattie (1979), Howe (1982), Nieswiadomy (1992), and Epsey et al. (1997). Most of these studies use model specifications which are consistent with economic theory but the theoretical demand models are applied to aggregate data.

1.10.3.2 Empirical Models of Water Use Empirical modeling of water demand consists of the search for variables that help explain water demand and the determination of their relationships to water quantities used. The results of previous studies contain important information about the principal explanatory variables and their mathematical relationship to water demand. Because researchers have defined water demand in many different ways, numerous empirical models appear in the literature. In this section, special emphasis is placed on the models of aggregate demands, which consider total demands of (1) a group of water users who use water for a similar set of purposes or (2) various often dissimilar users within a defined geographical area. Although this section does not present a comprehensive review of the entire literature of water-demand modeling, a sufficient number of studies are included to provide a representative sample of the approaches that have been employed to explore water demand.

1.10.3.2.1 Configuration of data sets In econometric studies, data on economic activities are collected at either micro- or macro-levels. Observations on individual households, families, or firms are referred to as microdata. Regional- or national-level accounts and observations of entire industries are called macrodata. In the analysis of water use, the corresponding types of data are often referred to as disaggregate and aggregate data. Levels of data aggregation by purpose of water use range from the most disaggregate level of end-uses (e.g., toilet flushing or cooling tower makeup) to aggregated sector-wide totals, including domestic, industrial, or other uses. Water use can also be aggregated by summing it over various time periods (day, month, or year) and geographical areas (townships, cities, counties, or states).

170

Predicting Future Demands for Water

For analytical purposes, observations of water use (and of the corresponding explanatory variables) can be obtained and organized in several ways. In mathematical terms, we can describe each data configuration by designating the water use of entity i during time period t as Qit. Depending on which type of arrangement is used, the following three types of data configurations can be distinguished:

•

• •

Time series data, Qit. Recorded or estimated water use of an individual water user, group of water users, or all uses within a defined geographical area, during each time period t in a time series, where i ¼ constant and t ¼ 1, 2, y, T. Cross-sectional data, Qit. Recorded or estimated water use of each individual water user, sector, or geographical area i during time period t, where i ¼ 1, 2, y, n, and t ¼ constant. Pooled time series and cross-sectional data, Qit. Recorded or estimated water use of each individual user, sector, or geographical area i, in each time period t, where i ¼ 1, 2, y, n and t ¼ 1, 2, y, T.

In time series data, observations of all variables in the data are taken at regular time intervals (e.g., daily, weekly, monthly, or annually). In cross-sectional data, observations are taken at one time (either a point in time or time interval) but for different entities (such as households, firms, sectors of water users, cities, counties, or states). Pooled data combine both the time series and cross-sectional observations to form a single data matrix. A special case of pooled data is known as panel (or longitudinal) data, which represent repeated surveys of the same cross-sectional sample at different periods of time. The above types of data configurations form an empirical basis for developing water-use relationships. The mathematical form (i.e., linear, multiplicative, and exponential) and the selection of the right-hand side (RHS) or independent (explanatory) variables depend on the type and aggregation of water demand represented by the left-hand side (LHS) or dependent variable. Depending on the purpose for which water-use estimations are to be used, different representations of the dependent variable may be employed. For example, in hydrologic studies of surface and groundwater resources, water use is usually represented as daily, monthly, or yearly withdrawals at a point such as a river intake or a groundwater well. Because the water withdrawn is typically used (or applied) over a larger land area, an equivalent hydrologic definition of water use would be the use of water within a defined geographical area (e.g., an urban area, a township, a county, or a river basin or subbasin) which is obtained from a single point of withdrawal. While the quantities of water withdrawn in time can be precisely measured at the withdrawal points, they cannot be modeled without an appropriate consideration of the nature of water demands for which the withdrawals are made. The nature of water demand depends on the aggregation of water uses and users. The aggregation levels can range from a single user at a point location to a diverse group of users within a geographical area. For an individual user, water use can be represented as a sum of quantities of water used for specific purposes (or enduses) as

Qjt ¼

X i

qijt

ð7Þ

where Qjt is the total water use of an individual user j during time period t and qijt the water used for a specific purpose of use i, such as garden watering, washing, or cooling during time period t. At a higher level of aggregation, water use within a larger geographical area such as a city, county, or river basin can be represented as a sum of water use for several groups of users within a number of subareas:

Qt ¼

XXX j

k

Qjkgt

ð8Þ

g

where Qt is the aggregate water use of all individual users j within k user sectors within geographical subareas g and Qjkgt designates water use by individual users in the area. In the case of an urban water-supply system, j would represent individual customers, k would represent major user sectors such as the single-family residential or commercial sectors, and g would represent sections of the city or pressure zones. Generally, water use at any level of aggregation Qt can be modeled as a function of explanatory variables Xi. However, because different components of aggregate water demand may be determined by different sets of explanatory variables and different functional forms, more precise models can be obtained by disaggregating demand Qt into its components, especially sectoral demands and modeling each component separately.

1.10.3.2.2 Functional forms and model parameters The most common approach to modeling water demand relies on multiple regression techniques. Usually, the dependent variable is assumed to be a linear function of several independent variables. For example, if there are three independent variables, the model can be written as

Q ¼ a þ b1 X1 þ b2 X2 þ b3 X3 þ e

ð9Þ

where a, b1, b2, b3 are the estimated regression coefficients, X1, X2, X3 the independent variables assumed to affect independent variable q, and e the random error term. The coefficients of a and bi are estimated by finding the values that minimize the sum of the squared deviations for the observed values of the dependent variable from the values of the dependent variable predicted by the regression equation. In order for the ordinary least-squares (OLS) regression analysis to yield valid results, the error term is assumed to be normally distributed and to have zero mean, common variance across all observations and to be independent of all explanatory variables. Also, there should be no correlation between independent variables and the distribution of the dependent variable should be approximately normal. Two additional conditions are that (1) none of the independent variables can be an exact multiple (or linear) combination of any other independent variable and (2) the number of observations must exceed the number of coefficients being estimated. When the five basic assumptions of regression model are satisfied, the OLS procedure would produce unbiased estimates of the regression coefficients a and bi , which have minimum variance among all unbiased estimates.

Predicting Future Demands for Water

Alternatives to the linear model include the log–log model (which is linear in the logarithms of the dependent and independent variables) and the semi-log models (in which either only the dependent variable is or only the independent variables are transformed into logarithms). For example, the log-linear (or double log) model with k explanatory variables can be written as

Qit ¼ ea0

Y

Xit bk  eeit

X

Z¼

bk ln Xkit þ eit

ð10bÞ

k

In some empirical models, Qit is converted into per capita withdrawals, for example, public-supply withdrawals are divided by the population served in the study area. Equation (10a) can be used to represent an aggregate demand curve or a market demand curve. According to a priori (theory-based) expectations, the demand curve would be negatively sloped. The shape and position of the demand curve are determined by the values of other explanatory variables. For residential demand, these variables may include income, household size, temperature, and rainfall. According to expectations, the effect of increasing income would be to shift the curve to the right, so that the same price would result in progressively larger quantities of water being used. The effect of increasing precipitation during the growing season would shift the curve to the left. The entire demand curve does not have to be known in order to determine the effect of independent variables on water demand (Boland et al., 1984). It is usually sufficient to know how specified incremental changes in explanatory variables will affect water use. In the case of price, this information is contained in the slope of demand curve. The slope gives the incremental change in water use for an incremental change in price, at some position on the curve. Because of the units chosen for axes of the demand curve (dollar per unit of water use and units of water use), the slope of the curve has an inconvenient dimension (dollars per unit of water use squared) (Boland et al., 1984). It is customary, therefore, to use a dimensionless measure of the relationship, calculated by dividing fractional (instead of incremental) change in water use by fractional change in price. This dimensionless measure is known as elasticity. For the price variable, it is called the price elasticity of water demand. It is defined for an arc of the curve as

Q2  Q1 Q Z¼ P2  P1 P

dQ P dP Q

ð12Þ

where water use is a function of price and other variables, the ordinary derivative in Equation (12) is replaced with a partial derivative:

Z¼

where variables and coefficients are as defined in Equation (9). By taking the natural logarithm of both sides of Equation (10a), the following log-linear model is obtained:

ln Qit ¼ a0 þ

specific point on the curve as follows:

ð10aÞ

k

qQ P qP Q

ð13Þ

Both arc and point definitions give a dimensionless elasticity, which is expected to be negative (because the demand curve is negatively sloped). Price elasticity may be interpreted as the percentage change in quantity which would result from a 1% change in price. A price elasticity of  0.3, therefore, indicates that 1.0% increase in price would be expected to result in a 0.3% decrease in quantity demanded (use). Conversely, a 1.0% decrease in price would produce a 0.3% increase in quantity demanded. Depending on the magnitude of the calculated elasticity, the demand is said to be perfectly inelastic when Z ¼ 0.0; relatively inelastic when 0.0 4 Z 4 1.0; unitary elastic when Z ¼  1.0; relatively elastic when  1.0 4 Z 4 –N; and perfectly elastic when Z ¼ –N. In other words, demand is said to be relatively inelastic when quantity changes less than proportionately with price, and relatively elastic when quantity changes more than proportionately with price. Elasticity can be calculated for all other explanatory variables using the following formulas for the most common functional forms. For any variable X, the method of calculating the elasticity value from the estimated regression coefficients in four different functional forms is as follows: 1. For the linear form with no log transformation of any variables, the elasticity with respect to X is calculated in terms of the means of the variables as

Q ¼ a þ bX and Z ¼ b

X Q

ð14Þ

2. For the log–log form (also called double-log), the elasticity with respect to X is constant and equal to the regression coefficient of X:

lnQ ¼ a þ b ln X and Z ¼ b

ð15Þ

3. For the semilog form where the dependent variable is untransformed and the dependent variables are transformed into their natural logarithms, the elasticity with respect to X is

ð11Þ

where Q  ¼ ðQ1 þ Q2 Þ=2 and P  ¼ ðP1 þ P2 Þ=2. A more frequently used definition is based on the derivative of the demand function, and yields the elasticity at a

171

Q ¼ a þ b ln X and Z ¼

b X

ð16Þ

4. For a semilog form where only the dependent variable is transformed into logarithm, the elasticity with respect to X

172

Predicting Future Demands for Water Table 4 studies

is directly proportional to X:

lnQ ¼ a þ bX and Z ¼ bX

ð17Þ

To compare empirical studies of water demand, it is both simple and instructive to compare only the elasticities with respect to the explanatory variables, thus getting around the difficulty of interpreting direct comparison of regression coefficients. The following sections compare the results of past studies in terms of the elasticities of key explanatory variables.

Variables used in municipal and residential water-use

Explanatory variable

Variable definitions

Population

Number of users per account Population density Average number of residents per water meter Marginal price of the last unit of water used Average price or total water billed divided by total use Ratio of average price to marginal price Monthly income per capita per dwelling unit Residential property value Per capita income Average household income Median household income Percent of families in low income bracket Imputed rent derived from home value Number of housing units Percent of units by housing type Percent of units occupied by owners Average housing units per acre Number of rooms per dwelling Number of bathrooms House size Lawn size Outdoor irrigable area Median number of rooms Building age Number of persons per dwelling unit Age of the head of household or spouse Educational attainment of the head of household Median age of householders Percent in family households Percent of married households Summer average evapotranspiration rate Monthly effective evapotranspiration rate Monthly rainfall Average monthly rainfall between last spring freeze month and first fall freeze month Precipitation per billing period Mean annual rainfall Precipitation during growing seasons Average temperature for months between last spring freeze month and first fall freeze month Monthly average temperature Monthly average of maximum daily temperatures Number of days without significant rainfall (Z 0.04 in) times the month’s average temperature Number of retail establishments Value added in manufacturing Number of employees in all sectors Number of production workers in manufacturing

Water price

Income

1.10.3.2.3 Elasticities with respect to major determinants of water use Past empirical studies of water demand have used a broad array of possible explanatory variables. Table 4 lists variables in seven major categories and their definitions. These were found in empirical studies of municipal and residential water use. Estimated elasticities of explanatory variables can be obtained from the published empirical equations either as constant elasticities in double-log models or as calculated elasticity values at the mean values of the dependent and independent variables in linear and semilog models. The empirically derived elasticities of key explanatory are summarized in the following. Price elasticity. Economic theory assumes that consumers respond to economic incentives by adopting behaviors that maximize their well-being. In one of the earliest studies of urban water demand, Metcalf (1926) documented a relationship between water use and price that implied price elasticity of demand in the range of  0.40 to  0.65. A substantial body of literature has been published since to confirm that consumers respond to changes in the price of water (Boland et al., 1984; Epsey et al., 1997). Empirical estimates of the price response (elasticity) generally range between  0.1 and  0.9 with higher (absolute) values in industrial and agricultural uses. These values of price elasticity indicate that a 1.0% increase in price would result in a 0.1– 0.9% decrease in water use. Table 5 shows the range and most likely values of price elasticities of water demand for several types of water users. The price elasticity values were obtained from 60 empirical studies of water demands. While these elasticity coefficients indicate that demand is relatively inelastic with respect to price, significant increases in price are expected to result in major reductions in demand. During water shortages, rationing through pricing has proved to be an effective strategy for achieving significant reductions in demand. For example, during the 1988 water shortages in Santa Barbara, California, the price was raised to 27 times the normal level (from $1.09/100 cubic feet or $0.39/m3 to $29.43/100 cubic feet or $10.40/m3) to deter all but the most essential uses of water in the city (Ferguson and Whitney, 1993). As a result, the sector-wide demands were reduced by 56% in the single family, 41% in multifamily, and 20% in commercial sector. Also, average wastewater flows were reduced by 45%. Although price increases were accompanied by the implementation of a sprinkling ban and other conservation measures, the effect of the price increase on indoor

Housing

Family composition

Weather

Othera

a

Used in municipal demand models only.

use (unaffected by sprinkling restrictions) implies a short-term elasticity of  0.22. Income elasticity. Economic theory indicates that together with price, income is a key determinant of residential water demand, because the latter determines the consumer’s ability to pay for water. Consumers decide on what features and

Predicting Future Demands for Water Table 5

173

Empirical price elasticities of water demand

Demand category

No. of studies

No. of estimates

Range of price elasticitiesa

Median value

Combined urban demand Residential demand Single-family only Nonresidential demand Commercial Industrial Institutional Agricultural irrigation

25 58 24 15 6 19 3 10

93 256 94 160 53 101 54 34

– – – – – – – –

– – – – – – – –

0.11 0.18 0.22 0.27 0.24 0.33 0.24 0.24

to to to to to to to to

– – – – – – – –

0.58 0.50 0.48 0.87 0.92 0.88 0.94 0.97

0.40 0.33 0.31 0.54 0.34 0.58 0.47 0.46

a

The range shows the 25th and 75th percentile in the distribution of reported estimates.

Table 6

Empirical income elasticities of residential water demand

Demand category

No. of studies

No. of estimates

Range of income elasticitiesa

Median value

Residential demand Single-family only Multifamily only Municipal demand All sectors

37 24 2 23 86

137 82 2 38 259

0.20 0.10 0.22 0.19 0.14

0.37 0.18 0.22 0.31 0.31

to to to to to

– – – – –

0.61 0.39 0.23 0.58 0.55

a

The range shows the 25th and 75th percentile in the distribution of reported estimates.

conveniences they want in their residences, and what technology they want to employ to achieve them, considering the required investments and the price of water. Table 6 compares income elasticity estimates which were derived from 86 studies of residential (and combined municipal) water demand. For all studies, the range of 25th and 75th percentile estimates is between 0.14 and 0.55 with a median value of 0.31. Elasticities with respect to air temperature and precipitation. Weather conditions influence water demand because some uses of water such as landscape or crop irrigation are sensitive to variables such as precipitation, air temperature, or evapotranspiration. Table 7 shows elasticity estimates for air temperature and precipitation which were derived from 30 studies of residential and nonresidential water demand. For all studies, the range of 25th and 75th percentile estimates of the elasticity of temperature is between 0.43 and 2.15 with a median value of 1.15. The corresponding percentile values of the elasticities of precipitation are between  0.03 and  0.19 with a median value of  0.07. These elasticities indicate that on average demand is more than 10 times more responsive to changes in air temperature than to changes in precipitation. Elasticity with respect to production output. Output in manufacturing activities is often used as an explanatory variable in industrial water-use studies. However, different studies used different proxies for output, such as output value, number of employment hours, and number of employees. Table 8 lists 10 estimates of elasticities for output variables. All estimates are positive, and range from 0.48 to 1.94. Output elasticities 41 indicate that the use of water increases faster than output. The values in Table 8 show that this is the case for some industrial end-uses of water (i.e., cooling, processing, and steam generation) and some industrial categories. Elasticities of output for stone products, photographic

equipment, heavy industry, and paper production are o1, thus indicating that water use increases slower than output.

1.10.3.3 Other Water-Use Relationships The preceding sections described water-use relationships that derive from the economic theory of water demand. However, the limited availability of data on economic variables and the aggregate nature of data on water use often preclude the development of econometric models of water demand. A noneconomic approach is sometimes used to develop wateruse relationships which represent water requirements for different types of water users. Such approaches are explicitly or implicitly based on the assumption that the quantities of water used relate to a technical or physical requirement and are unaffected by economic choice. Nevertheless, the requirements approach remains an option for quantifying water use where econometric models cannot be developed. Two examples of the requirements models are described here.

1.10.3.3.1 Cooling water requirements In once-through cooling systems in steam-based thermoelectric power plants, theoretical water requirements are a function of the amount of waste heat that has to be removed in the process of condensing steam. According to Backus and Brown (1975), the amount of water for 1 megawatt (MW) of electric generation capacity can be calculated:

L¼

6823ð1  eÞ Te

ð18Þ

where L is the amount of water flow in gallons per minute (gpm) per MW of generating capacity; 6823 the units conversion factor; T the temperature rise of the cooling water in F;

174 Table 7

Predicting Future Demands for Water Empirical elasticities of two weather variables

Demand category

No. of studies

Air temperature Residential demand Single-family only Multifamily only Municipal demand Nonresidential All sectors Precipitation

7 6 1 5 2 21

43 11 1 16 3 74

Residential demand Single-family only Multifamily only Municipal demand Nonresidential All sectors

11 7 2 10 1 30

57 22 3 43 6 131

Range of income elasticitiesa

No. of estimates

0.44–3.58 0.88–2.00 0.35–0.35 0.53–1.58 0.02–0.81 0.43–2.15  0.04  0.01  0.04  0.05  0.01  0.03

to to to to to to

Median value

1.37 0.88 0.35 1.31 0.02 1.15

 0.24  0.09  0.12  0.21  0.15  0.19

 0.09  0.02  0.12  0.09  0.03  0.07

a

The range shows the 25th and 75th percentile in the distribution of reported estimates.

Table 8

Examples of elasticities of output variables in industrial water-use models

Study/author

Measure of output

Elasticity

Notes

De Rooy (1974)

Output value

Dziegielewski et al. (1990)

Total employment

Renzetti (1988)

Total number of employee hours

Renzetti (1993)

Output value

1.21 1.22 1.19 0.48 0.60 1.11 0.69 1.94 0.72 0.61

Cooling water use only Processing only Steam generation only SIC328 (cut stone and stone products) SIC386 (photographic equipment and supplies) SIC334 (secondary nonferrous metals) Heavy industry Light industry Self-supplied paper industry Self-supplied textile industry

and e the thermodynamic efficiency of the power plant, expressed as decimal fraction. For example, in a coal-fired plant with thermal efficiency e of 40% and the condenser temperature rise of 20F, the water flow rate obtained from Equation (18) would be 512 gpm per MW. For a typical 650 MW plant, operating at 90% of capacity, the theoretical flow rate would be nearly 300 000 gpm, or 431.3 mgd. The daily volume of cooling water is equivalent to approximately 31 gallons per 1 kWh of generation. According to Croley et al. (1975), in recirculating systems with cooling towers, theoretical makeup water requirements are determined using the following relationship:

W ¼E

1 1

c c0

ð19Þ

where c/c0 is the concentration ratio which compares the concentration of solids in makeup water to their concentration in the recirculating cooling water and E the evaporative water loss which, for a typical mean water temperature of 80 F, can be calculated as

E ¼ ð1:91145  10 6 Þ  aQ

ð20Þ

where a is the fraction of heat dissipated as latent heat of evaporation (for evaporative towers a ¼ 75  85%); 1.911 45  106 the units conversion factor; and Q the rate of heat rejection by the plant in Btu h1, which can be calculated as

Q ¼ 3414426  P 

1e e

ð21Þ

where P is the rated capacity of the plant in MW; 3 414 426 the units conversion factor; and e the thermodynamic efficiency of plant expressed as a fraction. Again, for a typical 650 MW coal-fired plant with 40% efficiency, the heat rejection would be 3329 million Btu h1 and the evaporative water loss would be 5091 gpm. At the concentration ratio c/c0 of 0.25, the makeup water flow would be 6788 gpm or 0.63 gallons per 1 kWh of generation.

1.10.3.3.2 Supplemental irrigation water requirements Water required for supplemental irrigation depends on soil moisture deficit during the growing season which can be derived based on rainfall data. The total seasonal application depth can be determined according the method developed by the Illinois State Water Survey. The method is based on weekly precipitation records for the growing season from 1 May to 31

Predicting Future Demands for Water

August (Dziegielewski and Chowdhury, 2008). Rainfall deficit is calculated by accumulating weekly deficits or surpluses over the consecutive weeks of the growing season. If more than 1.25 inches (in) of rain falls during the first week of the growing season, one-half the amount of rain exceeding 1.25 in is added to the rain amount during the following week. If o1.25 in of rain falls during the first week, the difference between the actual rainfall and 1.25 in is the rainfall deficit that is assumed to be the quantity of water (in inches) applied by irrigation that week. For each subsequent week during the growing season, one-half of the cumulative rainfall during the previous week in excess of 1.25 in is added to the rainfall amount for the week. If the cumulative rainfall amount for a week is less than 1.25 in, then the difference between the actual rainfall and 1.25 in is the rainfall deficit that is assumed to be the quantity of water (in inches) applied by irrigation that week. The rainfall deficits for each week are then added to determine the total irrigation water use during the growing season. This procedure can be expressed in mathematical terms as follows: 1. If the total rainfall in the first week r1o1.25 in, then

d1 ¼ r1  1:25

ð22Þ

2. If the total rainfall in the first week r141.25 in, then

r2

e

d1 ¼ 0 ¼ r2 þ ðr1  1:25Þ=2 d2 ¼ r2 e  1:25

ð23Þ

where r2 e is the effective rainfall in week 2. In week 2, again, the precipitation deficit will be zero if r2 e 41.25 in, and one-half of the precipitation surplus will carry to the next week. The total seasonal rainfall deficit for the 18 weeks (i.e., 4 months) which make up the irrigation season is calculated as

dt ¼

18 X

di

ð24Þ

i¼1

Here, the values of precipitation deficit represent the total depth of water application in inches during the growing season. Thus, the requirements for supplemental irrigation water can be determined using the following formula:

Qt ¼

325; 851 At  dt 12  365

ð25Þ

where Qt is the annual (seasonal) volume of irrigation water in mgd in year t; At the irrigated land area in acres in year t; dt the depth of water application in inches in year t; and the conversion factors represent: 325 851 gallons/acre-foot, 12 inches/foot, and 365 days/year.

175

1.10.4 Demand Forecasting Techniques 1.10.4.1 Forecasting Principles and Criteria A basis for forecasting future quantities of water demanded is required if any type of planning is to be undertaken. As Gardiner and Herrington (1986) simply state: ‘‘y planning, of virtually any kind, requires forecasting’’ (p. 7). In planning for sustainable future water supply, forecasts of water demand which include predictions of improvements in efficiency of water use form a basis of long-term plans for balancing water demand with supply. Other planning activities which require forecasts of water demand include expansion of the capacity of water-supply infrastructure, allocation of limited water supplies among different users, as well as short-term operational and financial planning. Boland (1998) provides an excellent exposition on the basic premises and principles of forecasting. He defines a forecast as a statement about the future. In his earlier writings, he used a more detailed definition which described a forecast as a conditional statement about the future which is likely to materialize if the forecasting assumptions are proved to be correct (Boland, 1985). Boland (1998) also distinguishes the term forecast from the related terms prediction, projection, and extrapolation in terms of implied method or procedure for preparing the statement about the future. Thus, prediction implies nothing about the method and could be an entirely subjective and judgmental statement; extrapolation represents a continuation of past trends and projection suggests a prediction which is influenced (indirectly) by past trends (Boland, 1998: 81). Recently, the International Panel on Climate Change (IPCC) considered some clarifications to this terminology and appears to prefer an alternative term scenario defined as ‘‘ya coherent, internally consistent and plausible description of a possible future state of the world. It is not a forecast; rather, each scenario is one alternative image of how the future can unfold’’ (IPCC, 2008). In a recent study of regional water demands in Northeastern Illinois (Dziegielewski and Chowdhury, 2008), the IPCC definition was adopted in describing water-demand scenarios. The adopted definition ensured that the scenarios would not represent most likely forecasts or predictions, nor would they set upper and lower bounds of future water use but instead they would only describe three alternative paths in demand growth because different assumptions or conditions could result in water demands that are within or outside of the range represented by the three scenarios. In essence, a forecast or a scenario is a translation of a set of assumptions into a future outcome (i.e., a quantity of water used at some point in time). These assumptions are the basis and the main component of any forecast. Other forecast components such as its structure (i.e., time step or level of disaggregation) or computational algorithms and empirically derived models are important but not as critical as the forecasting assumptions. Forecasting assumptions are a part of each forecasting method, regardless of the level of its sophistication. Typically, the simpler methods have only a few assumptions which are likely to be crude and difficult to verify. More elaborate methods, such as the IWR-MAIN model (to be discussed later in this chapter), rely on hundreds of assumptions in deriving a

176

Predicting Future Demands for Water

forecast. Although it is possible to document and make all such assumptions explicit, it is rarely done. Only some of the assumptions, usually those judged by the analyst to be important, are made explicit. However, because the accuracy of the forecast cannot be assessed until the forecast period has passed, its putative validity can be established only by assessing the plausibility of forecast assumptions. Boland (1998) discusses the objective and subjective components of the forecast in the context of a two-step process which consist of explanation and prediction. The explanation step involves the analyses of historical water use and represents the objective part of the forecast. The prediction step applies the factors and relationships which explained water use in the past to generate a forecast. This step necessarily reflects the subjective judgment of the analyst. This subjective judgment is present in all the forecasting assumptions mentioned above. Therefore, a forecast should attempt to make a clear and credible portrayal of the determinants and assumptions behind future water demand. In order to increase acceptability of the forecast by decision makers as well as by other analysts, while preparing the forecast it is important to ensure that: 1. historical water-use data are presented and analyzed for trends and underlying causes and relationships; 2. historical trends and causes are differentiated across user sectors and geographical parts of the study area; 3. major factors influencing water usage rates are considered (e.g., prices, income, and housing densities), and the estimated models are correctly specified; 4. all assumptions are explicit and supported by analysis of past trends or a consensus on future trends; and 5. forecasts utilize an official or a consensus forecast of population and economic growth data. The above elements of the forecast can help ensure that the forecast is understood and accepted by decision makers. The following section describes the specific analytical methods which can be used in constructing a forecast.

1.10.4.2 Forecasting Models and Procedures All forecasts attempt to predict the future value of water use, Qt, as a function of one or more explanatory variables and associated assumptions about the forecasting method and related parameters. The methods differ in terms of the number of explanatory variables and the form of the functional relationship. The forecasting methods also differ with respect to the structure of the forecast, especially in terms of separation of demands into more homogeneous categories of water use. This section describes a range of methods which can be found among the past forecasts of water demand.

In fitting trend lines, several alternatives for functional form exist, including linear, exponential, and logarithmic. When a linear trend line is selected, the future value of water use would be calculated as

Qf ¼ Qt þ bðf  tÞ

ð27Þ

where Qt is water use during the base year t (or the last year with known water demand), b the annual increment in water use Q, and f a future year of the forecast. In an exponential model where the annual (fractional) percentage change r is (b  1)100% when b40, the exponential growth of the future water use is calculated as

Qf ¼ Qt ð1 þ rÞðf tÞ

ð28Þ

This is a well-known equation for compounding interest in financial calculations. Finally, in a logarithmic (log) model, future water use would be increasing at a decreasing rate. A variant of the logarithmic function is a form linear in its logarithms. Using the multiplicative form the log-linear equation, the future water use can be calculated as

Qf ¼ Qt

 b f t

ð29Þ

where t is the time period of the forecast assuming (t – n) ¼ 1 for the first time period of the historical data series. In this trend function, the percent rate of growth is inversely proportional to time. The trend extrapolation method is often used with aggregate data. The greatest difficulty of this method is deciding on what type of function represents the best fit to the historical data. Both linear and nonlinear trends (e.g., exponential or logarithmic) can show equally good fit to the historical data, but their extrapolation into the future may produce considerably different results. Another significant problem is the main forecasting assumption behind this method: that the historical rate of growth will continue into the future and produce the same effects. Boland (1998) considers this method to be ‘‘yto simplistic for virtually any application’’ (p. 85) because of its implicit assumption that water use is explained by the passage of time and that all other variables which are known to affect water use are perfectly correlated with time or their effects cancel each other. However, the use of time trend as one of the explanatory variables in water-use models can be helpful in capturing the residual effects of unspecified variables once the effect of known factors which affect water use is accounted for. Trend extrapolation is also used in deriving future values of some explanatory variables.

1.10.4.2.1 Time trend forecasting 1.10.4.2.2 Water requirement forecasts

Future water demand can be determined by extending the historical trend in the past records of water use. This method would be termed extrapolation as defined by Boland (1998). The only analytical step in this method is finding the functional form g in the equation

A longstanding forecasting practice is to assume that water use is proportional to the size of a water-using activity or the number of water users. This proportionality can be expressed using a linear model

Qf ¼ gðQt ; Qt1 ; :::; Qtn Þ

Q ¼ a þ bN

ð26Þ

ð30Þ

Predicting Future Demands for Water 250

or a multiplicative (log-linear) model

ð31Þ

150 100 50

The assumption of strict proportionality is often criticized by the economists because it ignores other factors that can affect water use, and in its application to derived demands it treats the size variable which is often the production output or employment as exogenous to the firm’s decision on water use. However, the representation of water demands as the number of users (or a driver of water use) times average rate of usage (or intensity of water use) is a convenient and practical forecasting approach. A well-known example of this approach is the per capita requirements method, which has been widely used in forecasting urban water use. According to this method, future values of the volume of publicly supplied water for a city or municipality are often obtained by multiplying future population (which represents N in Equation (32)) with an assumed per capita rate of water use (as b in Equation (32)) This method was widely used in the past because the metric of per capita water use can be easily obtained from the production records (see Section 1.10.2.2.1) and total population served can be obtained from the population census data. However, the validity of this method depends on the constancy of the per capita use rate over time. In most urban water-supply systems, per capita use rate is not constant; it fluctuates from year to year and often exhibits long-term increasing or declining trends. For example, Figure 1 illustrates the long-term changes in per capita water use in New York City. The per capita rate had been increasing from 101 gpcd in 1915 to 208 gpcd in 1988 but had declined to 134 gpcd by 2006. A per capita forecast prepared in any year in the past would not have captured the changes in per capita rates of use. Because the per capita approach relies on the assumption that the per capita rate will remain constant throughout the forecast period, the validity of per capita forecasts is doubtful. The main problem is that it ignores the changes in the structure and composition of urban demands over time and also ignores the effects of future changes in the determinants of different component demands. However, the unit-use coefficient methods can have some validity when applied to disaggregate demands of more homogeneous sectors or categories of water use. Examples of disaggregate requirement models are given in the following. Disaggregate requirement forecasts. When total demand in a geographical area is disaggregated by sectors, demand in sector k can be represented as a product of the number of water users

50 19 60 19 70 19 80 19 90 20 00 20 10

40

19

30

19

20

19

19

19

ð32Þ

19

Q ¼ bN

10

0 00

where a, b, b, and g are treated as constants at a given level of aggregate demand. Although this simple approach is criticized by economists (Hanemann, 1998), it is widely used in forecasting water demands for industrial, commercial, and agricultural categories. Aggregate requirement forecasts. By forcing the intercept in Equation (30) to be zero or assuming that the exponent g in Equation (31) is 1, we obtain a simple requirements model in which total water use is strictly proportional to N:

200

GPCD

Q ¼ bN g

177

Year Figure 1 Historical per capita rates of water use in New York City.

and the average rate of use within the sector:

Qkt ¼ S Nkt  qkt k

ð33Þ

where Nkt represents the number of users (or other units) in sector k at time t and qkt the unit use coefficient (or average rate of water use per user) in that sector. Similarly, the total demand within the geographical area can be represented as

Qt ¼ S S S Nkgjt  qkgjt k g j

ð34Þ

where k denotes the disaggregation of water use into homogeneous sectors of water users (e.g., residential, commercial, industrial, and institutional) and g represents the spatial disaggregation of water use into various geographical subareas that are relevant for planning purposes. An example of a unituse coefficient qkgjt could be average use in single-family buildings of the residential sector in a suburban section of a city. This single-coefficient model can be extended by expressing the average rates of water use within each sector as a function of one or more explanatory variables. Usually, the dependent variable is assumed to be a linear function of more than one independent variable. For example, if there are two independent variables, the theoretical model is similar to Equation (8)and can be written as

q ¼ a þ b1 X1 þ b2 X2 þ e

ð35Þ

where a, b1, b2 are estimated regression coefficients, X1, X2 the independent variables assumed to affect independent variable q, and, e the random error term. There are also forecasting methods that are based on landuse categories. Such approaches typically represent a particular case of single coefficient requirement models. The land-usebased models display data in a way that is convenient for infrastructure planning and city master planning. For example, DCSE, a California-based software product and consulting firm, has developed a geographic information system (GIS)based water-demand forecasting procedure for estimating

178

Predicting Future Demands for Water

water demands in response to changes in land-use and related use factors. Water-demand projections are based on distribution of land-use categories and the corresponding water-use factors. Because land use in a service area can change due to conversion of land to residential, commercial, industrial, and other urban uses, the result will be a change in total water use.

1.10.4.2.3 Demand forecasts In practice, these methods require that regression models are constructed using the proper specification of the economic variables such as price and income and the level of output in the case of derived demand (see Section 1.10.3.1). It should be noted that econometric models are derived from observations which represent the points of intersection of demand and supply curves, and they usually represent reduced forms, as opposed to the structural equations which represent the true demand functions. Unfortunately, demand models are generally available only for single-family residential water use. Studies of multifamily residential sector and major industrial, commercial, and agricultural sectors are very limited in number. Nevertheless, forecasters should consider developing econometric models in order to probe the validity of estimated (or assumed) wateruse relationships.

1.10.4.3 Dealing with Forecast Uncertainty As mentioned earlier, all forecasts of future water demand are inherently uncertain. Generally, the uncertainty associated with the analytically derived future values of water demand can come from a combination of the following distinct sources: 1. Random error. The random nature of the additive error process in a linear (or log-linear) regression model which is estimated based on historical data guarantees that future estimates will deviate from true values even if the model was specified correctly and its parameter values (i.e., regression coefficients) were known with certainty. 2. Error in model parameters. The process of estimating the regression coefficients introduces error because estimated parameter values are random variables which may deviate from the true values. 3. Specification error. Errors may be introduced because the model specification may not be an accurate representation of the true underlying relationship. 4. Scenario uncertainty. No future values of any model variables can be known with certainty. Various assumptions must be introduced when projections are made for the water-demand drivers (e.g., population, employment or irrigated acreage, income, price, precipitation, and other explanatory variables). The first three sources of error can be addressed by a careful analysis of the data and model parameters. The fourth source of error – the scenario error (or assumption error) – requires an explicit evaluation of assumptions through sensitivity analysis, Monte Carlo analysis, or the use of scenario-based forecasts.

1.10.4.3.1 Model-dependent prediction intervals In econometric forecasts, each empirically derived model can be tested for specification error by using Ramsey’s specification tests (Ramsey (1969), the Breusch–Pagan–Godfrey test Breusch and Pagan (1979)), Glejser’s test (Glejser, 1969), and Harvey’s test and White’s test (White, 1980) for heteroscedasticity. The specification and heteroscedasticity tests allow the analysts to develop predictive equations which minimize the errors from misspecification of the model and biases in model parameters. Other model-dependent errors can be quantified using confidence intervals (Dziegielewski et al., 2005). For example, assuming that the errors are normally distributed in a log-linear model in which Y designates water use, it can be shown that

EðYjexplanatory variablesÞ ¼ e se

2

=2

ðe ln Y Þ

ð36Þ

Thus, in log-linear models, the predicted value denoted as Y˜ is given by 2 Y˜ ¼ e s^e =2 ðe ln YÞ

ð37Þ

^e 2 is the mean square error of the log-linear model and where s lnY^it the predicted value obtained from the log-linear models. It is straightforward to obtain the in-sample prediction confidence intervals in a linear model. However, in a log-linear model, the in-sample prediction intervals are obtained under the assumption that the errors are normally distributed. Thus, for normally distributed errors the variance of Y˜ in Equation (37) is estimated by

  ˜ ¼ exp 2lnY þ Vii s ^2e VarðYÞ "  m=2  m #   ^2e s 2^ s2e 2 ^e 1   1  exp Vii s m m ð38Þ ^e 2 is the square of the standard error of the logawhere Vii s rithmic prediction (i.e., lnYit ), m the degrees of freedom, ^e 2 the mean square error of the log-linear model. The and s standard error of Y^ is denoted as

˜ ¼ SEðYÞ

qffiffiffiffiffiffiffiffiffiffiffiffiffiffi ˜ VarðYÞ

ð39Þ

0 Assuming that the Y˜ s are asymptotically normally distributed, the confidence interval for the prediction can be obtained as ˜ Y˜ þ za=2  SEðYÞ, ˜ ½Y˜  za=2  SEðYÞ; where za=2 is the critical value from a normal distribution for a prespecified a. However, for the out-of-sample predictions, the square of the ^e 2 ) is not standard error of the logarithmic prediction (i.e., Vii s available. To rectify this, one can use the average standard error of predictions (average over all observations in the historical data).

1.10.4.3.2 Dealing with forecast assumptions error The uncertainty caused by forecasting assumptions can be addressed by using scenario forecasts, sensitivity analysis, or probability forecasts. The scenario approach requires the

Predicting Future Demands for Water

development of scenario narratives and selection of a set of key forecasting variables and assumptions for each scenario. Usually, only a small number of scenarios are developed. The purpose of the scenarios is to capture future water demands under different sets of possible future conditions. Although the scenarios convey a sense of the range of future water use, they are not constructed to set upper and lower bounds of future values. Different sets of assumptions or conditions could result in demands that are within or outside of the range represented by the defined scenarios. Sensitivity analysis evaluates the sensitivity of the forecast values to changes in forecasting assumptions – one assumption at a time. In the case of the values of explanatory variables, alternative forecast values are generated by changing the value of the variable of interest while keeping the values of all other variables and forecasting assumptions unchanged. For key explanatory variables, the same information can be obtained by using the values of constant elasticities of water demand with respect to each variable and percent change in variable value. However, the effects of other assumptions are often less explicit and can be best determined by generating the alternative forecast values. Finally, probabilistic forecasts can be obtained by simultaneously varying two or more assumptions and generating a probability distribution of the values of water-demand forecast. Often, Monte Carlo simulation is used to generate very large numbers of outcomes based on distribution functions for the independent variables. Table 9 (IWR, 2001) shows an example of a probabilistic forecast for a water demand in public systems in a region in Virginia. The useful feature of such a forecast is the exceedance probability of the forecast values. The results in the table show that the mean forecast value for 2050 is 90.95 mgd and that there is only a 5% chance that the value of 96.94 mgd will be exceeded.

Table 9

179

Example of a probabilistic forecast

Year/water demand

2010

2050

Minimum Maximum Mean Std dev. Variance Skewness Kurtosis Distribution percentiles 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% 55% 60% 65% 70% 75% 80% 85% 90% 95%

62.71 71.67 67.16 1.34 1.81 0.04 2.73

80.23 105.22 90.95 3.50 12.28 0.25 3.02

64.98 65.43 65.76 66.01 66.22 66.42 66.61 66.78 66.97 67.14 67.32 67.50 67.69 67.89 68.10 68.34 68.61 68.93 69.38

85.45 86.58 87.30 87.89 88.44 88.98 89.46 89.93 90.40 90.84 91.28 91.75 92.20 92.71 93.23 93.85 94.56 95.49 96.94

From IWR (Institute for Water Resources) (2001) An evaluation of the risk of water shortages in the Lower Peninsula, Virginia. Prepared by Werick WJ, Boland JJ, Gilbert J, Dziegielewski B, Kiefer J, Massmann J, and Palmer RN. IWR Special Report. Alexandria, VA: US Army Corps of Engineers.

1.10.4.4.1 End-use accounting system

1.10.4.4 Forecasts with Conservation The increasingly important role of demand-side options in water-supply planning creates needs for methods to estimate the effects of various water conservation programs on future water demands. However, the effects of demand-side programs cannot be assessed without a detailed knowledge of water uses in a study area and without an understanding of the important factors that influence them now and will influence them in the future. Thus, the most important feature of forecasting the impacts of water conservation is a high level of disaggregation of water demands. In order to estimate the effects of water conservation measures on future demand, it is usually necessary to disaggregate water demand into the specific end-uses of water. The conservation impacts (i.e., water savings) are usually estimated as the reduction in average rates of water use for specific purposes. For example, in the residential sector average household water use can be represented as a summation of average water used for toilets, showers, kitchen faucets, washing machines, and landscape irrigation. These end-uses may each be affected by water conservation measures which would result in a lower average per household use.

Forecasting methods which focus on impacts of long- and short-term demand management measures usually employ a highly disaggregated end-use accounting approach. Dziegielewski et al. (1993) proposed the following equation for estimating water use for each end-use:

" qe ¼

X

! Mu Su

# UþK F A

ð40Þ

u

where qe is the quantity of water used by a given end-use e (gpd per unit); Mu the mechanical end-use parameter for efficiency class u (e.g., gpm and gallons per flush); Su the fraction of the sectoral end-use within each efficiency class (such as nonconserving, conserving, and ultraconserving); U the intensity of use parameters (e.g., flushes per day per unit, minutes of use per day per unit); K the mechanical parameter representing the rate of leakage; F the fraction of end-uses with leakage; and A the fraction of water-using entities in which end-use is present. Because the structure of end-use demands in a study area at any point in time will not remain constant over the forecast horizon, it is necessary to quantify the effects of various external factors on the parameters of Equation (41)(e.g., Su, U, F, A). For example, future increases in the price of water could

180

Predicting Future Demands for Water

decrease the incidence of leaks (F) in the short run and would also affect the distribution of end-uses among the efficiency classes (Su) in the long run. The other two parameters of the end-use equation (intensity U and presence A) will also be affected by changes in price. Changes in other explanatory variables will also affect the end-use parameters. For example, in the residential sector, in addition to price, the end-use parameters will be affected by variables such as income, household size, housing density, and weather.

5. nonurban; 6. user-specified sector; and 7. total residential. These categories correspond to the housing types used by the US Bureau of the Census. Average rates of water use within each residential subsector are estimated using econometric water-demand models. For the single-family sector, a log–log equation in the form of a multiplicative function of seven independent variables is used:

1.10.4.4.2 Baseline and restricted forecasts Impacts of conservation have to be assessed for different types of conservation measures, including passive, active, and emergency measures as well as future changes in water prices. It is a common forecasting practice to generate one forecast which assumes no active intervention to change future water demands. This forecast is usually referred to as a baseline or unrestricted forecast. Then one or more alternative forecasts are prepared based on assumed active efforts to reduce demands. These forecasts are referred to as restricted forecasts or forecasts with conservation. Some of the forecasting procedures and models described in this and the previous subsections have been incorporated into computer software programs, which structure the input data, provide computational algorithms, and generate forecast outputs. One example of the forecasting program is IWRMAIN.

1.10.4.5 Forecasting Software: The IWR-MAIN Program One of the first known software programs for forecasting urban water use MAIN II (Municipal And Industrial Needs) was developed by Hittman Associates Inc. (1969). The model was based on the residential and commercial water-use research projects carried on at the Johns Hopkins University (Linaweaver et al., 1966; Wolff et al., 1966; Howe and Linaweaver, 1967). During the 1980s, the Institute for Water Resources (IWR) of the US Army Corps of Engineers undertook a substantial research effort to update the model and modify it for easy access on personal computers. The product of this effort was a public-domain software package called IWR-MAIN version 5.1 (Dziegielewski and Boland, 1989) and the IWR-MAIN Water Demand Analysis Software, version 6.0 (Dziegielewski, 1993; Dziegielewski et al., 1996).

1.10.4.5.1 Model structure and procedures The model disaggregates total urban water use by customer sectors, time periods, spatial study areas, and end-use purposes. Through sectoral disaggregation, forecasts of water use can be prepared for major sectors of water users, including residential, nonresidential (constituting manufacturing, commercial, and governmental), public, and other. Demands can be further disaggregated within each sector. Within the residential sector, there are seven subsectors available for forecasting water demands. These include 1. 2. 3. 4.

single-family – 1 attached, 1 detached units; multifamily low density – 2, 3, 4 units per structure; multifamily high density – 5 or more units per structure; mobile homes;

Qr ¼ aIb1 Hb2 Lb3 T b4 Rb5 P b6 eb7 B

ð41Þ

where Qr is predicted residential water demand in gallons per housing unit per day; I the median household income in $ per year; H the average household size (persons); L the average housing density (units per acre); T the daily-maximum air temperature in farenheit; R the rainfall in inches; P the marginal price of water (including sewer) in $/1000 gallons; B the fixed charge or rate premium (i.e., Nordin’s bill difference) of the water/wastewater tariff in $ per month; a the constant; bi the constant elasticities of explanatory variables; b7 the coefficient of the rate premium (representing the water tariff structure); and e the base of the natural logarithm. The model can use generic water-use equations with default elasticities for explanatory variables in the residential sector which were derived through a meta analysis of empirical literature. Once average water use per household has been estimated by Equation (41), total water demand for a given subsector, season, and year is calculated by multiplying the average use rate Qr by the driver variable (number of households). Nonresidential water use is disaggregated by the model into the following major industry groups: (1) construction; (2) manufacturing; (3) transportation, communications, and utilities; (4) wholesale trade; (5) retail trade; (6) finance, insurance, and real estate; (7) services; and (8) public administration. These eight major industry groups are classified according to the US Department of Commerce Standard Industrial Classification (SIC) codes. Within each major industry group, SIC codes distinguish more homogeneous groups at the twodigit SIC level and even further at the three-digit SIC level. Because no generally applicable demand models exist that contain elasticities for price, labor productivity, cooling degree days, or the other variable for nonresidential water use, IWRMAIN estimates nonresidential water use by multiplying employment with water-use coefficient. Water use per employee coefficients are available for each category and were derived from a sample of about 7000 nonresidential establishments (Table 10). Some estimates in Table 10 represent relatively high values of per employee usage rates (e.g., security and commodity brokers, real estate firms, or legal services), because they represent uses of water other than indoor domestic uses. For example, in large office buildings the component end-uses may include landscape irrigation or makeup water for cooling towers.

Predicting Future Demands for Water Table 10 Water use coefficient for industrial, commercial, and institutional categories of water use Description of SIC categories

Sample size

Construction (SIC 15–17) General building contractors Heavy construction Special trade contractors Manufacturing (SIC 20–39)

246 66 30 150 2790

31 118 20 25 164

252 20 91 62 83 93 174 211 23 116

469 784 26 49 36 2614 37 267 1045 119

10 83 80 395 304 409 182 147 55 226

148 202 178 194 68 95 84 66 36 50

3 32 100 1 10 17 13 31 19 751

68 26 85 5 353 171 40 55 51 43

518 233 1044

46 87 93

Building materials and garden supplies General merchandise stores Food stores Automotive dealers and service stations Apparel and accessory stores Furniture and home furnishing stores Eating and drinking places Miscellaneous retail Finance, insurance, and real estate (SIC 60–67)

56 50 90 198 48 100 341 161 238

35 45 100 49 68 42 156 132 71

Depository institutions Nondepository institutions Security and commodity brokers Insurance carriers Insurance agents, brokers, and service Real estate Holding and other investment offices

77 36 2 9 24 84 5

Food and kindred products Textile mill products Apparel and other textile products Lumber and wood products Furniture and fixtures Paper and allied products Printing and publishing Chemicals and allied products Petroleum and coal products Rubber and miscellaneous plastics products Leather and leather products Stone, clay, and glass products Primary metal industries Fabricated metal products Industrial machinery and equipment Electronic and other electrical equipment Transportation equipment Instruments and related products Miscellaneous manufacturing industries Transportation and public utilities (SIC 40–49) Railroad transportation Local and interurban passenger transit Trucking and warehousing US postal service Water transportation Transportation by air Transportation services Communications Electric, gas, and sanitary services Wholesale trade (SIC50-51) Wholesale trade – durable goods Wholesale trade – nondurable goods Retail trade (SIC 52–59)

Table 10

Continued

Description of SIC categories

Sample size

Services (SIC 70–89)

1878

137

197 300 243 108 42 40 105 353 15 300 55 9

230 462 73 217 69 110 429 91 821 117 106 208

45 5 60 25

212 58 73 106

2 4 6 6 5 2

155 18 87 101 274 112

gped coefficient

62 361 1240 136 89 609 290 (Continued )

181

Hotels and other lodging places Personal services Business services Auto repair, services, and parking Miscellaneous repair services Motion pictures Amusement and recreation services Health services Legal services Educational services Social services Museums, botanical and zoological gardens Membership organizations Engineering and management services Services, NEC Public administration (SIC 91–97) Executive, legislative, and general Justice, public order, and safety Administration of human resources Environmental quality and housing Administration of economic programs National security and International affairs

gped coefficient

gped, gallons per employee per day. Source: Dziegielewski et al. (1996), IWR-MAIN Version 6.0.

1.10.4.5.2 IWR-MAIN conservation forecasts The conservation subroutine of IWR-MAIN 6.0 disaggregates urban water use into 20 end-uses covering residential and nonresidential water uses, both indoor and outdoor. Using Equation (40), each end-use is divided into three classes of efficiency (nonconserving, conserving, and ultraconserving). The rate of use in each of these efficiency classes is defined by the mechanical parameters (M1, M2, and M3). For example, toilets have mechanical parameters of 5.5 gallons per flush (gpf), 3.5 gpf, and 1.6 gpf for nonconserving, conserving, and ultraconserving end-uses, respectively. The percent of sector entities in each efficiency class (S1, S2, and S3) must be determined based on local information. For example, 40% of single-family residential units have toilets with 5.5 gpf, 50% use toilets with 3.5 gpf, and 10% use toilets with 1.6 gpf. The intensity (U) for each fixture defines how frequently a particular end-use occurs on a per household, per employee, or other basis. In the residential sector, an event or flow rate that is defined on a per person basis is multiplied by the average number of persons per household to determine the intensity of the end-use at the household level. For example, a toilet is flushed an average of 5.0 times per person per day. This value, when multiplied by the average number of persons per household in the study area, gives the site-specific intensity value for the toilet end-use. The presence of a particular end-use may vary by forecast year and represents the fraction of units (housing units, employees, or other measures of size) in a water use sector that

182

Predicting Future Demands for Water

have that particular end-use. The ability to adjust the presence factor by forecast year allows the forecaster to account for potential changes in the presence of an end-use. For example, the presence of dishwashers for the single-family sector in the base year may be 76% but could be expected to increase to 84% in 10 years. The conservation savings procedures distinguish between active and passive demand management programs. Active programs include interventions by water providers or other entities. Passive demand reduction is a result of natural shifts toward higher-efficiency classes (e.g., from the standard 3.5 gallon per flush toilet to the ultraconserving 1.6 gallon per flush toilet). The shifts of end-uses toward higher classes of efficiency are brought about primarily by plumbing codes that require increased efficiency in water-using fixtures which affect new construction, remodeling, and customer-initiated retrofitting. Shifts from less efficient to more efficient pools of end-uses are also expected to occur naturally over time as technology continually improves. The natural shifts toward higher efficiency pools (often accelerated by plumbing or efficiency codes) are determined using appropriate rates of movement. These movements predict the form in which these shifts take place and additionally show the rate at which they occur. Conservation measures may be defined as long-term, shortterm, or emergency (restricted use) programs. Long-term conservation programs permanently shift the number of entities into higher efficiency classes. These include all plumbing codes (passive programs) as well as active programs. An example of an active program that would be considered a longterm conservation program is a retrofit campaign in which low-flow showerheads are installed. Showerheads usually last about 15 years, after which it is assumed that they will be replaced with units that are at least as efficient, if not more efficient. Short-term conservation measures are implemented during periods of water shortages. These measures target the end-use fixtures by temporarily shifting them into a higher efficiency class. An example would be the distribution of toilet dams. These devices would temporarily shift users into a higher efficiency level but are unlikely to remain in the toilets for a long period of time. Although the short-term conservation measures focus on increasing the efficiency of end-uses, the emergency or restricted use measures focus on temporarily altering the behavior of water-using entities. Like the short-term programs, the restricted use programs evaluate potential reductions in demand which could be enacted during periods of water shortages. A restricted use conservation program might entail, for example, restricting the days when the residents of a community can wash their cars. Restrictive programs target specific end-uses and achieve water savings by using alternate (restrictive) values for presence factors, leakage percents, and intensities.

1.10.5 Example of a Regional Multisector Forecast This example taken from Dziegielewski and Chowdhury (2008) illustrates the development of future water-demand

scenarios for geographical areas that encompass the 11-county regional planning area of Northeastern Illinois, including the counties of Boone, Cook, DeKalb, DuPage, Kane, Kankakee, Kendall, Grundy, Lake, McHenry, and Will (Figure 2). In 2005, total resident population of the area was estimated at 8 743 900 persons and a total of 4 355 200 persons were employed in the local economy. Nearly 96% of the population was served by about 530 public water-supply systems and nearly 400 000 residents relied on private wells. Other water users included 12 large power plants, 352 golf courses, and about 30 000 acres of irrigated cropland. The study generated three water-demand scenarios by major user sectors and geographical service areas within the region that were extended to the year 2050. The scenarios were formulated to represent growth assumptions under current trends (CTs or baseline scenario) as well as under less resource-intensive (LRI or low-growth scenario) and more resource-intensive (MRI or high-growth scenario).

1.10.5.1 Water-Use Relationships The historical data on water withdrawals were used to estimate water-demand relationships that expressed water demand as a function of relevant explanatory variables. Table 11 lists the demand drivers and estimated elasticities of the explanatory variables for each demand sector. Because the dependent and, in most cases, independent variables were converted to natural logarithms, the coefficients represent constant elasticities of water demand. Accordingly, the elasticity of marginal price in the public-supply sector was estimated to be  0.1458, indicating that a 1.0% increase in price is expected to result in a 0.145 8% decrease in demand. The three forecast scenarios were defined by different sets of assumed conditions regarding the future values of demand drivers and explanatory variables. Table 12 compares several key assumptions that were used in constructing the three scenarios. Table 13 provides a summary of the future scenarios of average day water withdrawals for six categories of users within the four major sectors. For 2005, both the reported values and weather-adjusted values are shown. The lower panel of Table 13 shows the sum of total withdrawals with and without the once-through flow for power generation and gross per capita withdrawals without once-through flows. This distinction is made because the very high volumes of water withdrawals for once-through cooling are not directly comparable to withdrawals by other sectors. The three future water-demand scenarios show that total water withdrawals in the 11-county area of Northeastern Illinois will continue to increase to meet the demands of growing population and the concomitant growth in the economy of the region. However, the growth in total water demand could be faster or slower depending on which assumptions and expectations about the future conditions will prevail. Under the baseline (CT) scenario, by 2050, total water withdrawals (excluding water withdrawn for once-through cooling in electric power plants) would increase above the 2005 level by 35.8%, or 530 mgd. During the same period of time, total population is projected to increase by nearly 3 370 000, or 38.5%. This implies that water demand would grow slightly

Predicting Future Demands for Water

183

W I S C O N S I N BOONE

MCHENRY

LAKE

I

L

L

Chicago

DUPAGE

KANE

I

N

O

I

S

e

r

DEKALB

Lake Michigan

Fox

Riv

COOK KENDALL

WILL Ka nk ak

N

*Groundwater is also used within these areas in some cases.

Ri

GRUNDY

ee

10 MILES

INDIANA

ve r

KANKAKEE

Source: Chicago Metropolitan Agency for Planning

Figure 2 Regional forecast area in Northeastern Illinois.

slower than the region’s population. Gross per capita water withdrawals (i.e., total withdrawals by all sectors divided by total population) during the dry year of 2005 were estimated at 182.8 gpcd. Under normal weather conditions, the 2005 demands would have been 169.3 gpcd. In the baseline (CT) scenario, the gross per capita usage would decrease to 166.0 gpcd by 2050. This relatively unchanged per capita rate is a result of assumptions about gradual increases in water prices and a continuation of the historical trend in water conservation. Under the key assumptions of the high-growth (MRI) scenario, future water demands would grow faster than total population if income grows at a somewhat higher rate than under the baseline scenario, if future prices of water do not grow faster than inflation, if no additional gains in water conservation are achieved, and if more population growth takes place in the collar counties of Kane, Kendall, and McHenry in single-family housing. Under these conditions, by 2050, total water withdrawals would increase above the 2005 level by 64.1%, or 949.1 mgd. The growth of water demand would exceed the rate of population growth because of the increasing gross per capita usage rate. By 2050, it would increase to about 200.6 gpcd, as compared to the 2005

weather-normalized rate of 169.3 gpcd. In a sense, the high-growth scenario could be viewed as a warning that there is a possibility of a large increase of water demands in the future. Under the key assumptions of the low-growth (LRI) scenario, future demands would grow significantly slower than population if income grows at a somewhat slower rate than under the baseline scenario, if future prices of water grow significantly faster than inflation, if additional gains in water conservation are achieved, and if more population growth takes place in the urbanized counties of Cook and DuPage in multifamily housing. Under these conditions, by 2050, total water withdrawals would increase above the 2005 level by about 7.2%, or 107.2 mgd. The growth of water demand would be much slower than the rate of population growth because of the decreasing gross per capita usage rate. By 2050, it would decrease to about 131.1 gpcd, as compared to the 2005 weather-normalized rate of 169.3 gpcd. This scenario could be interpreted as a future outcome which would require an intervention in order to maintain a slower growth of demand. This intervention would likely require monitoring and management of water demand and making investments in long-term efficiency of water use.

184

Predicting Future Demands for Water

Table 11

Drivers of water demand and estimated elasticities of explanatory variables

Demand sector

Demand driver

Explanatory variables

Elasticity/coefficient

Public supply

Population served

Power generation

Gross electric generation

Air temperature Precipitation (growing season) Employment fraction Marginal price of water Median household income Conservation trend Unit-use coefficients

Industrial and commercial

Employment

Agricultural and irrigation

Irrigated acres Livestock counts Population

1.0951 –0.0949 0.0931a –0.1458 0.2845 –0.0593 0.67–0.89b 10.8–78.9c 0.3298 –0.0896 0.0279a –0.1077a 0.0032a –0.0074a 1.000 0.03–35.0d 1.6238 –0.2186 0.3499 –0.0325

Domestic self-supplied

Cooling degree days Precipitation (growing season) Manufacturing employment (%) Transportation employment (%) Fraction of self-supplied (%) Conservation trend (linear) Rainfall deficit Unit-use coefficients Air temperature Precipitation (growing season) Median household income Conservation trend

a

Dependent variables are in linear form. All other coefficient represent constant elasticities. The values represent unit withdrawal coefficients in gallons per kilowatt-hour of gross generation in plants with closed-loop cooling systems. c The values represent unit withdrawal coefficients in plants with open-loop once through cooling systems. d The values represent unit use coefficient per animal type. From Dziegielewski B and Chowdhury FJ (2008) Regional water demand scenarios for Northeastern Illinois: 2005–2050. Project Completion Report. Prepared for the Chicago Metropolitan Agency for Planning, Chicago, IL, USA, 15 June 2008. b

Table 12

Scenario assumptions for factors affecting future water demands in the 11-county area of Northeastern Illinois

Factor

CT (baseline)

LRI scenario

MRI scenario

Distribution of population of growth Mix of commercial/industrial activities Median household income Demand for electricity

Official projections

Shift to Cook and DuPage counties Decrease in water-intensive activities Growth of 0.7% yr1 9.61 MWh per capita per year þ no growth No new power plants, 3 units retired, 2 plants convert to closed-loop cooling 50% higher rate than historical trend Growth of 2.5% yr1 Decreasing cropland þ no new golf courses

Shift to Kane, Kendall, and McHenry counties Increase in water-intensive activities Growth of 1.0% yr1 9.61 MWh per capita per year þ 0.56% per year growth Two new power plants in study area with closed-loop cooling

Power generation

Current trends Growth of 0.5% yr1 9.61 MWh per capita per year þ 0.56% per year growth No new plants within study area, 3 units retired

Water conservation

Historical trend

Future water prices Irrigated land

Growth of 0.9% yr1 Constant cropland þ new golf courses: 10 per decade

No extension of historical trend Constant in 2005$ Constant cropland þ new golf courses: 20 per decade

CT, current trends; LRI, less resource-intensive; MRI, more resource-intensive. From Dziegielewski B and Chowdhury FJ (2008) Regional water demand scenarios for Northeastern Illinois: 2005–2050. Project Completion Report. Prepared for the Chicago Metropolitan Agency for Planning, Chicago, IL, USA, 15 June 2008.

1.10.5.2 Effects of Key Forecast Assumptions The plausibility of the scenario forecasts depends on how reasonable the assumptions about future changes in the explanatory variables and the size of their effects on water demand are. In order to verify the forecasts of total or sectoral water use in the future, it is helpful to re-analyze the resultant

forecasts in terms of the underlying effects of demand drivers and the unit usage rates. The total change in demand is the result of the change in the value of the driver (e.g., population served, employment, and irrigated acreage) and the change in unit usage rate. Table 14 compares the resultant unit usage rates for the major demand sectors for each scenario. The changes in future values of these rates are a result of changes in

Predicting Future Demands for Water Table 13

185

Summary of water withdrawal scenarios for Northeastern Illinois (in mgd)

Sector

2005-R

2005-Na

2050-CT

2050-LRI

2050-MRI

Public supply Self-supplied I&C Self-supplied domestic Irrigation and agriculture Power plants (makeup) Power plants (through flow) Total – all sectors Total w/o through-flow power

1255.7 191.6 36.8 62.0 52.3 4207.2 5805.6 1598.4

1189.2 162.4 31.8 44.6 52.3 4207.2 5687.5 1480.3

1570.2 291.6 41.2 55.4 52.3 3830.2 5840.9 2010.7

1217.9 222.1 37.3 43.8 66.4 2472.3 4059.8 1587.5

1837.2 391.4 49.3 60.7 90.8 3830.2 6259.6 2429.4

a

For comparison with future values, the 2005 withdrawals were adjusted by the model to represent normal weather conditions. R, reported; N, normal weather; CT, current trends; LRI, less resource-intensive; MRI, more resource-intensive; mgd, gallons per capita per day; 1.0 mgd, 3784.4 m3. From Dziegielewski B and Chowdhury FJ (2008) Regional water demand scenarios for Northeastern Illinois: 2005–2050. Project Completion Report. Prepared for the Chicago Metropolitan Agency for Planning, Chicago, IL, USA, 15 June 2008.

Table 14

Summary of unit usage rates by sector

User sector

2005-R

2005-Na

2050-CT

2050-LRI

2050-MRI

Public supply (gpcd) Self-supplied I&C (gped) Self-supplied domestic (gpcd) Irrigation and agriculture (in yr1) Power plants (makeup)(gal. kWh1) Power plants (through-flow)(gal. kWh1) Gross per capita rates (gpcd)

150.4 130.6 93.6 15.3 0.86 41.4 182.8

142.1 109.3 81.1 11.0 0.86 41.4 169.3

134.9 120.1 86.4 11.5 0.86 40.2 166.0

104.7 90.6 78.3 9.1 0.81 33.3 131.1

157.9 155.2 103.5 12.5 0.77 40.2 200.6

a

For comparison with future values, the 2005 withdrawals were adjusted by the model to represent normal weather conditions. gpcd, gallons per capita per day; gped, gallons per employee per day; in yr1, irrigation water application depth in inches per year; gal. kWh1, water use for thermoelectric cooling in gallons per kilowatt-hour. From Dziegielewski B and Chowdhury FJ (2008) Regional water demand scenarios for Northeastern Illinois: 2005–2050. Project Completion Report. Prepared for the Chicago Metropolitan Agency for Planning, Chicago, IL, USA, 15 June 2008.

Table 15

Effects of driver and unit rate changes on future water demand (CT scenario)

Sector

Public supply Self-supplied I&C Self-supplied domestic Irrigation and agriculture Power plants (makeup) Power plants (through flow) Total (w/o through flow)

2005 demand (mgd)

1189.2 162.4 31.8 44.6 52.3 4207.2 1480.3

2005–2050 increase in demand

Effect of change in unit rates

Effect of change in driver count

mgd

%

mgd

%

mgd

%

381.0 129.2 9.4 10.8 0.0 –377.0 530.4

32.0 79.5 29.6 24.2 0.0 –9.0 35.8

–83.5 37.5 2.6 2.2 0.0 –109.8 –41.1

–7.0 23.1 8.2 4.9 0.0 –2.6 –2.8

464.5 91.7 6.8 8.6 0.0 –267.2 571.5

39.0 56.4 21.4 19.3 0.0 –6.4 38.6

CT, current trends; mgd, million gallons per day. From Dziegielewski B and Chowdhury FJ (2008) Regional water demand scenarios for Northeastern Illinois: 2005–2050. Project Completion Report. Prepared for the Chicago Metropolitan Agency for Planning, Chicago, IL, USA, 15 June 2008.

key explanatory variables such as income, price, or conservation trend. The effects of weather conditions are shown by comparing the observed and weather-normalized usage rates for 2005. Only normal-weather predictions are shown for the future years. Table 15 compares the effects of growth (or change) in demand drivers, which are separated from the effects of

projected changes in unit usage rates. The 2005–50 increments in demand are decomposed into the effect of growth (or change) of demand driver and the effect of projected changes in unit usage rates. For example, for the public-supply sector, the 32.0% increase in demand is a net result of a 39% increase in population served and a 7% decrease in per capita rate of water use.

186 Table 16

Predicting Future Demands for Water Effects of key assumptions on per capita rates in public supply sector

Forecast assumption

2005-Na

Public supply (gpcd) Total 2005–2050 effect Effects of individual variables

142.1

Retail price of water Median household income Conservation trend Employment/population ratio Differential growth

2050-CT

2050-LRI

2050-MRI

134.9 –7.2

104.7 –37.4

157.9 þ 15.8

 8.2 þ 11.5 –11.2 þ 1.2 –0.5

–18.4 þ 6.5 –23.3 þ 1.0 –3.2

0.0 þ 18.1 0.0 þ 1.4  2.7

a

For comparison with future values, the 2005 withdrawals were adjusted by the model to represent normal weather conditions. CT, current trends; LRI, less resource-intensive; MRI, more resource-intensive; N, normal weather; gpcd, gallons per capita per day. From Dziegielewski B and Chowdhury FJ (2008) Regional water demand scenarios for Northeastern Illinois: 2005–2050. Project Completion Report. Prepared for the Chicago Metropolitan Agency for Planning, Chicago, IL, USA, 15 June 2008.

The change in unit usage rates is a result of changes in the future values of explanatory variables. Table 16 shows the effects of future changes in the values of four variables on per capita rates in public-supply sector. For the low-growth scenario, the per capita rate of usage is projected to decrease by 37.4 gpcd. This value represents a net effect of the increase in marginal price, increase in median household income, effect of conservation, and effect of the changing ratio of employment to resident population. One additional component of the change in gpcd is the effect of differential growth in population across the 37 geographical areas used in the forecast. For the baseline scenario, this effect is  0.5 gpcd; it would be zero if population served in all areas was projected to grow at the same rate. Making the effects of these different factors explicit allows the stakeholders who participate in the planning process to assess the reasonableness of the forecast and to decide which scenarios should be taken into account when formulating water-supply plans.

1.10.5.3 Effects of Future Climate Climate models indicate that by 2050 in Illinois there may be an average annual normal temperature departure of up to þ 6 F, and a possible departure from normal annual precipitation in a range of  5 to þ 5 in yr1 from the 1971–2000 long-term normal values (ISWS, 2007). Due to the nature of climate scenarios, no probabilities can be placed on the possible ranges of future air temperature and precipitation. The changes in normal annual temperature and precipitation would also result in average-weather changes during the growing season. The temperature increase of 6 F will also apply to the summer growing season. The distribution of precipitation changes is expected to range from þ 2.5 to  3.5 in during the growing season. The effects of these changes on future demands will vary by user sector, depending on each sector’s sensitivity of water withdrawals to air temperature and precipitation. The results show that future demands in all sectors are likely to be higher if future annual average air temperature increases and/or annual precipitation decreases. If, by 2050, temperature increases by 6 F, total withdrawals would increase

by 178.0 mgd (9.1%) above the baseline (CT) scenario values. The largest increase in total withdrawals above the baseline scenario would be 229.5 mgd (or 11.7%) by 2050, resulting from the combined effect of the temperature increase and a decrease in summer precipitation.

1.10.5.4 Forecast Summary The scenario forecasts of future water demands in the 11county planning area in Northeastern Illinois revealed the possibility of potentially large increases in total water withdrawals by 2050. The baseline (CT) and high-growth (MRI) scenarios when viewed in the context of regional supply limitations make a compelling case for the need to manage regional water demands. Total withdrawals (excluding oncethrough flows in power plants) in 2050 under baseline scenario would increase by 530.4 mgd and the increase could reach as high as 949.1 mgd under high-growth scenario. Meeting these additional demands would require large capital outlays for water infrastructure and would likely have significant impacts on some of the regional sources of water supply, especially groundwater aquifers and local rivers, which would create increased demands on water from Lake Michigan.

1.10.6 Conclusion The prospect of climate change, which could alter meteorological and hydrological regimes and possibly result in lower water availability and higher demands, is challenging water managers to search for effective ways to satisfy future demands without jeopardizing the long-term sustainability of current water resources systems. Credible long-term forecasts of water demand can help water planners to achieve an efficient allocation of future water supplies among competing uses. When predicting future demand for water, it is important to recognize that water is an economic good and, therefore, its future use will be responsive to changes in future economic conditions. The economic considerations should help water planners to achieve efficient and equitable use of water supplies while encouraging conservation and protection of water resources.

Predicting Future Demands for Water

This chapter described data and methods for developing a clear and credible forecast of future water demand which would enhance its acceptability by decision makers. The main characteristics of such a forecast include a high level of disaggregation of demand, use of econometric water-use models which conform to the economic theories of production and consumption, and provision of explicit and plausible forecasting assumptions. A credible forecast cannot be developed without first examining the historical data on water use. Obtaining valid data on water use for a sufficiently long period of time (usually 15–20 years) is usually the most time-consuming task of a forecasting project. The data must allow for disaggregation of total water use into different categories of demand and also have appropriate geographical and temporal resolution to permit the development of water-use relationships which would capture the effects of variables which affect water demand and operate at different scales of space and time. Additional large effort is required in obtaining data on explanatory variables. Once appropriate data are assembled, the next challenge is the development of water-use relationships, usually by applying statistical methods such as multiple regression. The criteria for deriving a empirical model which is useful for forecasting are somewhat different than those in a typical econometric studies where researcher wishes to know which variables are statistically significant and if the resultant model conforms to economic theory. A useful forecasting model requires not only an appropriate model specification but also accurate estimates of the regression coefficients (or elasticities) for the explanatory variables. The forecaster must develop fairly strong expectations about the size of the regression coefficients for key explanatory variables such as price, income, air temperature, and precipitation. Expectations about the sign (positive or negative) of the explanatory variables come directly from economic theory and the underlying physical relationships. The next step in developing a credible forecast involves the development of forecasting assumptions which are the basis and the main component of the forecast. The most critical are the assumptions about the future values of explanatory variables. Where possible, forecasts of water demand should be based on official or consensus forecast of population and economic growth in the study area. The future values of other explanatory variables (e.g., price of water, income, or climate) should be derived based on explicit and plausible assumptions. The uncertainty caused by forecasting assumptions can be addressed by using forecast scenario, sensitivity analysis, or developing probability forecasts. Finally, once the forecasted quantities of future water demand are prepared, the forecaster should provide a postforecast analysis of results. In essence, such an analysis should identify and quantify the effects and relative contributions of individual forecasting assumptions on the final results. For example, the change in water demand during the forecast period can be separated into the effects of population growth, changes in the price of water, income, climate, and other explanatory variables. The post-forecast analysis can help ensure that the forecast is understood and accepted by decision makers.

187

References Avery C (1999) Estimated water withdrawals and use in Illinois, 1992. Open-File Report 99–97. Urbana, IL: US Geological Survey. Backus CE and Brown ML (1975) Water requirements for solar energy. In: Gloyna EF, Woodson HH, and Howard Drew R (eds.) Water Management for the Electric Power Industry, Water Resources Symposium #8, pp. 270–279. Austin, TX: Center for Research in Water Resources, The University of Texas at Austin. Baumann D, Boland JJ, and Hanemann WM (1998) Urban Water Demand Management and Planning. New York: McGraw-Hill. Baumann DD, Boland JJ, and Sims JH (1984) Water conservation: The struggle over definition. Water Resources Research 20(4): 428--434. Boland J (1998) Forecasting urban water use: Theory and principles. In: Baumann DD, Boland JJ, and Hanemann M (eds.) Urban Water Demand Management and Planning, ch. 3, pp. 77–94. New York: McGraw-Hill. Boland JJ (1985) Forecasting water use: A tutorial. In: Torno HC (ed.) Computer Applications in Water Resources, pp. 907--916. New York, NY: The Society. Boland JJ, Dziegielewski B, Baumann DD, and Opitz EM (1984) Influence of price and rate structures on municipal and industrial water use. IWR Report 84-C-2. Fort Belvoir, VA: US Army Engineer Institute for Water Resources. Boland JJ, Dziegielewski B, Baumann DD, and Turner C (1982) Analytical bibliography for water supply and conservation techniques. IWR Report 82-C-07. Fort Belvoir, VA: US Army Engineer Institute for Water Resources. Blake NM (1956) Water for the Cities: A History of the Urban Water Supply Problem in the United States. Syracuse, NY: Syracuse University Press. Breusch TS and Pagan AR (1979) A simple test for Heteroskedasticity and random coefficient variation. Econometrica 47: 1287--1294. Croley TE II, Patel VC, and Cheng MS (1975) The Water and Total Optimization of Wet and Dry–Wet Cooling Towers for Electric Power Plants. Iowa City, IA: Iowa Institute of Hydraulic Research, University of Iowa. Deaton A and Muellbauer J (1999) Economics and Consumer Behavior. Cambridge: Cambridge University Press. De Rooy J (1974) Price responsiveness of the industrial demand for water. Water Resources Research 10(3): 403--406. Dupond D and Renzetti S (1998) Water use in the Canadian food processing industry. Canadian Journal of Agricultural Economics 46: 1--10. Dziegielewski B (1993) IWR-MAIN 6.0: A tool for demand management and planning. Journal of American Water Works Association (Aqualink Department) 85(8): 24. Dziegielewski B and Baumann DD (1992) Benefits of managing urban water demands. Environment 34(9): 6--11. 35–41. Dziegielewski B, Baumann DD, and Boland JJ (1983) Prototypal application of a drought management optimization procedure to an urban water supply system. IWR Report 83-C-4. Fort Belvoir, VA: US Army Engineer Institute for Water Resources. http://stinet.dtic.mil/oai/oai?&verb=getRecord&metadataPrefix=html &identifier=ADA138473 (accessed April 2010). Dziegielewski B and Bik T (2006) Water use benchmarks for thermoelectric power generation. Prepared for the United States Geological Survey. Southern Illinois University at Carbondale, 15 August 2006. Dziegielewski B and Boland JJ (1989) Forecasting urban water use: The IWR-MAIN model. Water Resources Bulletin 25(1): 101--109. Dziegielewski B and Chowdhury FJ (2008) Regional water demand scenarios for Northeastern Illinois: 2005–2050. Project Completion Report. Prepared for the Chicago Metropolitan Agency for Planing, Chicago, IL, USA, 15 June 2008. Dziegielewski B and Kiefer JC (2006) U.S. water demand, supply and allocation: Trends and outlook. IWR Report 2007-R-3. A white paper prepared for the U.S. Army Corps of Engineers Institute for Water Resources, Alexandria, VA, USA, 22 December 2006. Dziegielewski B, Opitz EM, Kiefer JC, and Baumann DD (1993) Evaluation of Urban Water Conservation Programs: A Procedures Manual, xxi þ 274pp., Book No. 0-89867-676-2. Prepared for California Urban Water Agencies, Sacramento, CA. Denver, CO: American Water Works Association. Dziegielewski B, Opitz EM et al. (1996) IWR-MAIN Water Demand Analysis Software. Users Manual and System Description. Carbondale, IL: Planning and Management Consultants. Dziegielewski B, Rodrigo D, and Opitz E (1990) Commercial and Industrial Water Use in Southern California. Carbondale, IL: Planning and Management Consultants. Dziegielewski B, Sharma SC, and Margono H (2005) Models for long-term forecasts of public supply water use. In: Proceedings of the XIIth IWRA World Congress on Water Resources. New Delhi, India, 22–25 November 2005. New Delhi, India: Central Board of Irrigation and Power.

188

Predicting Future Demands for Water

Dziegielewski, B, Strus CA, and Hinckley RC (1993) End-use approach to estimating water conservation savings. In: CONSERV93: The New Water Agenda. American Water Works Association, Denver, Colorado. Grigg NS (1986) Urban Water Infrastructure: Planning, Management, and Operations. New York: Wiley. (original from the University of Michigan, MI). Epsey M, Epsey J, and Shaw WD (1997) Price elasticity of residential demand for water. Water Resources Research 33(6): 1369--1374. Ferguson B and Whitney A (1993) Demand reduction in response to drought: The city of Santa Barbara experience. In: CONSERV93 Proceedings, ASCE, AWWA, AWRA, pp. 429–437. Foster HS Jr and Beattie BR (1979) Urban residential demand for water in the United States. Land Economics 55(1): 43--58. Gardiner V and Herrington P (1986) Water Demand Forecasting: Proceedings of a Workshop. Exeter: Short Run Press. Glejser H (1969) A new test for heteroscedasticity. Journal of the American Statistical Association 64: 316--323. Hanemann WM (1998) Determinants of urban water use. In: Baumann DD, Boland JJ, and Hanemann M (eds.) Urban Water Demand Management and Planning, ch. 2, pp. 31–75. New York: McGraw-Hill. Hittman Associates, Inc. (1969) Forecasting Municipal Water Requirements, Volume 1: The MAIN II System, PB 190275. Columbia, MD: Hittman Associates, Inc. Howe CW (1982) The impact of price on residential water demand: Some new insights. Water Resources Research 18(4): 713--716. Howe CW and Linaweaver FP Jr (1967) The impact of price on residential water demand and its relation to system design and price structure. Water Resources Research 3(1): 13--32. Hutson SS, Barber NL, Kenny JF, Linsey KS, Lumia DS, and Maupin MA (2004) Estimated use of water in the United States in 2000. US Geological Survey. USGS Circular 1268. http://water.usgs.gov/pubs/circ/2004/circ1268/ (accessed April 2010). ICWE (International Conference on Water and the Environment) (1992) The Dublin statement on water and sustainable development. http://www.un-documents.net/ h2o-dub.htm (accessed April 2010). IPCC (2008) Definition of terms used with the DDC pages. http://www.ipcc-data.org/ ddc_definitions.html (accessed July 2010). ISWS (Illinois State Water Survey) (2007) Tomorrow’s climate – future scenarios: CO2 concentrations. http://www.sws.uiuc.edu/wsp/climate/ ClimateTom_scenarios_co2.asp (accessed April 2010).

IWR (Institute for Water Resources) (2001) An evaluation of the risk of water shortages in the Lower Peninsula, Virginia. Prepared by Werick WJ, Boland JJ, Gilbert J, Dziegielewski B, Kiefer J, Massmann J, and Palmer RN. IWR Special Report. Alexandria, VA: US Army Corps of Engineers. Linaweaver FP, Jr., Beebe JC, and Skrivan FA (1966) Data Report of the Residential Water Use Research Project, Johns Hopkins University, Department of Environmental Engineering Science, Baltimore, MD. Metcalf L (1926) Effects of water rates and growth in population upon per capita consumption. Journal of the American Water Works Association 15(1): 1--20. Nieswiadomy ML (1992) Estimating urban residential water demand: Effects of price structure, conservation, and education. Water Resources Research 28(3): 609--615. Platt RH (1993) Water demand management. Commentary. Environment 35(3): 2--3. Ramsey JB (1969) Tests for specification errors in classical linear least-squares regression analysis. Journal of the Royal Statistical Society Series B 31: 350--371. Renzetti S (1988) An econometric study of industrial water demands in British Columbia, Canada. Water Resources Research 24(10): 1569--1575. Renzetti S (1992) Estimating the structure of industrial water demands: The case of Canadian manufacturing. Land Economics 69(2): 181--188. Renzetti S (1993) Examining the difference in self- and publicly supplied firms’ water demands. Land Economics 69(2): 181--188. Renzetti S (2002) The Economics of Water Demands. Norwell, MA: Kluwer. Solley WB, Pierce RR, and Perlman HA (1998) Estimated Use of Water in the United States in 1995, U.S. Geological Survey Circular 1200. Denver, CO: US Geological Survey. UNCED (United Nations Conference on Environment and Development) (1992) Agenda 21, Protection of the Quality and Supply of Freshwater Resources: Application of Integrated Approaches to the Development, Management and Use of Water Resources, ch. 18. Brazil: Rio de Janeiro. USDA (United States Department of Agriculture) (2009) 2007 Census of Agriculture, Summary and State Data Volume 1, Geographic Area Series Part 51. Washington, DC: National Agricultural Statistics Service. White H (1980) A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48: 817--838. WHPA and Dziegielewski B (2008) Water Demand Scenarios for the East-Central Illinois Planning Region. Bloomington, IN: Wittman Hydro Planning Associates. Wolff JB, Linaweaver FP, Jr., and Geyer JC (1966) Water Use in Selected Commercial and Institutional Establishments in the Baltimore Metropolitan Area. Johns Hopkins University.

1.11 Risk Assessment, Risk Management, and Communication: Methods for Climate Variability and Change C Brown, University of Massachusetts, Amherst, MA, USA KM Baroang, International Research Institute for Climate and Society, Palisades, NY, USA & 2011 Elsevier B.V. All rights reserved.

1.11.1 1.11.2 1.11.2.1 1.11.2.2 1.11.2.3 1.11.3 1.11.4 1.11.4.1 1.11.4.2 1.11.4.2.1 1.11.4.2.2 1.11.4.2.3 1.11.5 References

Introduction Background on Risk Assessment and Management Hazard Characterization Risk Assessment Risk Management Risk Management versus Consequence Management: The Upside of Risk Climate Risk Risk and Nonstationarity: Uncertain Information and Unreliable Probability Estimates Climate Consequence Management with Decision Scaling: An Approach Designed for Uncertain Information Step 1: Vulnerability and uncertainty identification Step 2: Consequence assessment Step 3: Consequence and uncertainty management Conclusion

1.11.1 Introduction The subject of risk is a familiar one for the water resources engineer and the water planner. When a harmful event is possible but not certain, there is risk. In the world of water resources harmful events are largely, but not solely, associated with precipitation: too much, too little, and too variable. Other aspects of the hydrologic cycle modulate this basic signal as do various human activities, but we largely depend on precipitation and it is uncertain. Risk is inherent to the profession. Yet there does not seem to be a predominant method for addressing risk in the profession. This fact gains prominence due to climate change, nonstationarity of the hydrologic record, and the potential for changing and growing risks in the future. In the field of water resources, research has largely proceeded from a scenario-based rather than a riskbased framework of analysis. The primary direction of analysis has been simulation of a few possible future scenarios and assessment of impacts corresponding to these scenarios. Unfortunately, the uncertainty of these scenarios in comparison to other possible futures makes this approach unhelpful for risk assessment. It may be that the approach to risk is so institutionalized that all remains is discourse on the probabilities of extreme events and how to produce better estimates of them. The cost of this institutionalization is an accompanying ossification of the way in which risk is addressed. Climate change implies that our tried and true methods to reduce uncertainties to produce estimates of risk may no longer serve us well. For this reason, it is an opportune moment to review the treatment of risk in water resources, from assessment to management and finally to communication. Interestingly, the rise in importance of risk management as a framework to plan for a changing climate is accompanied by a questioning as to

189 190 190 191 191 193 193 194 195 195 196 197 198 199

whether the prevailing approaches to risk management are still valid. As we shall see, standard approaches to risk assessment and risk management as employed in the water sector utilize estimates of probability density functions (PDFs) associated with hydrologic variables. There is reasonable concern as to whether we have the ability to assign such probabilities with any confidence given to the small sample size associated with extreme events, which are the events of interest. There is further concern as to whether use of the historic record to estimate hydrologic design variables is appropriate for designs prepared for the future climate. In order to explore these issues, we require an understanding of what risk assessment and management mean and then explore why climate change may necessitate a refinement of this approach. In this chapter, we will review basic risk concepts, some history of its development in water resources engineering and planning, visit some emerging approaches to risk associated with deep uncertainties such as those associated with climate change, and finally propose a process for addressing climate risks to water resources systems. In doing so, we find that a process for climate risk is a special application for the general risk assessment and management framework as presented here. Because the terminology associated with risk is used to mean different things in different communities, it is useful to begin with some basic definitions. The definitions listed here are generally consistent with usage in water resources and hydrology and to some extent with the natural hazards community (Plate, 2002; UNDRO Office of the United Nations Disaster Relief Coordinator, 1991): Hazard. The probability of an event that causes failure or a negative effect on a community or system. Risk. The product of the probability of a hazardous event occurring and the impact or consequence of that event; risk

189

190

Risk Assessment, Risk Management, and Communication: Methods for Climate Variability and Change

can increase if either the probability increases or the consequences of a hazard become more severe. Vulnerability. The characteristics of a community or system that cause them to be susceptible to adverse outcomes when exposed to a particular event. Resilience. The capacity of a community or system to recover from an adverse outcome due to a hazard and return to an acceptable level of functioning. Careful consideration of the definition of risk raises a topic of interest that is often not addressed in water resources. This is the idea that there are opportunities presented by uncertain events that have positive consequences. Our typical focus is on the tail of the event probability distribution that is associated with negative consequences, such as droughts or floods. However, there can also be beneficial opportunities associated with the tails of probability distributions. Consider the case of a farmer engaged in rain-fed agriculture planning the crop pattern for the season ahead. If that farmer plans a pattern optimal for the average seasonal rainfall total but not ideal for conditions that are below normal, then there is a risk associated with below-normal rainfall. If the rainfall is below normal then the crop yield will be reduced. The farmer could reduce his risk exposure to drought by choosing a crop pattern that is well suited for below-normal rainfall. However, the farmer would now forgo the opportunities associated with above-average rainfall, say a bumper crop of a water-intensive, high yielding cash crop. The relationship between risk and opportunity will be discussed later in this chapter.

yielding events, when they occur. Note that the consequences of an event are contingent on the risk management steps taken or not taken, and so C(x) can be specified as C(x|D), where D represents a decision. For example, the consequences associated with a flood of stage x is a function of the decision, D, to build a levee of some level or none at all. In the typical approach, the estimation of risk then requires the estimation of two functions, the probability of occurrence of the events of interest, f(x), and the estimation of the consequences of that event for a given decision, C(x|D). The identification of hazards and the estimation of f(x) associated with those hazards is the first step of addressing risks, which is called hazard characterization. This process is discussed in more detail below. While f(x) is typically thought of as acts of nature, in the case of hydrology human action can influence the characteristics of stream flow. For example, significant development of natural land surfaces to conditions of partially or completely impervious surface would likely change the frequency of high flow events. Decisions to preserve land surfaces in natural states or to develop them with low-impactdevelopment designs could similarly change the frequency of high flow events. Thus, one could consider a PDF of stream flow that is conditional on decisions, such as land use, that are made that affect stream flow characteristics, f(x|d). The next step is risk assessment which consists of the estimation of the expected value of the consequences, C(x|D). This is accomplished by integrating over the PDF of x:

 ¼ RðxÞ

Z

N

CðxjDÞf ðxÞdx

ð1Þ

0

1.11.2 Background on Risk Assessment and Management There are a variety of processes with different steps that have been proposed for risk assessment and management. Some place risk assessment as a single step within a larger risk management framework. Here, we describe the process of three general steps which cover the various processes that are consistent with the themes in most approaches. We also highlight particular topics not typically covered in depth, namely opportunities, residual risk, and surprise. The three general steps can be described as a hazard characterization step, a risk assessment step, and a risk management step. The mathematical description presented here draws from previous formalizations of risk, for example, Plate (2004) and Lund (2002). Risk consists of impacts or consequences that result from an event that may occur with some probability. It can be summarized in the following equation:

RðxÞ ¼

Z

where R is the expected risk for a given decision, D. The process of risk assessment produces a risk associated with a hazard or range of hazards. In the final step of risk management, one can then attempt to make a decision or decisions that reduce the unwanted risks. One can then evaluate the expected risks for the set of possible decisions, Di, i ¼ 1,y, N, for N possible decisions and select the risk-minimizing decision:

min Z ¼

Z

D

N

CðxjDÞf ðxÞdx

ð2Þ

0

Considering decisions related to changes to stream flow characteristics and thus the PDF of the stream flow, the equation can be more generally specified as

min Z ¼ D;d

Z

N

CðxjDÞf ðxjdÞdx

ð3Þ

0

N

CðxÞf ðxÞdx

0

where f(x) is the PDF of the event and C(x) is the consequences associated with that event on the system of interest. For the hydrologic risks, f(x) is the PDF of a hydrologic event such as a flood of the stage (height), x. The consequence function yields the costs or damages associated with a flood of stage x. We use the term consequence function in place of the more common damage function to highlight the potential for benefits that may arise from opportunities, uncertain benefit

This is one possible mathematical specification of the quantification of risks. One could consider this as an optimization, either formally with an optimization model that can be used to evaluate risk management alternatives or informally through a decision process to achieve a similar objective. The actual processes involved in hazard characterization, risk assessment, and risk management are more difficult. While there is much focus within the research community on the description of f(x) in terms of the selection of the distribution type and methods employed for the fitting of

Risk Assessment, Risk Management, and Communication: Methods for Climate Variability and Change

parameters, the importance of correctly characterizing CðxjDÞ and of developing a wide range of options for managing risks, Di, are perhaps of greater importance given to our growing awareness of the limitation in correctly specifying f(x).

1.11.2.1 Hazard Characterization The first step is one of identifying and characterizing hazards. This is an exploratory phase in which the performance of the system of interest is reviewed and its past and possible vulnerabilities are identified. Often past experiences with extreme hydrologic events offer the strongest evidence of the hazards that should be addressed. Each water-resource system is unique and has exposure to events of particular characteristics. For example, in a general way Vogel et al. (1997) classified reservoir systems into over-year or under-year storage categories based on their vulnerability to drought durations of months but less than years. Such systems would be interested in very different definitions of drought. It is also useful to consider the past experiences of other water-resource systems, especially those with similar size, similar climate, or similar operating constraints such as those posed by multiuse water systems and environmental requirements. Once the events of concern have been identified, the next part of this step is the characterization of the hazard through the estimation of probabilities associated with it. For waterresource systems, the events of interest are typically related to stream flow and so a hydrologic record that relates to the water system is genrally used to estimate these probabilities. Thus, a hazard is an event of interest (typically due to its potential to cause danger) with an associated probability of occurrence. Here, we focus on hydrologic hazards but the methodology is consistent with other hazards such as earthquakes or terrorist attacks. Note that the methodology also applies to events with positive consequences. These events opportunities are termed an event of interest (with potential for benefit) and its associated probability of occurrence. In the case of floods, the process is one of estimating the exceedance probability of certain high-flow events that were deemed to be dangerous. Similarly, the exceedance probability of droughts of specific duration and intensity (e.g., 20% reduction of inflows over a 6-month period) can be estimated. Traditionally, these probabilities were estimated using the historical record. For example, for floods, the annual maximum time series for the location of interest was used to estimate the parameters of an extreme value probability distribution. This distribution was then used to calculate the estimated exceedance probability for a flood of a given magnitude. The techniques for calculating parameters, choosing distributions, and using the distributions have been the focus of much attention and are fairly mature. A useful summary is found in Stedinger et al. (1993). Methods are also available for using data from other locations in a region when the location of interest has a limited historical record. For droughts, the process is similar, with the added dimension of the duration as well as the intensity of the event. Engineers have long been aware of the uncertainties that accompany the estimation of rare events such as flood events with low exceedance probabilities. These are addressed through risk management and communication. Nonetheless,

191

a great deal of effort is expended in reducing the uncertainty of hazard estimates. This has yielded better understanding of the physical causes of hydrologic extremes and in some cases identified temporal structure in their occurrence. Our growing knowledge of the physical causes and temporal structure of the frequency domain of hydrologic extremes has important implications for the estimation of hazard probabilities. This leads to significant implications for our management of risks as well and perhaps the need for significant reconsideration of how climate related risk in water-resource systems is addressed. The most prominent example is the case of the El Nino/ Southern Oscillation (ENSO). ENSO is a coupling of seasurface temperature anomalies, winds, and atmospheric pressure in the equatorial Pacific that influences temperature and precipitation patterns around the world through teleconnections, with the largest effect in the tropics. The ENSO phenomenon has two phases, one warm (El Nino) and one cold (La Nina), each with partly but not entirely opposite directions of influence (warmer vs. colder; dryer vs. wetter). ENSO is described as quasi-periodic and these events occur on an irregular cycle of about 3–5 years. Where the teleconnections are strong, an ENSO event implies that the hazard associated with a particular hydrologic event may be elevated in some locations and reduced in others, relative to the longterm mean rate of occurrence. For example, the drought hazard in the Philippines is elevated during an El Nino, while it is reduced during a La Nina. The presence of such temporal structure has implications for risk assessment and management, since foreknowledge of year-to-year changes in risk was not previously considered.

1.11.2.2 Risk Assessment In this step the concept of risk is quantified in terms of the consequences of an event and its probability of occurrence, in accordance with Equation (1). Here the description of consequences and the specification of the consequence function are central. A challenge in calculating the consequences is that they are not necessarily measured in commensurate units. Many consequences are fairly straightforward to quantify in economic terms, and methods for doing so are mature. In the field of water resources, however, in many cases damages may be more difficult to describe in monetary terms. For example, flood events often involve the potential loss of life. Floods and droughts may also have substantial impact on the environment through the destruction of habitat or insufficient stream flows for aquatic ecosystems. This issue is more prominent in the risk management step, where the alternatives for managing hydrologic risks are evaluated in terms of their costs and the benefits provided and it is addressed further below. As before, the involvement of stakeholders is a critical aspect of risk assessment. Stakeholders are knowledgeable about the consequences of historical events and may be the best source of conjecture on the impact of events not yet seen. Their opinions will also be valuable in the ranking of risks that cannot be easily quantified, such as specific environmental

192

Risk Assessment, Risk Management, and Communication: Methods for Climate Variability and Change

damages or the relative value of structural damages in comparison with lost life.

1.11.2.3 Risk Management The final step in this process entails making decisions and systematically exploring the consequences of those decisions as well as the uncertainties that will accompany any risk assessment process. There are a variety of approaches that have been proposed for risk management. Here, we present risk management as a systematic process that can be described in three parts: (1) systematic evaluation of the risks previously quantified and the cost of measures that address those risks; (2) development of a plan to address those risks and accompanying uncertainty; and (3) consideration and planning for residual risks and performance of the system when failure occurs. The systematic evaluation of risks begins with a review of the risks quantified during hazard characterization and risk assessment. These risks can then be evaluated in terms of the expected losses that they represent. As described above, the expected losses may be expressed in economic terms, such as monetary units of damage to structures, or in units of human injuries and deaths, or in terms of damage to the environment. Since these expected losses are compared with the economic costs of the measures considered for mitigating the risks, expression in terms of economic units simplifies the development of a risk management plan. Methods are available for calculating the economic costs of injuries, deaths, and environmental damage, although none are without controversy. Environmental impacts are similarly difficult to quantify in economic terms, although methods have been proposed. At the least these potential impacts must be tracked and incorporated informally into the decision process. Next, the candidate measures for reducing risk are evaluated in terms of their cost and their effectiveness. Often, engineers limit themselves to consideration of structural approaches to managing hydrologic risks. For example, dikes and levees are typically prime candidates for reducing flood risk, whereas increasing the storage capacity of a water-supply system is the main method for addressing drought concerns. However, there exist a suite of nonstructural measures that should be considered as alternatives or complements for structural measures. For droughts, examples include economic mechanisms such as water markets, option contracts, and even insurance (Characklis et al., 2006; Brown and Carriquiry, 2007). For flood risk reduction, examples include designated flooding areas, improved flood warning systems, and land-use planning. An appealing aspect of some nonstructural measures is that they may be used as options, meaning that they are not executed or fully paid for until they are needed. For example, a levee can be heightened through the addition of sand bags which are not filled and put into place until they are needed (Lund, 2002). The option approach becomes more advantageous as the uncertainty related to the risk analysis increases, such as due to possible changing climate conditions. Also, the advantage of the option approach increases with improved forecasts that can result in better decisions related to

execution of the options. Unfortunately, our experience with these measures is limited as is the number from which to choose in comparison to traditional structural approaches. It is an area of needed research. When risks and the costs of risk reduction measures are well characterized, optimization and decision analysis can be an effective means for developing and evaluating the risk management plan. The objective function for a plan that considers permanent measures and option measures can be specified as follows:

min Z ¼

X þ

Z

permanent measures  cost ðoption measures  costs þ damagesÞf ðxÞdx ð4Þ

An example of the use of optimization using a form of this equation to develop a flood risk reduction plan is described in Lund (2002). Optimization techniques such as used in that analysis allow a quantitative approach to the development of a risk management plan. However, the result of that approach, say the optimal plan, is only as useful as the quality of the data that was used to develop the plan. In the case of estimating probabilities of rare events or equivalently the magnitude of events with very low exceedance probabilities, the quality of these estimations may not be very reliable. Furthermore, according to the engineer, the process of going from the optimal plan to a plan that is actually implemented by a public agency with diverse and conflicted stakeholders is quite difficult and not well understood. The latter subject may be the most important to solve in order to achieve efficient reduction of hydrologic risks to society. The subject is explored in Fiering and Matalas (1993). The uncertainty related to the estimation of magnitudes of rare events can be addressed in several ways during the development of the risk management plan. The first approach is simply the application of decision analysis and should be a standard part of the risk management plan development. Using a decision analysis approach, it might be determined that the decision is not sensitive to specification of these very rare events. For example, it may become clear that a particular flood risk reduction candidate measure dominates other measures regardless of the magnitude of floods with exceedance probabilities of 103 or 104. Or the decision to provide protection for a structure may be decided based upon what a society is willing to pay rather than the estimated return period of an event that is protected against (Bondi (1985) cited in Plate (2004)). In other cases, the choice of a given plan and its subsequent performance is sensitive to these estimations. In such cases where uncertainties are significant and influential, a risk management plan should strive to achieve aspects of robustness and flexibility (de Neufville, 2004). While it can be defined in a variety of ways, here we define as robust the quality of a system that can continue to operate effectively under unexpected conditions. Redundancy is a key component of achieving robustness. For example, two pumps should allow water to be moved even if one unexpectedly fails. A groundwater well field allows a water-supply system to

Risk Assessment, Risk Management, and Communication: Methods for Climate Variability and Change

continue to operate if the reservoir runs dry. The quality of flexibility is a quality of a water system that provides the engineer or manager with options to react to changing conditions as they evolve. For example, a regulation plan for a reservoir could allow different release rules based on the prevailing climate conditions in place of the standard static regulation plan that is unchanging. Brown et al. described the development of a dynamic regulation plan for the management of outflows from Lake Superior that demonstrates how instilling flexibility in normally static plans is an adaptation strategy for climate change. Flexibility also allows the water manager to make use of forecast information. With only static operating procedures, foresight regarding future water supplies or demand is of little value since there are no changes that can be made in response to the new information. Once a risk reduction design has been formulated, the final step consists of addressing the full spectrum of possible risks that a system faces. There are two aspects to this. The first aspect is the consideration of residual risk. Residual risk is the risk that remains once a particular decision has been made. The concept applies to operational or planning timescales. For example, a planned flood risk reduction design may include the construction of a levee that is designed to withstand a flood magnitude with an estimated 500 year return interval. The residual risk consists of the risk that is not addressed by the design, in this case floods that are greater than the 500year flood. Such floods are certainly possible and must be planned for even if they are not part of the structural design considerations. On an operational timescale, it is the risk that remains once a specific operational design has been made under uncertainty. The residual risk for decision i may be specified as

RRi ¼

Z

N

CðxjDi Þf ðxÞdx

ð5Þ

0

The residual risk may be addressed in a variety of ways. It may be considered insignificant and no further action is taken. A decision analysis approach could be used. Using this method, the costs of addressing the residual risk could be estimated and compared with the risk. In most cases the residual risk estimate will be highly uncertain since by nature it involves the tails of distributions. The decision will be largely a subjective one and may or may not be influenced by the precautionary principle. Since in most cases it will not be economically acceptable to address all residual risks, there will be risks that are known but are not addressed. There will also be risks to the system that have been underestimated. The uncertainty associated with climate change, land-use change, and technological change compromises our ability to estimate future probabilities with confidence. As a result, we should not be surprised by underestimation and overestimation of risk. We should be surprised by risks that are unknown and cannot be reasonably anticipated. These are surprises or even black swans as popularized by Taleb (2007). From the water engineer standpoint, these three considerations make clear that any system design that incorporated risk reduction planning may still fail. This brings us to the second aspect of addressing the full spectrum of risk. Although risk reduction plans may

193

not cover all risks, we can plan for how our system operates during an inevitable failure. This may seem to fall outside the traditional realm of engineering. But since the engineer is ultimately trying to reduce impacts of hydrologic events on society (or rather reduce negative impacts and increase positive impacts), this represents more options for doing so successfully. For a water-supply system, a plan can be developed for rationing water, purchasing tanker water deliveries, and other measures if a drought were to shut down the reservoir supply. Similarly, emergency response, evacuation planning, designated flood areas, and secondary protections beyond primary levees should be incorporated into a flood risk management plan. The key is for managing surprises such that although they lead to failure of the system, they do not lead to catastrophe.

1.11.3 Risk Management versus Consequence Management: The Upside of Risk Risk management focuses on uncertainties related to events that have the potential to cause damages or harm. These are typically events that differ from the average conditions. Therefore, droughts are brought on abnormally by low water supply availability and floods by water that exceeds normal conditions. In some cases or for some stakeholders, these relatively rare events represent opportunities instead of risks. For example, excess water represents the opportunity to generate additional energy for a hydroelectricity generator or to provide high flow for ecosystems that require it. Droughts may provide opportunities for offline inspection and maintenance work. The concept of planning for the opportunities as well as for the risks associated with uncertainty has been described as uncertainty management (de Neufville, 2004). Here, the term consequence management is preferred as this is a neutral term of risk (positive or negative) and is distinct from uncertainty which derives from a variety of sources and does not necessarily have consequences. With consequence management, we include the positive consequences of possible hydrologic conditions in addition to the negative consequences (risks). Management plans are developed and evaluated for all consequences in the same way as done for risks alone. Interestingly, the management of risk (negative consequences) and the ability to exploit opportunities (positive consequences) are connected in many cases. The reason is that if risks are not being effectively managed, it is more difficult to capitalize on the opportunities associated with favorable side of the distribution. As a result, there can be substantial lost opportunities that result from ineffective risk management. Examples from the use of seasonal climate forecasts demonstrate this issue. With the forecast for above-average reservoir inflows, a multi-use reservoir can consider making additional releases for hydroelectricity production while still expecting to meet its storage requirement for water-supply purposes. However, the typical water manager would prefer not to make the additional releases because of the concern that belowaverage inflows may occur even if the probability is very low. As a result, the excess water and the opportunity it represents are lost as spills from a full reservoir. This decision is perfectly rational if the consequences associated with a shortage of

194

Risk Assessment, Risk Management, and Communication: Methods for Climate Variability and Change

water in the reservoir are disastrous even if the odds are low. Without an effective plan for managing low water conditions the risk associated with additional reservoir releases outweigh the benefit of extra hydroelectricity production. If there is an effective way to manage the downside risk, then the upside consequences could be better captured. So a system that has a backup water supply or the means to buy-out major water users (such as agriculture) is in a good position to take advantage of such an opportunity. These issues are explored in a case study based on the multipurpose reservoir system that supplies water to the city of Manila, the Philippines (Brown et al., 2008; Brown and Carriquiry, 2007).

1.11.4 Climate Risk The subject of climate change and nonstationarity has stirred interest in risk-based approaches within water resources planning and management. Specifically, there is an emerging question as to whether our traditional approaches to assessing and managing risk are appropriate with the recognition of nonstationarity. While there is naturally strong interest in improving our ability to estimate the return periods or exceedance probabilities of hazardous events, there is also a cause to reconsider more generally how we assess and manage risk. Initially, the sense of the water community was that there was too much uncertainty to take much action in response to climate change. The resulting focus emphasized the use of output from general circulation models (GCSMs) to reduce that uncertainty. This style of analysis typically utilized a scenario approach whereby a few simulations of possible future conditions were generated from a particular GCM run representing a possible future or scenario. The approach generally consisted of the following steps (see, e.g., Brekke et al., 2009): (1) choose a GCM or a few GCMs from which to derive the climate change model projection; (2) downscale the coarse-scale GCM output to the scale of a higher-resolution hydrologic model; (3) use the hydrologic model to produce estimates of stream flow for the climate change scenario; and (4) use the stream flow estimates to estimate a water resources impact of interest, such as reliability of a reservoir system, change in production of hydroelectricity, and implications for low stream flows. While seemingly straightforward, there are a number of drawbacks to this approach when used for planning and decision making. There are a variety of difficulties associated with the use of the GCM information itself. Typically mean conditions from the GCM are utilized. As a result, climate change is depicted as a shift in mean conditions. However, the water resources sector is usually more concerned in variability and extremes. For the water-supply sector, changes in the serial correlation of rainfall or stream flow are of concern, since several months or years of continuous below-normal conditions are the greatest risks. For risk reduction, extreme hydrologic events are more important than the mean conditions. The scenario approach does not attempt to address changes in variability or in the distribution of extremes. A more fundamental critique of this approach is that it is not risk based. The approach produces possible consequences that result from a particular scenario of climate change. There

are no probabilities associated with the particular scenarios or the consequences. As a result, there was no way to quantify the risk that climate change posed. The focus of research efforts turned to attempts to improve the ability to simulate future climate conditions. Emphasis was placed on increasing the resolution of the climate models and incorporating more and better understanding of the processes that comprise the Earth’s climate system. While this emphasis continues, the difficulty in reducing the uncertainty significantly due to the complexity of the Earth’s climate system and predicting the evolution of civilization over the next 100 years has led to dissatisfaction with this emphasis for use in water resources planning. On the topic of climate change, until recently there has been more focus on risk-based approaches within the adaptation segment of the climate change community generally than within water. This is reflected in the Intergovernmental Panel on Climate Change (IPCC) reports and broadly in the literature. While there are a wide variety of approaches, one might place them in one of two general categories. The first is the top-down approach. In this approach, the analysis begins with some attempt to predict future conditions and those future conditions are used to estimate impacts on society, ecosystems, hydrologic systems, etc. This approach does not typically broach the subject of probabilities. In some sense they may not be truly risk based for that reason. The notion of top down refers to the direction of analysis beginning from climate change impacts through various physical processes and finally considering the potentially impacted systems. The second category is characterized by the opposite direction of analysis and, for that reason, is referred to as a bottom-up or vulnerability-based approach. The analysis begins with the vulnerability, that is, consideration of how a system, such as a community, is susceptible to harmful effects from changes in climate. Thresholds can be set where systems are particularly vulnerable or where significant impacts begin to set in that are meant to be avoided. Since we define risk as a product of consequences and probability of those consequences, the estimation of probabilities of consequences is ostensibly a primary concern of risk management. We are severely limited in our ability to estimate the probability of future events. For many extreme events, this is true whether one considers climate change or not. We have very few observations of very rare events that the magnitude of particular percentile events is a function of the data which consists primarily of much more common events (of lower magnitude) and the choice of the extreme value distribution used to model them. A review of the methodologies explored in the estimation of probabilities of extreme events such as floods is beyond the scope of this chapter. The reader is referred to Stedinger et al. (1993). Here, it is noted that the prospects of getting the answer correct seem to be slim. From a risk perspective, a risk that must be managed is the very real risk that we are unable to accurately estimate the magnitude of a design flood or drought.

1.11.4.1 Risk and Nonstationarity: Uncertain Information and Unreliable Probability Estimates The prospect of climate change has raised to new prominence the specter that our estimates of probabilities associated with

Risk Assessment, Risk Management, and Communication: Methods for Climate Variability and Change

hydrologic events based on historical records may not be reliable in the future. Simultaneously, the projections of future climate conditions from GCMs are not reliable enough to be used to replace the historical record (Brown et al., 2009). As a result, water engineers face a future which has more uncertainty than they have previously been aware. Risk management is an applicable framework for addressing these issues. The methodology described here still relies on estimates of probabilities that are most likely to come from the historical record. Due to uncertainties related to climate change, among others, the final step of risk management, especially consideration of residual risk, surprise, and operation in failure mode, is essential. In addition, methods for decision making under so-called deep uncertainty may be explored. Robust decision making is representative of these methodologies (Lempert and Collins, 2003). The approach begins from a traditional statistical decision analysis approach but then evaluates decisions free of probability estimation. Instead, a candidate decision is evaluated for the conditions of the state variables under which that decision is compromised or performs poorly. Cluster analysis is used to identify the major areas of concern. Trade-off analysis is then used to evaluate alternatives to the original candidate that perform in these areas of concern. The approach appears computationally intensive in practice but the authors report good results in a few case studies (Lempert et al., 2003). Robust decision making is useful when little can be said of future conditions. However, there is some insight available regarding climate change. It may be possible to utilize the decision analysis if the process is tailored for the specific issues related to the use of uncertain information from GCM and historical data. Such an approach is presented here. The approach is called climate consequence management to highlight the possible opportunities that climate variability and change may present.

1.11.4.2 Climate Consequence Management with Decision Scaling: An Approach Designed for Uncertain Information Climate consequence management is a version of risk management that has been designed for the specific case of uncertainties associated with climate variability and change. The fundamental breakthrough is that a decision-analytic approach is employed in a manner that allows the use of climate information in a bottom-up or vulnerability-based way. It emphasizes identification of the specific climate information required that a decision hinges on. This allows the tailoring of the climate information to fulfill decision needs. The premise is that climate information can be useful in certain conditions by setting the bar fairly low in terms of what is needed. In some cases, for example, where climate information is deemed fairly reliable and projections are consistent in direction, this allows for probabilistic estimates of risk and risk-weighted decision making. There will be other cases where projections are contradictory and the process enables the identification of climate sensitivities and provides a framework for addressing them. In other cases, it will be found that some decisions are not sensitive to the projections of

195

climate change due to limited options or limited effects. The term ‘consequence’ is used to highlight the concept that variations and changes in climate bring opportunities as well as risks. The process consists of three steps that are analogous to the three-step process described above. In the first step, vulnerability and uncertainty identification, the hazards, and opportunities associated with climate are identified. Next, consequence assessment is the process of estimating probable consequences from the hazards and opportunities identified in step 1 and tailored climate information produced through the process of decision scaling to assign probabilities to the hazards and opportunities. In the final step, consequence and uncertainty management, a strategy for addressing probable consequences and key uncertainties, is being developed using a decision analytic framework. Decision scaling is again employed to tailor climate information to aid in the analysis of different decision options. Finally, residual risk and surprise are addressed through the incorporation of robust and resilient design, which may in some cases include adaptive management.

1.11.4.2.1 Step 1: Vulnerability and uncertainty identification The first step of climate consequence management is to assess the impacts of changes in climate across all timescales on water resources. This necessitates knowledge of both historical climate information and the resulting local consequences. The historical record is a useful starting point for identifying how climate has impacted the system in the past and the particular climate episodes that are challenging. A general overview of climate change information for the region being studied is accessed to prompt consideration of potential climate impacts that have not been observed in the historical record. At present, the IPCC reports on regional impacts are logical starting point for a summary of literature. The brainstorming process of identifying future potential impacts is not limited to these projections since they are not certain to describe the full range of future climate possibilities. It can also be very useful to identify thresholds in the system performance that when exceeded signify the need for adaptive actions. These thresholds are used in the consequence assessment and consequence management steps that follow to determine whether action is necessary in response to projected risks. Developing the appropriate knowledge requires a dialog with the stakeholders affected by or engaged in the water sector. Engaging stakeholders can both ensure that relevant impacts are considered and keep stakeholders aware of the process. A dialog with climate scientists and meteorological agencies can help supplement and interpret relevant climate information. By gaining a more robust understanding of these hazards and impacts, one can begin to determine the hydroclimatic risk and opportunity for a given system. While the focus naturally turns to the negative climate events in the historical record, it is important to consider also the effect that uncertainty in general has on the system. For example, preparations for a rare negative event may cause considerable lost opportunities in the many years the event does not occur. This can represent a significant opportunity cost due to

196

Risk Assessment, Risk Management, and Communication: Methods for Climate Variability and Change

uncertainty. Better management of the vulnerability to the rare event may allow the exploitation of more opportunities. While this process focuses on climate risks, it is important to recognize that climate is one of many factors affecting the system. When projecting future risk scenarios for a given system, possible changes in population growth, water demand, land use, and even values should also be considered and integrated into any comprehensive risk assessment. The following questions provide a general guideline for what to consider when performing vulnerability and uncertainty identification. What key climate-related challenges does the system currently face? These challenges might include moderate or severe droughts, flood events, variable flows, or others that are particularly disruptive to the system. This assessment is based on historical climate and current system characteristics, such as land use, population, and economic factors. It is important to identify the hazards historically associated with climate variability for the system while understanding also that the same type of climate event might have a more or less severe impact based on evolving nonclimate characteristics of the system. What damages occur as functions of these events? Once one has identified the climate-related hazards, one should assess the local impacts on the system. This includes an analysis of the spatial distribution of impacts and a determination of whether there are distributional effects from these events. Distributional effects cause some populations, such as those at lower economic status, to be more vulnerable to hazards (Rees, 2002). One should also determine impacts on both the human and environmental systems and seek to assess their vulnerabilities. The method of valuing consequences may differ. For example, economic valuation of consequences (e.g., foregone profits, direct costs associated with switching to another water source) is appropriate in some cases. However, in the case of severe consequences (e.g., famine), economic valuation alone may not be sufficient, as the social consequences may far outweigh direct economic costs. It may be important to determine local thresholds that determine the extent of climate-related consequences. While some water users can easily adapt to small reductions in water supply with little or no adverse effects, others may face significant damages from even the smallest supply variations. This may also be true for floods. The vulnerability across different users might lead to an aggregate threshold level and expected reliability for the system. Are there potential opportunities due to climate variability and change? Although the emphasis is generally on the possible negative impacts from climate variability and long-term change, changes in the system might also bring benefits. For example, a shift in phase in multi-decadal variability within a system could lead to improved average climate conditions for some sectors. Consider the apparent upswing in West African rainfall recently. It is important to remember interactions between climate variability and the possible impact of long-term climate change. The latter might also offer some opportunities (e.g., increased average precipitation in arid regions). Assessments should take into account the varying opportunities and risks across sectors and across or within regions, along with their uncertainties.

Are there opportunity losses due to decisions made to avoid current climate risks? Water-resources managers are typically quite risk averse, meaning that they would prefer an option with less uncertainty but possibly a lower net benefit over an option with greater uncertainty but a higher possible net benefit. Thus, decisions that minimize climate risks may also decrease the potential benefit and result in lost opportunities. Identifying these lost opportunities reveals increased possible benefits from improved climate consequence management. Have the occurrences of hazard events over the historical record followed identifiable patterns? The initial step is to determine recurrence periods for relevant climate events over the historical record. For example, the analysis might reveal how frequently the system has experienced severe droughts. In some cases, there is spatial or temporal structure (i.e., a pattern) in the historical hazard occurrence. This might include variability across various timescales (such as inter-annual variability due to ENSO) or longer-term trends. The main purpose at this point is to understand variability in the climate system and how it has affected hazard probabilities in the past. One is not yet making forecasts or projections about future scenarios. This analysis reveals the probabilities that have determined system risk up to the current period. The understanding of historical climate variability at different timescales also suggests the key components to consider in developing projections in future steps. This may include identifying appropriate predictors that can help one make simple forecasts of possible shifts in the probability distribution of supply in the system (e.g., shifts due to ENSO phases). Can thresholds be identified that represent changes in system performance that require action? Hydro-climatic conditions affect a water system’s ability to meet performance objectives. Climate variability and change have a significant impact on whether the system fails or is able to meet stakeholder needs. Given the uncertainty related to climate change, it is useful for decision purposes to identify thresholds that signal where a system performance is no longer acceptable and adaptive action is necessary. For example, a reservoir may be designed for a long-term reliability of water delivery of no less than 95%. A threshold of 95% reliability would then be appropriate, for if the reliability fell below this level action would need to be taken. Analysis and answers to the previous questions in this section provide data on historical climate variability and probabilities associated with various climate outcomes. If climate conditions and the historical variability were expected to continue into the future without any changes, one could model the expected reliability based on past experiences. However, this assumes that one is aware of all forms of variability in the past and has the ability to model the future with a high degree of accuracy. If the historical record is too short to capture the full range of climate variability (and this is not uncommon), the results of the analysis can be significantly biased due to sampling variability. In addition, this does not take into account the possible nonstationarity of the system. In order to address these concerns and appropriately assess the sensitivity of the system, it is best to model system performance based on both historical data and scenarios of

Risk Assessment, Risk Management, and Communication: Methods for Climate Variability and Change

possible future climate conditions. A procedure for doing so, decision scaling, is described below.

1.11.4.2.2 Step 2: Consequence assessment This step corresponds to the risk assessment described earlier. In the general sense, this is the product of the probability of some event and the consequences of the event. The process described here is designed for the special case of using climate information for the estimation of probable consequences. This process is termed decision scaling and described below. Decision scaling: tailoring climate information for risk assessment. Decision scaling is the process of tailoring climate information to aid in decision making related to climate risks and opportunities. It uses a decision analytic framework to identify the needed climate information, which is then produced through tailoring climate information from all possible, reliable sources. In general, this means that first the decision must be identified and next the points at which the decision changes as a function of climate information are specified. By doing so, the specific climate information can be tailored from various sources, including GCMs and paleoclimatological data. For example, if it is known that a reservoir system is vulnerable to multi-year droughts and inter-annual variability, but not vulnerable to within year variability, then the effort to produce climate information focuses on trends and temporal structure in inter-annual variability. The premise of this approach is that there are significant and irreducible uncertainties associated with projections of future climate and the resulting hydrologic conditions. Therefore, the emphasis should shift away from attempting to predict the future or provide a few examples of possible future climates. The usual process of analysis starts with GCMs and all their uncertainties, a few scenarios are generated and then those are input to a system model to review the impacts. In decision scaling, the process begins with the system or decision model which is perturbed with parametrically varied climate to generate a climate response function, that is, a representation of the systems’ sensitivity to climate changes. Then the GCM or other climate data are used to estimate the subjective probability of the climates of interest – these climate changes that cause significant impacts, exceed thresholds identified in step 1, or are related to a point where a decision changes. The most effective way to do this is through the use of multiple run, multi model ensembles to utilize the fullest representation of the model projection uncertainty. Decision scaling is used during the consequence assessment to describe probable consequences or the probability of exceeded adaptation action thresholds. It is used again during Step 3: Consequence and uncertainty management to aid in the decision making process. Also, the focus of recent interest in risk assessment and management is largely related to climate change implications. However, the process described here is appropriate for climate variability as well. Some particular considerations related to climate variability are described below. Consider uncertainty in climate forecasts. Based on location and climate characteristics, there may be significant variation in the ability to make climate predictions. For the same system, forecast skill might vary significantly across time scales.

197

It is critical to be aware of the predictive capacity for the given system and the uncertainty associated with any predictions. The probabilistic nature of climate forecasts reinforces the idea that they are neither guaranteed nor absolute. This uncertainty plays a significant role when integrating the climate information into decision making and one should explicitly assess the uncertainty of any forecasts you consult. The approach to assessing the forecast uncertainty depends on the techniques used to create the forecast and the projected time scale. For example, if a seasonal forecast has been developed using a statistical model, a cross-validation technique can be used to understand and quantify the uncertainty in the model. With complex dynamical and GCM-based models and projections over longer timescales, it is best to consult climate professionals to determine the uncertainty and errors present in the model. Some of the key discussion points regarding longer-term climate projections that include the effects of increasing greenhouse gases and other anthropogenic influences include 1. the climate model’s ability to reproduce climatology in the region; 2. whether the model captures the observed regional trend in twentieth century climate; 3. the extent to which there is a well-established physical basis for the model’s forecasts; 4. the degree of agreement between different models; and 5. the extent to which natural multi-decadal variability impacts the region. Asking these questions and validating forecast models can show where the model made errors and help understand that possible model weaknesses. This demonstrates the remaining uncertainty that must be addressed through management options, as discussed in the next step. In the case of climate change, uncertainty may be considered too significant to assign probabilities to specific climate events. Even when probabilities are assigned, they must be considered fairly broad estimations and due consideration given to surprise, resilient design, and performance in failure mode.

1.11.4.2.3 Step 3: Consequence and uncertainty management The final step consists of developing a plan to manage probable consequences and uncertainty that have been identified and described in the previous steps of this process. The process proceeds in the three parts described in Section 1.11.2. Decision scaling is used to provide climate information that is needed in the decision analytic framework used to evaluate alternatives. Here, decision scaling provides the probabilities of the scenarios that cause a particular option to be preferred over another. The estimated hydro-climatic risk determined in the previous two steps serves as the foundation for developing a portfolio of options to mitigate the risk and take advantage of possible opportunities. It is critical to realize that climate information provides information about the probability of particular climate events (such as droughts), but anything can still happen, even if it is very unlikely. For this reason, we again highlight critical aspects of consequence and uncertainty

198

Risk Assessment, Risk Management, and Communication: Methods for Climate Variability and Change

management that are vital to managing risks and opportunities associated with climate variability and change. These are residual risk, surprise, and performance of the system when failure occurs. In the context of climate uncertainty, systematically evaluating risks and opportunities due to climate and comparisons with costs of various actions will benefit from a decision analytic framework. Decision scaling facilitates the production of relevant climate information for use in such a framework. For example, decision analysis might determine that decision A is optimal for a future in which mean annual stream flow decreases by 10% or more, while decision B is optimal in all other cases. Decision scaling is used to estimate the probability of decision A being the better choice using the full range of climate data available. While there is uncertainty related to that probability, if it is a large number relative to the probability associated with the optimality of B, say more than 60%, then we have some confidence that A is the right choice even with the uncertainty. Through such analyses, a plan is developed for addressing risk, opportunities, and uncertainty. The degree of uncertainty associated with future climate variability and change may favor delay in decisions where appropriate. It may also favor the methods for managing the impacts of climate events that do not necessitate new investments in infrastructure. The reasoning is as follows: if an event is not very likely to occur, it is typically not worth making major investments to manage the impact. However, given the possibility that we have underestimated a risk, we should consider ways to avoid the negative impacts of that event, if possible. Finding solutions that can be called upon only when needed is an efficient way to manage the impacts of unlikely events. Residual risk and surprise may also be addressed through redundancy. If a water supply system consists of a single source, any impact on that source leaves the system vulnerable. While it may not be economically efficient to build new infrastructure to tap new sources, other opportunities may exist. The suite of risk management options might include economic instruments (such as insurance or water banks), infrastructure modifications, or integrating seasonal forecasts into decision making, among many others. Together, these approaches are termed portfolio of options because they consist not of a single solution, but rather a basket of possibilities – each of which may be the best choice in a particular circumstance. The final consideration is the ability of the system to perform when the primary risk management approaches fail. For a water-supply system this situation may occur due to the loss of the supply due to drought. For flood protection this is the overtopping of the levees. While the primary system has failed this does not mean the plan has failed and should not mean catastrophe. Through residual risk management and redundancies as described above, the system may be able to continue performing. Water-supply systems may use boil water orders or may contract with neighbor utilities for trucked water. In floods, the planning of evacuations and local emergency response can ensure that an extreme hydrologic event does not have extreme human impact.

Below are some additional considerations when developing the portfolio and determining the most appropriate solutions. Consider planning and operational approaches. The risk management solutions available depend partly on the time frame for action. Near-term operational options will most likely assume fixed infrastructure and some level of sunk costs (those that have already been allocated and cannot be recovered). Possible planning solutions, on the other hand, can include decisions regarding infrastructure and system design. Climate information should be integrated into decision making at the appropriate time scale to inform options most effectively. Projections of long-term climate change may have little value at the operational level for current practices. However, such projections might inform planning decisions as well as the framework under which operational decisions are made in the future (i.e., whether expected climate changes necessitate more flexible operational policies). Assess possible trade-offs. Limited human, financial, and natural resources lead to trade-offs in almost all decisions in water resources management. Water managers must seek to understand and assess possible benefits or consequences of their decisions within the context of these resource constraints. Uncertainty makes such assessment even more difficult, but can also increase the importance of decision outcomes. There is often a trade-off between increasing expected reliability for a system and increasing possible benefits from water allocation. Improved climate information and projections of likely futures may help shift the reliability scenarios. While this does not eliminate the necessary tradeoff, it can improve the possibility of achieving positive outcomes. Integrating thresholds of acceptable costs into decision making can help water managers balance trade-offs. Consider the impact of uncertainty. It is necessary to understand the uncertain nature of probabilistic forecasts in order to assess the suite of options appropriately. Rather than planning for a specific outcome, the most appropriate approach often requires planning for a set of scenarios. While the likelihood of a specific outcome might be higher than the likelihood of another, both are possible and should be considered in decision making. This uncertainty may lead to more flexible approaches and policies, with less emphasis on rigid options that leave little room for alternative outcomes. A flexible, adaptive plan might also increase the capacity to take advantage of possible opportunities from better than expected outcomes. Of particular importance is to consider the effects of low-probability but high-impact events on the system when actions are taken based on a forecast. For example, if the forecast leads one to expect more water, are there ways to mitigate the effects of an unlikely severe drought? This is important to consider because sometimes the anticipatory actions based on a forecast may leave a system more exposed to the down-side risk, or the risk associated with the less likely, but still possible, climate extreme.

1.11.5 Conclusion In this chapter an approach to risk assessment, management, and communication is presented that attempts to reconcile

Risk Assessment, Risk Management, and Communication: Methods for Climate Variability and Change

traditional approaches with our growing knowledge of uncertainty and nonstationarity that mark the hydrologic record. In describing these steps, it becomes clear to the authors at least that the research emphasis within the water community has focused primarily on the means to reduce the uncertainty related to hydrologic events and better prescribe the distributions used to estimate their probabilities. Relatively little effort has been devoted to develop innovative means of reducing hydrologic risk to society or of communicating risk in order to promote risk-reducing behavior. The effects on the hydrologic record of climate change and land-use change calls this research orientation into question. Irreducible uncertainties that hamper our ability to estimate hydrologic design variables imply that greater effort is needed for the development of designs and strategies that perform well under a range of possible future conditions. By developing a wider range of options, water engineers and planners will have greater opportunity to design systems that are resilient as conditions evolve into the future. To develop these options, new ideas related to the use of information technology, emergency planning, and economic incentives should be explored. The challenge of managing risks which cannot be effectively constrained with our traditional approaches means we must be willing to expand beyond our comfortable set of tools. If we are not willing to manage the full range of hydrologic risks, who do we expect will?

References Bondi H (1985) Risk in perspective. In: Cooper, MG (ed.) Risk: Man-Made Hazards to Man. New York: Oxford University Press. Brekke LD, Kiang JE, Olsen JR, et al. (2009) Climate Change and Water Resources Management – a Federal Perspective, U.S. Geological Survey Circular 1331, 65p (ISBN 978-1-4113-2325-4). http://pubs.usgs.gov/circ/1331 (accessed April 2010). Brown C and Carriquiry M (2007) Managing hydroclimatological risk to water supply with option contracts and reservoir index insurance. Water Resource Research 43: W11423 (doi:10.1029/2007WR006093).

199

Brown C, Conrad E, Sankarasubramanian A, and Someshwar S (2009) The use of seasonal climate forecasts within a shared reservoir system: The case of Angat reservoir, Philippines. In: Ludwig F, Kabat P, van Schaik H, and van der Valk M (eds.) Climate Change Adaptation in the Water Sector, 274pp. London: Earthscan. Brown C, Greene A, Block P, and Giannini A (2008) Review of downscaling methodologies for Africa applications. IRI Technical Report 08-05, 31 pp. Characklis GW, Kirsch BR, Ramsey J, Dillard KEM, and Kelley CT (2006) Developing portfolios of water supply transfers. Water Resources Research 42: W05403 (doi:10.1029/2005WR004424). de Neufville R (2004) Uncertainty Management for Engineering Systems Planning and Design, Engineering Systems Monograph. p. 18. Cambridge, MA: MIT Press. Dessai S, Hulme M, Lempert R, and Pielke R, Jr. (2009) Do we need better predictions to adapt to a changing climate? EOS Transactions AGU 90(13) (doi:10.1029/ 2009EO130003). Fiering MB and Matalas NC (1990) Decision-making under uncertainty. In: Waggoner PE (ed.) Climate Change and U.S. Water Resources, pp. 75–83. New York: John Wiley & Sons. International Research Institute for Climate and Society (IRI) (2006) A gap analysis for the implementation of the Global Climate Observing System Programme in Africa. IRI Technical Report Number IRI-TR/06/1. New York: IRI. Lempert R and Collins M (2007) Managing the risk of uncertain threshold responses: Comparison of robust, optimum, and precautionary approaches. Risk Analysis 27(4): 1009–1026. Lund JR (2002) Floodplain planning with risk-based optimization. Journal of Water Resources Planning and Management 3: 202–207. Plate EJ (2004) Risk management for hydraulic systems under hydrological loads. In: Bogardi JJ and Kundzewicz ZW (eds.) Risk, Reliability, Uncertainty and Robustness of Water Resources Systems, 220pp. New York, NY: UNESCO and Cambridge University Press. Plate EJ (2002) Risk management for hydraulic systems under hydrological loads. In: Bogardi JJ and Kundzewicz ZW (eds.) Risk, Reliability, Uncertainty and Robustness of Water Resources Systems, chap. 23, pp. 209–220. Cambridge, UK: Cambridge University Press. Rees J (2002) Risk and Integrated Water Resources Management. Global Water Partnership TEC Background Paper No. 6. Stedinger JR, Vogel RM, and Foufoula-Georgiou E (1993) Frequency analysis of extreme events In: Maidment DR (ed.) Handbook of Hydrology, chap. 18. New York: McGraw-Hill. Taleb NN (2007) The black swan: The impact of the highly improbable. New York: Random House, 400pp. UNDRO Office of the United Nations Disaster Relief Coordinator (1991) Mitigating Natural Disasters: Phenomena, Effects and Options. A Manual for Policy Makers and Planners. New York: United Nations. van Aalst M, Hellmuth M, and Ponzi D (2007) Come rain or shine: Integrating climate risk management into African development bank operations, Working Paper No. 89. Tunis: African Development Bank.

Preface – The Science of Hydrology S Uhlenbrook, Department of Water Engineering, DA Delft, The Netherlands & 2011 Elsevier B.V. All rights reserved.

The world is changing and it seems that the speed of changes is accelerating. In the overall introduction of the Treatise on Water Sciences, the editor-in-chief Peter Wilderer (The Importance of Water Science in a World of Rapid Change: A Preface to the Treatise on Water Science) discusses the prevailing changes, its drivers, and possible impacts on different water disciplines. A major challenge is that all changes and their various impacts are interacting with each other, although how and to what extent is often poorly understood. For scientists and practitioners, this makes the problem identification and the development of sustainable solutions for water problems a very difficult task. Therefore, it is very timely to summarize the contemporary state of the knowledge in the different fields of water sciences and technology, and to provide a platform for innovative research and development. I am pleased to conclude that this volume on hydrology is an important piece of the complex puzzle. What is hydrology? The International Association of Hydrological Sciences (IAHS) in collaboration with UNESCO defined hydrology as the ‘‘science that deals with the water of the earth, their occurrence, circulation and distribution, their chemical and physical properties, and their reaction with their environment, including their relation to living beings.’’ In addition, it states that hydrology is the ‘‘science that deals with the processes governing the depletion and replenishment of the water resources of the land areas of the earth, and various phases of the hydrological cycle.’’ This is indeed a very wide definition. Many aspects of the chemical properties and interactions with the environment are part of Volume 3 of this treatise. Topics that are directly related to the management of the water resources are part of (Preface – Management of Water Resources) of this treatise. However, this volume (The Science of Hydrology) deals with all major components of the water cycle and key water-quality aspects. It also discusses the linkages to closely related disciplines. The aims of the science of hydrology were well summarized by the Dutch Foresight Committee on Hydrological Science (KNAW, 2005) as follows:

1. to understand the mechanisms and underlying processes of the hydrological cycle and its interactions with the lithosphere, atmosphere, and biosphere; 2. to enhance our knowledge of interactions between the hydrosphere and atmosphere, the hydrosphere and lithosphere, and the hydrosphere and biosphere, thereby increasing our understanding of the role that water plays in the Earth system; 3. to quantify human impact on the past, present, and future conditions of hydrological systems; and 4. to develop strategies for sustainable use and protection of water resources, hydrological systems, and the associated environmental conditions.

The science of hydrology is special, as it holds a place, on the one hand, in the field of Earth System Sciences, where it is directly linked to earth science disciplines, such as atmospheric sciences, geomorphology, geology, soil sciences, geobiology, and ecology. On the other hand, hydrology is an applied science and, as such, a part of engineering. This makes the discipline highly relevant to the management and development of the water resources and the prediction and mitigation of water-related natural hazards (floods, droughts, landslides, etc.) to finally support life, civilization, and sustainable development. These complementary aspects of hydrology (Earth System Sciences and the basis for water management/engineering) make it an exciting and very relevant discipline. It is quite a dynamic discipline given the significant developments of the past decades; many of them are reviewed in this volume. The volume starts with a comprehensive overview of global hydrology and the spatio-temporal variability of hydrological fluxes and water resources on a large scale. It continues with several chapters on the main variables of the water balance, such as precipitation, evaporation and interception, and stream discharge; then it goes on to discuss the storage components of groundwater, soil water, lakes, and reservoirs. Unfortunately, a chapter on snow and ice, the globally largest and regionally/locally often very important water storage component, was withdrawn at a late stage and could not be replaced in time. The volume continues with several chapters discussing the state of the art and the possible future developments of observation methods for ground-based techniques (i.e., fieldbased methods, tracer techniques, and hydrogeophysics) and remote-sensing techniques. Key data analysis and modeling techniques as well as theoretical considerations are reviewed in four, mainly theoretical, chapters on scaling and regionalization, statistical methods, hydrological modeling, and uncertainty estimation techniques. The linkages between hydrology and aquatic ecology and biogeochemistry are discussed in two comprehensive chapters. Two chapters are related to the processes and issues of erosion and sedimentation as well as surface water–groundwater interactions. The inclusion of all these topics results in a sizable volume with 20 chapters, exceeding 500 pages. However, several hydrology-related topics are not or could be only partly covered (e.g., urban hydrology, snow and ice, coastal hydrological systems, landscape evolution, and hydrogeomorphology). Perhaps this can be seen as an invitation to redo the exercise in a few years from now, and to review the latest developments in this dynamic field and strive for more completeness.

References KNAW (2005) Turning the Water Wheel Inside Out. Foresight Study on Hydrological Science in the Netherlands. Amsterdam: Royal Academy of Arts and Sciences.

1

2.01 Global Hydrology T Oki, The University of Tokyo, Tokyo, Japan & 2011 Elsevier B.V. All rights reserved.

2.01.1 Introduction 2.01.1.1 The Earth System and Water 2.01.1.2 Water Reserves, Fluxes, and Residence Time 2.01.2 Global Water Cycle 2.01.2.1 Existence of Water on Earth 2.01.2.2 Water Cycle on the Earth 2.01.3 Global Water-Balance Requirements 2.01.3.1 Water Balance at Land Surface 2.01.3.2 Water Balance in the Atmosphere 2.01.3.3 Combined Atmosphere–River Basin Water Balance 2.01.3.3.1 Estimation of large-scale evapotranspiration 2.01.3.3.2 Estimation of total water storage in a river basin 2.01.3.3.3 Estimation of zonally averaged net transport of freshwater 2.01.3.4 Bottom Line of Global Water Balance 2.01.4 Global Water Balance 2.01.4.1 Uncertainties in Global Water-Balance Estimates 2.01.4.2 Water Balance and Climate 2.01.4.3 Annual Water Balance in Climatic Regions 2.01.5 Challenges in the Global Hydrology and Research Gaps 2.01.5.1 Macroscale Hydrological Modeling 2.01.5.2 Global Changes and Global Hydrology 2.01.5.3 Global Trace of Water Cycles 2.01.5.4 Interactions of Global- and Local-Scale Hydrology 2.01.5.5 Research Opportunities in Global Hydrology 2.01.5.6 Research Gaps in Global Hydrology Acknowledgments References

2.01.1 Introduction 2.01.1.1 The Earth System and Water The Earth system is unique in that water exists in all three phases, that is, water vapor, liquid water, and solid ice, when compared with the forms of water on other planets. The transport of water vapor is regarded as energy transport because of the large amount of latent-heat exchange that occurs during its phase change to liquid water (approximately 2.5 106 J kg1); therefore, the water cycle is closely linked to the energy cycle. Even though the energy cycle on the Earth is an open system driven by solar radiation, the amount of water on the Earth does not change during shorter than geological timescales (Oki, 1999), and the water cycle itself is a closed system. On a global scale, hydrologic cycles are associated with atmospheric circulation, which is driven by the unequal heating of the Earth’s surface and atmosphere in latitude (Peixo¨to and Oort, 1992). Annual mean absorbed solar energy at the top of the atmosphere is highest near the equator with approximately 300 W m2, and decreases rapidly at higher latitudes, and is

3 3 3 4 4 5 8 9 9 11 11 11 12 12 12 12 14 15 16 16 19 19 20 22 22 24 24

approximately 60 W m2 at the Arctic and Antarctic regions. Emitted terrestrial radiative energy from the Earth at the top of the atmosphere is approximately 250 W m2 for the areas between 201 N and 201 S, gradually decreases at higher latitudes, and is approximately 175 W m2 at the Arctic region and 150 W m2 at the Antarctic region. As a consequence, the net annual energy balance is positive (absorbing) over tropical and subtropical regions between 301 N and 301 S, and negative in higher latitudes (Dingman, 2002). Without atmospheric and oceanic circulations on the Earth, temperature differences on the Earth would have been more drastic. Temperatures in the equatorial zone would have been much higher such that the outgoing terrestrial radiation balances the absorbed solar energy, and the temperatures in the polar regions would have been much lower as well. Both the atmosphere and the ocean carry much energy from the equatorial regions toward the polar regions. In the case of atmosphere, the energy transport consists of sensible heat and latent-heat fluxes (Masuda, 1988). The global water circulation includes the latent-heat transport in which water vapor plays an active role in the atmospheric circulation. Water vapor is not a passive component of the atmosphere system;

3

4

Global Hydrology

rather, it affects atmospheric circulation by both radiative transfer and latent-heat release of phase change.

2.01.1.2 Water Reserves, Fluxes, and Residence Time The total volume of water on the Earth is estimated as approximately 1.4 1018 m3, and it corresponds to a mass of 1.4 1021 kg (Figure 1, revised from Oki and Kanae (2006)). Compared with the total mass of the Earth (5.974 1024 kg), the mass of water constitutes only 0.02% of the planet, but it is critical for the survival of life on the Earth, and the Earth is called the Blue Planet and the Living Planet. There are various forms of water on the Earth’s surface. Approximately 70% of its surface is covered with salty water, the oceans. Some of the remaining areas (continents) are covered by freshwater (lakes and rivers), solid water (ice and snow), and vegetation (which implies the existence of water). Even though the water content of the atmosphere is comparatively small (approximately 0.3% by mass and 0.5% by volume of the atmosphere), approximately 60% of the area of the Earth is always covered by cloud (Rossow et al., 1993). The Earth’s surface is dominated by the various phases of water. Water on the Earth is stored in various reserves, and various water flows transport water from one to another. Water flow

Water vapor over sea 10

Evaporation over ocean 436.5

(mass or volume) per unit time is also called water flux. The mean residence time in each reserve can be simply estimated from total storage volume in the reserve and the mean flux rate to and from the reserve:

Tm ¼ V=F

ð1Þ

where Tm, V, and F are mean residence time, total storage, and the mean flux rate, respectively. We can also represent the distribution of flux rate of water flow that comes in and goes out from the storage Chapman, 1972). The last column of Table 1 (simplified from the table in Korzun (1978)) presents global values of the mean residence time of water. Evidently, the water cycle on the Earth is a stiff differential system with variability on many timescales, from a few weeks to thousands of years. The mean residence time is also important when considering water-quality deterioration and restoration, since it can be an index of how much water is turned over. Apparently, river water or surface water is more vulnerable to pollution than groundwater; however, any measure to increase waterquality recovery tends to be more efficient for river water than groundwater, and, as suggested from Table 1, the mean

Total terrestrial precipitation 111 Snowfall Rainfall 12.5 98.5

Net water-vapor flux transport 45.5

Water vapor over land 3

Glaciers and snow 24 064

Total terrestrial evapotranspiration 65.5 21 Precipitation over ocean 391

6.4 11.7 Others (29.3)

7.6 11.6 Cropland (12.6) 2.66

Unirrigated

0.38 Domestic Sea 1 338 000

0.77

Irrigated

45.5

River 2

Industry

54

Forest (40.1)

0.2 0.3 Wetland (0.2) Wetland 11

31

Grassland (48.9)

29

Biological Permafrost water 300 1 Surface runoff 15.3

1.3 2.4 Lake (2.7) Soil moisture 17

Subsurface runoff 30.2

Lake 176

Groundwater 23 400 Flux, 103 km3 yr−1 Storage, 103 km3

The terrestrial water balance does not include Antarctica

( )

Area 106 km2

Figure 1 Global hydrological fluxes (1000 km3 yr1) and storages (1000 km3) with natural and anthropogenic cycles are synthesized from various sources. Big vertical arrows show total annual precipitation and evapotranspiration over land and ocean (1000 km3 yr1), which include annual precipitation and evapotranspiration in major landscapes (1000 km3 yr1) presented by small vertical arrows; parentheses indicate area (million km2). The direct groundwater discharge, which is estimated to be about 10% of the total river discharge globally, is included in river discharge. The values of area sizes for cropland and others are corrected from original ones. From Oki T, Nishimura T, and Dirmeyer P (1999) Assessment of annual runoff from land surface models using total runoff integrating pathways (TRIP). Journal of the Meteorological Society of Japan 77: 235–255 and Ok T and Kanae S (2006) Global hydrological cycles and world water resources. Science 313(5790): 1068–1072.

Global Hydrology Table 1

5

World water reservesa

Form of water

Covering area (km2)

Total volume (km3)

Mean depth (m)

World ocean Glaciers and permanent snow cover Ground water Ground ice in zones of permafrost strata Water in lakes Soil moisture Atmospheric water Marsh water Water in rivers Biological water Artificial reservoirs Total water reserves

361 300 000 16 227 500 134 800 000 21 000 000 2 058 700 82 000 000 510 000 000 2 682 600 148 800 000 510 000 000

1 338 000 000 24 064 100 23 400 000 300 000 176 400 16 500 12 900 11 470 2120 1120 8000 1 385 984 610

3700 1463 174 14 85.7 0.2 0.025 4.28 0.014 0.002

510 000 000

2718

Share (%) 96.539 1.736 1.688 0.0216 0.0127 0.0012 0.0009 0.0008 0.0002 0.0001

Mean residence time 2500 years 1600 years 1400 years 10 000 years 17 years 1 year 8 days 5 years 16 days A few hours 72 days

100.00

a

Simplified from Table 9 of ‘‘World water balance and water resources of the earth’’ by UNESCO Korzun, 1978. The last column, mean residence time, is from Table 34 of the report.

residence time of river water is shorter than that of groundwater. Since the major interests of hydrologists have been the assessment of volume, inflow, outflow, and the chemical and isotopic composition of water, the estimation of mean residence time of a certain domain has been one of the major targets of hydrology. It should be recalled that the residence time estimated with isotope tracers often differs from the hydrological residence time derived from Equation (1) (Uhlenbrook et al., 2002, 2004). This is due to the fact that in the subsurface system, the diffusive exchange processes between mobile and immobile parts make the residence time usually much longer. This process is particularly important for hydrochemical processes.

2.01.2 Global Water Cycle 2.01.2.1 Existence of Water on Earth Table 1 and Figure 1 denote the quantity of water stored in each of the reserves on the Earth. Most of the storage values given in Table 1 are taken from Korzun (1978), except for water vapor in the atmosphere which is calculated from atmospheric data (Oki et al., 1995). The various reserves of water on the Earth are discussed in the following: The proportion of water in the ocean is large (96.5%). Even though in classical hydrology, ocean processes are traditionally excluded, the global hydrological cycle is never closed without including them. The ocean circulations carry large amounts of energy and water. The surface ocean currents are driven by surface wind stress, and the atmosphere itself is sensitive to the sea-surface temperature. Temperature and salinity determine the density of ocean water, and both factors contribute to the overturning and deep-ocean general circulation Other major reserves are solid waters on the continent (glaciers and permanent snow cover) and groundwater. Glaciers are the accumulation of ice originated from the atmosphere, and they generally move slowly on land over a long period. Glaciers form a discriminative U-shaped valley over land, and remain moraine when they retreat. If a glacier flows into an ocean, its terminated end often forms an iceberg. Glaciers react in comparatively longer timescales against

climatic change, and they also induce isostatic responses of continental-scale upheavals or subsidence in even longer timescales. Even though it is predicted that the thermal expansion of oceanic water dominates the anticipated sea-level rise due to global warming, glaciers over land are also a major concern as the cause for sea-level rise associated with global warming. Groundwater is the subsurface water occupying the saturated zone. It contributes to runoff in the low-flow regime between storm events. Deep groundwater may also reflect the long-term climatological situation. Groundwater in Table 1 includes both gravitational and capillary water, but groundwater in the Antarctica (roughly estimated as 2 106 km3) is excluded. Gravitational water is the water in the unsaturated zone (vadose zone) which moves under the influence of gravity. Capillary water is water found in the soil above the water table by capillary diffusion, a phenomenon associated with the surface tension of water in soils acting as porous media. In terms of groundwater recharge, Do¨ll and Fiedler (2008) estimated the global groundwater recharge flux to be 12 666 km3 yr1 and approximately 1000 mm yr1 in the Amazon region. They assumed the recharge flux is a fraction of the total runoff. Koirala (2010) estimated groundwater recharge flux by coupling a land-surface scheme, namely Minimal Advanced Treatments of Surface Interaction and RunOff (MATSIRO; see Takata et al., 2003), with a macro-scale groundwater representation Yeh et al., 2005). The global distribution of model-simulated groundwater recharge is illustrated in Figure 2(a). Total groundwater recharge flux is estimated as 31 789 km3 yr1 and the value is close to the flux of subsurface runoff in Figure 1 (30 200 km3 yr1). Soil moisture is the water that is held above the groundwater table. It influences the energy balance at the land surface, such that a lack of available moisture suppresses evapotranspiration (which consists of soil evaporation, plant transpiration, and interception loss), and changes surface albedo. Soil moisture also alters the fraction of precipitation partitioned into direct runoff and infiltration. The precipitation water becoming direct runoff cannot be evaporated from the same place, while the water infiltrated into soil may be taken up by hydraulic suction and evaporated back into the atmosphere. The global distribution of model-estimated mean

6

Global Hydrology 90° N

60° N

30° N

0°

30° S

60° S 180°

120° W

0

10

20

60° W

30

50

0°

100

60° E

300

500

1000

120° E

1500

2000

180°

3000

4000

(mm yr−1)

(a)

90°

60°

30°

0°

−30°

−60°

−90° −180°

(b)

0.0

−150°

0.1

−120°

0.2

90°

−60°

0.3

−30°

0.4

0°

30°

0.5 1

0.6

60°

90°

0.7

120°

0.8

150°

0.9

180°

1.0

Figure 2 (a) Global map of long-term mean net groundwater recharge (mm yr ) estimated by an LSM coupled with macroscale groundwater model (Yeh and Eltahir, 2005; S. Koirala, 2010); (b) global distribution of annual mean soil wetness index estimated by 13 LSMs averaged for 1986–95 through the second phase of the Global Soil Wetness Project; (c) same as (b) but for annual precipitation (mm yr1) used in the GSWP2 based on observation; (d) same as (b) but for estimated annual runoff (mm yr1) by LSMs; (e) same as (b) but for mean river discharge (106 m3 yr1); (f) annual vapor-flux convergence (mm year1) for 1989–92 (Oki et al., 1995). (g) Annual mean evapotranspiration (mm yr1) estimated as a residual of (f) and precipitation corresponding to the period; (h) same as (b) but for annual mean evapotranspiration (mm yr1) by LSMs for 1986–95. Data of (a) from Takata K, Emori S, and Watanabe T (2003) Development of minimal advanced treatments of surface interaction and runoff. Global and Planetary Change 38: 209–222 and Koirala S (2010) Explicit Representation of Groundwater Process in a Global-Scale Land Surface Model to Improve Hydrological Predictions. PhD Thesis, The University of Tokyo; (b) from Dirmeyer PA, Gao XA, Zhao M, Guo ZC, Oki T, and Hanasaki N (2006) GSWP-2 multimodel anlysis and implications for our perception of the land surface. Bulletin of the American Meteorological Society 87: 1381–1397; and (f) from Oki T, Musiake K, Matsuyama H, and Masuda K (1995) Global atmospheric water balance and runoff from large river basins. Hydrological Processes 9: 655–678.

Global Hydrology

7

90°

60°

30°

0°

−30°

−60°

−90° −180°

(c)

0

−150°

20

−120°

50

−90°

100

−60°

150

−30°

200

0°

300

30°

500

60°

800

1400

90°

1800

120°

2400

150°

180°

3000

5000

Figure 2 Continued.

soil wetness index is shown in Figure 2(b). Generally, the distribution is correlated with precipitation distribution (Figure 2(c)) as well as with runoff distribution (Figure 2(d)), but the global distribution of river discharge (Figure 2(e)) accumulates total runoff generated in the upper watershed, and the shape of the river channels can be seen in the distribution. The atmosphere carries water vapor, which influences the heat budget via latent-heat-exchange processes. Condensation of water vapor releases latent heat, which warms the atmosphere and affects the atmospheric general circulation. Liquid water (droplets, clouds, etc.) in the atmosphere is another result of condensation. Clouds significantly change the radiation in the atmosphere and at the Earth’s surface. However, the volume of liquid (and solid) water contained in the atmosphere is relatively small, as most of the water in the atmosphere exists as water vapor. Water vapor is also the major absorber in the atmosphere of both short-wave and long-wave radiation. Precipitable water is the total water vapor in the atmospheric column integrated from land surface to the top of the atmosphere. Vertically integrated water-vapor flux convergence is a useful tool to diagnose global water balance (see Figure 2(f) for its global distribution). The amount of water stored in rivers (Figure 2(e)) is rather tiny compared to other reserves at any time; however, the recycling speed, which can be estimated as the inverse of the mean residence time (Equation (1)), of river water (river discharge) is relatively high, and it is important because most societal applications ultimately depend on river water as a renewable and sustainable resource. The amount of water stored transiently in a soil layer, in the atmosphere, and in the river channels is relatively minute,

and the time spent through these subsystems is relatively short. However, they play a dominant role in the global hydrological cycle.

2.01.2.2 Water Cycle on the Earth The water cycle plays many important roles in the climate system, and Figure 1 schematically illustrates various flow paths of water in the global hydrologic system (Oki and Kanae, 2006). Precipitation is calculated from global estimates based on observations from the forcing data of the Global Soil Wetness Project (GSWP2, the second phase the project, see the discussion in Section 2.01.4) over land, and data from Climate Modeling Analysis and Prediction (CMAP; Xie and Arkin, 1996) over ocean. Land-surface fluxes, such as evapotranspiration and surface and subsurface runoff, are the estimated results from GSWP2. Differentiation of precipitation between snow and rain over land is also either estimated by land-surface models that participated in the GSWP2, or given by individual forcing determined by temperature. Values on human water withdrawals for irrigation, industry, and households are taken from Shiklomanov (1997). Water-vapor transport and its convergence are estimated using the European Centre for Medium-Range Weather Forecast (ECMWF) objective analyses, obtained as the 4-year mean from 1989 to 1992 (Oki et al., 1995). The roles of these water fluxes in the global hydrologic system are now briefly reviewed:

•

Precipitation is the water flux from atmosphere to land or ocean surface. It drives the hydrological cycle over the land surface, also changes surface salinity (and temperature) over the ocean, and affects its thermohaline circulation. Rainfall

8

Global Hydrology

Annual runoff 210°

240°

270°

300°

330°

0°

30°

60°

90°

120°

150°

180°

60°

30°

0°

−30°

(mm yr−1)

−60° 0

(d)

10

5

50

100

200

500

1000

2000

5000

Annual river discharge 210°

240°

270°

300°

330°

0°

30°

60°

90°

120°

150°

180°

60°

30°

0°

−30°

(106 m3 yr−1)

−60° 0 (e)

10

100

200

500

1000

2000

5000

10 000

20 000

50 000

Water scarcity index

Figure 2 Continued.

•

refers to the liquid phase of precipitation. A part of it is intercepted by canopy over vegetated areas, and the remaining part reaches the Earth‘s surface as through-fall. The highly variable, intermittent, and concentrated behavior of precipitation in time and space domain compared to other major hydrological fluxes mentioned below makes the observation of this quantity and the aggregation of the process complex and difficult. Global distribution of precipitation is presented in Figure 2(c). Currently, satellitebased estimates merged with in situ observational data have been produced and revealed to the public (e.g., Kubota et al., 2007). Snow has special characteristics compared with rainfall. Snow may be accumulated and the surface temperature will not rise above 0 1C until the completion of snowmelt. The albedo of snow is quite high (as high as clouds). Consequently, the existence of snow changes the surface energy and water budget enormously. A snow surface typically

•

reduces the aerodynamic roughness, and therefore may also have a dynamical effect on the atmospheric circulation and hydrologic cycle. Evaporation is the return flow of water from the surface to the atmosphere and the latent-heat flux from the surface. The amount of evaporation is determined by both atmospheric and hydrological conditions. From the atmospheric point of view, the partition of incoming solar energy to the surface between latent and sensible heat flux is important. Wetness at the surface influences this partition significantly because the ratio of actual evapotranspiration to the potential evaporation is reduced due to drying stress. The stress is sometimes formulated as a resistance under which evaporation is classified as hydrology driven (soil controlled). If the land surface is wet enough compared to available energy for evaporation, the condition is classified as radiation driven (atmosphere controlled).

Global Hydrology

9

Annual vapor flux convergence (mm yr−1) 0

60° E

120° E

180

120° W

60° W

0

90° N

60° N

30° N

EQ.

30° S

60° S

90° S

(f)

−2000

−1600

−1200

−800

−400

.00

400

800

1200

1600

2000

Figure 2 Continued.

•

•

Transpiration is the evaporation of water through stomata of leaves. It has two special characteristics different from evaporation from soil surfaces. One is that the resistance of stomata is related not only to the dryness of soil moisture but also to the physiological conditions of vegetation, through the opening and closing of stomata. Another feature is that roots of plants can uptake water from deeper soil layers and transpirate the water, compared to the case of evaporation from bare soil without plants. Vegetation also modifies the surface energy and water balance by altering surface albedo and by intercepting some precipitation and evaporating this rainwater. The global distribution of total evapotranspiration is shown in Figure 2(g), which is estimated using the atmospheric water-balance computation (Equation (7) in Section 2.01.3.3.1), and in Figure 2(h), estimated by land-surface models (in Section 2.01.4.1). Runoff returns water from the land to the ocean, which may otherwise be transported in vapor phase by evaporation and atmospheric advection. The runoff into the ocean is also important for the freshwater balance and the salinity of the ocean. Rivers carry not only water mass but also sediment, chemicals, and various nutritional matters from continents to seas. Without rivers, global hydrologic cycles on the Earth are not closed. Runoff at hillslope scale is a nonlinear and complex process. Surface runoff can be generated when the intensity of rainfall or snowmelt exceeds the infiltration rate of the soil, or when precipitation falls on the saturated land surface. Saturation at land surface can be formed mostly by topographic-concentration mechanism along the hillslopes. Infiltrated water in the upper part of the hillslope flows down the slope and discharges at the bottom of the hillslope. Due to the highly

variable heterogeneity of topography, soil properties (such as hydraulic conductivity and porosity) and precipitation, basic equations such as Richard’s equation, which are fairly valid at a point scale or hillslope scale, cannot be directly applied in the macroscale because of the nonlinearity involved. The global distributions of runoff and river discharge are illustrated in Figures 2(d) and 2(e). The global water cycle integrates these components, which consist of the state variables (precipitable water, soil moisture, etc.) and the fluxes (precipitation, evaporation, etc.).

2.01.3 Global Water-Balance Requirements The conservation law of water mass in any arbitrary control volume indicates the water balance. In this section, the water balance over land, for an atmospheric column, and its combination is presented (Oki, 1999). Some applications of these water-balance equations for estimating some of the waterbalance components are introduced as well.

2.01.3.1 Water Balance at Land Surface In the field of hydrology, river basins have commonly been selected for study, and water balance has been estimated using ground observations, such as precipitation, runoff, and storage in lakes and/or groundwater. The water balance over the land is described as

q S=q t ¼ P  E  Ro  Ru

ð2Þ

where S represents the water storage within the area, t is the time, (q S/q t) is the change of total water storage with time,

10

Global Hydrology Annual evapotranspiration (mm yr−1) 0

60° E

120° E

180°

120° W

60° W

0

90° N

60° N

30° N

EQ.

30° S

60° S

90° S

(g)

0.00

300

900

600

1200

1500

1800

2100

2400

2700

3000

90°

60°

30°

0°

−30°

−60°

−90° −180°

(h)

−200

−150°

0

−120°

20

−90°

50

−60

100

−30°

150

0°

200

30°

300

500

60°

800

90°

1400

120°

1800

150°

180°

2400

3000

Figure 2 Continued.

P is the precipitation, E is the evapotranspiration, Ro is the surface runoff, and Ru is the groundwater movement (all fluxes above are given in the unit volume per time step). S includes snow accumulation in addition to soil moisture, groundwater, and surface-water storage including retention

water within the control volume. The control volume is defined by the area of interest over the land with its bottom generally at the impermeable bedrock. These terms are shown in Figure 3(a). Equation (2) implies that water storage over land is increased by precipitation, and decreased

Global Hydrology

by evapotranspiration, surface runoff, and groundwater movement. If the considered area of water balance is set within an arbitrary boundary, Ro represents the net outflow of water from this area (i.e., the total outflow minus total inflow from surrounding areas). Although, in general, it is not easy to estimate groundwater movement Ru, the net flux per unit area within a large area is expected to be comparatively small. If all groundwater movement is considered to be that observed at the gauging point of a river (Ru ¼ 0), then Equation (2) becomes

q S=q t ¼ P  E  Ro

2.01.3.2 Water Balance in the Atmosphere Atmospheric water-vapor flux convergence contains waterbalance information that can complement the traditional hydrological elements such as precipitation, evapotranspiration, and discharge. The basic concepts as well as the application of atmospheric data to estimate terrestrial water balance were first presented by Starr and Peixo¨to (1958). The atmospheric water balance for a column of atmosphere from the bottom at land surface to the top of the atmosphere is described by

q W=q t ¼ Q þ ðE  PÞ

2.01.3.3 Combined Atmosphere–River Basin Water Balance Since there are common terms in Equations (3) and (4), they can be combined as

q W=q t þ Q ¼ ðP  EÞ ¼ q S=q t þ Ro

• •

Annual change of atmospheric water-vapor storage is negligible ((q W/q t) ¼ 0). Annual change of water storage at the land is negligible ((q S/q t) ¼ 0).

With these assumptions, Equation (5) simplifies into

Q ¼ ðP  EÞ ¼ Ro

(a) Water balance

in the basin

Groundwater movement

ð6Þ

If a river basin is considered as the water-balance region, Ro is simply the discharge from the basin. The simplified Equation (6) demands that the water-vapor convergence, that is, precipitation minus evaporation and net runoff should balance over the annual period when the temporal change of all storage terms can be neglected.

Precipitable water

Vapor flux

Precipitable water

Runoff Basin storage

ð5Þ

Figure 3(c) illustrates the balance in Equation (5), and shows that the difference of precipitation and evapotranspiration is equal to the sum of the decrease of atmospheric water-vapor storage and lateral (horizontal) convergence, and also to the sum of the increase of water storage over the land and runoff. Theoretically, Equation (5) can be applied for any control volume of land area combined with the atmosphere above, even though the practical applicability depends on the accuracy and availability of atmospheric and hydrologic information. The following further assumptions are often employed in annual water-balance computations:

ð4Þ

where W represents the precipitable water (i.e., column storage of water vapor) and Q is the convergence of water-vapor flux in the atmosphere (all fluxes given in the unit volume per time step). Since the atmospheric water content in both solid and liquid phases are generally small, only the water vapor is considered in Equation (4). The balance is schematically illustrated in Figure 3(b), which describes that the water storage

Precipitation Evapotranspiration

in an atmospheric column is increased by the lateral convergence of water vapor and evapotranspiration through the bottom of the column (i.e., land surface), and decreases by the precipitation falling out from the bottom of the atmosphere column to the land.

ð3Þ

This assumption is generally valid at the outlet of a catchment. In most cases, surface runoff Ro becomes river discharge through the transport of river-channel network. The river discharge is an integrated quantity over the whole catchment and can be observed at a downstream point in contrast to other fluxes, such as P and E, which have to be spatially measured.

11

Vapor flux

Runoff

Precipitation

Basin storage

Evapotranspiration (b) Water balance in the atmosphere

Groundwater movement

(c) Combined water balance

Figure 3 (a) Terrestrial water balance, (b) atmospheric water balance, and (c) combined atmosphere–land surface water balance. (a), (b), and (c) correspond to Equations (2), (4), and (5), respectively.

12

Global Hydrology

2.01.3.3.1 Estimation of large-scale evapotranspiration Generally, it is not an easy task to obtain large-scale evapotranspiration E based on observations except for the annual timescale in which E can be estimated as the residual of P and Ro. However, the combined water balance can help estimate E at a shorter timescale, for example, monthly. Note that Equation (4) can be rewritten as

E ¼ q W=q t  Q þ P

ð7Þ

which can be applicable over a period shorter than a year, unlike the assumption in Equation (6). If atmospheric and precipitation data are available over a short timescale such as a month or a day, evapotranspiration can be estimated at the corresponding timescales; however, it is also subject to severe limitations imposed by the data accuracy. The region over which the evapotranspiration is estimated is not limited to a river basin; rather, it depends on the scale and the associated accuracy of the available atmospheric and precipitation data. The global distributions of precipitation P, integrated water-vapor convergence Q, and evapotranspiration E estimated by Equation (7) are presented in Figures 2(c), 2(f), and 2(g), respectively. The zonal mean precipitation P, integrated water-vapor convergence Q, and evapotranspiration E estimated by Equation (7) are presented in Figures 4(a)–4(c). As can be seen from the zonal mean precipitation along the midlatitude, storm tracks over the North Pacific and Atlantic oceans are stronger in December–January–February (DJF) than in June–July–August (JJA). The Intertropical Convergence Zone (ITCZ) is enhanced in JJA, when the southeastern part of the Asian continent is covered by the southwest monsoon rainfall. The distribution of E is less dependent on the latitude and has smaller seasonal changes compared to P (see also Trenberth and Guillemot, 1998). The mean evapotranspiration in tropical areas is approximately 4 mm d1.

2.01.3.3.2 Estimation of total water storage in a river basin

2.01.3.4 Bottom Line of Global Water Balance Water balance on the global scale with consideration of land and ocean areas separately can be expressed as

Total terrestrial water storage, as the sum of surface water (such as river water, snow water, and water in lakes), soil moisture, and groundwater, is generally difficult to estimate on the global scale. Combining Equations (3) and (4) yields

q S=q t ¼ q W=q t þ Q  Ro

the cycles of energy and water are closely related. Wijffels et al. (1992) used values of atmospheric water-vapor convergence Q from Bryan and Oort (1984) and discharge data from Baumgartner and Reichel (1975) to estimate the freshwater transport by oceans and atmosphere, but their results seem to have large uncertainties and they did not present the freshwater transport by rivers. The annual freshwater transport in the meridional (north– south) direction can be estimated from Q and river discharge with geographical information such as the location of river mouths and basin boundaries (Oki et al., 1995). The estimated result is shown in Figure 5, in which it shows that in the case of oceans net transport is the residual of northward and southward freshwater flux by all ocean currents globally, and it cannot be compared directly with individual ocean currents such as the Kuroshio and the Gulf Stream. Transports by the atmosphere and by the ocean have almost the same absolute values for most of the latitudes, but with a different sign. The transport by rivers is about 10% of these other fluxes globally (there may be an underestimation because average Q tends to be smaller than average river discharge observed at the global land surface). The negative (southward) peak by rivers at 301 S is mainly due to the Parana River in South America, and the peaks at the equator and 101 N are due to rivers in South America, such as the Magdalena and Orinoco. Large Russian rivers, such as the Ob, Yenisey, and Lena, carry freshwater toward the north between 50 and 701 N. These results indicate that the hydrological processes over land play non-negligible roles in the climate system, not only by the exchange of energy and water at the land surface, but also through the transport of freshwater by rivers which affects water balance of the oceans and forms a part of the hydrological circulation on the Earth among the atmosphere, land, and oceans.

ð8Þ

which indicates that the change of water storage in the control volume over the land can in principle be estimated from the atmospheric and runoff data. Although an initial value is required to obtain the absolute value of storage, the atmospheric water balance can be useful in estimating the seasonal change of total water storages in large river basins.

2.01.3.3.3 Estimation of zonally averaged net transport of freshwater The meridional (north–south direction) distribution of the zonally averaged annual energy transports by the atmosphere and the ocean has been evaluated, even though there are quantitative problems in estimating such values (Trenberth and Solomon, 1994). However, the corresponding distribution of water transport has not often been studied, although

P1  E1  q S1 =q t ¼ R ¼ ðPo  Eo Þ þ q So =q t

ð9Þ

where P, E, S, and R represent precipitation, evapotranspiration, total water storage, and continental runoff, respectively, with the subscript l indicating values for land and o for ocean. For the steady state, the temporal changes of Sl and So can be neglected and Equation (9) becomes

P1  E1 ¼ R ¼ ðPo  Eo Þ

ð10Þ

which indicates that continental runoff can be estimated as a residual of total evapotranspiration (Eo) and precipitation (Po) over the ocean. It could be an effective method to estimate continental runoff since a macroscale estimation of precipitation and evaporation is relatively easier over ocean than over the land. If precise estimates of the long-term trend of global mean precipitation and evapotranspiration over both land and ocean are available, there is a potential to infer the trend of the water stored over the land or ocean as suggested by Equation (9).

Global Hydrology

13

2.01.4 Global Water Balance

From Equation (10)

Pe ¼ Ee

ð11Þ

can be derived which states that precipitation all over the Earth Pe ¼ Pl þ Po and evapotranspiration all over the Earth Ee ¼ Pl þ Po should be identical under the conditions when the temporal changes of water storage over land and ocean are negligible.

The values quoted in Table 1 and Figure 1 are estimated based on various observations with some assumptions in order to obtain global perspectives. These values are sometimes different in other references probably because the source of observed data, methodology of estimation, and assumptions are different. In some cases, global water balances are estimated using empirical relationship of evapotranspiration to precipitation in each latitude belt (Baumgartner and Reichel, 1975).

Zonal mean precipitation (overall, land and sea) 10.0 Annual DJF JJA

(mm d−1)

7.5

5

2.5

0 60° S (a)

40° S

20° S

90° S

EQ.

20° N

40° N

60° N

Latitude

90° N

Vapor-flux convergence (overall, land and sea) 6 Annual DJF JJA

(mm d−1)

3

0

−3

−6 (b)

60° S 90° S

40° S

20° S

EQ. Latitude

20° N

40° N

60° N 90° N

Figure 4 (a) Meridional distribution of precipitation (P ) for mean over land and sea; (b) same as (a) but for vapor-flux convergence (Q ); (c) same as (a) but for evapotranspiration (E ) calculated as a residual of P and Q. Data of (a) from Xie P and Arkin PA (1996) Analyses of global monthly precipitation using gauge observations, satellite estimates, and numerical model predictions. Journal of Climate 9: 840–858 and (b) from Oki T, Musiake K, Matsuyama H, and Masuda K (1995) Global atmospheric water balance and runoff from large river basins. Hydrological Processes 9: 655–678.

14

Global Hydrology Zonal mean evaporation (overall, land and sea) 9 Annual DJF JJA

(mm d−1)

6

3

0

−3 (c)

60° S 90° S

40° S

20° S

EQ. Latitude

20° N

40° N

60° N 90° N

Figure 4 Continued.

Under an international research project, the land-surface models (LSMs) were used to estimate global water and energy balances for 1986–95 in order to obtain global distribution of surface soil moisture, which is not easy to obtain but relevant for understanding the land–atmosphere interactions (Dirmeyer et al., 2006). The project was called the second phase of GSWP2 and its goal was to produce state-of-the-art global data sets of land-surface fluxes, state variables, and related hydrologic quantities.

2.01.4.1 Uncertainties in Global Water-Balance Estimates In GSWP2, meteorological forcing data are hybrid products of the National Center for Environmental Prediction (NCEP)/ Department of Energy (DOE) reanalysis data and observational data based on in-situ and satellite monitoring, provided at a 3-hourly time step for a period of 13.5 years from July 1982 to December 1995. The first 3.5 years’ data are used for spin up. The land-surface parameters are specified from the Earth Resources Observation and Science Data Center (EDC) for the land-cover data and the International Geosphere–Biosphere Programme Data Information System (IGBP-DIS) for the soil data. Both land-surface parameters and meteorological forcing are at 11 resolution for all land grids excluding Antarctica. Figure 6(a) illustrates the model-derived global water balance over the global land excluding ice, glacier, and lake. The numeric in the box corresponds to the 10-year mean annual value of eight LSMs participated in the GSWP2 project (Oki et al., 2005). All the simulations were performed using identical forcing data given by 11 11 longitudinal and latitudinal grid boxes, and typical time steps of the calculations are 5 min to 3 h. The vertical ranges shown above and below the boxes indicate the maximum and minimum values in the interannual variation of mean annual value among the eight LSMs. The horizontal ranges shown left and right to the

boxes indicate the maximum and minimum values of the intermodel variation of the 10-year mean value of eight LSMs. Generally, intermodel variation exceeds interannual variation, which suggests that the uncertainty associated with model selection is larger than the sampling error of estimating global water balance. In the case of rainfall, intermodel variation is small because identical precipitation forcing was given to LSMs so that the differences among LSM estimates were merely caused by the rain/snow judgment made by each modeling group. The advantage of using models to estimate global water balance is the capability to have more detailed insights than using observations. For example, snow over the land excluding ice and glacier areas is approximately 10% of total precipitation, and the ratio of surface runoff and subsurface runoff is approximately 2:3 in Figure 6(a), but the latter is approximately 1:2 in Figure 1. This is because the model results used for calculating average value are different for Figures 1 and 6(a). In some LSMs, neither surface nor subsurface runoff process is considered, which is the reason why the minimum values are zero. Even though at present it is difficult to assess the validity of these breakdowns, due to the lack of observations on the partitions between snow and rain or between surface and subsurface runoff, such model-based estimates will stimulate scientific interest to collect and compile global information on these important hydrological quantities in the future. Further, evapotranspiration was estimated separately by bare-soil evaporation (Es), evaporation from intercepted water on leaves (Ei), evaporation from open water (Ew), and transpiration from vegetation (Et), as shown in Figure 6(b), even though the intermodel variations are quite large partially because some LSMs do not consider all of these components of evapotranspiration. Even though the values in Figure 6(b) are not definitive, it is interesting to see that bare-soil evaporation

Global Hydrology

15

Toward north (mean 1989−92) 40

100

Atmosphere Continent

30

Ocean

10

0

50

Land/sea (%)

Water flux (1012 m3 yr−1)

20

−10 −20 −30 −40

0 60° S 90° S

30° S

EQ. Latitude

30° N

60° N 90° N

Figure 5 The annual freshwater transport in the meridional (north–south) direction by atmosphere, ocean, and rivers (land). Water-vapor flux transport of 20  1012 m3 yr1 corresponds to approximately 1.6  1015 W of latent-heat transport. Shaded bars behind the lines indicate the fraction of land at each latitudinal belt.

and transpiration from vegetation are closely comparative, and interception loss is approximately 10% of the total evapotranspiration. It would be interesting if these estimates can be revised and validated by certain observation-based measures, and intermodel discrepancies can therefore be reduced.

2.01.4.2 Water Balance and Climate Based on observed precipitation (Xie and Arkin, 1996) and river-discharge records archived at the Global Runoff Data Centre, annual water balance for most river basins worldwide was estimated (Oki et al., 1999), and this is presented in Figure 7. The ordinate in Figure 7 is the residual of long-term mean annual precipitation and runoff, and this annual loss should correspond to long-term mean annual evapotranspiration. Different symbols are used for plotting: red stars indicate water balance of the river basins where gauging stations are located between 201 S and 201 N. The plus symbols indicate the river basins where gauging stations are located between 201 and 401 in both hemispheres, and the blue circles are 401 or higher. The line connects the mean precipitation and annual loss for each 51 latitudinal belt. As seen, even though the scatter is large, approximately 70% of precipitation is evapotranspirated in high-latitude river basins. On the other hand, mean evapotranspiration in tropical river basins is approximately 1000 mm yr1 with less dependency on annual precipitation. Such analyses on the relationship between P and E have long been used for estimation of global water balance, for example, in Baumgartner and Reichel (1975). It is also

clear from Figure 7 that river basins with annual precipitation less than 800 mm yr1 have marginal amounts of river runoff since most precipitation is used for evapotranspiration. In these river basins, evapotranspiration is mainly controlled by the availability of water (water controlled), and this is in contrast to tropical river basins with precipitation higher than 1000 mm yr1 where annual evapotranspiration is limited by the available energy (energy- or radiation controlled). Budyko (1974) proposed an equation

E=P ¼ ½zðtanh 1=zÞð1  cosh z þ sinh zÞ ð1=2Þ

ð12Þ

where E and P are annual evapotranspiration and precipitation respectively, and the Budyko’s dryness index is defined as

z ¼ Rn =lP

ð13Þ

where Rn and l are net radiation and the coefficient of latent heat, respectively. This equation is derived by considering that the E/P should be asymptotic to 1.0 for dry regions (large z) since E should be less than P, and E/P should be asymptotic to Rn/lP for wet regions (small z) since E should be less than Rn/l. Budyko’s equation (12) is conceptual, but it can provide a realistic water balance as shown in Figure 8. Mean water balance averaged for each 0.2z( ¼ Rn/lP) bin estimated by an LSM corresponds fairly well with the curve according to the Budyko’s equation (12), even though large scatters are found in the plots of each 11 longitudinal and latitudinal grid box. Yang et al.

16

Global Hydrology

Global terrestrial water budget Unit: mm yr−1 91 58

86

793

102

80

Legend

759

742

Snow

508

414

726

742

574

499

Rainfall

0

Intermodel range

142

302

133

Surface runoff

432

338 308

Total discharge

138 60

303

150

Soil-water storage

149

* Except lake and ice/glacier, 14 409 grids or 1.302 59e + 8 km2

Average of eight models (1986−95)

364 270

152

Subsurface runoff

788

726

302

196

0

759

ET 151

214

Interannual range

793

524

788

Balance = −1 mm

(a)

Global composition of evapotranspiration Unit: mm yr−1

Legend

574

793

Interannual range

524 414

508 499

742

ET Es 0

65

62

137

0

229

544

Average of eight models (1986−95)

248 0 7

238

Intermodel range

Et

59

11

788

726

Ew

Ei

759

6

47

210

531

235

1

223

(b)

Figure 6 (a) Global terrestrial water balance averaged for 1986–95 estimated by eight LSMs in boxes. Interannual variation range (vertical) for 1986–95 and intermodel discrepancies (horizontal) among eight models are presented for the annual mean estimates; (b) same as (a) but for global composition of evapotranspiration.

(2009) analyzed annual water balance in 99 river basins in China and concluded that this scatter can at least partially be explained by the vegetation cover in the river basin.

2.01.4.3 Annual Water Balance in Climatic Regions Annual water balances estimated by GSWP2 were analyzed and each 11 11 grid box was classified into one of the

following six climatic regions according to the Budyko’s dryness index z and annual precipitation:

• • • • •

arid region: z44.0; semiarid region: 4.04z4 ¼ 2.0; semi-humid region: 2.04z4 ¼1.2; humid region: 1.24z4 ¼ 0.7; tropical humid region: 0.74 ¼ z and annual precipitation larger than 2000 mm yr1; and

Global Hydrology

17

Precipitation and annual loss over the world 2000 Low latitude (20°S−20°N) Mid-latitude (20°S−40°S, 20°N−40°N) High latitude (40°S−90°S, 40°N−90°N)

Annual loss (mm yr−1)

1600

1200 5 15 20 25

800

10

0

30 40 45 35 50 55 60 65

400 70 0 0

400

800

1200

1600

Precipitation (mm

2000

2400

2800

yr−1)

Figure 7 Annual water balance in major river basins. Annual loss is estimated as a residual of annual precipitation and observed runoff in catchments of 250 gauging stations of river discharge. From Oki T, Nishimura T, and Dirmeyer P (1999) Assessment of annual runoff from land surface models using total runoff integrating pathways (TRIP). Journal of the Meteorological Society of Japan 77: 235–255.

•

very humid region: 0.74 ¼ z and annual precipitation less than 2000 mm yr1.

The six classified regions are illustrated in Figure 9 along with the ice-covered region. The differentiation is difficult between tropical humid region and very humid region only by using the Budyko’s dryness index z. Therefore, annual precipitation is considered in the classification. It is interesting to see that z is similar in both tropical and high-latitude regions in addition to the Asian monsoon region. Perhaps, it is also necessary to consider the seasonal change of major waterbalance terms for better differentiation of these regions. Long-term mean water balance for each climatic region classified by z is presented in Figure 10(a). Separation of mean annual precipitation into evapotranspiration, surface runoff, and subsurface runoff is also illustrated. The sum of these corresponds to annual precipitation, and the ratio of annual evapotranspiration to precipitation is close to the mean z of each region. Slightly negative evapotranspiration in the ice region indicates net sublimation in the region, that is, the land surface obtains energy from the atmosphere through sublimation. Evapotranspiration is also divided into four components: bare-soil evaporation, transpiration, evaporation from intercepted water, and evaporation from open water, as presented in Figure 10(b). Note that not all the LSMs that participated in GSWP2 have considered all of these four components, and some of these values could be underestimated. Transpiration

and evaporation from intercepted water are proportional to the vegetation biomass in each region, and it is interesting to note that the magnitude of bare-soil evaporation is relatively uniform globally than other components, except for arid, very humid, and ice regions.

2.01.5 Challenges in the Global Hydrology and Research Gaps 2.01.5.1 Macroscale Hydrological Modeling The development of macroscale hydrological models was a serious topic of discussion among Japanese scientists researching land–atmosphere interaction studies in the early 1990s when Global Energy and Water Experiment (GEWEX) Asian Monsoon Experiment (GAME) was under preparation. Two approaches were identified: one to expand a conventional microscale rainfall-runoff hydrological model into a macroscale model, which can run on the continental scale with a detailed energy balance and vegetation representation, and the other to enhance hydrological processes in LSMs and couple them with horizontal water-flow processes, particularly with river flow. A river-routing scheme was hence developed with a globalflow direction map, and named as the total runoff integrating pathways (TRIPs) (Oki and Sud, 1998). Such a river-routing scheme can be coupled with any LSM, and can also be used as a post-processor integrating the runoff estimated by LSMs into

18

Global Hydrology Budyko’s diagram for GSFC (Mosaic) (1987)

E/P; evapotranspiration/precipitation

1.0

0.8

0.6

0.4

0.2

0.0 0.0

0.5

1.0

1.5 2.0 2.5 3.0 3.5 4.0 Rn/lP; net radiation/(latent heat * precipitation)

4.5

5.0

5.5

6.0

Figure 8 Annual energy and water balance of 11 longitude and latitudinal grids calculated by an LSM for 1987 (Koster et al., 1999). Plots indicate the energy and water balances in each grid box and green line presents the mean E/P for each 0.2 z ¼ Rn/lP bin. The red curve indicates Equation (12). From Budyko MI (1974) Climate and Life, Miller DH (trans.). San Diego, CA: Academic Press.

Arid Semiarid Semi-humid Humid Tropical humid Very humid Iee cover Figure 9 Classification of energy and water-balance regime using Budyko’s dryness index z ¼ Rn/lP.

Mean water balance in climatic regions (1986−95)

l ba

e

G lo

Ic

id m

id Ve

ry

hu

um

id

lh

Tr o

pi

Se

m

ca

i- h

H um

r id m

ia

Ar Se

(a)

id

Rsub Rsurf ET

um

3000 2500 2000 1500 1000 500 0 −500

id

(mm y−1)

Global Hydrology

1200 Ew Ei Et Es

1000 800 600

19

seasonal pattern of observed TWS variation by Gravity Recovery and Climate Experiment (GRACE; see Tapley et al., 2004) without an appropriate representation of river-storage component. The dominant role of river storage was already indicated in a pilot study which compared total TWS changes estimated by the atmospheric water-balance method and a GCM simulation coupled with TRIP in the Amazon river basin (Oki et al., 1996). However, the message was not undoubtedly convincing until recent years when satellite-observed GRACE data became available. Using a geodesy approach, Han et al. (2009) employed a fixed-velocity version of TRIP in the Amazon river basin and its vicinity, and compared the model simulations to the residual of GRACE raw measurements derived from removing all the gravity-influencing factors except for the horizontally moving water. They demonstrated that the optimal flow velocity of TRIP in the Amazon varies between rising and falling water levels.

2.01.5.2 Global Changes and Global Hydrology

400 200 l G

lo

ba

e Ic

id hu

m

id ry

ic op

Ve

al

H

hu

um

m

id

id

Tr

(b)

um i-h

m Se

Se

m

ia

Ar

rid

id

0

Figure 10 Components of (a) the surface hydrologic balance, and (b) total evapotranspiration estimated under the second phase of the Global Soil Wetness project. From Dirmeyer PA, Gao XA, Zhao M, Guo ZC, Oki T, and Hanasaki N (2006) GSWP-2 multimodel anlysis and implications for our perception of the land surface. Bulletin of the American Meteorological Society 87: 1381–1397.

river discharge (Oki et al., 1999). The first version of TRIP adopted a primitive fixed-velocity scheme (Miller et al., 1994), while the variable-velocity version was also developed (NgoDuc et al., 2007). TRIP was coupled in some general circulation model (GCM) projections used in the Intergovernmental Panel on Climate Change (IPCC) Assessment Report 4 (AR4) to identify the impact of climate change on hydrological cycles (Falloon and Betts, 2006), and there have been some studies of future assessment on the world water resources and global flood disasters utilizing the TRIP model, as well (Oki and Kanae, 2006; Hirabayashi and Kanae, 2009). Further, Kim et al. (2009) underscored the importance of river component in terrestrial water storage (TWS) variation over global river basins. To reduce simulation uncertainty, ensemble simulations were performed with multiple precipitation data, and a localized Bayesian model averaging technique was applied to TRIP simulation. Figure 11 shows that river storage not only explains different portions of total TWS variations, but also plays different roles in different climatic regions. It is the most dominant water-storage component in wet basins (e.g., the Amazon) in terms of amplitude, and it acts as a buffer which smoothes the seasonal variation of total TWS especially in snow-dominated basins (e.g., the Amur). It signifies that model simulation of TWS may not be able to properly reproduce the amplitude and

Macroscale hydrological models have also been developed in response to the societal expectations for solving current and future world water issues. There has been a concrete demand for the information on how much water resources are available now and what kinds of changes are projected for the future. Conventionally, available freshwater resources are commonly defined as annual runoff estimated by historical river-discharge data or water-balance calculation (Lvovitch, 1973; Baumgartner and Reichel, 1975; Korzun, 1978). Such an approach has been used to provide valuable information on the annual freshwater resources in many countries. Atmospheric water balance using the water-vapor flux-convergence data could be alternatively used to estimate global distribution of runoff owing to the advent of atmospheric reanalysis and data-assimilation system (Oki et al., 1995). Simple analytical water-balance models have been widely used to estimate global-scale available freshwater resources in the world since the beginning of this century (Alcamo et al., 2000; Vo¨ro¨smarty et al., 2000; Do¨ll et al., 2003; Rockstro¨m et al., 2009). Later, LSMs were used to simulate global water cycles (Oki et al., 2001; Dirmeyer et al., 2006). Changes of hydrological cycles during the twenty-first century associated with anticipated climate change are projected (Milly et al., 2005; Nohara et al., 2006), and their impacts on the demands and supplies of global water resources are estimated assuming future climatic and social-change scenarios (Arnell, 2004; Alcamo et al., 2007; Shen et al., 2008). Some of those estimations were calibrated by multiplying an empirical factor for the river basins, where observed river-discharge data are available. However, recent model simulations with advanced climate forcing data can estimate global runoff distribution with adequate accuracy without the need of calibration (Hanasaki et al., 2008a). Changes in extreme river discharge are also of interest now (Hirabayashi et al., 2008). Several recently developed macroscale hydrological models for water-resource assessment also include a reservoir-operation scheme (Haddeland et al., 2006; Hanasaki et al., 2006) in order to simulate the real hydrological cycles, which are significantly influenced by anthropogenic activities and modified from natural hydrological cycles even on the global

20

Global Hydrology (b) Storage anomaly (mm per month) (c)

Discharge (mm per month)

(a)

Relative TWS (mm per month)

40

150 100 50 0 −50 −100 −150

AMAZON

700 600 500 400 300 200 100

40 35 30 25 20 15 10 5

40 30 20 10 0 −10 −20 −30

AMUR

160 140 120 100 80 60 40 20

120 100 80 60

1 2 3 4 5 6 7 8 9 10 11 12

1 2 3 4 5 6 7 8 9 10 11 12 2002

2003

2004

2005

2006

2007

Figure 11 (a) Seasonal variations of gauged discharge (black solid line), discharge routed by TRIP (red solid line) and runoff without routing (gray dashed line). (b) Seasonal variations of GRACE observed TWSA (black solid line), simulated TWSA with river storage (red solid line), simulated TWSA without river storage (gray dashed line), and the major water storage components in TWS. Gray crosses ( þ ), green circles (K), and blue triangles (m) represent the individual storage component of snow water, soil moisture, and river storage, respectively. (c) Interannual variations of relative TWS: the GRACE observation (black dot), simulation with river storage (red solid line), and simulation without river storage (gray dashed line). Each area shaded by blue, gray, and green indicates the portion of river storage, snow water, and soil moisture in the simulated relative TWS, respectively. From Kim H, Yeh P, Oki T, and Kanae S (2009) The role of river storage in the seasonal variation of terrestrial water storage over global river basins. Geophysical Research Letters 36: L17402 (doi:10.1029/2009GL039006).

scale in the Anthropocene (Crutzen, 2002). An integrated water-resources model is further coupled with a crop-growth submodel, which can simulate the timing and quantity of irrigation requirement, and a submodel, which can estimate environmental flow requirement (Hanasaki et al., 2008a). Such an approach is able to assess the balances of water demand and supply on a daily timescale, and a gap in the subannual distribution of water availability and water use can be detected in the Sahel, the Asian monsoon region, and southern Africa, where conventional water-scarcity indices such as the ratio of annual water withdrawal to water availability and available annual water resources per capita (Falkenmark and Rockstro¨m, 2004) cannot properly detect the stringent balance between demand and supply (Hanasaki et al., 2008b).

2.01.5.3 Global Trace of Water Cycles Numerical models can be associated with a scheme tracing the origin and flow path as if tracing the isotopic ratio of water (Yoshimura et al., 2004; Fekete et al., 2006). Such a flow-tracing function of water in the integrated water-resources model (Hanasaki et al., 2008a) considering the sources of water withdrawal from stream flow, medium-size reservoirs, and nonrenewable groundwater, in addition to precipitation to croplands, enabled the assessment of the origin of water producing major crops (Hanasaki et al., 2010). Figure 12(a) illustrates the ratio of blue water to total evapotranspiration during cropping period in irrigated croplands. Here, the blue water is defined as that part of evapotranspiration originating from irrigation, whereas the green water is from precipitation (see Falkenmark and Rockstro¨m, 2004). Figure 12(a) shows a distinctive geographical distribution in the dependence on blue water. In addition, the ratios of the source of blue water for stream flow including the influence of large reservoirs,

medium-size reservoirs, and nonrenewable and nonlocal blue water are shown in Figures 12(b)–12(d). Areas highly dependent on nonrenewable and nonlocal blue water were detected in Pakistan, Bangladesh, western part of India, north and western parts of China, some regions in the Arabian Peninsula, and the western part of the United States through Mexico. Cumulative nonrenewable and nonlocal blue-water withdrawals estimated by the model correspond fairly well with the country statistics of total groundwater withdrawals (Hanasaki, 2009, personal communication), and such an integrated model has the ability to quantify the global virtual water flow (Allan, 1998; Oki and Kanae, 2004) or water footprint Hoekstra and Chapagain, 2007) through major crop consumption (Hanasaki et al., 2010). It is apparent that these achievements illustrate how the framework of global off-line simulation of LSMs, coupled with lateral river-flow model and/or anthropogenic activities, driven by the best-available meteorological forcing data, such as precipitation and downward radiation, is relevant for estimating global energy and water cycles, validating the estimates and sometimes the quality of forcing data with independent observations, and improving the models themselves. There are attempts to utilize this framework for assessing the impacts of climate change on future hydrological cycles which would demand adaptation measures in waterresources management, flood management, and food production. For such purposes, it is necessary to develop reliable forcing data for the future, based on GCM projections probably with bias corrections and spatial and temporal downscaling, as well as developing best estimates for the future boundary conditions for hydrological simulations such as vegetation type and land use/land cover. Figure 13 summarizes the concept of how the forcing data have a large impact on the accuracy of the output from theories, equations, and numerical models. Certainly, the spatial

Blue/(blue+green)

(a)

0

Streamflow/blue

1

(b)

0

Medium-size reservoirs/blue

(c)

0

1 NNBW/blue

1

(d)

0

1

Figure 12 (a) The ratio of blue water to the total evapotranspiration during a cropping period from irrigated cropland (the total of green and blue water). The ratios of (b) streamflow, (c) medium-size reservoirs, and (d) nonrenewable and nonlocal blue-water withdrawals to blue water. From Hanasaki N, Inuzuka T, Kanae S, and Oki T (2010) An estimation of global virtual water flow and sources of water withdrawal for major crops and livestock products using a global hydrological model. Journal of Hydrology 384: 232–244.

22

Global Hydrology

and temporal boundary conditions as well as the field information characterizing the target region, such as land cover and land use, are critically important to obtain reasonable estimates. Therefore, it is recommended to examine the quality of forcing data and boundary conditions along with revising the core theory, principle equations, and model code, or tuning model parameters, for hydrological modeling. It is particularly important for global-scale studies since uncertainties in the forcing data and field information are relatively large compared to those at the local scale. It should also be recalled that the applicability and accuracy of the model are highly dependent on the specific temporal and spatial scales.

Owing to recent advancements in global earth-observation technology and macroscale modeling capacity, global hydrology can now provide basic information on the regional hydrological cycle which may support the decision-making process in the integrated water-resources management. It should also be examined to what extent such a framework of off-line simulation of LSMs can be applied to finer spatial and temporal scales, such as 1-km grid spacing and hourly time interval. For such research efforts, observational data from regional studies can provide significant information, and efforts to integrate data sets from various regional studies should be promoted.

2.01.5.5 Research Opportunities in Global Hydrology 2.01.5.4 Interactions of Global- and Local-Scale Hydrology As described in Oki et al. (2006), water, as one of the major components of the global climate system, is one of the major cross-cutting axes in the Earth system science. The water cycle transports various materials, such as sediments and nutrients, from land to the oceans. Water resources are closely related to energy, industry, and agricultural production. Of course, water is indispensable for life and supports health. Water issues are related to poverty, and providing access to safe drinking water is one of the key necessities for sustainable development. In the past, water issues remained local issues; however, due to the increase in international trade and mutual interdependence among countries, water issues now often need to be dealt with on the global scale, and require information on global hydrology for their solutions. Sharing hydrological information relating to the transboundary river basins and shared aquifers will help reduce conflict between relevant countries, and quantitative estimates of recharge amounts or potentially available water resources will assist in implementing sustainable water use. Global hydrology is not merely concerned with global monitoring, modeling, and world water-resources assessment.

There are still challenging scientific issues to be resolved in global hydrology. For instance, separation of rain and snow on the global scale, as illustrated in Figure 1, is of interest. However, this is based on numerical-model estimates and it is quite uncertain about the accuracy of the numbers or even the ratio between rain and snow (here, approximately 8:1). The situation is similar for the ratio between surface and sub-surface runoff (here, approximately 1:2). It is reasonable to infer direct groundwater discharge from land to oceans to be a residual between total runoff estimated by numerical models subtracted by observation-based total discharge; however, the accuracy of such an estimate is yet to be determined. As described in Section 2.01.1, it is believed that total amount of the water on the Earth is conserved on a timescale shorter than geological timescales; however, do we have any reliable observational evidence for it? Our knowledge seems to be incomplete on the water exchange between Earth’s surface and mantle, although a recent report suggested that the lower mantle may store 5 times more water than the ocean (Murakami et al., 2002). The direct groundwater discharge to the ocean, estimated to be about 10% of total river discharge globally (Church, 1996),

Sources of uncertainties • Quantity • Runoff, ET, … • Quality • Solutes • Isotopic ratio • Sediments, …

Environmental info. Precip., radiation, temp., humidity, wind,… (atmospheric forcing)

Regional info. soil, vegetation, topography, basin area, … (parameters)

Theory, equation, or model

Initial/boundary conditions Applicability and accuracy would differ by temporal and spatial scales, and the characteristics of the target region.

Output with various temporal scales • Annual mean • Daily, hourly, … • Duration curve, … • Extremes, …

Figure 13 Sources and causes of uncertainties and errors in the hydrological simulations and estimations.

Global Hydrology

is included in the river discharge plotted in Figure 1. According to the model estimates by Koirala (2010), groundwater recharge in the grid boxes of 11 11 in longitude and latitude near the coastal line is totally 5890 km3 yr1, and it is approximately 13% of the annual total runoff from the continents to oceans of 45 500 km3 yr1. Even though the total amount of groundwater recharge depends on the grid size defining the coastal region, is it just a coincidence or is it that groundwater recharge in the coastal areas provides a good proxy of direct groundwater runoff from continents to oceans? Scientific interest in global hydrology has increased since the 1980s as public awareness of global environmental issues and process interactions, such as El Nin˜o events and anthropogenic climate change, has risen. Further work is required for the detection and attribution of the present-day hydrological changes – in particular, changes in the intra-seasonal water availability and in the frequency and magnitude of extreme events. Uncertainties in the future projections of hydrological quantities and water qualities should be reduced and quantitative accuracy should be enhanced, particularly for runoff regime changes. It is further anticipated that feedbacks of mitigation and adaptation measures for the concerned climate change on water sectors will be assessed for proper policymaking.

23

In addition to anthropogenic climate change, it is of interest in global hydrology to assess the impact of land-use change, such as deforestation and urbanization, human activities, such as reservoir construction and water withdrawals for irrigation, industry, and domestic water uses, and emission of air pollutants which would have been suppressing weak rainfall and modulate precipitation occurrence weekly.

2.01.5.6 Research Gaps in Global Hydrology Hydro-meteorological monitoring networks need to be maintained and further expanded to enable the analysis of hydro-climatic trends at the local level and the improvement in the accuracy of predictions, forecasts, and early warnings. As clearly illustrated in Figure 14 (from Oki et al., 1999), global hydrological simulations are relatively poor in areas with little in-situ observations. Basic observational networks on the ground are critically indispensable for proper monitoring and modeling of global hydrology; however, it is also required to utilize remotely sensed information in order to fill the gaps of in-situ observations. One of the current trends in the utilization of remote sensing technique is the so-called data

Runoff estimation error (mean 11 LSMs) and the density of raingauges in each river Basin Y1987 600

Y1988

Density of raingauges (/106km2)

500

400

300

200

100

Before routing

0 −2000

−1500

−1000

−500

0

500

1000

1500

2000

Runof (model observation) (mm yr−1) Figure 14 Comparisons between the density of rain gauge (/106 km2) used in preparing the forcing precipitation and the mean bias error (mm yr1) of 11 LSMs for 150 major river basins in the world in 1987 and 1988. From Oki T, Nishimura T, and Dirmeyer P (1999) Assessment of annual runoff from land surface models using total runoff integrating pathways (TRIP). Journal of the Meteorological Society of Japan 77: 235–255.

24

Global Hydrology

assimilation, which optimally merges observations (not limited to remote sensing) with numerical-model estimates. Reliable observational data are essentially necessary not only as the forcing data for global hydrological modeling, but also for the validation of model estimates. River discharge and soil-moisture data are critically important for global hydrological studies. However, contributions from the operational agencies in the world are not yet well established and need to be enhanced. Isotopic ratio of rainwater is collected and the data set is available through the Global Network of Isotopes in Precipitation (GNIP) by the International Atomic Energy Agency (IAEA). These observational data can be used to investigate the routes and mechanisms of how the evaporated waters from the ocean surface are transported and precipitated at particular locations, which can be estimated by water-vapor transfer models with the consideration of isotopic processes (Yoshimura et al., 2008). The information on the isotopic ratio of river waters is not well organized and cannot be used easily on the global scale. Current global hydrological modeling has not yet integrated most of the latest achievements in process understanding and regional- or local-scale modeling studies. Global simulation of solutes and sediments are emerging. Both natural and anthropogenic sources should be considered, as for nutrients, and probably such models should be coupled with agricultural models which simulate crop growth. Moreover, it is rather difficult that river ice jams can be simulated properly by current hydrologic models. For both problems of water quality and ice jams, a proper simulation of water temperature in rivers and lakes are requisite. Moreover, the representation of groundwater has been rather simple in global hydrological modeling. Some of the above issues have not been emphasized in the current global hydrological modeling due to their relatively minor impact on the climatic feedbacks from the land surface to the atmosphere. It is meaningful to recall that global hydrology has been developed in cooperation with global climate modeling; however, it is time to develop global hydrological models primarily for responding to the demands of understanding the hydrological cycle on the land surface and for supporting better water-resources management. From this point of view, integrated hydrological and water-resource models, which consider natural and anthropogenic water cycles and are coupled with crop models and reservoir operation models in order to provide a more realistic impact assessment and support the design of practical adaptation measures, should be developed and implemented.

Acknowledgments This study was funded by Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (19106008 4), the Global Environment Research Fund of the Ministry of Environment (S-5), and the Innovation Program of Climate Change Projection for 21st Century of the Ministry of Education, Culture, Sports, Science, and Technology of Japan.

References Alcamo J, Floerke M, and Maerker M (2007) Future long-term changes in global water resources driven by socio-economic and climatic changes. Hydrological Sciences Journal 52(2): 247--275. Alcamo J, Henrichs T, and Ro¨sch T (2000) World water in 2025 – global modeling and scenario analysis for the World Commission on water for the 21st century. Kassel World Water Series, Technical Report No. 2. Kassel, Germany: Centre for Environmental Systems Research, University of Kassel. Allan JA (1998) Virtual water: A strategic resource. Global solution to regional deficits. Groundwater 36(4): 545--546. Arnell NW (2004) Climate change and global water resources: SRES scenarios and socio-economic scenarios. Global Environmental Change 14: 31--52. Baumgartner A and Reichel E (1975) The World Water Balance: Mean Annual Global, Continental and Maritime Precipitation, Evaporation and Runoff. 179pp. Munich: Ordenbourg. Bryan F and Oort A (1984) Seasonal variation of the global water balance based on aerological data. Journal of Geophysical Research 89: 11717--11730. Budyko MI (1974) Climate and Life. Miller DH (trans.). San Diego, CA: Academic Press. Chapman TG (1972) Estimating the frequency distribution of hydrologic residence time. In: World Water Balance, vol. 1, pp. 136–152. IASH-UNESCO-WMO. Church TM (1996) An underground route for the water cycle. Nature 380: 579--580. Crutzen PJ (2002) Geology of mankind-the Anthropocene. Nature 415(23). doi:10.1038/415023a Dingman SL (2002) Physical Hydrology, 2nd edn. Upper Saddle River, NJ: PrenticeHall. 646pp Dirmeyer PA, Gao XA, Zhao M, Guo ZC, Oki T, and Hanasaki N (2006) GSWP-2 multimodel anlysis and implications for our perception of the land surface. Bulletin of the American Meteorological Society 87: 1381--1397. Do¨ll P and Fiedler K (2008) Global-scale modeling of groundwater recharge. Hydrology and Earth System Sciences 12: 863--865. Do¨ll P, Kaspar F, and Lehner B (2003) A global hydrological model for deriving water availability indicators: Model tuning and validation. Journal of Hydrology 270(1): 105--134. Falkenmark M and Rockstro¨m J (2004) Balancing Water for Humans and Nature. 247pp. London: Earthscan. Falloon PD and Betts RA (2006) The impact of climate change on global river flow in HadGEM1 simulations. Atmospheric Science Letters 7: 62--68. Fekete BM, Gibson JA, Aggarwal P, and Vo¨ro¨smarty CJ (2006) Application of isotope tracers in continental scale hydrological modeling. Journal of Hydrology 330: 444--456. Haddeland I, Lettenmaier DP, and Skaugen T (2006) Effects of irrigation on the water and energy balances of the Colorado and Mekong river basins. Journal of Hydrology 324: 210--223. Han S-C, Kim H, Yeo I-Y, et al. (2009) Dynamics of surface water storage in the Amazon inferred from measurements of inter-satellite distance change. Geophysical Research Letters 36: L09403 (doi:10.1029/2009GL037910) Hanasaki N, Inuzuka T, Kanae S, and Oki T (2010) An estimation of global virtual water flow and sources of water withdrawal for major crops and livestock products using a global hydrological model. Journal of Hydrology 384: 232--244. Hanasaki N, Kanae S, and Oki T (2006) A reservoir operation scheme for global river routing models. Journal of Hydrology 327: 22--41. Hanasaki N, Kanae S, Oki T, et al. (2008a) An integrated model for the assessment of global water resources – part 1: Model description and input meteorological forcing. Hydrology and Earth System Sciences 12: 1007--1025. Hanasaki N, Kanae S, Oki T, et al. (2008b) An integrated model for the assessment of global water resources – part 2: Applications and assessments. Hydrology and Earth System Sciences 12: 1027--1037. Hirabayashi Y and Kanae S (2009) First estimate of the future global population at risk of flooding. Hydrological Research Letters 3: 6--9. Hirabayashi Y, Kanae S, Emori S, Oki T, and Kimoto M (2008) Global projections of changing risks of floods and droughts in a changing climate. Hydrological Sciences Journal 53(4): 754--772. Hoekstra AY and Chapagain AK (2007) Water footprints of nations: Water use by people as a function of their consumption pattern. Water Resource Management 21: 35--48. Kim H, Yeh P, Oki T, and Kanae S (2009) The role of river storage in the seasonal variation of terrestrial water storage over global river basins. Geophysical Research Letters 36: L17402 (doi:10.1029/2009GL039006) Koirala S (2010) Explicit Representation of Groundwater Process in a Global-Scale Land Surface Model to Improve Hydrological Predictions. PhD Thesis, Graduate School of Engineering, The University of Tokyo, Tokyo, Japan.

Global Hydrology

Korzun VI (1978) World Water Balance and Water Resources of the Earth – Studies and Reports in Hydrology, 25, Paris: UNESCO. Koster RD, Oki T, and Suarez MJ (1999) Assessing success in the offline validation of land surface models. Journal of the Meteorological Society of Japan 77: 257--263. Kubota T, Shige S, Hashizume H, et al. (2007) Global precipitation map using satelliteborne microwave radiometers by the GSMaP project. IEEE Transactions on Geoscience and Remote Sensing 45(7): 2259--2275. Lvovitch MI (1973) The global water balance. Transactions – American Geophysical Union 54: 28--42. Masuda K (1988) Meridional heat transport by the atmosphere and the ocean: Analysis of FGGE data. Tellus 40A: 285--302. Miller JR, Russell GL, and Caliri G (1994) Continental-scale river flow in climate models. Journal of Climate 7: 914--928. Milly PCD, Dunne KA, and Vecchia AV (2005) Global pattern of trends in streamflow and water availability in a changing climate. Nature 483: 347--350. Murakami M, Hirose K, Yurimoto H, Nakashima S, and Takafuji N (2002) Water in Earth’s lower mantle. Science 295(5561): 1885--1887. Ngo-Duc T, Oki T, and Kanae S (2007) A variable streamflow velocity method for global river routing model: Model description and preliminary results. Hydrology and Earth System Sciences Discussions 4: 4389--4414. Nohara D, Kitoh A, Hosaka M, and Oki T (2006) Impact of climate change on river discharge projected by multi-model ensemble. Journal of Hydrometeorology 7: 1076--1089. Oki T (1999) The global water cycle. In: Browning K and Gurney R (eds.) Global Energy and Water Cycles, pp. 10--27. Cambridge: Cambridge University Press. Oki T, Agata Y, Kanae S, Saruhashi T, Yang DW, and Musiake K (2001) Global assessment of current water resources using total runoff integrating pathways. Hydrological Sciences Journal 46: 983--995. Oki T, Hanasaki N, Shen Y, Kanae S, Masuda K, and Dirmeyer PA (2005) Global water balance estimated by land surface models participated in the GSWP2. In: The 19th Conference on Hydrology, AMS Annual Meeting. San Diego, CA, USA, January 2005. Oki T and Kanae S (2004) Virtual water trade and world water resources. Water Science and Technology 49(7): 203--209. Oki T and Kanae S (2006) Global hydrological cycles and world water resources. Science 313(5790): 1068--1072. Oki T, Kanae S, and Musiake K (1996) River routing in the global water cycle, GEWEX News, WCRP, International GEWEX Project Office, 6: 4–5. Oki T, Musiake K, Matsuyama H, and Masuda K (1995) Global atmospheric water balance and runoff from large river basins. Hydrological Processes 9: 655--678. Oki T, Nishimura T, and Dirmeyer P (1999) Assessment of annual runoff from land surface models using total runoff integrating pathways (TRIP). Journal of the Meteorological Society of Japan 77: 235--255. Oki T and Sud YC (1998) Design of total runoff integrating pathways (TRIP) – a global river channel network. Earth Interactions 2: 1--37. Oki T., Valeo C., and Heal K. (eds.) (2006) Hydrology 2020, ISSN 0144-7815. Wallingford: IAHS. Peixo¨to JP and Oort AH (1992) Physics of Climate. 520pp. New York: American Institute of Physics. Rockstro¨m J, Falkenmark M, Karlberg L, Hoff H, Rost S, and Gerten D (2009) Future water availability for global food production: The potential of green water for

25

increasing resilience to global change. Water Resources Research 45: W00A12 (doi:10.1029/2007WR006767) Rossow WB, Walker AW, and Garder LC (1993) Comparison of ISCCP and other cloud amounts. Journal of Climate 6: 2394--2418. Shen Y, Oki T, Utsumi N, Kanae S, and Hanasaki N (2008) Projection of future world water resources under SRES scenarios: Water withdrawal. Hydrological Sciences Journal 53(1): 11--33. Shiklomanov IA (ed.) (1997) Assessment of water resources and water availability in the world. Background Report for the Comprehensive Assessment of the Freshwater Resources of the World. Geneva: WMO/SEI. Starr VP and Peixo¨to J (1958) On the global balance of water vapor and the hydrology of deserts. Tellus 10: 189--194. Takata K, Emori S, and Watanabe T (2003) Development of minimal advanced treatments of surface interaction and runoff. Global and Planetary Change 38: 209--222. Tapley BD, Bettadpur S, Ries JC, Thompson RF, and Watkins MM (2004) GRACE measurements of mass variability in the Earth system. Science 305(5683): 503--505. Trenberth KE and Guillemot CJ (1998) Evaluation of the atmospheric moisture and hydrological cycle in the NCEP/NCAR reanalyses. Climate Dynamics 14: 213--231. Trenberth KE and Solomon A (1994) The global heat balance: Heat transports in the atmosphere and ocean. Climate Dynamics 10: 107--134. Uhlenbrook S, Frey M, Leibundgut Ch, and Maloszewski P (2002) Hydrograph separations in a mesoscale mountainous basin at event and seasonal timescales. Water Resources Research 38(1096). doi:10.1029/2001WR000398 Uhlenbrook S, Roser S, and Tilch N (2004) Development of a distributed, but conceptual catchment model to represent hydrological processes adequately at the meso scale. Journal of Hydrology 291: 278--296. Vo¨ro¨smarty CJ, Green P, Salisbury J, and Lammers RB (2000) Global water resources: Vulnerability from climate change and population growth. Science 289: 284--288. Wijffels SE, Schmitt RW, Bryden HL, and Stigebrandt A (1992) Transport of freshwater by the oceans. Journal of Physical Oceanography 22: 155--162. Xie P and Arkin PA (1996) Analyses of global monthly precipitation using gauge observations, satellite estimates, and numerical model predictions. Journal of Climate 9: 840--858. Yang D, Shao W, Yeh PJ-F, Yang H, Kanae S, and Oki T (2009) Impact of vegetation coverage on regional water balance in the nonhumid regions of China. Water Resources Research 45: W00A14. Yeh PJ-F and Eltahir EAB (2005) Representation of water table dynamics in a land surface scheme. Part I: Model development. Journal of Climate 18: 1861--1880. Yoshimura K, Kanamitsu M, Noone D, and Oki T (2008) Historical isotope simulation using reanalysis atmospheric data. Journal of Geophysical Research 113: doi:1029.10/2008JD010074 Yoshimura K, Oki T, Ohte N, and Kanae S (2004) Colored moisture analysis estimates of variations in 1998 Asian monsoon water sources. Journal of the Meteorological Society of Japan 82: 1315--1329.

2.02 Precipitation D Koutsoyiannis, National Technical University of Athens, Athens, Greece A Langousis, Massachusetts Institute of Technology (MIT), Cambridge, MA, USA & 2011 Elsevier B.V. All rights reserved.

2.02.1 2.02.1.1 2.02.1.2 2.02.1.3 2.02.1.4 2.02.1.5 2.02.1.5.1 2.02.1.5.2 2.02.1.5.3 2.02.1.5.4 2.02.1.5.5 2.02.2 2.02.2.1 2.02.2.2 2.02.2.3 2.02.2.3.1 2.02.2.3.2 2.02.2.4 2.02.2.4.1 2.02.2.4.2 2.02.2.5 2.02.2.5.1 2.02.2.5.2 2.02.2.5.3 2.02.2.5.4 2.02.2.5.5 2.02.2.5.6 2.02.2.5.7 2.02.3 2.02.3.1 2.02.3.1.1 2.02.3.1.2 2.02.3.1.3 2.02.3.2 2.02.3.2.1 2.02.3.2.2 2.02.3.3 2.02.3.3.1 2.02.3.3.2 2.02.3.3.3 2.02.4 2.02.4.1 2.02.4.2 2.02.4.3 2.02.4.4 2.02.4.5 2.02.5 2.02.5.1 2.02.5.2 2.02.5.3 References

Introduction The Entrancement of Precipitation Forms of Precipitation Precipitation Metrics The Enormous Variability of Precipitation Probability and Stochastic Processes as Tools for Understanding and Modeling Precipitation Basic concepts of probability Stochastic processes Stationarity Ergodicity Some characteristic stochastic properties of precipitation Physical and Meteorological Framework Basics of Moist Air Thermodynamics Formation and Growth of Precipitation Particles Properties of Precipitation Particles Terminal velocity Size distribution Clouds and Precipitation Types Cumulus cloud systems Stratus cloud systems Precipitation-Generating Weather Systems Fronts Mechanical lifting and orographic precipitation Extratropical cyclones Isolated extratropical convective storms Extratropical squall lines and rainbands Monsoons Tropical cyclones Precipitation Observation and Measurement Point Measurement of Precipitation Measuring devices Typical processing of rain gauge data Interpolation and integration of rainfall fields Radar Estimates of Precipitation Basics of radar observation and measurement Radar observation of distributed targets and the estimation of precipitation Spaceborne Estimates of Precipitation The IR signature of cloud tops The visible reflectivity of clouds The microwave signature of precipitation Precipitation modeling Rainfall Occurrence Rainfall Quantity Space–Time Models Rainfall Disaggregation and Downscaling Multifractal Models Precipitation and Engineering Design Probabilistic versus Deterministic Design Tools Extreme Rainfall Distribution Ombrian Relationships

27 27 27 28 29 30 32 33 35 37 37 39 42 43 44 44 45 45 46 46 46 46 46 47 47 48 48 50 52 52 52 53 54 57 57 58 59 60 60 61 61 61 62 64 65 66 68 68 69 71 71

27

28

Precipitation

2.02.1 Introduction 2.02.1.1 The Entrancement of Precipitation Precipitation and its related phenomena, such as cloud formation and movement, thunders, and rainbow, are spectacular (Figure 1) due to their huge diversity and complexity. This complexity makes them difficult to comprehend, model, and predict. Hence, it is understandable that ancient civilizations explained these phenomena in a hyperphysical manner, assuming that deities were responsible for their creation. For example, in Greek mythology, some of the phenomena were deified (e.g., Iris is the name of a goddess as well as of the rainbow), whereas the most impressive among them, thunders in particular, were attributed to the action of the King of the Gods, Zeus (Jupiter in Roman mythology; similar deities are Indra in Hinduism, Thor in Norse, etc.). Demystification of these processes and formation of the physical concept of the hydrological cycle was closely related to the birth of science, by the turn of the seventh century BC. While the hydrological cycle was founded as a concept in the sixth century BC by

Anaximander, Anaximenes, and Xenophanes, and was later advanced by Aristotle (Koutsoyiannis et al., 2007), certain aspects related to precipitation can be understood only within the frame of modern science. The fact that a solid or liquid hydrometeor resists gravity and remains suspended in the atmosphere in a cloud is counterintuitive, and needs advanced knowledge of physics, fluid dynamics, and statistical thermodynamics to be understood and modeled. The complexity of the processes involved in precipitation and their enormous sensitivity to the initial conditions (where tiny initial differences produce great differences in the final phenomena), retain, to this day, some of the ancient mythical and magical magnificence of the societal perception of precipitation. People still believe in hyperphysical interventions in matters concerning precipitation. As put by Poincare´ (1908), father of the notion of chaos: Why do the rains, the tempests themselves seem to us to come by chance, so that many persons find it quite natural to pray for rain or shine, when they would think it ridiculous to pray for an eclipse?

Figure 1 Precipitation and related phenomena (from upper-left to lower-right): Monsoon rainfall (Pune, India, September 2009; photo by D Koutsoyiannis); snowy mountainous landscape (Mesounta, Greece, December 2008; from http://www.mesounta.gr/mesounta/ist_eik1/ 07_xion_03.htm); thunder (Athens, Greece, November 2005; from the photo gallery of Kostas Mafounis); rainbow (Mystras, Greece, April 2008, from laspistasteria.wordpress.com/2008/04/08/rainbow-3/).

Precipitation

Amazingly, however, and at the very same time, there is little disbelief in some climate modelers’ prophecies (or outputs of global circulation models (GCMs)) of the precipitation regimes over the globe in the next 100 years or more. This indicates an interesting conflict between perceptions of precipitation – that it is so unstable, uncertain, and unpredictable that prayers are needed to invoke precipitation, and that for some scientists, the future evolution of precipitation on Earth is still predictable in the long term. The latter belief concerns not only the general public, but also the scientific community. For example, a Google Scholar search with either of the keywords ‘precipitation’ or ‘rainfall’, plus the keywords ‘climate change’ and ‘GCM’, locates 21 700 publications (as of August 2009), of which about 200 have been cited 100 times or more. This huge list of results appears despite the fact that climate modelers themselves admit to the performance of their models being low, as far as precipitation is concerned (Randall et al., 2007). An independent study by Koutsoyiannis et al. (2008), which compares model results for the twentieth century with historical time series, has shown that the models are not credible at local scales and do not provide any basis for assessment of future conditions. These findings demonstrate that, even today, the perception of precipitation, not only by the general public, but even by scientists specialized in the study of precipitation, meteorologists, climatologists, and hydrologists, continues to be contradictory, problematic, and, in some sense, mysterious.

2.02.1.2 Forms of Precipitation Precipitation occurs in a number of forms, either liquid or solid, or even mixed (sleet). Liquid precipitation includes rainfall and drizzle, where the former is the most common and most significant, and the latter is characterized by much smaller drop sizes and lighter intensity. Dew is another liquid form, formed by condensation of water vapor (mostly at night) on cold surfaces (e.g., on tree leaves). Most important among the solid forms of precipitation are snow and hail. At high latitudes or at high altitudes, snow is the predominant form of precipitation. Snowfall may occur when the temperature is low and snow accumulates on the ground until the temperature rises sufficiently for it to melt. On the other hand, hail may fall in relatively high temperature and usually melts rapidly. While hailstones are amorphous and usually large (one to several centimeters in diameter), snowflakes are symmetrical and visually appealing with a tremendous variety of shapes, so that no two snowflakes are the same. Occult precipitation is induced when clouds or fog is formed in forested areas, and it includes liquid (fog drip) and solid (rime) forms. Fog drip occurs when water droplets are deposited on vegetative surfaces, and the water drips to the ground. Rime is formed when supercooled air masses encounter exposed objects, such as trees, that provide nucleation sites (see Sections 2.02.2.1 and 2.02.2.2) for formation and buildup of ice, much of which may fall to the ground in solid or liquid form. In some places (e.g., in humid forested areas), precipitation of this type may reach significant amounts; for example, rime constitutes about 30% of the annual precipitation in a Douglas fir forest in Oregon (Harr, 1982; Dingman,

29

1994) and about 30% of total precipitation in fir-forested mountainous areas of Greece (Baloutsos et al., 2005), and it is the sole precipitation type on the rainless coast of Peru (Lull, 1964; Dingman, 1994).

2.02.1.3 Precipitation Metrics The principal metric of precipitation is the rainfall depth h (commonly expressed in millimeters) that falls at a specified point in a specified period of time t; this can be easily perceived and measured by a bucket exposed to precipitation. A derivative quantity is the precipitation intensity

i :¼

dh dt

ð1Þ

with units of length divided by time (typically mm h1, mm d1, and mm yr1). Since it cannot be measured directly (at an instantaneous time basis), it is typically approximated as

i¼

Dh Dt

ð2Þ

where Dh is the change of the depth in a finite time interval Dt. The intensity derived from Equation (2) is a time-averaged value – but at a point basis. Spatial averaging at various scales is always very useful as can be seen in Section 2.02.1.4. This averaging needs precipitation measurement at several points, followed by appropriate numerical integration methods (see Section 2.02.3.1). While traditional precipitation-measurement networks are sparse, thus making the estimation of areal precipitation uncertain, in recent decades, new measurement techniques have been developed implementing radar and satellite technologies (Sections 2.02.3.2 and 2.02.3.3). These provide a detailed description of the spatial distribution of precipitation, thus enabling a more accurate estimation. The latter techniques inherently involve the study of other metrics of precipitation such as the distribution of the size, velocity, and kinetic energy of the precipitation particles, and the socalled radar reflectivity (Section 2.02.3.2). Furthermore, the quantitative description of the processes related to the fall, accumulation, and melting of snow involves a number of additional metrics, such as the snowfall depth (new snow falling), the snowcover depth or snowpack depth (the depth of snow accumulated at a certain point at a particular time), the snow density rs, and the water equivalent of snowfall or of snowpack, defined as h ¼ h0 rs/rw, where h0 is the snowfall or the snowcover depth, and rw ¼ 1000 kg m3 is the liquid water density. Typical values of rs for snowfall range between 0.07 and 0.15rw (e.g., Dingman, 1994) but a commonly used value is rs ¼ 0.1rw ¼ 100 kg m3. For this value, a snowfall depth of say, 10 cm, corresponds to a precipitation water equivalent of 10 mm. The density of snowpack is generally larger than 0.1rw (because of compaction due to gravity and other mechanisms) and depends on the elapsed time and snowpack depth. After a few days, it is about 0.2rw, whereas after some months it may become about 0.4rw.

30

Precipitation

2.02.1.4 The Enormous Variability of Precipitation The different phases and forms of precipitation, and the different shapes of precipitation particles (drops, flakes, and hailstones) are just a first indicator of the great diversity of the precipitation phenomena. At a macroscopic level and in quantitative terms, this diversity is expressed by the enormous variability of the precipitation process, in space and time, at all spatial and temporal scales. Intermittency is one of the aspects of variability, but even in areas or time periods in which precipitation is nonzero, the precipitation depth or average intensity is highly variable. Figure 2 shows the spatial variability of precipitation over the globe in mm d1 at the climatic scale (average for the 30year period 1979–2008) and at an annual scale (average for the year 2006), based mostly on satellite data (see Figure 2 caption and Section 2.02.3.3). While the average precipitation rate over the globe and over the specified 30-year period is 2.67 mm d1 or 977 mm yr1, we observe huge differences in different areas of the globe. In some areas, mostly in tropical seas and in equatorial areas of South America and Indonesia, this rate exceeds 10 mm d1 or 3.65 m yr1. On the other hand, in large areas in the subtropics, where climate is dominated by semi-permanent anticyclones, precipitation is lower than 1 mm d1 or 365 mm yr1. Significant portions of these areas in Africa, Australia, and America are deserts, where the average precipitation is much lower than 1 mm d1. In addition, in polar regions, where the available atmospheric moisture content is very low due to low temperature (see Section 2.02.2.1 and Figure 14), the amounts of precipitation are very small or even zero. For example, it is believed that certain dry valleys in the interior of Antarctica have not received any precipitation during the last 2 million years (Uijlenhoet, 2008). Figure 3 depicts the zonal precipitation profile and shows that the climatic precipitation rate at an annual basis is highest at a latitude of 51 N, reaching almost 2000 mm yr1 and has a second peak of about 1500 mm yr1 at 51 S. Around the Tropics of Cancer and Capricorn, at 23.41 N and S, respectively, the rainfall rate displays troughs of about 600 mm yr1, whereas at mid-latitudes, between 351 and 601 both N and S, rainfall increases again and remains fairly constant, close to the global average of 977 mm yr1. Then, toward the poles, it decreases to about 100 mm in Antarctica and slightly more, to 150 mm yr1, in the Arctic. Figure 3 also shows monthly climatic profiles for the months of January and July. It can be seen that the rainfall conditions for the 2 months are quite different, with the largest differences appearing at about 151 N and S and the smallest at about 301 N and S. Below 301 in the Northern Hemisphere, as well as above the Arctic Circle (66.61), rainfall is higher during summer (July) than during winter (January), but at mid-latitudes, this relationship is reversed. Similar conditions are met in the Southern Hemisphere (where January and July are summer and winter months, respectively). In both Figures 2 and 3, apart from climatic averages, the specific values for a certain year, namely 2006, are also shown. We observe that there are differences in the climatic values, manifesting temporal variability over the different years. This variability seems to be lower in comparison to the spatial

variability over the globe, as well as to the seasonal variability reflected in the profiles of different months. However, while the spatial variability over the globe and the seasonal variability are well comprehended and roughly explainable in terms of basic physical and astronomical knowledge (i.e., solar radiation, relationship of temperature and atmospheric moisture content, and motion of Earth), in other words, they are regular, the interannual variability is irregular, and difficult or even impossible to predict. Such irregular variability appears at finer timescales as well as at finer spatial scales. In fact, as easily understood from elementary statistics, as the spatial and/or temporal scale becomes finer, the variability increases. Figure 4 demonstrates how the variability of the spatial distribution of rainfall at a monthly temporal scale (January 2006) increases when the spatial scale decreases from 2.51  2.51 (upper panel) to 0.251  0.251 (lower panel). Clearly, the areas of equal rainfall amount (including areas of negligible rainfall, i.e., o1 mm d1E0.04 mm h1), which are smooth in the upper panel become rough and erratic in the lower panel. Moreover, the maximum observed rainfall is 21 mm d1 (monthly 651 mm) in the upper panel and 1.2 mm h1 (monthly 893 mm) in the lower panel. Figure 5 demonstrates the increasing variability with the decreasing timescale. Specifically, it depicts how the image of the rainfall distribution changes at a daily scale (9 January 2006) and at a sub-daily scale, at 3-hourly intervals of the same day. The differences between Figure 5 and Figure 4 are prominent. Especially at the 3-hourly scale, a vast part of the globe receives no rainfall, and the part that receives rainfall, it is irregularly distributed, yet not showing a totally random pattern. The maximum observed rate during this 3-hourly interval is 22 mm h1, about 18 times higher than the maximum rate at the monthly scale shown in Figure 4. The lowest panels of Figure 5 provide a zoom-in over the area lying between 91 N–51 S and 78–921 E, which is located in the Indian Ocean, south-east of Sri Lanka. This area received a large amount of rainfall on this particular day, with a rate that is nonuniform in space and time. Figures 6–8 focus on the temporal variability of precipitation. Figure 6 depicts the monthly and annual variation of the average precipitation over the globe. We can see that at both scales, the variability is remarkable. Thus, the annual precipitation in the last 30 years has varied between 957 and 996 mm and obviously much higher variation should have occurred in the past – but data of this type covering the entire globe do not exist for earlier periods. However, we can get an idea of earlier variation using rain gauge data (see Section 2.02.3.1) at certain locations. Perhaps, the oldest systematic observations of rainfall quantity in the world were made in Korea, in the fifteenth century. Rainfall records for the city of Seoul (37.571 N, 126.971 E, and 85 m) exist for the period since 1770, and are considered to be reliable (Arakawa, 1956; Wang et al., 2006, 2007). The recorded annual rainfall in Seoul is plotted in Figure 7 along with the running climatic averages at 10-year and 30-year timescales. The data are now available at a monthly scale from the climatic database of the Dutch Royal Netherlands Meteorological Institute (KNMI), while the monthly data for 1770–1907 appear also in Arakawa (1956).

Precipitation

31

90N

60N

30N

EQ

30S

60S

90S 180

120W

60W

0

60E

120E

180

90N

60N

30N

EQ

30S

60S

90S

1

2

3

4

5

6

7

8

9

10

11

12

1

Figure 2 Precipitation distribution over the globe in mm d , (upper) at a climatic scale (average for the 30-year period 1979–2008) and (lower) at an annual scale (average for year 2006). Data and image generation due to the Global Precipitation Climatology Project (GPCP) made available by NASA at http://disc2.nascom.nasa.gov/Giovanni/tovas/rain.GPCP.2.shtml; resolution 2.51  2.51.

Comparisons show that the two time series are generally consistent, but not identical. The more modern data series has a few missing values, which generally correspond to high values of the older version (and it has been common practice in hydrometeorological data processing to delete very high

values or outliers, which are regarded suspect, see Section 2.02.3.1.2). In the time series plotted in Figure 7, these gaps have been filled in using the values of the older time series, and a few other missing values have been filled in with the average of the four nearest monthly values of the same month

32

Precipitation Precipitation (mm yr−1) 0

500

1000

1500

2000

90

3000

3500

45

Annual, average 1979−2008 Global, average 1979−2008 Annual, 2006 January, average 1979−2008 July, average 1979−2008 January 2006 July 2006

30

Tropic of Cancer

75

Arctic Circle

60

Latitude (°)

2500

15 0 −15

Tropic of Capricorn

−30 −45 −60

Antarctic Circle

−75 −90

0

1

2

3

4

5

6

7

8

9

Precipitation (mm d−1) Figure 3 Zonal precipitation profile: precipitation averaged over all longitudes for different latitudes varying from 901N to 901S (–901N); data from GPCP.

(see justification in Section 2.02.3.1.3). The plot shows that during the 238 years of record, the annual rainfall varied between 634 and 3057 mm and the climatic 30-year average varied between 1139 and 1775 mm. These figures indicate a huge variability: the maximum observed annual rainfall is almost 5 times greater than the minimum and the maximum 30-year climatic rainfall is 55% higher than the minimum. Such observed changes underscore the ever-changing character of climate, and render future changes of precipitation predicted by climate modelers (which typically vary within 10– 20%; compare Fig. 10.12, upper left panel, in Meehl et al., 2007, with Figure 2 herein) to be unrealistically low and too unsafe to support planning. Figure 7 also includes a plot of another long time series, for Charleston City, USA (32.791 N, 79.941 W, and 3 m); the record begins in 1835. This time series is also available at the KNMI database, and a few missing monthly values have been filled in by the average of the four nearest monthly values of the same month. Here, the annual rainfall varied between 602 and 1992 mm (3.3 times higher than minimum) and the climatic 30-year average varied between 1135 and 1425 mm (25% higher than minimum). Finally, Figure 8 depicts the time series of a storm measured at unusually high temporal resolution, that is, 10 s. This storm, with duration 96 790 s or about 27 h starting at 199002-12T17:03:39, is one of several storms that were measured at the University of Iowa using devices that support high

sampling rates (Georgakakos et al., 1994). Figure 8 also includes plots at 5-min and hourly timescales. The minimum intensity was virtually zero at all three scales, whereas the maximum rainfall intensity was 118.7, 38.9, and 18.1 mm h1 at timescales of 10 s, 5 min, and 1 h, respectively. As the mean intensity during the storm is 3.89 mm h1, these maximum values are 30, 10, and 4.6 times higher than the mean. This example highlights the spectacular variability of rainfall, particularly at fine timescales (see also Uijlenhoet and SempereTorres, 2006). As the total rainfall amount of this storm event only slightly exceeds 100 mm, it could be thought of as a rather modest event. Storms with amounts much higher than this are often recorded even in semi-dry climates and, obviously, the variability of rainfall intensity during such storms is even higher.

2.02.1.5 Probability and Stochastic Processes as Tools for Understanding and Modeling Precipitation The high variability and the rough and irregular patterns in observed fields and time series are much more prominent in precipitation than in other meteorological variables such as atmospheric pressure or temperature. High variability implies high uncertainty and, unavoidably, this affects predictability in deterministic terms. Considering weather prediction as an example, it is well known that the forecasts of atmospheric pressure and temperature are much more reliable than those

Precipitation

33

50N 40N 30N 20N 10N EQ 10S 20S 30S 40S 50S 180

120W

60W

0

60E

120E

180

120W

60W

0

60E

120E

180

50N 40N 30N 20N 10N EQ 10S 20S 30S 40S 50S 180

0.04

0.1

0.2

0.3

0.4

0.5

0.7

0.8

0.9

1

1.1

1.2

1

Figure 4 Monthly rainfall distribution over the globe in January 2006 in mm h : (upper) data with resolution 2.51  2.51 from GPCP; (lower) data with resolution 0.251  0.251 from the Tropical Rainfall Measuring Mission (TRMM) and Other Rainfall Estimate (3B42 V6) archive, made available by NASA at http://disc2.nascom.nasa.gov/Giovanni/tovas/TRMM_V6.3B42.shtml.

of precipitation. Numerical weather prediction (NWP) uses current weather conditions as an input to mathematical models of the atmosphere, which solve the flow (Navier– Stokes) equations, the thermodynamic energy equation, the state equation of gases, and the equation for conservation of water vapor, over a grid covering the entire atmosphere. The processes related to cloud formation and precipitation (see Section 2.02.2.2) are less accurately represented in these models. While the continuous improvement of NWP models resulted in a considerable reduction of forecast errors on pressure and temperature, the improvement in the so-called quantitative precipitation forecast (QPF) has been slower (Olson et al., 1995). Further, although the advances in computing infrastructure permitted the increase in model resolution that leads generally to an improvement of precipitation forecasts, recently, many authors have highlighted the limitations of such an approach (e.g., Mass et al., 2002; Lagouvardos et al., 2003; Kotroni and Lagouvardos, 2004). The major advancement in QPF in the last decades was the abandonment of the pure deterministic approach, which seeks a unique prediction, and the adoption of a more probabilistic approach to precipitation forecast, based on earlier ideas of Epstein (1969) and Leith (1974). In this approach, known as ensemble forecasting, the same model produces many

forecasts. To produce different forecasts, perturbations are introduced, for example, in the initial conditions, and, because of the nonlinear dynamics with sensitive dependence on the initial conditions (e.g., Lorenz, 1963), these perturbations are magnified in time, thus giving very different precipitation amounts in a lead time of 1 or more days. The different model outputs can then be treated in a probabilistic manner, thus assigning probabilities to rainfall occurrence as well as to the exceedance of a specified rainfall threshold. In this manner, although the model uses deterministic dynamics, the entire framework is of the Monte Carlo or stochastic type. This method is satisfactory for a time horizon of forecast of a few days. In hydrology, this time horizon is relevant in realtime flood forecasting. However, in hydrological design, horizons as long as 50 or 100 years (the lifetimes of engineering constructions) are typically used. For such long horizons, the use of deterministic dynamics and of the related laborious models would not be of any help. However, a probabilistic approach is still meaningful – in fact the only effective approach – and, in this case, it can be formulated irrespective of the dynamics. Rather, the probabilistic approach should be based, in this case, on historical records of precipitation, such as those displayed in Figures 7 and 8. The reasoning behind neglecting the deterministic dynamics is

34

Precipitation 50N 40N 30N 20N 10N EQ 10S 20S 30S 40S 50S

180

120W

60W

120W

60W

0

60E

120E

180

60E

120E

180

50N 40N 30N 20N 10N EQ 10S 20S 30S 40S 50S 180

0

9N

9N

8N

8N

7N

7N

6N

6N

5N

5N

4N

4N

3N

3N

2N

2N

1N

1N

EQ

EQ

1S

1S

2S

2S

3S

3S

4S

4S

5S 5S 78E 79E 80E 81E 82E 83E 84E 85E 86E 87E 88E 89E 90E 91E 92E 78E 79E 80E 81E 82E 83E 84E 85E 86E 87E 88E 89E 90E 91E 92E

0

2

4

8

10

12

14

16

18

20

22

Figure 5 Spatial rainfall distribution at daily and sub-daily scale: (upper) daily rainfall over the zone between 501N and 501S on 9 January 2006; (middle) 3-hourly rainfall at 09:00 on the same day; (lower left) zoom-in of the upper panel for daily rainfall in the Indian Ocean south-east of Sri Lanka (shown in figure); (lower right) zoom-in of the middle panel for 3-hourly rainfall for the same area. Data in mm h1 with resolution 0.251  0.251 from the TRMM 3B42 V6 archive.

Precipitation

35

2.85

1020

2.75 995 2.7 970

2.65 2.6

945

Precipitation rate (mm yr−1)

Precipitation rate (mm d−1)

2.8

2.55 920

Monthly

2.5

Annual 2.45 1979

Climatic, 1979−2008 1984

1989

1994

1999

2004

895 2009

Year Figure 6 Evolution of the globally averaged monthly precipitation in the 30-year period 1979–2008, based on data from GPCP (see caption of Figure 2).

that, beyond a certain time horizon (which in precipitation is of the order of several days), even the simplest nonlinear systems tend to a statistical equilibrium state. In this state, the probability distribution of the system properties, conditioned on the initial state, is practically equal to the marginal (i.e., unconditional) probability distribution of the same properties (Koutsoyiannis, 2009). This equilibrium, which is different from the typical thermodynamic equilibrium, corresponds to the maximization of the entropy of the vector of random variables defining the system state.

2.02.1.5.1 Basic concepts of probability Probability is thus not only a mathematical tool to model precipitation uncertainty, but also a concept for understanding the behavior of precipitation. Probabilistic thinking provides insights into phenomena and their mathematical descriptions, which may not be achievable in deterministic terms. It should be recalled that, according to the Kolmogorov (1933) system, probability is a normalized measure, that is, a function P that maps sets (areas where unknown quantities lie) to real numbers (in the interval [0, 1]). Furthermore, a random variable x is a single-valued function of the set of all elementary events (so that to each event, it maps a real number) and is associated with a probability distribution function. The latter is defined as

Fx ðxÞ :¼ Pfx r xg

ð3Þ

where x is any real number, which should be distinguished from the random variable x. (Distinction of random variables from their values is usually done by denoting them with upper case and lower case letters, respectively. This convention has several problems – e.g., the Latin x and the Greek w, if put in

upper case, are the same symbol X – other texts do not distinguish the two at all, thus creating another type of ambiguity. Here, we follow a different convention, in which random variables are underscored and their values are not.) Fx ðxÞ is a nondecreasing function of x with the obvious properties Fx ðNÞ ¼ 0 and Fx ðþNÞ ¼ 1. For continuous random variables (as is, for instance, the representation of a nonzero rainfall depth), the probability that a random variable x would take any particular value x is Pfx ¼ xg ¼ 0. Thus, the question of whether one particular value (say x1 ¼10 mm, assuming that x denotes daily rainfall at a location) is more probable than another value (say x2 ¼10 m, which intuitively seems extremely improbable) cannot be answered in terms of the probability function P, as all particular values have probability equal to zero. The derivative of F, that is,

f x ðxÞ :¼

dFx ðxÞ dx

ð4Þ

termed the probability density function, can provide this answer, as the quantity f x ðxÞ dx is the probability that rainfall will lie in an interval of length dx around x. Apparently, then the ratio f x ðx1 Þ=f x ðx2 Þ equals the ratio of the probabilities at points x1 and x2. These rather simple notions allow quantification of uncertainty and enable the production of different type of predictions, which offer a concrete foundation of rational decisions for the design and management of water-resourses projects. This quantification is sometimes (mostly in Bayesian statistics) referred to as ‘probabilization of uncertainty’ that is meant to be the axiomatic reduction from the notion of unknown to the notion of a random variable (Robert, 2007).

36

Precipitation 3500 Annual 3000

10-year average 30-year average

Rainfall (mm)

2500 2000 1500 1000 500 0 1770

1790

1810

1830

1850

1870

1890

1910

1930

1950

1970

1990

2010

Year 2000

Rainfall (mm)

1500

1000

500

Annual 10-year average 30-year average

0 1830

1850

1870

1890

1910

1930

1950

1970

1990

2010

Year

Figure 7 Annual precipitation time series in two of the stations with the longest records worldwide: (upper) Seoul, Korea; (lower) Charleston City, USA. Data from the database of the Dutch Royal Netherlands Meteorological Institute (KNMI; http://climexp.knmi.nl) and additional information as shown in text.

2.02.1.5.2 Stochastic processes In the study of rainfall variation in time, the notion and the theory of stochastic processes provide the necessary theoretical framework. A stochastic process is defined as an arbitrarily (usually infinitely) large family of random variables xðtÞ (Papoulis, 1991). In most hydrological applications, time is discretized using an appropriate time step d; for integer i, the average of the continuous time process xðtÞ from t ¼ (i  1)d to t ¼ i d, is usually denoted xi and forms a discrete time stochastic process. The index set of the stochastic process (i.e., the set from which the index t or i takes its values) can also be a vector space, rather than the real line or the set of integers. This is the case, for instance, when we assign a random variable (e.g., rainfall depth) to each geographical location (a two-dimensional (2D) vector space) or to each location and time instance (a 3D vector space). Stochastic processes with a multidimensional index set are also known as random fields.

A realization x(t) (or xi) of a stochastic process xðtÞ (or xi ), which is a regular (numerical) function of the time t (or a numerical sequence in time i), is known as a sample function. Typically, a realization is observed at countable time instances (and not in continuous time, even if the process is of continuous-time type). This sequence of observations is also referred to as a time series. Clearly then, a time series is a sequence of numbers, whereas a stochastic process is a family of random variables. (Unfortunately, a large body of literature does not make this distinction and confuses stochastic processes with time series.) The distribution and the density functions of the random variable xi, that is,

Fi ðxÞ :¼ Pfxi r xg;

f i ðxÞ :¼

dFi ðxÞ dx

ð5Þ

Precipitation 3

6

9

12

18

181

21

24

37

27 h

120 10-s scale 5-min scale

100 Rainfall intensity (mm h−1)

Hourly scale 80

60

40

20

0 0

10 000

20 000

30 000

40 000

50 000

60 000

70 000

80 000

90 000

Time (s) Figure 8 Time series of a storm in Iowa, USA, measured at the University of Iowa with temporal resolution of 10 s; time zero corresponds to 199002-12T17:03:39. From Georgakakos KP, Carsteanu AA, Sturdevant PL, and Cramer JA (1994) Observation and analysis of Midwestern rain rates. Journal of Applied Meteorology 33: 1433–1444.

are called, respectively ‘first-order distribution function’ and ‘first-order density function’ of the process. Likewise, the second-order distribution function is Fi1 i2 ðx1 ; x2 Þ ¼ Pfxi1 r x1 ; xi2 r x2 g and this can be generalized to define the nth-order distribution function. It should be recalled that the expected value of a function g of one, two, or more random variables is the integral of g multiplied by the density f, that is,

E½gðxi Þ :¼

ZN

gðxÞf i ðxÞ dx

N

E½gðxi1 ; xi2 Þ ¼

ZN ZN

gðx1 ; x2 Þf i1 i2 ðx1 ; x2 Þ dx1 dx2

2.02.1.5.3 Stationarity As implied by the above notation, in the general setting, the statistics of a stochastic process, such as the mean and autocovariance, depend on time i and thus vary with time. However, the case where these statistical properties remain constant in time is most interesting. A process with this property is termed a ‘stationary’ process. More precisely, a process is called ‘strict-sense’ stationary, if all its statistical properties are invariant with a shift in the time origin. That is, the distribution function of any order of xiþj is identical to that of xi . A process is called ‘wide-sense stationary’ if its mean is constant and its autocovariance depends only on time differences (lags), that is,

E½Xi  ¼ m;

N N

ð6Þ The use of square brackets in E[ ] and the random variables xi rather than their values x signifies the fact that the expected value is not a function of the real number x; rather, it depends solely on the distribution function associated with the random variable xi. Of particular interest are the cases where gðxi Þ ¼ xi, where E½xi  ¼: mi is the mean value of xi , and gðxi1 ; xi2 Þ ¼ ðxi1  mi1 Þðxi2  mi2 Þ, where E½ðxi1  mi1 Þðxi2  mi2 Þ ¼: Ci1 i2 is the process autocovariance, that is, the covariance of the random variables xi1 and xi2 . The process variance (the variance of the variable xi ), is a special case of the latter, that is, Var½xi  ¼ Cii , whereas the standard devipffiffiffiffiffiffi ation is the square root of the latter, that is, si :¼ Cii . Consequently, the process autocorrelation (the correlation coefficient of the random variables xi1 and xi2 ) is ri1i2: ¼ Ci1i2/ (si1 si2).

E½ðXiþj  mÞðXi  mÞ ¼ Cj

ð7Þ

Evidently, the standard deviation is constant too, that is, si ¼ s, and the autocorrelation is a function of the time lag only, that is, riþj, i ¼ rj. A strict-sense stationary process is also wide-sense stationary, but the reverse is not true. A process that is not stationary is called nonstationary. In a nonstationary process, one or more statistical properties depend on time. A typical case of a nonstationary process is the cumulative rainfall depth whose mean obviously increases with time. For instance, let us assume that the instantaneous rainfall intensity iðtÞ at a geographical location and period of the year is a stationary process, with a mean m. Let us further denote by hðtÞ, the rainfall depth collected in a large container (a cumulative rain gauge) at time t, and assume that at the time origin, t ¼ 0, the container is empty. Clearly E½hðtÞ ¼ mt. Thus hðtÞ is a nonstationary process. It should be stressed that stationarity and nonstationarity are properties of a stochastic process, not of a sample function

38

Precipitation

or time series. There is some confusion in the literature about this, as there are several studies that refer to a time series as stationary or nonstationary. As a general rule, to characterize a process as nonstationary, it suffices to show that some statistical property is a deterministic function of time (as in the above example of the cumulative rainfall), but this cannot be directly inferred merely from a time series. To understand this, let us consider the time series of annual rainfall in Seoul, plotted in the upper panel of Figure 7. Misled by the changing regime of precipitation at the climatic scale, as manifest in the plot of the 30-year average, it would be tempting to note (1) an increasing trend in the period 1770–90; (2) a constant climate with high precipitation during 1790–1870; (3) a decreasing trend between 1870 and 1900; and (4) a constant climate with low precipitation thereafter. It is then a matter of applying a fitting algorithm to determine, say, a broken-line type of function to the time series, which would be called a deterministic function of time. The conclusion would then be that the time series is nonstationary. However, this is a wrong ex-post argument, which interprets the long-term variability of the processes as a deterministic function. Had the function been indeed deterministic, it would also apply to future times, which obviously is not the case. Comparison with the previous example (cumulative rainfall), where the deterministic function E½xðtÞ ¼ mt was obtained by theoretical reasoning (deduction) rather than by inspection of the data, demonstrates the real basis of nonstationarity. Koutsoyiannis (2006b) has provided a more detailed study of this issue. Stochastic processes describing periodic phenomena, such as those affected by the annual cycle of the Earth, are clearly nonstationary. For instance, the daily rainfall at a mid-latitude location cannot be regarded as a stationary process. Rather, a special type of a nonstationary process, whose properties depend on time in a periodical manner (are periodic functions of time), should be used. Such processes are called ‘cyclostationary’ processes.

implies that the random variable has zero variance. This is precisely the condition that makes a process ergodic, a condition that does not hold true for every stochastic process.

2.02.1.5.4 Ergodicity

However, this law hardly holds in geophysical time series, including rainfall time series, whatever the scale is. This can be verified based on the examples presented in Section 2.02.1.4. A more plausible law is expressed by the elementary scaling (power-law) property

The concept of ergodicity (from the Greek words ergon, work; and odos, path) is central to the problem of determining the distribution function of a process from a single sample function (time series). A stationary stochastic process is ergodic if any statistical property can be determined from a sample function. Given that, in practice, the statistical properties are determined as time averages of time series, the above statement can be formulated alternatively – a stationary stochastic process is ergodic if time averages equal ensemble averages (i.e., expected values). For example, a stationary stochastic process is mean ergodic if

E½xi  :¼ limN-N

N 1X xi N i¼1

ð8Þ

The left-hand side in the above equation represents the ensemble average, whereas the right-hand side represents the time average, for the limiting case of infinite time. While the left-hand side is a parameter, rather than a random variable, the right-hand side is a random variable (as a sum of random variables). Equating a parameter with a random variable

2.02.1.5.5 Some characteristic stochastic properties of precipitation It has been widely accepted that rainfall exhibits some autocorrelation (or time dependence) if the timescale of study is daily or sub-daily, but this dependence vanishes at larger timescales, such as monthly or yearly. Thus, for timescales monthly and above, rainfall data series have been traditionally treated as independent samples. Mathematically, such a perception corresponds to a Markovian dependence at fine timescales, in which the autocorrelation decreases rapidly with time lag in an exponential manner, that is,

rj ¼ rj

ð9Þ

where r: ¼ r1. Then for a large lag j, or for a large scale of aggregation and even for the smallest lag (one), the autocorrelation is virtually zero (e.g., Koutsoyiannis, 2002). If xi denotes the stochastic process at an initial timescale, which is designated as scale 1, then the averaged process at an aggregated timescale k ¼ 2, 3, y, is ðkÞ

xi :¼

xði1Þkþ1 þ y þ xi k

ð10Þ

ð1Þ

(with xi  xi ). Let s(k) be the standard deviation at scale k. In processes xi independent of time, s(k) decreases with scale according to the well-known classical statistical law of inverse square-root, that is,

s s ðkÞ ¼ pffiffiffi k

s ðkÞ ¼

s k 1H

ð11Þ

ð12Þ

where H is the so-called Hurst exponent, named after Hurst (1951) who first studied this type of behavior in geophysical time series. Earlier, Kolmogorov (1940), when studying turbulence, had proposed a mathematical model to describe this behavior. This behavior has been known by several names, including the Hurst phenomenon, long-term persistence, and long-range dependence, and a simple stochastic model that reproduces it is known as a simple scaling stochastic model or fractional Gaussian noise (due to Mandelbrot and van Ness, 1968). Here, the behavior is referred to as the Hurst–Kolmogorov (HK) behavior or HK (stochastic) dynamics and the model, as the HK model. This behavior implies that the autocorrelation decreases slowly, that is, according to a power-type function,

Precipitation

with lag j: ðkÞ

ri ¼ rj ¼ ð1=2Þ½ðjj þ 1jÞ2H þ ðjj  1jÞ2H   jjj2H E Hð2H  1Þj2H2

ð13Þ

so that independence virtually never holds, unless H ¼ 0.5, a value which reinstates classical statistics including the law in Equation (11). Most often, natural processes including rainfall are positively correlated and H varies in the range (0.5, 1). The above framework is rather simple and allows easy exploration of data to detect whether they indicate consistence with classical statistics or with the HK behavior. A simple exploration tool is a double logarithmic plot of the estimates of standard deviation s(k) versus scale k, which is known as a ‘climacogram’. (ClimacogramoGreek Klımako´grammao(climax ´ (klımax) ¼ scale) þ (gramma (gramma) ¼ written).) In such a plot, the classical law and the HK law are manifested by a linear arrangement of points with slopes –0.5 and H  1, respectively. We must bear in mind, however, that a consequence of the HK law in Equation (12) is that the classical estimator of the variance

s2 ¼

n 1 X ðxi  xÞ2 n  1 i¼1

ð14Þ

ðnÞ

where n is the sample size, x  x1 is the estimator of the mean, and s the estimator of standard deviation, implies negative bias if there is temporal dependence. The bias becomes very high for HK processes with H approaching 1. Apparently then, s could be a highly biased estimator of s; an approximately unbiased estimator is (Koutsoyiannis, 2003a; Koutsoyiannis and Montanari, 2007):

rffiffiffiffiffiffiffiffiffiffiffiffiffi n0 s :¼ s n0  1

E

ð15Þ

39

where n0 is the equivalent (or effective) sample size, that is, the sample size that in the framework of classical statistics would ðnÞ lead to the same uncertainty (in the estimation of m by x  x1 ) as an HK series yields with sample size n. For an HK process, n0 is related to n by

n0 ¼ n 2ð1HÞ

ð16Þ

It can be seen that n0 can be very small even for high n if H pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi is high, and thus the correcting factor n0 =ðn0  1Þ in Equation (15) can be very large (see Koutsoyiannis and Montanari, 2007). Returning to the time series of globally averaged monthly precipitation in the 30-year period 1979–2008, which has been discussed earlier and is displayed in Figure 6, we may now study its statistical properties for several timescales. As the precipitation amounts are averaged over the entire globe, the effect of seasonality is diminished and the time series can be modeled using a stationary process rather than a cyclostationary one. Figure 9 depicts the climacogram, that is, a logarithmic plot of standard deviation versus scale. Empirical estimates of standard deviations have been calculated using both the classical estimator in Equation (14) and the HK estimator in Equation (15). Theoretical curves resulting from the classical statistical model (assuming independence), the Markovian model, and the HK model have also been plotted. For the Markovian model, the lag one autocorrelation coefficient, estimated from the monthly data, is r ¼ 0.256 and for the HK model, the estimate of the Hurst coefficient is H ¼ 0.70. This can be obtained readily from the slope of the straight line fitted to the group of empirical points in Figure 9, which should be H  1. Here, a slightly modified algorithm from Koutsoyiannis (2003a) has been used for the estimation of H. Overall, Figure 9 clearly demonstrates that the empirical

−1.2 Empirical, classical estimate Empirical, HK estimate Classical statistical model Markov model HK model

Log(standard deviation in mm d−1)

−1.3

−1.4

−1.5

−1.6

−1.7

−1.8 0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Log(scale in months) Figure 9 Climacogram of the time series of globally averaged monthly precipitation in the 30-year period 1979–2008 shown in Figure 6. The estimate of the Hurst coefficient for the HK model is H ¼ 0.70.

40

Precipitation

points are inconsistent with the classical and Markovian models and justify an assumption of HK behavior. Similar plots have been constructed, and are shown in Figure 10, for the annual precipitation time series from Seoul, Korea, and Charleston City, USA, displayed in Figure 7. Again, the empirical evidence from data precludes the applicability of the classical statistical model and favors the HK statistics. An additional plot for the 10-s precipitation time series in Iowa, USA, displayed in Figure 8, is depicted in Figure 11. Here, the Hurst coefficient is very high, H ¼ 0.96. The difference between the empirical points based on classical statistics on the

one hand and the HK statistics on the other hand is quite distinctive. Apparently, the classical model is completely inappropriate for the rainfall process. The HK stochastic processes can be readily extended in a 2D setting (or even a multidimensional one). The 2D version of Equation (12) is

s ðkÞ ¼

s k 22H

ð17Þ

This can be obtained by substituting k2 for k in Equation (12). Equations (15) and (16) still hold, provided that n is the

2.7

Empirical, classical estimate Empirical, HK estimate Classical statistical model

Log(standard deviation in mm)

2.6

HK model 2.5

2.4

2.3

2.2

2.1 0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Log(scale in years)

2.5 Empirical, classical estimate Empirical, HK estimate

Log(standard deviation in mm)

2.4

Classical statistical model HK model

2.3

2.2

2.1

2

1.9

1.8 0

0.2

0.4

0.6

0.8

1

1.2

1.4

Log(scale in years) Figure 10 Climacogram of the annual precipitation time series at: (upper) Seoul, Korea and (lower) Charleston City, USA, which are shown in Figure 7; the estimated Hurst coefficients are 0.76 and 0.74, respectively.

Precipitation

41

0.9

Log(standard deviation in mm h−1)

0.85 0.8 0.75 0.7 0.65 0.6 0.55 Empirical, classical estimate 0.5

Empirical, HK estimate Classical statistical model

0.45

HK model 0.4 0

0.5

1

1.5

2

2.5

3

Log(scale in 10-s intervals) Figure 11 Climacogram of the 10-s precipitation time series in Iowa, USA, displayed in Figure 8; the estimated Hurst coefficient is 0.96.

1.5

Log(standard deviation in mm h−1)

1.4

1.3

1.2

Empirical, classical estimate

1.1

Empirical, HK estimate Classical statistical model HK model

1 0

0.2

0.4

0.6

0.8

1

Log(scale in 0.25° of latitute and longitude) Figure 12 Climacogram of the spatial daily rainfall over the area 91 N–51 S and 78–921 E (Indian Ocean south-east of Sri Lanka) on 9 January 2006, as shown in the lower-left panel of Figure 5; the estimated Hurst coefficient is 0.94.

number of points, which is inversely proportional to k2. Figure 12 demonstrates this behavior by means of a climacogram for the spatial daily rainfall over the area 91 N–51 S and 78–921 E (Indian ocean south-east of Sri Lanka) on 9 January 2006, displayed in the lower-left panel of Figure 5. Here, the estimated Hurst coefficient is again very high,

H ¼ 0.94. As in all previous cases, the classical model is completely inappropriate, while the HK model seems reasonable for scales X4, which correspond to a resolution of 11 11 and beyond. Thus, the evidence presented using several examples of different spatial and temporal scales indicates that HK

42

Precipitation

dynamics is consistent with the nature of rainfall. These dynamics appear as scaling behavior, either in time or in space, which is either full, applicable to the entire range of scales, or asymptotic, applicable to large scales. Both these scaling behaviors are manifested as power laws of standard deviation versus temporal or spatial scale and of autocorrelation versus lag. There exists another type of scaling behavior in precipitation, the scaling in state, which is sometimes confused with the other two scaling behaviors, but is fundamentally different. Scaling in state is a property of the marginal distribution function of rainfall (it has no relation to the dependence structure of the process unlike other types of scaling) and is expressed by power laws of the tails of (1) the probability density function f(x), (2) the survival function (or exceedance probability) F  ðxÞ :¼ Pfx4 xg ¼ 1  FðxÞ, and (3) the return period T ¼ d/F*(x) where d is the length of the timescale examined. These scaling properties are expressed as

xp T k ;

F  ðxÞp x 1=k ;

f ðxÞp x 11=k

ð18Þ

and are equivalent to each other. All these are asymptotic, that is, they hold only for large values of x or, in other words, for the distribution tails. Such tails are known by several names, such as long, heavy, strong, power-type, overexponential, algebraic, or Pareto tails. The latter name comes from the Pareto distribution, which in its simplest form is given in Equation (18), although its generalized form is applicable to rainfall (see Section 2.02.5.2). As this is an asymptotic behavior, long records are needed to observe it. Figure 13 shows a logarithmic plot of the empirical distribution (expressed in terms of return period T) of a large data set of daily rainfall. This data set was created using records of 168 stations worldwide, each of which contained data of 100 years or more (Koutsoyiannis, 2004b). For each station with n years of record, n annual maximum values of daily rainfall were extracted. These values

were standardized by their mean and merged into one sample of length 17 922 station-years. From the theoretical distributions, also plotted in Figure 13, it is observed that the Pareto distribution (whose right tail appears as a straight line in the logarithmic plot; see Section 2.02.5.2) with k ¼ 0.15 provides the best fit, thus confirming the applicability of asymptotic scaling in state and the inappropriateness of the exponential-type tail. This has severe consequences, particularly in hydrological design, as distributions with exponential tails have been most common in hydrological practice, whereas it is apparent that the power-type tails are more consistent with reality. As shown in Figure 13, the difference between the two types can be substantial. Koutsoyiannis (2005a, 2005b) produced the aforesaid different types of scaling from the principle of maximum entropy. As entropy is a measure of uncertainty, the applicability of the principle of maximum entropy and its consistence with observed natural behaviors characterizing the precipitation process underscores the dominance of uncertainty in precipitation.

2.02.2 Physical and Meteorological Framework Atmospheric air is a heterogeneous mixture of gases, also containing suspended particles in liquid and solid phase. The most abundant gases are nitrogen (N2) and oxygen (O2) that account for about 78% and 21%, respectively, by volume of the atmospheric permanent gases, followed by argon (Ar) and traces of other noble gases. Their concentrations are almost constant worldwide and up to an altitude of about 90 km. Water vapor (H2O) appears in relatively low concentrations, which are highly variable. However, water vapor is very important for energy exchange on Earth (it accounts for 65% of the radiative transfer of energy in the atmosphere; Hemond

10 Empirical Pareto/least squares Pareto/L-moments Exponential/L-moments

Rescaled rainfall depth

5

2

1

0.5 1

2

5 10

100

1000

10 000

100 000

Return period Figure 13 Logarithmic plot of rescaled daily rainfall depth vs. return period: empirical estimates from a unified sample over threshold, formed using rainfall data from 168 stations worldwide (17 922 station-years). The unified sample was rescaled by the mean of each station, and fitted using a Pareto and an exponential distribution model. Adapted from Koutsoyiannis D (2004b) Statistics of extremes and estimation of extreme rainfall: 2. Empirical investigation of long rainfall records. Hydrological Sciences Journal 49(4): 591–610.

Precipitation

and Fechner-Levy, 2000), as well as for mass transfer processes in the hydrological cycle. Under certain conditions (i.e., pressure and temperature), water vapor can transform into droplets or ice crystals with subsequent release of latent heat (see Sections 2.02.2.1 and 2.02.2.2). More generally, the water-vapor content of atmospheric air affects its density, and it is of central importance in atmospheric thermodynamics (Section 2.02.2.1). The varying content and importance of water vapor in precipitation processes and thermodynamics has led to the study of atmospheric air as a mixture of two (ideal) gases: dry air and water vapor. This mixture is usually referred to as ‘moist air’ and has thermodynamic properties determined by its constituents (e.g., Rogers and Yau, 1996; Cotton and Anthes, 1989). The particles of solid and liquid material suspended in air are called ‘aerosols’. Common examples of aerosols are water droplets and ice crystals (called ‘hydrometeors’), smoke, sea salt (NaCl), dust, and pollen. The size distribution of solid aerosols depends strongly on their location. For example, the size spectrum of aerosols over land is narrow with high concentrations of small particles (e.g., kaolinite, dust, and pollen), whereas the size spectrum of aerosols over sea is wider with small concentrations of larger particles (e.g., sea salt; Ryan et al., 1972). Existence of aerosols in the atmosphere is of major importance, since a select group of aerosols called ‘hydroscopic nuclei’, is crucial for the nucleation of liquid water and initiation of rain (e.g., Brock, 1972, and Section 2.02.2.2). When moist air is cooled (i.e., below its dew point; see Section 2.02.2.1), an amount of water vapor condenses and a cloud forms, but precipitation may or may not occur. Initiation of rain requires the formation of hydrometeors (i.e., water droplets and ice crystals) of precipitable size (e.g., Gunn and Kinzer, 1949; Twomey, 1964, 1966; Brock, 1972). Formation and growth of these particles are governed by processes that take place at scales comparable to their size (micrometers to millimeters). The latter processes form the core of cloud microphysics, whereas large-scale processes, related to thermodynamics of moist air and motion of air masses, form the core of cloud dynamics. Importantly,

precipitation is the combined effect of both large- and microscale processes, and both processes are equally important and necessary for precipitation to occur.

2.02.2.1 Basics of Moist Air Thermodynamics In a parcel of moist air at temperature T with volume V and mass M ¼ Md þ Mv, with the two components denoting mass of dry air and water vapor, respectively, the density is r ¼ M/V and the concentration of water vapor, known as specific humidity, is q: ¼ Mv/M. The quantity r ¼ Mv/Md: ¼ q/(1  q) is usually referred to as the mixing ratio. The total pressure of the moist air in the parcel, p (the atmospheric pressure), equals the sum of the partial pressures of dry air pd and that of water vapor e (i.e., p ¼ pd þ e). Specific humidity and vapor pressure are interrelated through

q¼

ee p  ð1  eÞe

ð19Þ

where e ¼ 0.622 is the ratio of molar masses of water vapor and dry air. Air cannot hold an arbitrarily high quantity of vapor. Rather, there is an upper limit of the vapor pressure e*, called the saturation vapor pressure, which depends on the temperature T and is given by the Clausius–Clapeyron equation. A useful approximation to this equation is

e  ðTÞ ¼ 6:11 exp



 17:67T T þ 237:3

Saturation specific humidity, q* (g kg−1)

10

1

−30

−20

−10

0

10

Temperature, T (°C) Figure 14 Saturation specific humidity as a function of air temperature.

ð20Þ

where e* is in hPa and T is in 1C. Consequently, from Equations (19) and (20) we can calculate the saturation specific humidity q*, which is a function of T, and expresses the water vapor holding capacity of air. As shown in Figure 14, this capacity changes drastically, almost exponentially, with temperature, so that a change of temperature from  40 to 40 1C increases this capacity by 2.5 orders of magnitude. The ratio of the actual to saturation vapor pressure, that is, e/e* ¼ : U, called the relative humidity, is normally lesser than 1. When an air parcel cools, while e remains constant, e*

100

0.1 −40

43

20

30

40

44

Precipitation

decreases and hence U increases, up to the saturation value 1 or 100%. The temperature Td at which saturation occurs is called the dew point temperature and is calculated using Equation (20) by setting e*(Td) ¼ e. Therefore, cooling of the air parcel below the dew point temperature results in condensation, or transformation of the excess water vapor into liquid water in the form of droplets. During this change of phase, the relative humidity remains 100%. Condensation releases heat at a fairly constant rate (LE2.5 MJ kg1); this rate equals that of evaporation of water at a constant temperature and is thus called latent heat. For an air parcel to ascend and expand spontaneously, so that condensation and cloud formation can occur, the ambient (atmospheric) temperature gradient g : ¼  dT/dz, where z denotes altitude, also known as lapse rate, must be high (otherwise, an uplifted air parcel will sink again). While the parcel ascends and expands adiabatically (i.e., in a way that no heat transfer takes place between the air parcel and its ambient air), its own lapse rate is gd ¼ 9.8 1C km1 if the expansion is dry adiabatic (i.e., if it takes place without condensation of water vapor) and somewhat smaller, g*, if the expansion is moist adiabatic (i.e., if the temperature has fallen below dew point, so that some of the water vapor in the parcel condenses to liquid form). The gradient g* is not constant but varies with temperature T and air pressure p so that g* ¼ 4 1C km1 for T ¼ 25 1C and p ¼ 1000 hPa, whereas g* ¼ 9 1C km1 for T ¼  25 1C and p ¼ 1000 hPa; an average value is g* ¼ 6.5 1C km1 (Koutsoyiannis, 2000b; see also Wallace and Hobbs, 1977). When the ambient lapse rate g is smaller than g*, the atmosphere is stable, and no spontaneous lift occurs and no clouds are formed. When g4gd, the atmosphere is unstable and favors air lift and formation of clouds. The case g*ogogd is known as conditional instability and it serves as an important mechanism for mesoscale precipitation processes (see Sections 2.02.2.4 and 2.02.2.5).

2.02.2.2 Formation and Growth of Precipitation Particles The Clausius–Clapeyron equation describes the equilibrium condition of a thermodynamic system consisting of bulk water and vapor. A state out of the equilibrium, in which e4e* (U41) is possible, but is thermodynamically unstable, and is called supersaturation. Detailed study of the transition of water vapor to liquid or ice at or above saturation is associated with certain free-energy barriers. An example of such an energy barrier is the dynamic energy associated with the surface tension, s, of a water droplet. For a spherical droplet, s is proportional to the pressure of water within the droplet p and inversely proportional to its radius r (i.e., s ¼ p/2r). This means that a high vapor pressure is needed for a very small droplet to be maintained and not evaporate. In essence, the free-energy barrier of surface tension makes droplet formation solely by condensation of water vapor (a process usually referred to as ‘homogeneous nucleation’), almost impossible in nature. However, if the surface-tension barrier is bypassed, common supersaturations of the order of 1–2% (i.e., U ¼ 1.01–1.02) are sufficient for water vapor to diffuse toward the surface of the droplet. The rate of diffusional growth is proportional to the supersaturation U  1 of the ambient air,

and inversely proportional to the radius r of the droplet: that is, dr/dtp(U  1)/r (Mason, 1971; Rogers and Yau, 1996). While homogeneous nucleation requires large supersaturations, formation of droplets is drastically facilitated by particulated matter of the size of micrometers or lower, the aerosols, some of which, called condensation nuclei, are hydrophilic and serve as centers for droplet condensation (Brock, 1972; Slinn, 1975; Hobbs et al., 1985). This process is usually referred to as ‘heterogeneous nucleation’ and it is almost exclusively the process that governs water vapor condensation in the atmosphere (Houze, 1993). When the temperature in the cloud drops below the freezing point, water droplets are said to be supercooled, and they may or may not freeze. For pure water droplets, homogeneous freezing does not occur until the temperature drops below  40 1C (Rogers and Yau, 1996). However, the presence of certain condensation nuclei, called ice nuclei, may allow freezing of water droplets at temperatures a few degrees below 0 1C. These nuclei are particles of the size of micrometers, or lower, which form strong bonds with water and closely match the crystallic structure of ice. Different particles serve as condensation nuclei at different subfreezing temperatures. For example, silver iodide (AgI) serves as an ice nucleator at  4 1C and kaolinite at  9 1C (e.g., Houghton, 1985). Evidently, a cloud is an assembly of tiny droplets with usually met concentrations of several hundreds per cubic centimeter, and radii of several micrometers. This structure is very stable and the only dominant process is vapor diffusion, which accounts for the size-growth evolution of the whole droplet population (Telford and Chai, 1980; Telford and Wagner, 1981). Precipitation develops when the cloud population becomes unstable and some droplets grow faster relative to others. In general, two main mechanisms account for the cloud microstructure becoming unstable. The first mechanism is the collision and coalescence (i.e., sticking) of larger (and fastermoving) collector drops with smaller (and slower-moving) collected droplets. This mechanism is particularly important for precipitation development in warm clouds (i.e., at temperatures in excess of 0 1C; see, e.g., Houze, 1993) and, for a long time, it has formed an active research area in cloud and precipitation physics (e.g., Langmuir, 1948; Bowen, 1950; Telford, 1955; Scott, 1968, 1972; Long, 1971; Drake, 1972a, 1972b; Gillespie, 1972, 1975; Robertson, 1974; Berry and Reinhardt, 1974a, 1974b; Vohl et al., 1999; Pinsky et al., 1999, 2000; Pinsky and Khain, 2004; and review in Testik and Barros (2007)). Its significance for precipitation processes depends considerably on the droplet-size spectra, with larger effectiveness for wider spectra with small concentrations of larger particles (Berry and Reinhardt, 1974a, 1974b). The second mechanism is related to interaction between water droplets and ice crystals, and is limited to clouds with tops that extend to subfreezing temperatures (i.e., cold clouds). In particular, when an ice crystal develops in the presence of a large number of supercooled droplets, the situation becomes immediately unstable and the ice crystal grows due to diffusion of water vapor from the droplets toward the crystal. This is due to the fact that the equilibrium vapor pressure over ice is less than that over water at the same subfreezing temperature. Thus, the ice crystal grows by

Precipitation

diffusion of water vapor and the supercooled droplets evaporate to compensate for this. The transfer rate of water vapor depends on the difference between the equilibrium vapor pressure of water and ice, a quantity that becomes sufficiently large at about  15 1C (Uijlenhoet, 2008). The latter process is called the Bergeron–Findeisen mechanism, named after the scientists who first studied it (Bergeron, 1935; Findeisen, 1938). Once the ice crystals have grown by vapor diffusion to sizes sufficiently large for gravitational settling to dominate, they start falling and colliding with their ambient droplets and ice crystals, a process usually referred to as ‘accretional growth’. In the first case (i.e., when ice crystals collide with droplets), graupel or hail may form, whereas in the second case, snowflakes are likely to form. As the frozen particles fall, it is possible to enter layers with temperatures higher than 0 1C and start melting. If the particles have relatively small terminal velocities (or equivalently small size; see Section 2.02.2.3), they may reach the ground as raindrops indistinguishable from those formed by coalescence. Alternatively, in cold weather, or when large hailstones are formed, the precipitation particles may reach the ground unmelted. Additional discussion on the mechanisms of formation and growth of precipitation particles, and the potential human intervention on the mechanisms by technological means are discussed in Chapter 4.05 Abstraction of Atmospheric Humidity.

2.02.2.3 Properties of Precipitation Particles 2.02.2.3.1 Terminal velocity The terminal velocity UX(D) of a precipitable particle of type X ¼ R (rain), H (hail), S (snow), and effective diameter D is the maximum velocity this particle may develop under gravitational settling relative to its ambient air. In theory, UX(D) can be obtained by balancing the weight of the particle with the sum of the static and dynamic buoyancy (i.e., drag forces) on the particle. For a rigid spherical raindrop, one obtains pffiffiffiffi UR ðDÞp D (e.g., Rogers and Yau, 1996). Theoretical calculation of UX(D) becomes more complicated when the dynamical characteristics of the falling particles depend on their linear size D and the ambient temperature T. For example, droplets with diameters D smaller than about 0.35 mm are approximately spherical, drops with diameters in the range 0.35–1 mm tend to deform by the aerodynamic shear receiving a more elliptical shape, whereas larger drops frequently break down into smaller droplets due to excessive elongation or surface vibrations (e.g., Testik and Barros, 2007; Uijlenhoet, 2008). Moreover, the crystallic structure, shape, size, and, hence, the aerodynamic properties of snowflakes depend on the ambient temperature T (Fletcher, 1962; Locatelli and Hobbs, 1974; Houghton, 1985; Rogers and Yau, 1996). In the absence of exact theoretical solutions for the terminal velocity UX(D) of precipitation particles under complex atmospheric conditions, several empirical formulae have been developed (e.g., Gunn and Kinzer, 1949; Liu and Orville, 1969; Wisner et al., 1972; Locatelli and Hobbs, 1974; Atlas and Ulbrich, 1977; Lin et al., 1983). According to Liu and

45

Orville (1969), who performed a least squares analysis of Gunn and Kinzer’s (1949) data, the terminal velocity of raindrops of diameter D can be approximated by a power-law type relationship:

UR ðDÞ ¼ a Db

ð21Þ

where a ¼ 2115 cm1b s1 and b ¼ 0.8 are empirical constants. For raindrops with diameters in the range 0.5pDp5 mm, Atlas and Ulbrich (1977 (see also Uijlenhoet, 2008) suggest the use of Equation (21) with parameters a ¼ 1767 cm1b s1 and b ¼ 0.67. For hail, Wisner et al. (1972) suggest

UH ðDÞ ¼ D1=2



4grH 3CD r

1=2 ð22Þ

where g ¼ 9.81 m s2 is the acceleration of gravity, rE1.2 kg m3 is the density of air, rH ¼ 800–900 kg m3 is the density of the hailstone, and CD ¼ 0.6 is a drag coefficient. For graupel-like snow of hexagonal type, Locatelli and Hobbs (1974) suggest:

US ðDÞ ¼ c Dd

ð23Þ

where c ¼ 153 cm1d s1 and d ¼ 0.25 are empirical constants that, in general, depend on the shape of the snowflakes (e.g., Stoelinga et al., 2005). UX(D) relationships other than power laws have also been suggested (e.g., Beard (1976) and review by Testik and Barros (2007)). However, the power-law form in Equations (21)–(23) is the only functional form that is consistent with the powerlaw relations between the radar reflectivity factor Z (see Section 2.02.3.2) and the rainfall intensity i (Uijlenhoet, 1999, 2008).

2.02.2.3.2 Size distribution A commonly used parametrization for the size distributions of precipitation particles is that introduced by Marshall and Palmer (1948). According to this parametrization, precipitation particles have exponential size distributions of the type

nX ðDÞ ¼ n0X expðbX DÞ;

X ¼ R; H; S

ð24Þ

where the subscript X denotes the type of the particle: rain (R), hail (H), or snow (S); D is the effective diameter of the particle; bX is a distribution scale parameter with units of (length1) (see below); and n0X is an intercept parameter that depends on the type of the particle with units of (length4): that is, number of particles per unit diameter and per unit volume of air (see below). To determine the parameters n0R and bR in Equation (24) for rainfall, Marshall and Palmer (1948) used observations from summer storms in Canada. The study reported a constant value of the intercept parameter n0R ¼ 8 102 cm4, whereas the scale parameter bR was found to vary with the rainfall intensity i at ground level as: bR ¼ 41 i0.21 cm1, where i is in millimeters per hour. Clearly, the mean raindrop size 1/bR increases with increasing rainfall intensity i.

46

Precipitation

Gunn and Marshall (1958) used snowfall observations from Canada to determine the parameters n0S and bS for snow. The study concluded that both n0S and bS depend on the precipitation rate as:

n0S ¼ 0:038 i0:87 cm4 ;

bS ¼ 25:5 i0:48 cm1

2.02.2.4 Clouds and Precipitation Types ð25Þ

where i is the water equivalent (in millimeters per hour) of the accumulated snow at ground level. Similar to the mean raindrop size, the mean snowflake size 1/bS increases with increasing i. A modification to the distribution model of Gunn and Marshall (1958) has been proposed by Houze et al. (1979) and Ryan (1996). According to these authors, the intercept parameter for snow, n0S, is better approximated as a decreasing function of the temperature T of the ambient air. The latter is responsible for the properties and structures of ice crystals (see Section 2.02.2.2). Federer and Waldvogel (1975) used observations from a multicell hailstorm in Switzerland to determine the parameters n0H and bH for hail. The study showed pronounced variability of the intercept parameter n0H ¼ 15 106 to 5.2 104 cm4, moderate variability of the scale parameter bH ¼ 3.3–6.4 cm1, and concluded showing an exponential mean size distribution for hailstones with constant parameters: n0HE1.2 104 cm4 and bHE4.2 cm1. Alternative models, where the size distributions of precipitation particles are taken to be either gamma or lognormal, have also been suggested (e.g., Ulbrich, 1983; Feingold and Levin, 1986; Joss and Waldvogel, 1990). However, the exponential distribution model introduced by Marshall and Palmer (1948) has been empirically validated by a number of studies (see, e.g., Kessler, 1969; Federer and Waldvogel, 1975; Joss and Gori, 1978; Houze et al., 1979; Ryan, 1996; Ulbrich and Atlas, 1998; Hong et al., 2004), and has found the widest application by being used in the cloud-resolving schemes of many state-of-the-art NWP models (e.g., Cotton et al., 1994; Grell et al., 1995; Reisner et al., 1998; Thompson et al., 2004; Skamarock et al., 2005). A more general formulation for the size distribution of precipitation particles, which includes the exponential model of Marshall and Palmer (1948), and the gamma and lognormal models as special cases, was suggested by SempereTorres et al. (1994, 1998). According to their formulation, the size distribution of precipitation particles can be parametrized as

nX ðDÞp if gðDX =iz Þ;

X ¼ R; H; S

ð26Þ

where f and z are constant exponents, i is the precipitation rate, and g(x) is a scalar function with parameter vector a. For a certain form of g, the functional dependence of the parameters f, z, and a is obtained by satisfying the equation for the theoretical precipitation rate originating from particles with size distribution nX(D) (Sempere-Torres et al., 1994, 1998; Uijlenhoet, 2008):

i¼

p 6

ZN 0

nX ðDÞUX ðDÞD3 dD;

X ¼ R; H; S

where UX(D) is given by Equations (21)–(23). Note, however, that the units of nX depend on those used for D and i and, of course, the functional form of g(x).

ð27Þ

Clouds owe their existence to the process of condensation, which occurs in response to several dynamical processes associated with motions of air masses, such as orographic or frontal lifting (see Section 2.02.2.5), convection, and mixing. At the same time, clouds and the resulting precipitation influence the dynamical and thermodynamical processes in the atmosphere. For example, clouds affect air motions through physical processes, such as the redistribution of atmospheric water and water vapor, the release of latent heat by condensation, and the modulation of the transfer of solar and infrared (IR) radiation in the atmosphere. A cloud system is formed by a number of recognizable isolated cloud elements that are identifiable by their shape and size (e.g., Scorer and Wexler, 1963; Austin and Houze, 1972; Orlanski, 1975). On the lowest extreme, cloud systems with a scale of about 1 km or less are classified as microscale systems. On the highest extreme, atmospheric phenomena of linear extent of 1000 km and upward are classified in the synoptic scale and include the cloud systems associated with baroclinic instabilities, and extratropical cyclones (i.e., lowpressure centers). In between these two extreme scales, atmospheric phenomena with linear extent between a few kilometers and several hundred kilometers are the so-called mesoscale phenomena. These phenomena are more likely associated with atmospheric instabilities, as well as frontal and topographic lifting. Mesoscale phenomena include many types of clouds and cloud systems that are usually classified into two main categories: stratiform and convective (cumulus) cloud systems. In general, stratiform cloud systems have the shape of a flat appearing layer and produce widespread precipitation associated with large-scale ascent, produced by frontal or topographic lifting, or large-scale horizontal convergence. By contrast, convective cloud systems have large vertical development, produce localized showery precipitation, and are associated with cumulus-scale convection in unstable air. Next, we focus on the structure of these systems and the forms of precipitation they produce.

2.02.2.4.1 Cumulus cloud systems Cumulus clouds are formed by small thermals (upwardmoving air parcels heated by contact to the warm ground) where condensation occurs and they grow to extend vertically throughout the troposphere. Their vertical extent is controlled by the depth of the unstable layer, while their horizontal extent is comparable to their vertical extent. A typical linear dimension of a cumulus cloud is 3–10 km, with updraft velocities of a few meters per second (Rogers and Yau, 1996). Observations performed by Byers and Braham (1949; see also Weisman and Klemp, 1986) revealed that convective storms are formed by a number of cells, each one of which passes through a characteristic cycle of stages (Figure 15). The cumulus stage of a cell is characterized by an updraft throughout most of the cell. At this stage, which lasts approximately 10–20 min, the cell develops and expands

Precipitation

47

z ≈ 10−12 km

Rain

Rain z ≈ 6 km

New cell development Updraft

Downdrafting air

Updraft

z = 0 km 6−8 km

8−16 km

6−11 km

Cumulus stage 10−20 min

Mature stage 15−30 min

Dissipating stage ≈ 30 min

Figure 15 Stages of development of convective cells. Adapted from Weisman ML and Klemp JB (1986) Characteristics of isolated convective storms. In: Ray PS (ed.) Mesoscale Meteorology and Forecasting, ch. 15, pp. 331–358. Boston, MA: American Meteorological Society.

vertically while the air becomes saturated and hydrometeors grow due to vapor condensation and turbulent coalescence (see Section 2.02.2.2). Some ice and water particles grow large enough to fall relative to the ambient updraft and initiate a downdraft within the cell. The downdraft is initially in saturated condition, but as it moves toward the lower troposphere and mixes with subsaturated air, evaporational cooling occurs, which introduces negative buoyancy and accelerates the downdraft. This is the start of the mature stage of the cell, which lasts for approximately 15–30 min. The air of the downdraft reaches the ground, as a cold core, and changes the surface wind pattern. This change may initiate a new thermal at a neighboring location, which might grow to a new cell. The downdraft interferes with the updraft at the lower levels of the cloud and finally cuts off the updraft from its source region. At this point, the cell enters its dissipating stage. At this stage, which lasts for about 30 min, the updraft decays and consequently, the precipitation source is eliminated.

whereas thick stratus clouds (i.e., 1–2 km vertical extent) are capable of producing substantial widespread rain or snow. Although the classification of cloud systems in stratiform and convective is useful for observation purposes, it cannot be considered sharp (Harrold and Austin, 1974). Observations from radars or rain gauges show that widespread precipitation has a fine-scale structure with intense precipitation regions confined to elements with size of a few kilometers, while rainfall features of convective origin (e.g., cells) can grow and/ or cluster over a large region producing continuous precipitation similar to that of stratiform formations. In general, convective rainfall patterns are nonuniform and are associated with locally intense rainfall regions ranging in size from 3 to 10 km. The latter evolve rapidly in time and are separated by areas free of precipitation. By contrast, stratiform patterns are associated with less-pronounced small-scale structures and a wider overall extent that persists in time.

2.02.2.5 Precipitation-Generating Weather Systems 2.02.2.5.1 Fronts

2.02.2.4.2 Stratus cloud systems Stratus clouds are associated with mesoscale, or even synoptic, vertical air motions that arise from large-scale horizontal convergence and frontal or orographic lifting of moist air masses. The ascending motion of air is weak (i.e., a few tens of centimeters per second) relative to cumulus convection, but it extends over large areas and durations to produce widespread rain or snow. The lifetime of a stratus formation is of the order of days, and its size may extend over hundreds of kilometers horizontally. The ascended air masses, having the form of a flat appearing layer, remain convectively stable even after they are lifted to higher altitudes. Since atmospheric turbulence is not intense, initiation of rain is mainly dominated by the ice particle growth due to vapor deposition (the Bergeron–Findeisen mechanism; Section 2.02.2.2), when the ascended air masses are thick enough to reach subfreezing temperatures. In general, thin stratus clouds are usually nonprecipitating,

Atmospheric circulation is formed by advecting air masses with fairly uniform characteristics. Depending on their source of origin, different air masses may have different temperatures and moisture contents. For example, continental air masses are drier and their temperatures vary in a wider range relative to maritime air masses. The interface of two opposing air masses with different temperatures and moisture contents is usually referred to as a front. Along this interface, the warmer and lighter air rises above the colder and denser air. The vertical lifting causes the warmer air to cool adiabatically, the water vapor to condense, and, hence, precipitation to form. A cold front occurs when advancing cold air wedges itself under warmer air and lifts it (Figure 16(a)), whereas a warm front develops when faster-moving warm air overrides a colder and denser air mass (Figure 16(b); Koutsoyiannis and Xanthopoulos, 1999). An occluded front forms when warm air is trapped between two colder and denser air masses. An example of an occluded front is shown in Figure 16(c), where a cold front catches up a slower-moving warm front.

48

Precipitation

Upper-level winds

Cold front Warm air

Advancing cold air (a)

Warm front

Advancing warm air

Retreated cold air Cold air

(b)

Warm air Warm front Cold front Cold air Advancing cold air

Occluded front

(c)

Figure 16 Schematic illustration of different types of fronts: (a) cold front, (b) warm front, and (c) occluded front. Adapted from Koutsoyiannis D and Xanthopoulos T (1999) Engineering Hydrology, 3rd edn., p. 418. Athens: National Technical University of Athens (in Greek).

Fronts may extend over hundreds of kilometers in the horizontal direction and are associated with vertical wind speeds of the order of a few tens of centimeters per second. This range of values is in accordance with vertical motions caused by the horizontal wind convergence of synoptic-scale low-level flow. Hence, frontal precipitation is mostly stratiform with widespread rain or snow over large areas and durations. Note, however, that embedded within the areas of frontal precipitation there are mesoscale regions that exhibit cellular activity.

2.02.2.5.2 Mechanical lifting and orographic precipitation Orographic precipitation occurs when horizontally moving warm and humid air meets a barrier such as a mountain range. In this case, the barrier causes uplift of the incoming air. As the moist air moves upslope, it cools adiabatically, water vapor condenses to liquid water or ice (depending on the altitude where the dew point temperature occurs), and precipitation is likely to form (e.g., Smith, 1993; Hemond and Fechner-Levy, 2000). In general, orographic precipitation (unless combined with other mechanisms such as cyclonic activity and fronts) is narrow banded since it occurs in association with water-vapor condensation by mechanical lifting, a process that becomes

effective at a certain elevation along the topography. After surpassing the top of the mountain range, on the lee side, the air moves downward and this causes adiabatic warming, which tends to dissipate the clouds and stop the precipitation, thus producing a rain shadow.

2.02.2.5.3 Extratropical cyclones Extratropical cyclones are synoptic scale low-pressure systems that occur in the middle latitudes (i.e., pole-ward of about 301 latitude) and have length scales of the order of 500–2500 km (e.g., Hakim, 2003). They usually form when two air masses with different temperatures and moisture contents that flow in parallel, or are stationary, become coupled by a preexisting upper-level disturbance (usually a low-pressure center) near their interface. An example is the formation of extratropical cyclones along the interface of mid-latitude westerlies (i.e., winds that flow from West to East; e.g., Lutgens and Tarbuck, 1992), with the equator-ward-moving polar, and thus colder, air masses (i.e., polar easterlies). As shown in Figure 17, which refers to the Northern Hemisphere, the motion of both warm and cold air masses is caused by pressure gradients and their direction is south–north and north–south, respectively (Koutsoyiannis

Precipitation

49

er lie s

High ld

air

la r

ea

st

Co

Low

Po

Cold air

Warm

t

on d fr

`

Col

rli es

Warm air

te

r

.w es M id la t

(a)

front

m ar

ai

W

(b)

High

Cold air Low

Low

Co

ld

m

ar

air

W

Occluded front air

Warm front

(c)

(d)

m

ar

air

W

Wid stra espre a tifo rm d reg io

n

ar m W

Co l

d

air

ai

r

Great Britain

Band of precipitation

France

(e) Figure 17 (a)–(d) Schematic illustration of the evolution of an extratropical cyclone at the interface of mid-latitude westerlies and the equator-wardmoving polar easterlies; and (e) extra-tropical cyclone over the British Isles on 17 January 2009: motion of air masses, fronts, and characteristic precipitation regions. (a)–(d), Adapted from Koutsoyiannis D and Xanthopoulos T (1999) Engineering Hydrology, 3rd edn., p. 418. Athens: National Technical University of Athens (in Greek); and (e) from http://www.ncdc.noaa.gov/sotc/index.php?report ¼ hazards&year ¼ 2009&month ¼ jan).

50

Precipitation

Over shooting top

Storm

motion

12−16 km Upper-level winds anvil vau lt

9 km

Rain and hail

3−6 km Mid-level winds

Forward flank downdraft

Rear flank downdraft Tornado

Gust front Figure 18 Schematic illustration of the wind circulation in a super-cell storm. Adapted from http://www.nssl.noaa.gov/primer/tornado/.

and Xanthopoulos, 1999). However, these directions are diverted to the right (in the Northern Hemisphere) by Coriolis forces. The initial disturbance formed by the shear along the interface of the two air masses (Figure 17(a)) grows as the warmer and lighter air rises above the colder air and starts rotating in an emerging spiral called the cyclone (Figure 17(b)). As the cyclone evolves, the cold front approaches the slower-moving warm front (Figure 17(c)) and then catches up with it forming an occluded front (Figure 17(d)). Finally, mixing between the two air masses causes the fronts to lose their identities and the cyclone to dissipate. The adiabatic cooling of the warm and moist air results in a widespread region of stratiform precipitation that propagates with the upper-level flow far beyond the fronts (Figure 17(e)).

2.02.2.5.4 Isolated extratropical convective storms

The super-cell storm is the most intense of all isolated convective storms. It has a lifetime of several hours, it exhibits large vertical development, and produces strong winds, heavy rainfall, or hail, and long-lived tornadoes, that is, intense vortices with diameters of the order of 100–500 m (e.g., Browning and Ludlam, 1962; Rotunno, 1986; Weisman and Klemp, 1986; Bluestein, 2003), where the updrafts and downdrafts are displaced horizontally and interact mutually to sustain a long-lived circulation (Figure 18). The updraft enters at low levels and ascends in a region called the vault, which might penetrate into the stratosphere. Super-cell storms usually evolve from multicell formations when the magnitude of the vertical wind shear, defined as the difference between the density-weighted mean wind over the lowest 6 km and a representative surface layer wind (e.g., 500 m mean wind), suffices to produce a long-lived rotating updraft that mutually interacts with the downdraft (e.g., Weisman and Klemp, 1982, 1984, 1986).

A short-lived single-cell is the simplest storm of convective origin. Single cells have horizontal cross sections of the order of 10–100 km2 and move with the mean environmental flow over the lowest 5–7 km of the troposphere. The stages of development of a single-cell storm were discussed in Section 2.02.2.4. The multicell storm is a cluster of short-lived single cells with cold outflows (i.e., downdrafts) that combine to form a large gust front (Weisman and Klemp, 1986). The convergence along the leading edge of the front triggers new updraft development and subsequent formation of new cells. Because of the new cell development, multicell storms may last several days and span over large areas with linear extents of hundreds of kilometers.

Intense rainfall events are usually organized in lines (i.e., squall lines) and bands (i.e., rainbands) with characteristic scales of hundreds of kilometers. According to Hane (1986), rainbands are sufficiently elongated rainfall areas that are nonconvective or weakly convective, and squall lines include all linear convective structures stronger than rainbands. These large-scale features are considered to be manifestations of the large mesoscale horizontal circulation, in association with spatial fluctuations of the surface temperature and moisture content of atmospheric air masses.

2.02.2.5.5 Extratropical squall lines and rainbands

Precipitation

The conditions for squall line formation are (1) a convectively unstable near-surface environment (i.e., moist and warm near-surface air with relatively cold air aloft) to maintain the development of convective cells, (2) a layer of dry air directly above the near-surface moist air to enhance development of an intense and wide cold downflow by evaporative cooling (i.e., the dry middle-level air causes precipitation particles to evaporate and a negatively buoyant cold front to form), and (3) a triggering mechanism for release of the convective instability (e.g., frontal or orographic lifting). Once the squall line has formed, it feeds itself through convergence along the cold gust front. This convergence produces strong ascent and forms new cells ahead of the storm. Rainbands in extratropical regions occur primarily in association with well-organized extratropical cyclones (Hane, 1986). In this case, precipitation is maintained by the ascent resulting from the warm advection of the advancing cyclone, with subsequent formation of a widespread region of stratiform precipitation (Section 2.02.2.5.3). Extratropical rainbands can also be formed in synoptic-scale environments other than those associated with cyclonic circulation. An example is the environment associated with the development of symmetric instabilities (e.g., Bennetts and Sharp, 1982; Seltzer et al., 1985).

2.02.2.5.6 Monsoons The term monsoon generally applies to climates that exhibit long, distinct, and remarkably regular rainy and dry periods associated with the spatial distribution of solar heating during summer and winter. According to a definition proposed by Ramage (1971), a monsoon climate is characterized by (1) prevailing wind directions that shift by at least 1201 between January and July, (2) prevailing wind direction that persists at least 40% of the time in January and July, (3) mean wind speeds that exceed 3 m s1 in either January or July, and (4) fewer than one cyclone–anticyclone alternation every 2 years in either January or July in a 51 latitude–longitude rectangle. In essence, Ramage’s (1971) criteria exclude most extratropical regions with prevailing synoptic-scale cyclonic and anticyclonic circulations and, in addition, require the mean wind direction to be driven and sustained exclusively by the seasonally varying temperature contrast between continental and oceanic masses. Under these constraints, only India, SouthEastern Asia, Northern Australia, and West and central Africa have monsoon climates (Slingo, 2003). For example, in India, about 80% of the mean annual rainfall accumulation (about 2 m) occurs during the months of June, July, and August (Smith, 1993). The main driving mechanism for monsoons is the temperature contrast between continental and oceanic masses due to the seasonal cycle of solar heating. More precisely, the lower thermal inertia of continental masses relative to oceans causes the former to heat up more rapidly during spring and summer by the solar radiation. This results in a sharp temperature gradient, which causes a humid flow of oceanic near-surface air to move toward the land (something similar to a massive sea breeze). As it reaches the land, the humid air warms up and rises, water vapor condenses to liquid water, and rain falls. A similar process occurs during winter, when the continental

51

air masses cool up more rapidly than the surrounding ocean water, with subsequent formation of a cold and dry massive low-level flux toward the ocean. An important factor that determines the intensity of monsoon rainfall is the geographical orientation of continents and oceans relative to the equator (Slingo, 2003). For example, the north–south orientation of the South-Eastern Asian and Northern Australian monsoon system allows the dry outflow from the winter continent to warm up and load moisture from the ocean, flow across the equator toward the summer hemisphere, and, eventually, feed the monsoon rains over the summer continent. This is also the reason why the largest rainfall accumulations for durations larger than 24 h are associated with the Asian–Australian monsoon system (Smith, 1993).

2.02.2.5.7 Tropical cyclones Tropical cyclones form a particular class of synoptic-scale lowpressure rotating systems that develop over tropical or subtropical waters (Anthes, 1982; Landsea, 2000). These systems have linear extent of the order of 300–500 km and are characterized by well-organized convection and cyclonic (counterclockwise in the Northern Hemisphere) surface wind circulation around a relatively calm low-pressure region, called the eye of the storm (Figures 19 and 20). Tropical cyclones with sustained wind speeds in the range 17–32 m s1 are called tropical storms whereas stronger tropical cyclones are usually referred to as hurricanes (i.e., when observed in the North Atlantic Ocean, in the Northeast Pacific Ocean east of the dateline, and in the South Pacific Ocean east of 1601 E) or typhoons (i.e., when observed in the Northwest Pacific Ocean west of the dateline). Note, however, that extreme rainfall accumulations for durations of the order of a day, or higher, are usually produced by moderate or even low-intensity tropical cyclones (Langousis and Veneziano, 2009b). An example is the tropical storm Allison in 2001, which looped over the Houston area causing rainfall accumulations in excess of 850 mm. According to the US National Oceanic and Atmospheric Administration (NOAA; Stewart, 2002), Allison (2001) ranks as the costliest and deadliest tropical storm in the history of the US with 41 people killed, 27 of who were drowned, and more than $6.4 billion (2007 USD) in damages. The genesis and development of tropical cyclones require the following conditions to be maintained (e.g., Gray, 1968, 1979): (1) warm ocean waters (surface temperature T427 1C); (2) a conditionally unstable atmosphere where the air temperature decreases fast with height; (3) a relatively moist midtroposphere to allow the development of widespread thunderstorm activity; (4) a minimum distance of about 500 km from the equator in order for the Coriolis force to be sufficiently large to maintain cyclonic circulation; (5) a nearsurface disturbance with sufficient vorticity and low-level convergence to trigger and maintain cyclonic motion; and (6) low magnitude of vertical wind shear (less than 10 m s1), defined as the difference between the 200- and 850-hPa horizontal wind velocities in the annular region between 200 and 800 km from the tropical cyclone center (Chen et al., 2006). The latter condition is important for the maintenance of the deep convection around the center of the cyclone.

52

Precipitation Tropopause, Z ≈ 15 km Main vortex top Z ≈ 10 km

Divergence region

Outflow Eye

Cloud

Rain Main vortex

Vertical wind

Boundary layer top, H ≈ 1−2 km Surface boundary, Z=0

Inflow R = 0 R ≈ 15−40 km

Boundary layer

R ≈ 150−200 km

Figure 19 Schematic representation of the structure of a mature hurricane.

30

d

an

b ain r-r

Storm motion

te

Ou

29

≥18 16

Inner-rainbands

14

Latitude (deg)

28

12 10 27

8 6 4

26

25

Eyewall

2 0 (mm h−1)

Out

er-r ainb

and Katrina (2005)

24 −92

−91

−90

−89

−88

−87

−86

Longitude (deg) Figure 20 TRMM microwave imager (TMI) rainfall retrievals for Hurricane Katrina on 28 August (2005) at 21:00 UTC (frame 44373): different types of rainbands and their locations relative to the center of the storm.

At a first approximation, a tropical cyclone can be seen as a heat engine fueled by the buoyant motion of warm and saturated (hence convectively highly unstable) air masses that lie directly above the warm tropical and subtropical ocean waters (e.g., Emanuel, 1986, 1989; Renno and Ingersoll, 1996; Marks, 2003). By contrast, extratropical cyclones obtain their energy

from the horizontal temperature gradients in the atmosphere (Section 2.02.2.5.3). During its mature stage, a tropical cyclone includes four distinct flow regions (Yanai, 1964; Smith, 1968; Frank, 1977; Willoughby, 1990; Smith, 2000), as depicted in Figure 19:

Precipitation

1. Away from the surface boundary (in the altitude range from 2–3 km to about 10 km), frictional stresses are negligible and the horizontal winds are in approximate gradient balance (e.g., La Seur and Hawkins, 1963; Hawkins and Rubsam, 1968; Holland, 1980; Willoughby, 1990, 1991; Vickery et al., 2000). In this region, usually referred to as the main vortex, the radial inflow is negligible, whereas the tangential flow is maintained by the balance between the inward-directed pressure gradient force and the sum of the outward-directed centrifugal and Coriolis forces. 2. Within the boundary layer (in the altitude range below 1– 2 km), frictional stresses decelerate the tangential flow, reduce the magnitude of the Coriolis and centrifugal forces, and result in an inward net force that drives low-level convergence. Calculations performed by Smith (1968, 2003), Kepert (2001), Kepert and Wang (2001), and Langousis et al. (2008) show that the radial inflow in the boundary layer turns upward before it reaches the tropical cyclone center causing vertical fluxes of moisture. Langousis and Veneziano (2009a) showed that these fluxes can be used to obtain accurate estimates for the large-scale mean rainfall intensity field in tropical cyclones as a function of the tropical cyclone characteristics. 3. At altitudes in excess of about 10 km, the curved isobars, which are responsible for the tropical cyclone formation and maintenance, start to flatten. As a consequence, the inward-directed pressure gradient force that maintains the cyclonic circulation decreases with increasing height leading to an outward-directed net force that drives high-level divergence. 4. Finally, there is a core flow region, called the eye of the tropical cyclone, with diameters of the order of 15–40 km. This region is free of cloud with light tangential winds and a downflow close to the axis. The condensation of water vapor caused by the ascending motion of humid near-surface air leads to the formation of cloud systems. These systems, which are usually precipitating, are organized around the cyclone center into long quasi-circular formations usually referred to as rainbands. Despite variations of rainband characteristics from one storm to another, and during the evolution of a single storm (e.g., Miller, 1958; Barnes et al., 1983; Marks, 1985; Molinari et al., 1999), a number of studies (Willoughby et al., 1984; Powell, 1990; Molinari et al., 1994, 1999, among others) have shown that rainbands, depending on their location relative to the storm center, share similar structural characteristics and can be organized into three distinct classes: eyewall, inner-rainbands, and outer-rainbands (Figure 20): 1. The eyewall is a well-developed convective band that surrounds the eye of the tropical cyclone. This band has a width of approximately 10–15 km with upward-directed quasi-steady velocities in the range of 0.5–3 m s1 or more, with the larger values being associated with more intense systems. The quasi-steady updrafts mostly reflect the radial convergence of horizontal fluxes, which become maximum close to the eye of the tropical cyclone (Smith, 1968; Shapiro, 1983; Kepert, 2001). The eyewall almost always

53

has the highest cloud tops (Jorgensen, 1984a), contains the largest annular mean rainfall intensity (Marks, 1985; Houze et al., 1992), and exhibits weak cellular structure as evidenced by radar observations (e.g., Jorgensen, 1984b; Marks, 1985). 2. The inner-rainbands (Molinari et al., 1994, 1999) are a group of spiral bands located outside the eyewall at radial distances smaller than approximately 120 km, and are also referred to as a stationary band complex (Willoughby et al., 1984). This group moves slowly, if at all, and maintains a rather fixed position relative to the vortex. Rainfall inside the inner-rainband region is mostly stratiform, with active convection covering 5–10% of the total rainfall area and contributing 40–50% of the total rainfall volume (e.g., Marks, 1985; Marks and Houze, 1987; Marks et al., 1992). 3. Outer-rainbands typically occur at radial distances larger than approximately 150 km from the tropical cyclone center (e.g., Powell, 1990; Molinari et al., 1994). They develop by the increased convergence at the boundary of the vortex envelope, where the convectively unstable environmental air flows around the storm and gives rise to formation of convective cells (e.g., Beer and Giannini, 1980; Ooyama, 1982; Molinari et al., 1994). Consequently, outerrainbands have more cellular structure than inner ones, which develop in a less-unstable atmosphere.

2.02.3 Precipitation Observation and Measurement 2.02.3.1 Point Measurement of Precipitation 2.02.3.1.1 Measuring devices The measurement of precipitation at a point is as easy as placing a bucket at the point of observation and periodically measuring the quantity of water it collects. The collected volume divided by the area of the opening is the precipitation depth. Due to this simplicity, such gauges have been used systematically since many centuries, and must have been discovered independently in different times, perhaps even in the antiquity, and in different places in the world, such as in ancient Greece and ancient India (Kosambi, 2005). However, their records have not survived; so the oldest available records now are those in Seoul, Korea, already presented in Section 2.02.1.3 and Figure 7 (upper), which go back to 1770, even though measurements must have been taken in much earlier periods since 1441 (Arakawa, 1956). The traditional device for rainfall measurement, known as rain gauge or pluviometer, is still in use today and, in fact, remains the most accurate device also providing the calibration basis for new measurement devices and techniques. It is a simple cylinder whose opening has an area (e.g., 200– 500 cm2 according to World Meteorological Organization (1983)) larger than (e.g., 10-fold) the cross section of the cylinder, which allows a greater sensitivity of the reading of the rainfall depth in a millimetric ruler attached to the cylinder. In another type of instrument, known as cumulative gauge, which is placed in inaccessible areas, the diameter of the cylinder may be larger than that of the opening, to enable the storage of a large volume of precipitation between the times of two visits to the place.

54

Precipitation

In an autographic (or recording) rain gauge, also known as a pluviograph, the water depth in the cylinder is recorded with the help of a mechanism involving a floating device. Another type of recording gauge, known as a tipping-bucket gauge, introduces the rainwater to one of a pair of vessels with a known small capacity (typically equivalent to 0.2 mm of rainfall) that is balanced on a fulcrum; when one vessel is filled, it tips and empties, while the time of this event is recorded, and the other vessel is brought into position for filling. In traditional autographic devices, these recordings are done on a paper tape attached to a revolving cylinder driven by a clockwork motor that is manually wound. In modern instruments, this device is often replaced by electronic systems, which provide digital recordings on a data logger and/or a computer connected by a cable or radio link. A rain gauge does not include all precipitation forms, snow in particular, except in light snowfalls when the temperature is not very low and the snow melts quickly. Generally, accurate measurement of snow precipitation (the water equivalent) needs specific instruments, equipped with a heating device to cause melting of snow. If such an instrument is not available, the snow precipitation is estimated as 1/10 of the snowfall depth (see justification in Section 2.02.1.3).

2.02.3.1.2 Typical processing of rain gauge data Measurement of precipitation in rain gauges is followed by several consistency checks to locate measurement errors and inconsistencies. Errors are caused due to numerous reasons, including human lapses and instrument faults, which may be systematic in case of inappropriate maintenance. Inconsistencies are caused by changes of installed instruments, changes in the environmental conditions (e.g., growing of a tree or building of a house near the rain gauge), or movement of the gauge to a new location. When errors are detected, corrections of the measurements are attempted. The standard meteorological practices include checks of outliers (a measured value is rejected if it is out of preset limits), internal consistency (checks are made whether different variables, e.g., precipitation and incoming solar radiation are compatible with each other), temporal consistency (the consistency of consecutive measurements is checked), and spatial consistency (the consistency of simultaneous measurements in neighboring stations is checked). Such checks are done in the timescale of measurement (e.g., daily for pluviometers or hourly for pluviographs) but systematic errors can only be located at aggregated (e.g., annual) timescales. The most popular method applied at an aggregated timescale for consistency check and correction of inconsistent precipitation data is that of the double-mass curve, which is illustrated in Figure 21. The method has a rather weak statistical background and is rather empirical and graphical (but there is a more statistically sound version in the method by Worsley (1983)). The double-mass curve is a plot of the successive cumulative annual precipitation Syi at the gauge that is checked versus the successive cumulative annual precipitation Sxi for the same period of a control gauge (or the average of several gauges in the same region). If the stations are close to each other and lie in a climatically homogeneous region, the annual values should correlate to each other. A fortiori, if the

two series are consistent with each other, the cumulated values Syi and Sxi are expected to follow a proportionality relationship. A departure from this proportionality can be interpreted as a systematic error or inconsistency, which should be corrected. Such a departure is usually reflected in a change in the slope of the trend of the plotted points. The aggregation of annual values xi and yi to calculate Sxi and Syi is typically done from the latest to the oldest year. Figure 21 (upper) shows the double-mass curve for 50 pairs of values representing annual precipitation at two points, whose cross-correlation (between xi and yi) is 0.82. The newest 25 points form a slope of m ¼ 0.70, whereas the oldest 25 form a much greater slope, m0 ¼ 0.95. Assuming that the newest points are the correct ones (with the optimistic outlook that things are better now than they were some years before), we can correct the older 25 annual yi by multiplying them with the ratio of slopes, l ¼ m/ m0 : ¼ 0.737. A second double-mass curve, constructed from the corrected measurements, that is, from y0i :¼ yi for ip25 (the newest years) and y0i :¼ lyi for i425 (the oldest years) is also shown in Figure 21 (upper). In fact, the data values used in Figure 21 are not real rainfall data but rather are generated from a stochastic model (Koutsoyiannis, 2000a, 2002) so that both stations have equal mean and standard deviation (1000 and 250 mm, respectively), be correlated to each other (with correlation coefficient 0.71) and, most importantly, exhibit HK behavior (with H ¼ 0.75, compatible with the values found in the real-world examples of Section 2.02.1.5). Hence, evidently, all values are correct, consistent, and homogeneous, because they were produced by the same model assuming no change in its parameters. Thus, the example illustrates that the method can be dangerous, as it can modify measurements, seemingly inconsistent, which however are correct. While this risk inheres even in time-independent series, it is largely magnified in the presence of HK behavior. Figure 21 (lower) provides a normal probability plot of the departure of the ratio l from unity (where the horizontal axis z is the standard normal distribution quantile and the distributions were calculated by the Monte Carlo method) for two cases: assuming independence in time and assuming HK behavior with H ¼ 0.75 as in the above example. The plots clearly show that, for the same probability, the departure of l from unity in the HK case is twice as high as in the classical independence case. For the HK case, departures of 70.25 from unity appear to be quite normal for 25-year trends and even more so for finer timescales, that is, 70.35 to 70.40 for 10-year to 5-year consecutive trends (not shown in Figure 21). Note that the method is typically applied even for corrections of as short as 5-year trends (Dingman, 1994), and so its application most probably results in distortion rather than correction of rainfall records. Apparently, the correction of the series using the doublemass curve method removes these trends that appear in one of the two time series. Removal of trends results in reduction of the estimated Hurst coefficient or even elimination of the exhibited HK behavior (Koutsoyiannis, 2003a, 2006b). Thus, if we hypothesize that the HK behavior is common in precipitation, application of methods such as the double-mass curves may have a net effect of distortion of correct data, based on a vicious circle logic: (1) we assume time independence of

Precipitation

55

50 000 Data Fitted broken line Adjusted data Adjusted straight line

40 000

Generation model Σyi (mm)

30 000

20 000

10 000

0 0

10 000

20 000 30 000 Σ xi (mm)

40 000

50 000

0.4 H = 0.5 (classical independence case) H = 0.75 (HK case)

Departure of slope ratio from 1, λ−1

0.3

0.2

0.1

0

−0.1

−0.2

−0.3 −3

−2

−1

0

1

2

3

Standard normal variate, z Figure 21 Illustration of the double-mass curve method and the associated risks in applying it. (Upper): Typical double-mass curve for 50 pairs of points, where the first 25 (newest) and the last 25 (oldest) form slopes m ¼ 0.7 and m0 ¼ 0.95, respectively; the adjusted points with l ¼ m/m0 ¼ 0.737 are also shown. (Lower): Comparison of probability distributions of the departure of the ratio l from unity for series independent in time or with HK behavior with H ¼ 0.75; the distributions were calculated using the Monte Carlo method based on synthetic series with a total size of 1000.

the rainfall process; (2) we interpret manifestation of dependence (the HK behavior in particular) as incorrectness of data; (3) we modify the data so as to remove the influence of dependence; and (d) we obtain a series that is much closer to our faulty assumption of independence. The widespread use of the double-mass curve method in routine processing of

precipitation time series may thus have caused enormous distortion of the real history of precipitation at numerous stations worldwide, in addition to masking HK behavior. The above discourse aims to issue a warning against unjustified use of consistency check and correction methods that could eliminate the extreme values (see, e.g., the note about

56

Precipitation

the Seoul station in Section 2.02.1.3) and the long-term variability implied by the HK behavior; the effect of both these mistreatments of data is a serious underestimation of design precipitation and flow in engineering constructions and management decisions. As a general advice for their correct application, we can stress that all methods of this type should never be applied blindly. An inspection of local conditions (environment of the rain gauge station and practices followed by the observer) as well as of the station’s archive history is necessary before any action is taken toward altering the data. Unless information on local conditions and archive history justify that inconsistencies or errors exist, corrections of data should be avoided.

2.02.3.1.3 Interpolation and integration of rainfall fields The interpolation problem, that is, the estimation of an unmeasured precipitation amount y from related precipitation quantities xi (i ¼ 1, y, n) is encountered very often in routine hydrologic tasks, such as the infilling of missing values of recorded precipitation at a station or the estimation of precipitation at an ungauged location. The integration problem refers to the estimation of an average quantity y over a specified area (or time period) based on measurements xi (i ¼ 1, y, n) of the same quantity and the same time period at different points (or respectively, at different time periods at the same point). The literature provides a huge diversity of methods, most of which, however, could be reduced to a linear statistical relationship applicable to both the interpolation and the integration problem:

y ¼ w1 x1 þ y þ wn xn þ e

ð28Þ

where wi denotes a numerical coefficient (weighting factor) and e denotes the estimation error. The same could be written in vector form:

y ¼ wTx þ e

ð29Þ

with w :¼ [w1, y, wn]T and x :¼ ½x1 ; ? ; xn T , and the superscript T denotes the transpose of a vector (or a matrix). The notation in Equations (28) and (29) suggests that x, y, and e are treated as random variables, even though this may not be necessary in some of the existing methods. All interpolation techniques provide a means for estimating the numerical coefficients wi, either conceptually or statistically, whereas the statistical methods provide, in addition, information about the error. Most commonly, the latter information includes the expected value me :¼ E½e and its standard deviation se :¼ ðVar½eÞ1=2 . A statistical estimation in which E½e ¼ 0 is called unbiased, and one in which the mean square error MSE :¼ E½e 2  ¼ s2e þ m2e , is the smallest possible, is called best; if both these happen, the estimation is called best linear unbiased estimation (BLUE). While the BLUE solution is in principle quite simple (see below), the estimation of its weighting factors is not always straightforward. Hence, several simplified statistical methods as well as empirical conceptual methods are in common use. Another reason that explains why such a diversity of methods has emerged is the different type of objects that each of the elements of the vector x represents. For instance, in temporal interpolation, these

elements can be observed values at times before and after the time of interpolation. In spatial interpolation these can be simultaneously observed values for stations lying in the neighborhood of the point of interpolation. Simultaneous temporal and spatial interpolation, although unusual, may be very useful. For example, an optimal way to infill a missing value in a time series at a specific time would be to include in x measurements taken in neighboring gauges at this specific time, as well as measurements taken at the point of interest at preceding and subsequent times. Let us first examine the different methods in which the estimation of y is based on a single observation x  xi at one neighboring (in space or time) point only. Here is a list of options, in which the following notation has been used: mx :¼ E½x and my :¼ E½y are the expected values of x and y, respectively; s2x :¼ E½ðx  mx Þ2  and a2y :¼ E½ðy  my Þ2  are the variances of x and y, respectively; sxy :¼ E½ðx  mx Þðy  my Þ is the covariance of x and y; and rxy :¼ sxy/(sx sy) is the correlation of x and y. 1. Equality: y ¼ x. The single point of observation considered in this naive type of interpolation is the station i nearest to the interpolation point, with x  xi. As discussed below, this simple interpolation forms the background of the Thiessen method of spatial integration. It is generally biased, with bias me ¼ my  mx and its MSE is s2y þ s2x  2sxy þ m2e . However, if the precipitation field is stationary (so that the means and variances at all points are equal to global parameters m and s2, respectively), it becomes unbiased, with MSE ¼ 2s2 (1  rxy). Evidently, for rxyo0.5, the method results in MSE 4 s2 and therefore there is no meaning in adopting it for low correlation coefficients (an estimate x ¼ m would be more effective). 2. Normal ratio: y ¼ w x with w ¼ my/mx. This is a better alternative to the equality case, but it requires a sample of measurements to be available for y in order to estimate the average my. This estimation is unbiased (me ¼ 0) but not best (MSE ¼ sy2 þ sx2 my2/mx2  2 sxy my/mx). 3. Homogenous linear regression: y ¼ wx with w ¼ E½y x= E½x 2  ¼ ðsxy þ mx my Þ=ðs2x þ m2x Þ. This is a biased estimation (me ¼ my  w mx) albeit best ðMSE ¼ s2y þ ðm2y s2x  2mx my sxy  s2xy Þ=ðm2x þ s2x ÞÞ. 4. Linear regression: y ¼ w x þ b with w ¼ Cov½y x=Var½x ¼ sxy =s2x and b ¼ my  w mx. This can be derived from Equation (28) by adding an auxiliary variable whose values are always 1 (i.e., y ¼ w x þ b 1). It has the properties of being both unbiased and best, with MSE ¼ s2y ð1  r2xy Þ. However, it has the deficiency of potentially resulting in negative values, if bo0, or of excluding values between 0 and b if b 4 0. Another drawback emerges when many values of y are estimated in an attempt to extend a record of y based on a longer record of x. In this case, the resulting extended record has negatively biased variance, because the method does not preserve variance. To remedy this, a random error e should be added (using the probability distribution of e), which however is not determined in a unique manner and makes the method no longer best. 5. Organic correlation: y ¼ wx þ b with w ¼ sign [rxy] sy/sx and b ¼ my  w mx. This preserves both mean (i.e., it is unbiased) and variance, but it is not best ðMSE ¼ 2s2y ð1  jrxy jÞ.

Precipitation Evidently, for |rxy|o0.5, the method results in MSE 4 s2y and therefore adopting it is pointless for low correlation coefficients. Similar to the standard linear regression, the organic correlation retains the deficiency of producing negative values or excluding some positive values. Coming to the interpolation based on multiple xi, in the simplest case, all weights wi are assumed equal for all i, that is, wi ¼ 1/n so that y is none other than the average of xi (the arithmetic mean method). This simple version is used very often to fill in sparse missing values of rain gauge records. The quantities xi could be simultaneous measurements at neighboring points (say, within a radius of 100 km), or at neighboring times, or both. Here, neighboring times should not necessarily be interpreted in the literal meaning, but with an emphasis on similarity of states. For example, a missing value of monthly precipitation in April 2000 could be estimated by, say, the average of the precipitation of the April months of 1998, 1999, 2001, and 2002. In another version, the average of all April months with available data are used, but a local average (as we have already discussed in Section 2.02.1.4) is preferable over an overall average, assuming that precipitation behaves like an HK process rather than a purely random one; this is similar to taking the average of points within a certain distance rather than a global average in spatial interpolation. This is not only intuitive but it can have a theoretical justification (D. Koutsoyiannis, personal notes), according to which for an HK process with H ¼ 0.7, a local average based on 3 time steps before and 3 after the interpolation time is optimal (produces lowest MSE); the optimal number of points becomes 2 þ 2 and 1 þ1 for H ¼ 0.75 and HX0.8, respectively. This simple method does not impose any requirement for calculation of statistical quantities for its application. Another method of this type, which takes account of the geographical locations and, in particular, the distances di between the interpolated stations, is the method of inverse distance weighting (IDW). In each of the basis stations, it assigns weights as

db wi ¼ Pk i b j¼1 dj

ð30Þ

where the constant b is typically assumed to be 2. Among methods whose application requires statistical quantities to be known, the simplest is a direct extension of the normal ratio method, in which wi ¼ ð1=nÞðmy =mxi Þ. The BLUE method itself belongs to this type. Initially, we can observe that a simple but biased solution for w in Equation (29) can be easily obtained as

w ¼ C 1 g;

me ¼ my  wT lx ;

s2e ¼ s2y  gT C1 g ¼ s2y  wT g

ð31Þ

where g :¼ Cov½y; x is the vector whose elements are the covariances of y with x (see Section 2.02.1.5) and C :¼ Cov½x; x is the positive definite symmetric matrix whose elements are the covariances of the vector x with itself. One way to make it unbiased is to add an auxiliary variable xnþ1 whose values are always 1. This is the multivariate extension of the typical linear regression described in point 4 of the previous list, and thus it retains the deficiency of potentially

57

producing negative values or excluding some positive values. A better way to make it unbiased is to add a constraint my ¼ wT lx (the bivariate analog of this is the equality case, described in point 1 of the list). In the latter case, the MSE becomes

MSE ¼ s2e ¼ s2y þ m2y þ wT ðC þ lx lTx Þw  2wT ðg þ my lx Þ

ð32Þ

Minimization of the MSE with the above constraint using a Lagrange multiplier –2l results in the system of equations

lTx w ¼ my

Cw þ lx l ¼ g;

ð33Þ

whose solution for the n þ 1 unknowns w1, y, wn, l is 0

w0 ¼ C 1 g0

ð34Þ

where

" 0

w :¼

# " w C 0 ; C :¼ l lTx

# " # lx g 0 ; g :¼ my 0

ð35Þ

The value of the error is then calculated as

MSE ¼ s2e ¼ s2y þ wT Cw  2wT g

ð36Þ

As seen in Equations (31) and (34), the application of the method requires a number of covariances to be estimated (specifically, this number is (n2 þ 3n)/2, given that C is symmetric). Not only does this restrict the method’s application to points where measurements exist, in order to estimate the covariances, but, when n is large, it is infeasible to reliably estimate so many parameters from data and to derive a positive definite C. The viable alternative is to assume a parametric stochastic model for the precipitation field. In the simplest case, the field could be assumed stationary and isotropic, where mxi ¼ my ¼ m; sxi ¼ sy ¼ s, and the covariance among any two points i, j is a function f of the geographical distance dij between these points, that is, sij : ¼ Cov[xi , xj ] ¼ f(dij). In this case, Equation (35) simplifies to

" 0

C :¼

C 1T

" # # " # 1 w g 0 0 ; w :¼ 0 ; g :¼ l 1 0

ð37Þ

where l0 ¼ l m and 1 is a vector with all its elements equal to 1. The last solution is widely known as ‘kriging’ (although kriging is sometimes formulated not in terms of covariance as in here, but in terms of the so-called semivariogram, a notion that is not appropriate for processes with HK behavior). We can observe from Equation (37) that the solution is now independent of m, as is also the error, which is still calculated from Equation (36). It only depends on the covariance function f(d). A function f(d) compatible with the HK behavior of precipitation, as discussed in Section 2.02.1.5, is of the form

f ðdÞ ¼ minðc; a d 4

H4

Þ

ð38Þ

where H is the Hurst coefficient and cb0 and a are parameters; in particular, c violates theoretical consistency but has

58

Precipitation

been introduced to avoid problems related to the infinite covariance for distance tending to zero. It can be observed that if the point of interpolation coincides with any one of the basis points i, then g is identical to one of the columns of C and g0 is identical to one of the columns of C0 . Thus, given the symmetry of C and C0 , from Equation (31) or (34), we obtain that w is a unit vector, that is, all elements are zero except one, which will be equal to 1. This shows the consistency of the method, that is, its property to reproduce the measurements at gauged points with zero error. All of the above methods that can interpolate at an arbitrary point (rather than only at a gauged one) provide a basis for numerical integration to find the average precipitation over a specific area A. Eventually, these methods result again in Equation (28) or (29), where now y is the areal average precipitation. In particular, in the arithmetic mean and the normal ratio methods, because they do not make any assumption about the position of the point to which interpolation refers, the estimate y is an interpolation at any point and a spatial average as well. The equality method works as follows: the geographical area of interest is divided into polygons, the so-called Thiessen polygons, each of which contains the points nearest to each of the stations. All points belonging to a specified polygon are regarded to have received a precipitation amount equal to that of the station corresponding to this polygon. Thus, in the integration, we use either Equation (28) or (29), where all gauged xi in the area are considered with weights wi ¼ Ai/A, whereas Ai and A are the areas of the polygon corresponding to xi. The remaining methods (IDW and BLUE) can be explicitly put in the form of Equation (29), but this is rather tedious if done analytically. A simpler alternative is to make interpolations to many points, for example, on a dense square grid. In turn, the gridded interpolations could be used for integration using equal weights for all grid points (i.e., arithmetic mean).

2.02.3.2 Radar Estimates of Precipitation Radio detection and ranging (radar) was developed at the beginning of World War II as a remote-sensing technique to measure the range and bearing of distant objects (such as ships and airplanes) by means of radio echoes (e.g., Battan, 1973). Since the early 1970s, radar techniques have also been used for the identification (i.e., shape, size, motion, and thermodynamic properties) of precipitation particles. The latter are weather-related distributed targets, which in contrast to ships and airplanes, have characteristics that evolve in time and depend on the atmospheric conditions. Because of their ability to provide estimates of areal precipitation quickly (i.e., at time intervals of about 5–15 min), at high resolutions (i.e., down to spatial scales of about 1 km) and over wide areas (i.e., with an effective range of about 200–400 km), radars have found wide application in atmospheric research, weather observation, and forecasting (e.g., Atlas et al., 1984; Doviak and Zrnic, 1993; Uijlenhoet, 1999, 2008; Bringi and Chandrasekar, 2001; Krajewski and Smith, 2002; Testik and Barros, 2007). An example is the next generation weather radar (NEXRAD) network with 159

operational weather surveillance radar 88 Doppler (WSR88D) units (as of February 2009), deployed throughout the continental United States and at selected locations overseas. According to NOAA’s weather service (US National Oceanic and Atmospheric Administration, 2009), since its establishment in 1988, the NEXRAD project has provided significant improvements in severe weather and flash flood warnings, air traffic control, and management of natural resources.

2.02.3.2.1 Basics of radar observation and measurement A typical weather radar has three main components (Battan, 1973): (1) the transmitter, which generates short pulses of energy in the microwave frequency portion of the electromagnetic spectrum, (2) the antenna, which focuses the transmitted energy into a narrow beam, and (3) the receiver, which receives the backscattered radiation from distant targets that intercept the transmitted pulses. Some important parameters, and their range of values, that characterize the radar equipment are (Rogers and Yau, 1996): (1) the instantaneous power of the pulse PtE10–103 kW (also referred to as peak power), (2) the duration of the pulse tE0.1–5 ms, (3) the frequency of the signal nE3–30 GHz, (4) the pulse repetition frequency (PRF) frE200–2000 Hz, defined as the reciprocal of the time interval tmax that separates two distinct pulses (i.e., tmax ¼ f 1 r E 0:5  5 ms), and (5) the beamwidth of the antenna yE11, defined as the angular separation between points where the power of the transmitted signal is reduced to half of its maximum value (or equivalently 3 dB below the maximum). The latter is attained at the beam axis. The wavelength l of the signal is defined as the distance between two sequential crests (or troughs) of the electromagnetic wave and it is related to its frequency as

ln ¼ c

ð39Þ

where c ¼ 3 108 m s1 is the velocity of light in vacuum. It follows from Equation (39) that typical frequencies n ¼ 3–30 GHz correspond to wavelengths l between 10 and 1 cm, but most weather radars operate at wavelengths l ¼ 3–10 cm (X-, C-, and S-band; see, e.g., Uijlenhoet and Berne, 2008). Shorter wavelengths are more effectively attenuated by atmospheric hydrometeors and precipitation particles (hence the transmitted signal has a small effective range), whereas for longer wavelengths the backscattered radiation from the precipitation particles does not have sufficient power to be detected by the receiver without noise induced by ground targets (e.g., Uijlenhoet, 2008). When conducting radar observations and measurements, the direction of the target is obtained from the azimuth and elevation of the antenna when the returning echo is received. The range r of the target is calculated from the relation

r ¼ c t=2

ð40Þ

where t is the time interval between the transmission of the pulse and the reception of the echo. If the target is moving, the radial velocity ur of the target (i.e., in the radar-pointing direction) can be obtained from the frequency shift Dn of the

Precipitation

received relative to the transmitted signal. The frequency shift is caused by the Doppler effect and it is related to ur as:

Dn ¼ 2ur =l

ð41Þ

with positive Dv being associated with targets that move toward the radar. If t (the time interval between transmission and reception) is larger than tmax (the reciprocal of the pulse repetition frequency, fr), the echo from the target will reach the receiver after a new pulse has been transmitted. Hence, targets that return enough energy to be detected by the receiver (see below) and are located at distances r4rmax ¼ c/(2fr), will appear unrealistically close to the antenna. Thus, rmax is the maximum range within which targets are indicated correctly on the radar screen and it is usually referred to as the unambiguous range (Battan, 1973; Rogers and Yau, 1996). The visibility of a target by the radar depends on whether the returning signal has sufficient power Pr to be detected by the receiver. As an example, we consider a point target (i.e., a target with linear dimension smaller than about 10% of l) with cross section At located at distance r from the radar. We suppose that the radar transmits pulses with peak power Pt that propagate isotropically in space (i.e., in a 3D sphere). It follows from simple geometric considerations that the power Pi intercepted by the target is

Pi ¼

Pt At 4pr 2

ð42Þ

where 4pr2 is the surface area of a sphere with radius r. If the transmitted signal is focused in a narrow beam by the antenna (as is commonly the case), Equation (42) becomes

Pt At Pi ¼ G 2 4pr

ð43Þ

2

where G ¼ (4p Ae)/l is a dimensionless constant called the antenna axial gain that depends on the characteristics (i.e., the wavelength l) of the signal and the aperture Ae of the antenna. Assuming that the target scatters the intercepted signal isotropically in space, the power Pr that reaches the radar is

Pr ¼

Pi Ae Pt At Ae l 2G 2 ¼G 2 ¼ Pt At ð4pÞ 3 r 4 2 4pr 2 ð4pr Þ

ð44Þ

If the power Pr is large enough to be detected by the receiver without unwanted echoes (e.g., noise from ground targets), the target is visible to the radar and it is indicated on the radar screen. For nonisotropic scatterers, the cross section of the target At should be replaced by the backscatter cross section s of the target. For spherical particles with diameter Dol/10, usually referred to as Rayleigh scatterers, s can be calculated from the relation (Battan, 1973)

s¼

p 5 jKj 2 D 6 l4

ð45Þ

59

where |K| is the amplitude of the complex refraction index (|K|2E0.93 for liquid water and 0.21 for ice), which characterizes the absorptive and refractive properties of the spherical scatterer. Due to the much higher value of |K|2 for liquid water relative to ice (about 4.5 times higher), the melting layer of ice particles in precipitation-generating weather systems appears on the radar screen as a bright band of high reflectivity.

2.02.3.2.2 Radar observation of distributed targets and the estimation of precipitation For a typical weather radar that operates in the C-band portion of the electromagnetic spectrum (l ¼ 3.75–7.5 cm), raindrops and snowflakes (i.e., particles with effective diameters Do5– 6 mm) can be approximated as Rayleigh spherical scatterers with backscatter cross section s given by Equation (45). However, there are reasons why atmospheric hydrometeors should not be treated as isolated point targets. One reason is that the pulse transmitted by the radar illuminates simultaneously, numerous precipitation particles that are included in a certain volume of air V, referred to as the resolution volume of the radar. Hence, the returned signal contains spatially averaged information from the whole population of raindrops and snowflakes in V. For parabolic antennas, where the beam pattern is approximately the same in all directions, an accurate estimate of V can be obtained by assuming that the resolution volume is a cylinder with effective height equal to half of the pulse length l ¼ c t and diameter dV ¼ r y, that is, the separation distance between points where the power of the transmitted signal is reduced to half of its maximum value. This gives

 2 ry ct V¼p 2 2

ð46Þ

where y is in radians. Equation (46) assumes that all energy in the radar transmitted pulse is contained within the half-power beamwidth; assuming a Gaussian shape of the beam pattern, the denominator of (46) (and, likewise, that of (49) below) should be multiplied by a factor 2 ln 2 (Probert-Jones, 1962). Another reason why raindrops and snowflakes cannot be treated as isolated point targets is that their turbulent motion that causes the power Pr of the returned signal to fluctuate in time. To this extent, an accurate approximation of the timeaveraged power Pr (over a sufficiently long interval of about 102 s), which accounts also for multiple backscattering cross sections, is given by (Rogers and Yau, 1996)

l 2G 2 X Pr ¼ Pt s ð4pÞ 3 r 4 V

ð47Þ

where r is the time-averaged range of the resolution volume V, and the summation is taken over all s in V. For Rayleigh scatterers, Equations (45) and (47) are combined to give

p 2 G 2 jKj 2 X 6 D Pr ¼ Pt 64r 4 l 2 V

ð48Þ

60

Precipitation

Assuming homogeneity of the population of hydrometeors in V, Equation (48) can be written as

Z N p 2 G 2 jKj 2 V nðDÞD6 dD Pr ¼ Pt 64r 4 l 2 0 p 3 G 2 jKj 2 y 2 ct jKj 2 Z Z¼C 2 ¼ Pt 2 2 512r l r

ð49Þ

where n(D) is the size distribution of precipitation particles in V (i.e., number of particles per unit diameter and per unit volume of air), C is the so-called radar constant that depends solely on the characteristics of the system under consideration, and

ZN Z :¼

nðDÞD 6 dD

ð50Þ

0

is the reflectivity factor with units (length3) that depend solely on the size distribution of the precipitation particles. For the Marshall and Palmer (1948) parametrization described by Equation (24), Equation (50) takes the form

Z ¼ 720 n0 b7

ð51Þ

where n0 and b are the intercept and scale parameters of the exponential size distribution. For the expressions given in Section 2.02.2.3.2, for rain and snow we obtain

ðaÞ Z ¼ 296 i 1:47 ðrainÞ ðbÞ Z ¼ 3902 i 2:49 ðsnowÞ

ð52Þ

where Z has units of mm6 m3 and i is the rainfall intensity (or the water equivalent of the accumulated snow at ground level) in millimeters per hour. For rain, Equation (52a) is very close to the empirical Z  i relationships (usually referred to as Z  R relationships, where Ri denotes the rainfall intensity) found in the literature (e.g., Marshall et al., 1955; Battan, 1973; Uijlenhoet, 1999, 2001, 2008), whereas for snow there is more variability and Equation (52b) should be seen only as an approximation. When combined, Equations (40), (49), and (52) allow conversion of radar measurements (i.e., Pr , t and r) to precipitation intensity i.

2.02.3.3 Spaceborne Estimates of Precipitation The history of observation of Earth from space started on 4 October 1957, when the Soviet Union successfully launched Sputnik-I, the first artificial satellite. Sputnik-I provided information on the density of the highest layers of the atmosphere and on the radio-signal distribution in the ionosphere. The first launch was immediately followed by the launch of Sputnik-II by the Soviet Union on 3 November 1957 and the launches of Explorer-I (1 February 1958), Vanguard-I (17 March 1958), Vanguard-II (17 February 1959), and TIROS-I (1 April 1960) by the United States of America. The success of TIROS-I in surveying atmospheric conditions (in particular, the cloud coverage of Earth) opened a new era for

meteorological research and development using spaceborne observations. Since the 1970s, meteorological satellites have become essential in studying the development and evolution of weatherrelated phenomena over the 71% of the Earth’s surface covered by sea, where other types of measurements are unavailable. For example, the Tropical Rainfall Measuring Mission (TRMM; Simpson et al., 1988; Kummerow et al., 1998), which started in November 1997 by the National Aeronautics and Space Administration (NASA) of the United States and the National Space Development Agency (NASDA) of Japan, has provided vast amounts of rainfall and energy estimates in tropical and subtropical regions and advanced the understanding and modeling of extreme rainfall events caused by tropical cyclones (e.g., Lonfat et al., 2004, 2007; Chen et al., 2006, 2007; Langousis and Veneziano, 2009a, 2009b). TRMM data have also been used to improve the accuracy of high-resolution weather forecasts produced by limited-area models (e.g., Lagouvardos and Kotroni, 2005) and to investigate the relationship between lighting activity, microwave brightness temperatures (see below), and spaceborne radar-reflectivity profiles (Katsanos et al., 2007). We can distinguish two types of sensing by satellites, passive and active. Passive sensing is based on measuring the radiative intensity emitted or reflected by particles in the atmosphere, such as cloud droplets and hydrometeors of precipitable size. Active sensing is conducted using radar equipment carried by the satellite. Next, we discuss some basic principles of passive remote sensing in the visible (V, lE0.39– 0.77 mm), IR (wavelengths lE0.77 mm–0.1 mm), and microwave (MW, lE0.1 mm–10 cm) portions (channels) of the electromagnetic spectrum. The basic principles of operation of active sensors are similar to those of radars, reviewed in Section 2.02.3.2. For a more detailed review on the principles and techniques of remote sensing, the reader is referred to Barrett and Martin (1981), Elachi (1987), Stephens (1994), and Kidder and Vonder Haar (1995).

2.02.3.3.1 The IR signature of cloud tops The high absorptivity of cloud droplets in the IR spectral range causes clouds to appear opaque in the IR channel. Hence, the IR radiation received by the satellite’s radiometer originates mostly from the cloud tops, which can be approximated with sufficient accuracy as black bodies, that is, as objects that absorb all incident radiation and emit it at a rate that depends solely on their temperature. In this case, we can use Stefan– Boltzman’s law of radiation (e.g., Barrett and Martin, 1981) to calculate the temperature Tb of the cloud tops from the intensity J of the received IR radiation:

Tb ¼ ðJ=sSB Þ1=4

ð53Þ

where sSB ¼ 5.7 108 W m2 K4 is the Stefan–Boltzman constant and Tb is in kelvins. Tb is usually referred to as brightness temperature (e.g., Smith, 1993) and, for a given atmospheric lapse rate g (see Section 2.02.2.1), it can be used to calculate cloud top heights. Evidently, lower brightness temperatures Tb correspond to clouds with higher tops and larger probabilities of rain.

Precipitation

Hence, we can develop regression equations to relate brightness temperatures to observed surface rainfall rates (e.g., Griffith et al., 1978; Stout et al., 1979; Arkin, 1979; Richards and Arkin, 1981; Arkin and Meisner, 1987; Adler and Negri, 1988). Two important limitations apply (Richards and Arkin, 1981; Liu, 2003): (1) due to the statistical character of the regressed quantities, the accuracy of the rainfall-retrieval algorithm increases with increasing scale of spatial or temporal averaging, and (2) the parameters of the regression depend on the climatology of the region and, therefore, cannot be used at regions with different climatic characteristics. An example of surface rainfall estimation from IR images is the temperature threshold method developed by Arkin (1979), Richards and Arkin (1981) and Arkin and Meisner (1987). Arkin (1979) used IR imagery from the Synchronous Meteorological Satellite-1 (SMS-1) and radar data from Global Atmosphere Research Program (GARP) Atlantic Tropical Experiment (GATE) to investigate the correlation between radarestimated precipitation rates and the fraction of areas with brightness temperature Tb below a certain threshold Tmin. The study found a maximum correlation (around 0.85) for a brightness temperature threshold TminE235 K (  38 1C). Richards and Arkin (1981) showed that a linear relationship is sufficient to describe the dependence between spatially averaged surface rainfall and the fraction of areas with Tbo235 K, with error variance that increases with decreasing scale of spatial averaging. Based on these results, Arkin and Meisner (1987) suggested the use of the Geostationary Operational Environmental Satellite (GOES) Precipitation Index (GPI) to calculate spatial rainfall averages in the tropics:

GPI ¼ 3ðmm=hÞFc H

ð54Þ

where GPI is the spatially averaged rainfall accumulation in a grid box of 2.51 latitude  2.51 longitude, Fc is the mean fraction (a dimensionless quantity between 0 and 1) of the grid box covered by brightness temperatures Tbo235 K, and H is the length of the observation period in hours. The temperature threshold method of Arkin (1979), Richards and Arkin (1981), and Arkin and Meisner (1987) produces accurate estimates of the spatially averaged rainfall in the tropical belt (301 S to 301 N), at grid scales larger than 2.51(E275 km) (Arkin and Meisner, 1987) and for averaging durations greater than about a month (Ba and Nicholson, 1998). The error increases significantly as we move to midlatitudes, especially during cold seasons (e.g., Liu, 2003). Extensions of the method include the use of the upper tropospheric humidity (UTH) in the vicinity of convective clouds as an additional predictive variable (Turpeinen et al., 1987), and the combination of IR and visible imagery (i.e., bi-spectral methods; see below) to exclude nonprecipitating clouds with high tops.

2.02.3.3.2 The visible reflectivity of clouds The signature of Earth in the visible (V) channel is due to the reflection of the sunlight by clouds and, when the sky is clear, the surface features. Consequently, visible imagery is available only during daylight hours. Due to its shorter wavelength, visible radiation can penetrate deeper into clouds than the

61

infrared portion of the electromagnetic spectrum, but similar to the IR channel, it still represents the upper portion of clouds and serves as an indirect signature of surface rainfall. However, visible reflectivity can complement the IR brightness temperatures to allow better classification of clouds and qualitative assessment of the probability of precipitation. This is the basis of the well-known bi-spectral methods (e.g., Lovejoy and Austin, 1979; Bellon et al., 1980; Tsonis and Isaac, 1985; Tsonis, 1987; O’Sullivan et al., 1990; Cheng et al., 1993; Cheng and Brown, 1995; King et al., 1995; Liu, 2003). The visible reflectivity of clouds increases fast with the increasing liquid water path, that is, the vertically integrated liquid water in the atmospheric column. Hence, we can use IR brightness temperatures to calculate the altitude of the cloud tops and visible reflectivities to obtain a qualitative estimate of the vertically averaged liquid water of the cloud, which is indicative of the rainfall potential. For example, low brightness temperatures (i.e., cold cloud tops) and high visible reflectivities (i.e., thick clouds) indicate cumulonimbus formations with high probability of precipitation (see Section 2.02.2.4.1), warm cloud tops and high visible reflectivities indicate stratiform rainfall (see Section 2.02.2.4.2), whereas cold cloud tops and low visible reflectivies indicate cirrus clouds, which are usually nonprecipitating. An example of bi-spectral methods is the RAINSAT technique developed by Lovejoy and Austin (1979) and Bellon et al. (1980). This technique uses visible reflectivities to reduce the number of false alarms obtained from the IR channel and more accurately estimate surface rainfall rates. The RAINSAT method was developed using GOES infrared and visible imagery and radar data from tropical (i.e., GATE) and mid-latitude (i.e., McGill weather radar, Quebec, Canada) locations as ground truth. The method was optimized by Cheng et al. (1993) and Cheng and Brown (1995) for the area of the UK, using IR and visible imagery from the European geostationary satellites Meteosat-2, Meteosat-3, and Meteosat-4 and rainfall retrievals from nine weather radars located in the United Kingdom and Ireland. A similar cloud classification technique has been proposed by Tsonis and Isaac (1985) and Tsonis (1987). This technique is based on cluster analysis of pixels with different brightness temperatures and visible reflectivities and has been developed using GOES satellite data and rainfall retrievals from the Woodbridge weather radar in Ontario, Canada.

2.02.3.3.3 The microwave signature of precipitation Contrary to the IR and visible spectral ranges, microwave radiation can effectively penetrate through cloud and rain layers and provide the signature of the integrated contribution of precipitation particles in the atmospheric column. Hence, brightness temperatures obtained from the MW channel are better linked to surface rainfall rates than the visible reflectivities and IR brightness temperatures. The type and size of the precipitation particles detected by the microwave radiometer depends on the frequency of the upwelling radiation. Above 80 GHz (i.e., wavelengths lo3.75 mm), ice crystals scatter the upwelling MW radiation and fade the signature of raindrops. Hence, above 80 GHz, the radiometer senses only ice, where lower brightness

62

Precipitation

temperatures are associated with more scattering, larger ice particles, and higher precipitation intensities at ground level. Below about 20 GHz (i.e., l41.5 cm), the radiative intensity of raindrops dominates the microwave signature of hydrometeors in the atmospheric column, whereas ice particles are virtually transparent. Thus, below 20 GHz, the microwave radiometer detects the vertically integrated signature of rainwater, where higher brightness temperatures are associated with more intense rainfall at ground level. Lowfrequency microwave imagery is especially useful when calculating surface rainfall rates over oceans, where the almost constant sea surface temperature and emissivity allow translation of the spatial and temporal variations of brightness temperatures to variations of sea-level rainfall rates (e.g., Liu, 2003). The same is not true over land, where the surface features cause the ground temperature and emissivity to vary significantly in space and time. Another limitation of lowfrequency microwave images is the saturation of the microwave channel at high rainfall rates, which causes negative biases of the obtained rainfall intensity (e.g., Liu, 2003; Viltard et al., 2006). Between 20 GHz and 80 GHz, scattering and emission by raindrops and ice particles occur simultaneously and the microwave radiation undergoes multiple transformations. Hence, the microwave radiometer detects different rain paths at different microwave frequency ranges. Combining brightness temperatures from different MW channels to more accurately assess surface rainfall rates is an open research problem and it has driven the development of many rainfall-estimation algorithms (Grody, 1991; Spencer et al., 1989; Alishouse et al., 1990; Berg and Chase, 1992; Hinton et al., 1992; Liu and Curry, 1992, 1993; Ferriday and Avery, 1994; Petty, 1994a, 1994b, 2001a, 2001b; Kummerow and Giglio, 1994a, 1994b; Ferraro and Marks, 1995; Kummerow et al., 1996, 2001; Berg et al., 1998; Aonashi and Liu, 2000; Levizzani et al., 2002). For a review of microwave methods of estimation over ocean and land, and their advantages and limitations, the reader is referred to Wilheit et al. (1994), Petty (1995), and Kidd et al. (1998) respectively.

2.02.4 Precipitation modeling As already clarified in Section 2.02.1.5, modeling of precipitation is not possible without using any type of a stochastic approach. Even the deterministic numerical weather forecast models, which determine the state and motion in the atmosphere by solving differential equations, to model precipitation, use parametrization schemes. These schemes, instead of describing the detailed dynamics of the precipitation process, establish and use equations of statistical type to quantify the output of the dynamical system. In addition, as mentioned in Section 2.02.1.5, the modern framework for predicting precipitation particularly as input to hydrological models (the ensemble forecasting), is of the Monte Carlo or stochastic type. The description of these stochastic techniques belongs to the sphere of weather forecasting and is not within the scope of this chapter. In more engineering-oriented applications, precipitation is typically modeled as an autonomous process, without particular reference to the atmospheric

dynamics. Next, we outline some of the most widespread modeling practices for precipitation but without details and mathematical formulations, which the interested reader can find in the listed references.

2.02.4.1 Rainfall Occurrence From the early stages of the analysis of precipitation intermittency, it was recognized that rainfall occurrences are not purely random. In other words, rainfall occurrence cannot be modeled (effectively) as a Bernoulli process in discrete time or, equivalently, as a Poisson process in continuous time. It should be recalled that in a Bernoulli process, an event (rainfall/wet state) occurs with a probability p (and does not occur with probability 1  p) constant in time, and each event is independent of all preceding and subsequent events. In a Poisson process, the times of occurrence of events (i.e., the starting times of rainfalls) are random points in time. In this process, the time differences between consecutive occurrences are independent identically distributed (IID) with exponential distribution. Both discrete time and continuous time representations of the rainfall occurrence process, which in fact are closely related (e.g., Foufoula-Georgiou and Lettenmaier, 1986; Small and Morgan, 1986), have been investigated. The most typical tool of the category of discrete time representations is the Markov chain model (Gabriel and Neumann, 1962; Feyerherm and Bark, 1964; Hershfield, 1970; Todorovic and Woolhiser, 1975; Haan et al., 1976; Chin, 1977; Katz, 1977a, 1977b; Kottegoda and Horder, 1980; Roldan and Woolhiser, 1982). In this model, any time interval (e.g., day) can be in one of two states, dry or wet, and it is assumed that the state in a time interval depends on the state in the previous interval. It was observed, however, that Markov chain models yield unsatisfactory results for rainfall occurrences, especially for dry intervals (De Bruin, 1980). Moreover, the interannual variance of monthly (or seasonal) total precipitation is greater than that predicted by Markov chain models, an effect usually referred to as overdispersion (Katz and Parlange, 1998). Extended versions of the binary state Markov chains using a higher number of past states may improve performance. Additional states in such model versions have been defined based on a combination of states of two consecutive periods (Hutchinson, 1990) or on accounting for the rainfall depth of each interval (Haan et al., 1976). A more effective enhancement is to use transition probabilities taking into account more than one previous interval, which leads to stochastic binary chains of order higher than one (Pegram, 1980; Katz and Parlange, 1998; Clarke, 1998). In more recent developments, to account for a long number of previous time intervals and simultaneously avoid an extremely high number of transition probabilities, it was proposed that, instead of the sequence of individual states of these intervals, one could use conditional probabilities based on aggregation of states of previous intervals (Sharma and O’Neill, 2002). Similarly, one could use a discrete wetness index based on the number of previous wet intervals (Harrold et al., 2003). An extension of the Markov chain approach to multiple sites has been studied by Pegram and Seed (1998).

Precipitation

In a more recent study, Koutsoyiannis (2006a) used the principle of maximum entropy, interpreted as maximum uncertainty, to explain the observed dependence properties of the rainfall-occurrence process, including the overdispersion or clustering behavior and persistence. He quantified intermittency by the probability p(1) that a time interval of length 1 h is dry, and dependence by the probability that two consecutive intervals are dry, that is by p(2), where in general p(k) denotes the probability that an interval of length k is dry. Using these two probabilities and a multiscale entropy-maximization framework, he was able to determine any conditional or unconditional probability of any sequence of dry and wet intervals at any timescale. Thus, he described the rainfall occurrence process including its dependence structure at all scales using only two parameters. The dependence structure appeared to be non-Markovian, yet not over-exponential. Application of this theoretical framework to the rainfall data set of Athens indicated good agreement of theoretical predictions and empirical data at the entire range of scales for which probabilities dry and wet can be estimated (from 1 h to several months). An illustration is given in Figure 22. In the continuous time representation of the rainfall occurrence process, the dominant tools are the cluster-based point processes (Waymire and Gupta, 1981a, 1981b, 1981c). These are essentially based on the prototype of the spatial distribution of galaxies devised by Neyman and Scott (1952) to describe their property of clustering relative to the Poisson process. With reference to storms, if they were regarded as instantaneous pulses positioned at random points in time, the logarithm of probability that the interarrival time exceeds a value x, or the log survival function, would be proportional to

63

x. However, empirical evidence suggests that the log survival function is a nonlinear concave function of x, which indicates a tendency for clustering of rainfall events relative to the Poisson model (Foufoula-Georgiou and Lettenmaier, 1986). This clustering has been modeled by a cascade of two Poisson processes, corresponding to two characteristic timescales of arrivals of storms and storm cells. The Neyman–Scott process with instantaneous pulses was the first one applied to rainfall occurrence (Kavvas and Delleur, 1981; Rodriguez-Iturbe et al., 1984), later succeeded by the Neyman–Scott rectangular pulses and the very similar Bartlett–Lewis rectangular pulse models (Rodriguez-Iturbe et al., 1987). The Bartlett–Lewis rectangular pulse model, which is the most typical and successfully applied model of this type, assumes that rainfall occurs in the form of storms of certain durations and that each storm is a cluster of random cells. The general assumptions of the rainfall occurrence process are: 1. Storm origins ti occur according to a Poisson process with rate l. 2. Origins tij of cells of each storm i arrive according to a Poisson process with rate b. 3. Arrivals of each storm i terminate after a time vi exponentially distributed with parameter g. 4. Each cell has a duration wij exponentially distributed with parameter Z. In the original version of the model, all model parameters are assumed constant. In a modified version, the parameter Z is randomly varied from storm to storm with a gamma distribution with shape parameter a and scale parameter n.

p(k)

1

0.1

0.01 1

10

100 k

1000

10 000

Figure 22 Probability dry p(k) vs. scale k (in h), as estimated from a hourly rainfall data set in Athens, Greece, and predicted by the maximum entropy model in Koutsoyiannis (2006a) for the entire year (circles and red full line) and the dry season (June–September; diamonds and blue full line). The model was fitted using two data points in each case (marked in full in the plot), that is, the probability dry for 1 h, pp(1), and 2 h, p(2), which are respectively 0.9440 and 0.9335 for the entire year and 0.9888 and 0.9860 for the dry season. The final model is expressed as 1=Z Z p ðk Þ ¼ p ½1þðx 1Þðk 1Þ , where the parameters are respectively Z ¼ 0.63 and x ¼ 0.816 for the entire year and Z ¼ 0.83 and x ¼ 0.801 for the dry season. For comparison, lines resulting from the Markov chain model are also plotted (dashed lines). From Koutsoyiannis D (2006a) An entropicstochastic representation of rainfall intermittency: The origin of clustering and persistence. Water Resources Research 42(1): W01401 (doi:10.1029/ 2005WR004175).

64

Precipitation Storms

Intensity (mm h−1)

4

Storm 1 Storm 2 Storm 3 Storm 4

3

2

1

0 2

1 Time (days)

Figure 23 Simulated realization of a series of four storms from the Bartlett–Lewis rectangular pulse model (modified version with randomly varying Z) occurring within three days (notice the overlap of storms 1 and 2, which is allowed by the model), implemented by the Hyetos software (see Section 2.02.4.4). The model parameters are l ¼ 0.94 d1, k ¼ b/Z ¼ 1.06, j ¼ g/Z ¼ 0.059, a ¼ 2.70, n ¼ 0.0068 d1, and mx ¼ sx ¼ 24.3 mm d1.

Subsequently, parameters b and g also vary so that the ratios k: ¼ b/Z and j: ¼ g/Z are constant. A major problem of these models was their inability to reproduce the probability of zero rainfall at multiple timescales (Velghe et al., 1994). In this respect, Foufoula-Georgiou and Guttorp (1986) noted that the Neyman–Scott model parameters are scale dependent and thus cannot be attributed a physical meaning. To ameliorate this, modifications of both the Neyman–Scott model (Entekhabi et al., 1989) and the Bartlett–Lewis model (Rodriguez-Iturbe et al., 1988; Onof and Wheater, 1993, 1994) were proposed. These are in fact based on the randomization of the mean interarrival time of one of the two Poisson processes. Evaluation and comparison of several cluster-based rectangular pulse models for rainfall were done by Velghe et al. (1994) and Verhoest et al. (1997), whereas a comprehensive review of Poisson-cluster models has been provided by Onof et al. (2000). An extension of the concept introducing a third Poisson process was proposed by Cowpertwait et al. (2007).

2.02.4.2 Rainfall Quantity In the discrete time representations of rainfall occurrence, the rainfall quantity in each wet interval is modeled separately from the occurrence process, usually based on statistical analysis of the observed record. In the point-process representations, the storms and cells are abstract quantities that do not fully correspond to real-world objects. Therefore, they cannot be identified in the recorded time series. An assumption is typically made that each cell has a uniform intensity xij with a specified distribution, and based on all assumptions, the statistical characteristics of the rainfall process at one or more timescales are derived analytically (Rodriguez-Iturbe et al., 1987, 1988). These statistical characteristics are compared to the empirically derived statistics, and, by minimizing the departures of the two, the model parameters are estimated. The distribution of the uniform intensity xij is typically assumed to be exponential with parameter 1/mx. Alternatively, one can choose a two-parameter gamma distribution with mean mx

and standard deviation sx. In this manner, the point-process models describe the entire rainfall process, including occurrence and quantity. A demonstration of the model is shown in Figure 23. However, in some cases (e.g., Gyasi-Agyei and Willgoose, 1997), point processes have been used to simulate merely rainfall occurrences and then have been combined with other models that simulate rainfall depths. Other modeling approaches for the rainfall process (including its intermittency) are reviewed in Srikanthan and McMahon (2001). With their typical assumptions, including those of the exponential or gamma distribution for rain-cell amount, the point-process models, despite providing satisfactory representation of the process at a specific timescale or a small range of timescales, cannot really perform satisfactorily over a wide range of scales and also lead to exponential distribution tails, whereas it has been recently recognized that the tails must be of power type (see Sections 2.02.1.5 and 2.02.5.2). Generally, the distribution function of rainfall varies among different timescales. At very fine scales, the density is J-shaped, that is, with a mode at zero, and perhaps with density tending to infinity as the rainfall depth or the intensity tends to zero. At coarse timescales such as monthly (for wet months) and annual, the distribution becomes bell-shaped and tends to become normal as the scale increases. However, its tail always departs from the exponential tail of the normal distribution. In fact, for theoretical reasons, if at the right tail, the survival function is a power function of the rainfall depth or intensity x, with exponent 1/k, that is, F*(x)px1/k (see Equation (18)), then it will be of the same type and will have precisely the same exponent 1/k at any timescale (the proof is omitted). This behavior of the tail is perhaps the only invariant distributional property across all scales, whereas the shape of the body of the distribution varies significantly across different scales. However, even this variation must have a simple and unique explanation, which is the principle of maximum entropy. Specifically, Koutsoyiannis (2005a) has shown that all diverse shapes of the distribution across different scales can be derived from the principle of maximum entropy constrained on known mean and variance.

Precipitation

Papalexiou and Koutsoyiannis (2008) proposed a single distribution (a power-transformed beta prime distribution, also known as generalized beta of second kind; see also Koutsoyiannis, 2005a) with four parameters, which provides good fits for rainfall intensity at timescales from hourly to annual. Only one of the four parameters (corresponding to the exponent of the tail) is invariant across scales. If the range of scales of interest is smaller, then specific special cases of this distribution can be used as good approximations. For example, the three-parameter Burr type VII distribution, which has the advantage of providing a closed form of the quantile function, can be used effectively for timescales from a few minutes to a couple of months (Papalexiou and Koutsoyiannis, 2009).

65

Storm area

Rainfall cells

2.02.4.3 Space–Time Models Space–time modeling of precipitation is one of the most demanding tasks of stochastic modeling in hydrology and geophysics. Rainfall intermittency should be modeled in both space and time, along with the motion of rainfall fields, the rainfall quantity, and its temporal and spatial structure. One of the relatively simple solutions has been provided by the extension of point-process models used for the rainfall process at a single site. This extension introduces a description of rainfall cells in space, in addition to that in time, and a motion of the cells. As an example, we summarize here the Gaussian displacement spatial–temporal rainfall model (GDSTM; Northrop, 1996, 1998). This model, is a spatial analog of a point-process model having a temporal structure similar to that of the Bartlett-Lewis rectangular pulse model described above and a spatial structure known as the Gaussian displacement structure, introduced by Cox and Isham (1988). Similar to its single-site analog, GDSTM assumes that rainfall is realized as a sequence of storms, each consisting of a number of cells. Both storms and cells are characterized by their centers, durations, and areal extents (see sketch in Figure 24) and, in addition, cells have certain uniform rainfall intensity. Specifically, the following assumptions characterize storms and cells. Storm centers arrive according to a homogeneous Poisson process of rate l in 2D space (denoted by x, y) and time (denoted by t) and move with a uniform velocity (Vx, Vy). Each storm has a finite duration L (assumed exponentially distributed with parameter b ¼ 1/mL) and an infinite areal extent, represented by an elliptical geometry with eccentricity E and orientation y, and incorporates a certain number of rainfall cells. However, a storm can be assigned a finite storm area, the area that contains a certain percentage of rainfall cells. The storm area varies randomly and in each storm, it is determined in terms of the realization of a random variable w, which determines uniquely (for the specific storm) a set of parameters s2x ; s2y , and r that determine the displacement of cell centers from the storm center. Specifically, w is Gammadistributed with shape and scale parameters determined in terms of the eccentricity e and the mean storm area ms. At the same time, the parameter r is determined in terms of the eccentricity e and the storm orientation, y. Following the generation of w, the parameters s2x and s2y are determined

Storm center Figure 24 Sketch of the spatial structure of the Gaussian displacement spatial–temporal rainfall model.

in terms of the eccentricity e, the storm orientation y, and the value of w. Each rainfall cell is assigned a center ðxc ; yc ; tc Þ. The time origin tc follows a Poisson process starting at the time ordinate of the storm origin t0 (with the first cell being located at this point) and ending at t0 þ L. The expected number of cells within that time interval is mc ¼ 1 þ b/g, where g is the cellgeneration Poisson process parameter. The spatial displacements from the storm center are random variables jointly normally distributed with zero means, variances s2x and s2y , and correlation r. Given these parameters, the displacement Dx of each cell is generated as a normal variate (0, sx) and the displacement Dy as a normal variate (my|x, sy|x). Furthermore, each cell has a finite duration D (assumed exponentially distributed with parameter 1/mD) and an elliptical area with major axis a, forming an angle y with the x-axis (west–east), ffi pffiffiffiffiffiffiffiffiffiffiffiffiffi and minor axis b ¼ 1  e 2 a. It is assumed that a is a random variable gamma distributed with shape and scale parameters depending on the mean storm area mA and the eccentricity e, respectively. Finally, each cell has an intensity x independent of any other variable, exponentially distributed with parameter 1/mx. The model is defined in terms of 11 independent parameters, namely: (1) the rate of storm arrival (number of storms per area per time), l; (2) the mean cell duration, mD; (3) the mean storm duration, mL; (4) the mean cell area, mA; (5) the mean storm area, ms; (6) the mean number of cells per storm, mc; (7) the mean cell intensity, mx; (8 and 9) the components of the cell and storm velocity in the x direction (east), Vx, and in the y direction (north), Vy; (10) the cell and storm eccentricity, e; and (11) the cell and storm orientation, y. Similar to its single-point analog, the entities of the spatial point-process model are abstract. To make the model outputs comparable to reality, integration from continuous time over a specific timescale and/or spatial scale is needed, from which the first- and second-order rainfall statistics are

66

Precipitation

calculated. The latter serve as the basis for parameter estimation using either rain gauge or radar data. Due to model complexity, the calculation of the statistics can be done only numerically; hence, the entire model application (and the parameter estimation in particular, which needs numerical optimization, e.g., using the generalized reduced-gradient method) is laborious.

2.02.4.4 Rainfall Disaggregation and Downscaling Both disaggregation and downscaling refer to the generation of a precipitation field at a specific temporal and/or spatial scale given a known precipitation field (measured or simulated) at a certain larger temporal and/or spatial scale (lower resolution). Disaggregation and downscaling are very useful procedures and have several applications, such as in the following cases: 1. Global-scale weather-prediction models provide rainfall forecasts at a low resolution, for example, grid size of 50 km. Hydrologic models require the description of the precipitation field at a much higher resolution, with grid size of the order of 1 km. 2. Satellite precipitation estimates are available at a spatial scale X0.251 (latitude and longitude), or about 28 km at the equator, and a temporal scale of 3 h. Again, hydrologic applications require higher resolutions. 3. The majority of historical point rainfall records come from daily rain gauges, which have often been operational for several decades. The number of rain gauges providing hourly or sub-hourly resolution data is smaller by about an order of magnitude. However, hydrologic applications, especially flood studies, usually need hourly or even subhourly data. 4. In complex problems of stochastic generation of precipitation time series or precipitation fields, it is difficult to reproduce simultaneously, the long-term and the shortterm stochastic structure of precipitation using a single model. A better approach is to couple several models, starting from a large-scale model to represent the long-term behavior. The outputs of the latter are then disaggregated into finer scales. Note, however, that in a recent study Langousis and Koutsoyiannis (2006) developed a stochastic framework capable of reproducing simultaneously the long-term and the short-term stochastic structure of hydrological processes, avoiding the use of disaggregation. While disaggregation and downscaling are similar in nature, they also have a difference that distinguishes them. Downscaling aims at solely producing a precipitation field y with the required statistics at the scale of interest, being statistically consistent with the given field x at the finer scale. Disaggregation demands full and precise consistency, which introduces an equality constraint in the problem of the form

Cy¼x

ð55Þ

where C is a matrix of coefficients. For example, assuming that x is an annual amount of precipitation at a station and y is the vector consisting of the 12 monthly precipitation values at the

same station, C will be a row vector with all its elements equal to 1, so that Equation (55) represents the requirement that the sum of all monthly precipitation amounts must equal the annual amount. Task 1 could be accomplished by running a second meteorological model at the limited area of interest. Such models, known as limited-area models, can have much higher resolution than global models. The description of this type of downscaling, known as dynamical downscaling, because it is based on the atmospheric dynamics, is not within the scope of this chapter. In contrast, a stochastic procedure need not refer to the dynamics, and is generic and appropriate for both downscaling and disaggregation and for all above tasks 1–4. This generic procedure resembles the interpolation procedure described in Section 2.02.3.1.3, but there are two important differences. First, it is necessary to include the error terms in the generation procedure (recall that in interpolation, which is a point estimation, knowing only the mean and variance of the error was sufficient). Second, the generated values y at the different points should be statistically consistent to each other. This precludes the separate application of an algorithm at each point of interest and demands simultaneous generation at all points. In turn, this demands that the error terms in different points should be correlated to each other. All these requirements could be summarized in the linear generation scheme

y ¼A xþB v

ð56Þ

where A and B are matrixes of coefficients and v is a vector of independent random variables, so that the term Bv ¼: e corresponds to the error term in interpolation (cf. Equation (29)). In disaggregation, Equation (56) should be considered simultaneously with Equation (55). For Gaussian random fields without intermittency, the application of Equations (55) and (56) is rather trivial. However, the intermittency of the rainfall processes and the much-skewed distributions at fine timescales are severe obstacles for rainfall disaggregation. To overcome such obstacles, several researchers have developed a plethora of rather ad hoc disaggregation models (see review by Koutsoyiannis (2003b)). However, the application of the above theoretically consistent scheme is still possible, if combined with a stochastic model, accounting for intermittency (e.g., a Bartlett– Lewis model), and if an appropriate strategy is used to implement Equation (55). Such a strategy includes recursive application of Equation (56) until the error in Equation (55) becomes relatively low, and is followed by correction of the error of the accepted final iteration by appropriate adjusting procedures, which should not alter the covariance structure of the precipitation field. The general strategy of stochastic disaggregation is described in Koutsoyiannis (2001) and two implementations for temporal rainfall disaggregation at a fine (hourly) scale at a single site and at multiple sites are described in Koutsoyiannis and Onof (2001) and Koutsoyiannis et al. (2003), respectively. The models described in the latter two papers, named Hyetos and MuDRain, respectively, are available online, and have been used in several applications worldwide. Typical results of the two models are shown in Figures 25 and 26, respectively.

Precipitation

0.906 0.902

0.85 0.843

1 Probability (P )

Intensity (mm h−1)

1

0 1

2

0.374 0.374

0

3

P (dry hour)

Time (days)

P (dry day)

P (dry hour|wet day)

1

9

7.629

8

0.9 6.455

0.8

7 4.671

6

Autocorrelation

Variation, skewness

67

4.365

5 4

0.7 0.6 0.5 0.4

3

0.3

2

0.2 0.1

1

0

0

1

2

3

Skewness

Variation

4

5 6 Lag (h)

7

8

9

10

Figure 25 Typical screens produced by the Hyetos software during disaggregation of daily to hourly rainfall data, where plots in green and red refer to disaggregated and original data respectively. Upper-left panel shows typical hyetographs, where the green (disaggregated) plot is the result of the storms shown in Figure 23 converted to a hyetograph at an hourly scale. Notice that while daily totals match, the temporal distribution of rainfall differs in the disaggregated and original hyetographs. However, in the statistical sense, the disaggregated series resembles the original, as shown in the other panels comparing statistics of disaggregated and original series.

Hourly rainfall depth (mm)

6 5

H2

H5

S2

S5

4 3 2 1 0 12/07/98

12/07/98

12/07/98

12/07/98

13/07/98

12/07/98

12/07/98

12/07/98

12/07/98

13/07/98

00:00

06:00

12:00

18:00

00:00

00:00

06:00

12:00

18:00

00:00

Figure 26 While, as shown in Figure 25 (upper-left panel), in single-variate disaggregation, the produced hyetographs resemble the actual ones only in a statistical sense, multivariate disaggregation reproduces the actual shapes of hyetographs provided that fine-scale (e.g., hourly) data exist in at least one of the stations. The two panels show a comparison of historical (marked H) and simulated (by the MuDRain disaggregation model; marked S) hyetographs on a day with relatively high rainfall (B16 mm) at two rain gauges (2 and 5) in the Brue catchment located in South-Western England. From Koutsoyiannis D, Onof C, and Wheater HS (2003) Multivariate rainfall disaggregation at a fine timescale. Water Resources Research 39(7): 1173 (doi:10.1029/2002WR001600).

2.02.4.5 Multifractal Models Rainfall models of multifractal type have for a long time been known to accurately reproduce several statistical properties of actual rainfall fields in finite but practically important ranges

of scales: typically from below 1 h to several days in time and from below 10 km to more than 100 km in space (Schertzer and Lovejoy, 1987, 1989; Tessier et al., 1993; Fraedrich and Larnder, 1993; Olsson, 1995; Lovejoy and Schertzer, 1995; Over and Gupta, 1996; Carvalho et al., 2002; Nykanen and

68

Precipitation

Harris, 2003; Kundu and Bell, 2003; Deidda et al., 2004, 2006; Gebremichael and Krajewski, 2004; Calenda et al., 2005; Gebremichael et al., 2006; Veneziano and Langousis, 2005; Garcı´a-Marı´n et al., 2007; Langousis and Veneziano, 2007). These properties include the scaling of the moments of different orders (Schertzer and Lovejoy, 1987; Menabde et al., 1997; Deidda et al., 1999; Deidda, 2000), the power law behavior of spatial and temporal spectral densities (Olsson, 1995; Tessier et al., 1996; Deidda et al., 2004, 2006), the alteration of wet and dry intervals (Over and Gupta, 1996; Schmitt et al., 1998; Olsson, 1998; Gu¨ntner et al., 2001; Langousis and Veneziano, 2007), and the distribution of extremes (Hubert et al., 1998; Veneziano and Furcolo, 2002; Veneziano and Langousis, 2005; Langousis and Veneziano, 2007; Langousis et al., 2007; Veneziano et al., 2009). Significant deviations of rainfall from multifractal scale invariance have also been pointed out. These deviations include breaks in the power-law behavior of the spectral density (Fraedrich and Larnder, 1993; Olsson, 1995; Menabde et al., 1997), lack of scaling of the non-rainy intervals in time series (Schmitt et al., 1998), differences in scaling during the intense and moderate phases of rainstorms (Venugopal et al., 2006), the power deficit at high frequencies relative to multifractal models (Perica and Foufoula-Georgiou, 1996a, 1996b; Menabde et al., 1997; Menabde and Sivapalan, 2000), and more complex deviations as described in Veneziano et al. (2006a). Next, we review some basic properties of stationary multifractal processes and discuss a simple procedure to construct discrete multifractal fields based on the concept of multiplicative cascades. For a detailed review on the generation of multifractal processes and their applications in hydrological modeling and forecasting, the reader is referred to Veneziano and Langousis (2010). Let i ðdÞ ðtÞ be the average rainfall intensity averaged over timescale d at time t. The stochastic process i ðdÞ ðtÞ is said to be stationary multifractal if, for any timescale d, its statistics remain unchanged when the time axis is contracted by a factor r41 and the intensity is multiplied by a random variable ar , that is, d

i ðd=rÞ ðtÞ ¼ ar iðdÞ ðtÞ d

has zero mean, it can be viewed as a special case of the multifractal process in which the random variable ar is replaced by a deterministic power function of resolution r. A property of stationary multifractal processes, which has been used to verify multifractality, is that the spectral density s(o) behaves like ob where o is the frequency, and bo1 is a constant (e.g., Fraedrich and Larnder, 1993; Olsson, 1995, Deidda et al., 2004; Hsu et al., 2006). More comprehensive checks of multifractality involve the dependence of statistical moments of different orders on scale. In particular, under perfect multifractality E½ði ðdÞ Þ q p E½ðar Þq p dKðqÞ p r KðqÞ, where K(q) is a convex function, usually referred to as moment-scaling function (Gupta and Waymire, 1990; Veneziano, 1999). All concepts and methods are readily extended to space–time rainfall (Veneziano et al., 2006b). A simple procedure to construct discrete stationary multifractal fields is based on iterative application of Equation (57) starting from a large timescale dpdmax and gradually decreasing the timescale (i.e., at resolutions rpmn, where m 4 1 and nX1 are integers). The contraction by the same factor r ¼ m at each step simplifies generation, since only the distribution of ar  am is needed. This forms the concept of socalled isotropic discrete multiplicative cascade. Its construction in the D-dimensional cube SD starts at level 0 with a single tile O01  SD with constant unit intensity inside O01 . At level n ¼ 1, 2, y (or equivalently at resolutions r ¼ mD, m2D, y) each tile at the previous level n  1 is partitioned into mD tiles where m 4 1 is the integer multiplicity of the cascade. The intensity inside each cascade tile Oni (i ¼ 1, y, mnD) is obtained by multiplying that of the parent tile at level n  1 by an independent copy yi of a unit-mean random variable y, called the generator of the cascade. Clearly, for r ¼ mnD, ar ¼ y1 y2 ? yn . For illustration, Figure 27 shows a simulated realization of a 2D binary (i.e., m ¼ 2) discrete multiplicative cascade developed to level n ¼ 8.

2.02.5 Precipitation and Engineering Design 2.02.5.1 Probabilistic versus Deterministic Design Tools

ð57Þ

where ¼ denotes equality in (any finite-dimensional) distribution. The notation implies that the distribution of ar depends only on r and not on time t or the intensity i ðdÞ. Obviously, the mean of ar is 1 and furthermore ar is assumed to be stochastically independent of i ðdÞ at the higher scale d. The distribution of ar characterizes the scaling properties as well as many other characteristics of the rainfall process including the marginal distribution, intermittency, distribution of extremes, etc. Equation (57) need not apply for arbitrarily large timescales but rather applies up to a maximum scale d ¼ dmax. In rainfall, dmax seems to be of the order of several days and it is representative of the mean interarrival time of rainfall events (Langousis and Veneziano, 2007; Langousis et al., 2007; Veneziano et al., 2007). We note for comparison that the related equation in the simple scaling (HK) repred sentation of Section 2.02.1.5. is ði ðd=rÞ  mÞ ¼ r 1H ði ðdÞ  mÞ or 1H 1H ðdÞ ðd=rÞ d i ¼ mð1  r Þþr i , so that, when the HK process

The design and management of flood protection works and measures require reliable estimation of flood probability and risk. A solid empirical basis for this estimation can be offered by flow-observation records with an appropriate length, sufficient to include a sample of representative floods. In practice, however, flow measurements are never enough to support flood modeling. The obvious alternative is the use of hydrologic models with rainfall input data to generate streamflow. Notably, even when flow records exist, rainfall probability still has a major role in hydrological practice; for instance, in major hydraulic structures, the design floods are estimated from appropriately synthesized design storms (e.g., US Department of the Interior, Bureau of Reclamation, 1977, 1987; Sutcliffe, 1978). The need to use rainfall data as the basis of hydrologic design becomes even more evident in the study of engineering structures and urban water-management systems that modify the natural environment, so that past flood records, even if they exist, are no longer representative of the future modified system.

Precipitation

69

Figure 27 Simulated realization of a 2D stationary multifractal field. The random variable y is taken to be lognormal with unit mean value and logvariance ðslny Þ2 ¼ 0:2lnð2Þ.

Hydrologic design does not necessarily require full modeling of the rainfall process, of the type discussed in Section 2.02.4. Usually, in design studies, the focus is on extreme rainfall, which, notably, may not be represented well in such models, which are better for the average behavior of rainfall. However, historically, the perception of intense rainfall and the methodologies devised to model it have suffered from several fallacies spanning from philosophical to practical issues, which we describe next to cast a warning against their acceptance and use. The first fallacy is of a rather philosophical type. As discussed in Section 2.02.1.5, the modeling of the rainfall process in pure deterministic terms has been proven to be problematic. However, deterministic thinking in science is strong enough, so that after the failure in providing full descriptions, it was headed to determining physical bounds to precipitation in an attempt to design risk-free constructions or practices. The resulting concept of probable maximum precipitation (PMP), that is, an upper bound of precipitation that is physically feasible (World Meteorological Organization, 1986), is perhaps one of the biggest failures in hydrology. Using elementary logic, we easily understand that even the terminology is self-contradictory, and thus not scientific. Namely, the word probable contradicts the existence of a deterministic limit. Several methods to determine PMP exist in literature and are described in World Meteorological Organization (1986). However, examination in depth of each of the specific methods separately will reveal that they are all affected by logical inconsistencies. While they are all based on the assumption of the existence of a deterministic upper limit, they determine this limit statistically. This is obvious in the so-called statistical approach by Hershfield (1961, 1965), who used 95 000 station-years of annual maximum daily rainfall belonging to 2645 stations, standardized each record, and found the maximum over the 95 000 standardized values. Naturally, one of the 95 000 standardized values would be the greatest of all others, but this is not a deterministic limit to call PMP (Koutsoyiannis, 1999). If one examined 95 000 additional measurements, one might have found an even higher value. Thus, the logical problem here is the incorrect interpretation that an observed maximum in precipitation is a physical upper limit.

The situation is perhaps even worse with the so-called moisture maximization approach of PMP estimation (World Meteorological Organization, 1986), which seemingly is more physically (hydrometeorologically) based than the statistical approach of Hershfield. In fact, however, it suffers twice by the incorrect interpretation that an observed maximum is a physical upper limit. It uses a record of observed dew point temperatures to determine an upper limit, which is the maximum observed value. Then it uses this limit for the so-called maximization of an observed sample of storms, and asserts the largest value among them as PMP. Clearly, this is a questionable statistical approach, because (1) it does not assign any probability to the value determined and (2) it is based only on one observed value (known in statistics as the highest-order statistic), rather than on the whole sample, and thus it is enormously sensitive to one particular observation of the entire sample (Papalexiou and Koutsoyiannis, 2006; Koutsoyiannis, 2007). Thus, not only does the determination of PMP use a statistical approach (rather than deterministic physics), but it uses bad statistics. The arbitrary assumptions of the approach extend beyond the confusion of maximum observed quantities with physical limits. For example, the logic of moisture maximization at a particular location is unsupported given that a large storm at this location depends on the convergence of atmospheric moisture from much greater areas. Rational thinking and fundamental philosophical and scientific principles can help identify and dispel such fallacies. In particular, the Aristotelian notions of potentia (Greek, dynamis) and of potential infinite (Greek, apeiron; Aristotle, Physics, 3.7, 206b16) that ‘‘exists in no other way, but ... potentially or by reduction’’ (and is different from mathematical complete infinite) would help us to avoid the PMP concept. In fact, this does not need a great deal of philosophical penetration. The same thing is more practically expressed as ‘‘conceptually, we can always imagine that a few more molecules of water could fall beyond any specified limit’’ (Dingman, 1994). Yet, the linkage to the Aristotelian notions of potentia and potential infinity may make us more sensitive in seeing the logical inconsistencies (see also Koutsoyiannis, 2007). According to Popper (1982) the extension of the Aristotelian idea of potentia in modern terms is the notion of

70

Precipitation

probability. Indeed, probability provides a different way to perceive the intense rainfall and flood and to assign to each value a certain probability of exceedance (see next session) avoiding the delusion of an upper bound of precipitation and the fooling of decision makers that they can build risk-free constructions. In this respect, the criticism of the PMP and the probable maximum flood (PMF) involves logical, technical, philosophical, and ethical issues (e.g., Benson, 1973). One typical argument against the use of probabilistic approaches, in favor of PMP, which is very old yet popular even today, has been stated by Horton (1931; from Klemes, 2000), ‘‘It is, however, important to recognize the nature of the physical processes involved and their limitations in connection with the use of statistical methods. y Rock Creek cannot produce a Mississippi River flood any more than a barnyard fowl can lay an ostrich egg.’’ However, this argument reveals an incorrect perception of probability and statistics. In a probability theoretic context, there is not a logical inconsistency. Assuming, for example, that the annual peak flood of the Mississippi river (xM) is on the average (mM), a million times larger than the average (mC) flood of a certain small creek (xC), and assuming that both xM and xC have a lognormal distribution with standard deviation slnx of logarithms of about 0.3 (which is roughly equal to the coefficient of variation of the annual flood peaks, assumed equal in the two streams), one can readily find that the probability that the flood in the creek xC in some year exceeds the mean annual flood mM of Mississippi is F*(z): ¼ 1  FG(z) where FG is the standard normal distribution function and z ¼ ln(mM/mC)/slnx or z ¼ ln(106)/0.3 ¼ 46. For large z, the approximation ln F*(z) ¼  (1/2)[ln(2pz2) þ z2] holds (e.g., Abramowitz and Stegun, 1965); hence ln F*(z) ¼  1062.75, so that the probability of exceedance is F*(z) ¼ 10462. That is, according to the probabilistic approach, the return period of the event that the small creek flood matches or exceeds the mean annual flood of the Mississippi is 10462 years. Assuming that the age of the universe is of the order of 1010 years, one would wait, on the average, 10452 times the age of the universe to see this event happen – if one foolishly hoped that the creek, the Mississippi, and the Earth would exist for such a long time. Evidently, such a low probability could be regarded as synonymous to impossibility, which shows that the probabilistic approach does not regard the floods of Mississippi equivalent to those of a small creek (see also an example about the age of a person by Feller (1950)).

2.02.5.2 Extreme Rainfall Distribution Having been exempted from the concept of an upper limit to precipitation and having adopted a probabilistic approach, the real problem is how the rainfall intensity grows as the probability of exceedance decreases. Clearly, as the probability of exceedance tends to zero, the intensity tends to infinity. There exists a mathematically proven lower limit to the rate of this growth, which is represented by an exponential decay of the probability of exceedance with intensity. The alternative is a power-low decay and, as already mentioned in Section 2.02.1.5, the two options may lead to substantial differences in design quantities for high return periods. In this respect, the

most important questions, which have not received definite answers yet, are again related to the notion of infinity. Accordingly, the distribution tails are important to know in engineering design. However, the study of the tails is difficult and uncertain because the tails refer to infrequent events that require very long records to appear. Traditionally, rainfall records are analyzed in two ways. The most frequent is to choose the highest of all recorded precipitation intensities (for a given averaging timescale) at each year and form a statistical sample with size equal to the number of years of the record. The other is to form a sample with all recorded intensities over a certain threshold irrespective of the year they occurred. Usually, the threshold is chosen high enough, so that the sample size is again equal to the number of years of the record. This however is not necessary: it can well be set equal to zero, so that all recorded intensities are included in the sample. However, the threshold simplifies the study and helps focus the attention on the distribution tail. If x1 ; x2 ; y; xn are random variables representing the recorded average intensities within a year at nonoverlapping time periods equal to a chosen timescale d, then the maximum among them y :¼ maxðx1 ; x2 ; y; xn Þ has a distribution function Hn(y) fully dependent on the joint distribution function of xi . Assuming that xi are IID with common distribution function F(x), then Hn(x) ¼ [F(x)]n. If n is not constant, but rather can be regarded as a realization of a random variable (corresponding to the fact that the number of rainfall events is not the same in each year) with Poisson distribution with mean n, then the distribution function H becomes (e.g., Todorovic and Zelenhasic, 1970; Rossi et al., 1984)

HðxÞ ¼ expfn½1  FðxÞg

ð58Þ

In particular, if the threshold has been chosen with the above rule (to make the sample size equal to the number of years of the record) then obviously n ¼ 1. Equation (58) expresses in a satisfactory approximation, the relationship between the above two methodologies and the respective distributions F and H. The two options discussed above are then represented as follows: 1. Exponential tail.

FðxÞ ¼ 1  expðx=l þ cÞ; HðxÞ ¼ exp½expðx=l þ cÞ; x  lc

ð59Þ

where l 4 0 and c 4 0 are parameters, so that lc represents the specified threshold. Here F is the exponential distribution and H is the Gumbel distribution, also known as extreme value type I (EV1) distribution. 2. Power tail.

x  FðxÞ ¼ 1½1 þ k  c 1=k ;  h l  i1=k  x ; HðxÞ ¼ exp  1 þ k  c l

x  lc

ð60Þ

where l 4 0, c 4 0 and k 4 0 are parameters, and lc represents the specified threshold. Here F is the generalized Pareto distribution (a generalized form of Equation (18)) and H is the generalized extreme value (GEV) distribution.

Precipitation

In the case k 4 0 considered here, GEV is also called the extreme value type II (EV2) distribution. The case ko0 is mathematically possible and is called the extreme value type III (EV3) distribution. However, this is inappropriate for rainfall as it puts an upper bound (lc) for x, which is inconsistent. The case k ¼ 0, corresponds precisely to the exponential tail (exponential and Gumbel distributions). For years, the exponential tail and the Gumbel distribution have been the prevailing models for rainfall extremes, despite the fact that they yield unsafe (the smallest possible) design rainfall values. Recently, however, their appropriateness for rainfall has been questioned. Koutsoyiannis (2004a, 2005a, 2007) discussed several theoretical reasons that favor the power/EV2 over the exponential/EV1 case. As already mentioned (Section 2.02.1.5.5), Koutsoyiannis (2004b, 2005a) compiled an ensemble of annual maximum daily rainfall series from 169 stations in the Northern Hemisphere (28 from Europe and 141 from the USA) roughly belonging to six major climatic zones and all having lengths from 100–154 years. The analysis provides sufficient support for the general applicability of the EV2 distribution model worldwide. Furthermore, the ensemble of all samples was analyzed in combination and it was found that several dimensionless statistics are virtually constant worldwide, except for an error that can be attributed to a pure statistical sampling effect. This enabled the formation of a compound series of annual maxima, after standardization by the mean, for all stations (see Figure 13, which shows the distribution of a compound sample over threshold of all stations, except one in which only annual maxima existed). The findings support the estimation of a unique k for all stations, which was found to be 0.15. Additional empirical evidence with the same conclusions is provided by the Hershfield’s (1961) data set, which was the basis of the formulation of Hershfield’s PMP method. Koutsoyiannis (1999) showed that this data set does not support the hypothesis of an upper bound in precipitation, that is, PMP. Rather, it is consistent with the EV2 distribution with k ¼ 0.13, while the value k ¼ 0.15 can be acceptable for that data set too (Koutsoyiannis, 2004b). This enhances the trust that an EV2 distribution with k ¼ 0.15 can be regarded as a generalized model appropriate for mid-latitude areas of the Northern Hemisphere. In a recent study, Veneziano et al. (2009) used multifractal analysis to show that the annual rainfall maximum for timescale d can be approximated by a GEV distribution and that typical values of k lie in the range 0.09–0.15 with the larger values being associated with more arid climates. This range of values agrees well with the findings of Koutsoyiannis (1999, 2004b, 2005a). Similar results were provided by Chaouche (2001) and Chaouche et al. (2002). Chaouche (2001) explored a database of 200 rainfall series of various time steps (month, day, hour, and minute) from the five continents, each including more than 100 years of data. Using multifractal analyses, it was found that (1) an EV2/Pareto type law describes the rainfall amounts for large return periods; (2) the exponent of this law is scale invariant over scales greater than an hour (as stated in Section 2.02.4.2, it cannot be otherwise

71

because this is dictated by theoretical reasons); and (3) this exponent is almost space invariant. Other studies have also expressed skepticism for the appropriateness of the Gumbel distribution for the case of rainfall extremes and suggested hyper-exponential tail behavior. Coles et al. (2003) and Coles and Pericchi (2003) concluded that inference based on the Gumbel model for annual maxima may result in unrealistically high return periods for certain observed events and suggested a number of modifications to standard methods, among which is the replacement of the Gumbel model with the GEV model. Mora et al. (2005) confirmed that rainfall in Marseille (a rain gauge included in the study by Koutsoyiannis (2004b)) shows hyper-exponential tail behavior. They also provided two regional studies in the Languedoc-Roussillon region (south of France) with 15 and 23 gauges, for which they found that a similar distribution with hyper-exponential tail could be fitted. This finding, when compared to previous estimations, leads to a significant increase in the depth of rare rainfall. On the same lines, Bacro and Chaouche (2006) showed that the distribution of extreme daily rainfall at Marseille is not in the Gumbel-law domain. Sisson et al. (2006) highlighted the fact that standard Gumbel analyses routinely assign near-zero probability to subsequently observed disasters, and that for San Juan, Puerto Rico, standard 100-year predicted rainfall estimates may be routinely underestimated by a factor of two. Schaefer et al. (2006) using the methodology by Hosking and Wallis (1997) for regional precipitation-frequency analysis and spatial mapping for 24-h and 2-h durations for the Washington State, USA, found that the distribution of rainfall maxima in this State generally follows the EV2 distribution type.

2.02.5.3 Ombrian Relationships One of the major tools in hydrologic design is the ombrian relationship, more widely known by the misnomer rainfall intensity-duration-frequency (IDF) curve. An ombrian relationship (from the Greek ombros, rainfall) is a mathematical relationship estimating the average rainfall intensity i over a given timescale d (sometimes incorrectly referred to as duration) for a given return period T (also commonly referred to as frequency, although frequency is generally understood as reciprocal to period). Several forms of ombrian relationships are found in the literature, most of which have been empirically derived and validated by the long use in hydrologic practice. Attempts to give them a theoretical basis have often used inappropriate assumptions and resulted in oversimplified relationships that are not good for engineering studies. In fact, an ombrian relationship is none other than a family of distribution functions of rainfall intensity for multiple timescales. This is because, the return period is tied to the distribution function, that is, T ¼ d/[1  F(x)], where d is the mean interarrival time of an event that is represented by the variable x, typically 1 year. Thus, a distribution function such as one of those described in Section 2.02.4.2, is at the same time an ombrian relationship. This has been made clear in Koutsoyiannis et al. (1998) who showed that the empirical considerations usually involved in the construction of

72

Precipitation

ombrian curves are not necessary at all, and create difficulties and confusion. However, the direct use in engineering design of a fully consistent multiscale distribution function may be too complicated. Simplifications are possible to provide satisfactory approximations, given that only the distribution tail is of interest and that the range of scales of interest in engineering studies is relatively narrow. Such simplifications, which were tested recently and were found to be reasonable (Papalexiou and Koutsoyiannis, 2009) are 1. The separability assumption, according to which the influences of return period and timescale are separable (Koutsoyiannis et al., 1998), that is,

iðd; TÞ ¼

aðTÞ bðdÞ

ð61Þ

where a(T) and b(d) are mathematical expressions to be determined. 2. The use of the Pareto distribution for the rainfall intensity over some threshold at any timescale, as discussed in Section 2.02.5.2; this readily provides a simple expression for a(T). 3. The expression of b(d) in the simple form

bðdÞ ¼ ð1 þ d=yÞ Z

ð62Þ

where y40 and Z40 are parameters. A justification of this relationship, which is a satisfactory approximation for timescales up to a few days, can be found in Koutsoyiannis (2006a). Based on assumptions 1–3, we easily deduce that the final form of the ombrian relationship is

iðd; TÞ ¼ l0

ðT=dÞ k  c0 ð1 þ d=yÞ Z

ð63Þ

where c0 40, l0 40 and k40 are parameters. In particular, as discussed in Section 2.02.5.2, k is the tail-determining parameter and unless a long record exists, which could support a different value, it should be assumed k ¼ 0.15. Equation (63) is dimensionally consistent, if y has units of time (as well as d), l0 has units of intensity, and k and c are dimensionless. The numerator of Equation (63) differs from a pure power law that has been commonly used in engineering practice, as well as in some multifractal analyses. Consistent parameter-estimation techniques for ombrian relationships have been discussed in Koutsoyiannis et al. (1998) as well as in Chapter 2.18 Statistical Hydrology.

References Abramowitz M and Stegun IA (1965) Handbook of Mathematical Functions. New York: Dover. Adler RF and Negri AJ (1988) A satellite infrared technique to estimate tropical convective and stratiform rainfall. Journal of Applied Meteorology 27: 30--51. Alishouse JC, Snyder SA, Vongsathorn J, and Ferraro RR (1990) Determination of oceanic total precipitable water from the SSM/I. IEEE Transactions on Geoscience and Remote Sensing 28(5): 811--816.

Anthes RA (1982) Tropical Cyclones: Their Evolution, Structure, and Effects. Meteorological Monographs Number 41. Boston, MA: American Meteorological Society. Aonashi K and Liu G (2000) Passive microwave precipitation retrievals using TMI during the Baiu period of 1998. Part I: Algorithm Description and Validation. Journal of Applied Meteorology 39: 2024--2037. Arakawa H (1956) On the secular variation of annual totals of rainfall at Seoul from 1770 to 1944. Theoretical and Applied Climatology 7(2): 205--211. Arkin PA (1979) The relationship between fractional coverage of high cloud and rainfall accumulations during GATE over the B-scale array. Monthly Weather Review 107: 1382--1387. Arkin PA and Meisner BN (1987) The relationship between large-scale convective rainfall and cold cloud over the Western Hemisphere during 1982–84. Monthly Weather Review 115: 51--74. Atlas D and Ulbrich CW (1977) Path and area integrated rainfall measurement by microwave attenuation in the 1–3 cm band. Journal of Applied Meteorology 16: 1322--1331. Atlas D, Ulbrich CW, and Meneghini R (1984) The multiparameter remote measurement of rainfall. Radio Science 19: 3--21. Austin PM and Houze RA (1972) Analysis of the structure of precipitation patterns in New England. Journal of Applied Meteorology 11: 926--935. Ba MB and Nicholson SE (1998) Analysis of convective activity and its relationship to the rainfall over the rift valley lakes of East Africa during 1983–90 using the Meteosat infrared channel. Journal of Applied Meteorology 37: 1250-1264. Bacro J-N and Chaouche A (2006) Incertitude d’estimation des pluies extreˆmes du pourtour me´diterrane´en: Illustration par les donne´es de Marseille. Hydrological Sciences Journal 51(3): 389--405. Baloutsos G, Bourletsikas A, and Kaoukis K (2005) Study and investigation of the fog precipitation characteristics in the fir forest of Agios Nikolaos, Evritania, Greece, in Greek. Geotechnica Epistemonica Themata 2(16): 34--45. Barnes GM, Zipser EJ, Jorgensen D, and Marks FD (1983) Mesoscale and convective structure of hurricane rainband. Journal of the Atmospheric Sciences 40: 2125--2137. Barrett EC and Martin DW (1981) The Use of Satellite Data in Rainfall Monitoring. London: Academic Press. Battan LJ (1973) Radar Observations of the Atmosphere. Chicago, IL: The University of Chicago Press. Beard KV (1976) Terminal velocity and shape of cloud and precipitation drops aloft. Journal of the Atmospheric Sciences 33: 851--864. Beer T and Giannini L (1980) Tropical cyclone cloud bands. Journal of the Atmospheric Sciences 37: 1511--1520. Bellon A, Lovejoy S, and Austin G (1980) Combining satellite and radar data for the short-range forecasting of precipitation. Monthly Weather Review 108: 1554--1566. Bennetts DA and Sharp JC (1982) The relevance of conditional symmetric instability to the prediction of mesoscale frontal rainbands. Quarterly Journal of the Royal Meteorological Society 108: 595--602. Benson MA (1973) Thoughts on the design of design floods Floods and Droughts, Proceedings of the 2nd International Symposium in Hydrology, pp. 27--33. Fort Collins, CO: Water Resources Publications. Berg W and Chase R (1992) Determination of mean rainfall from the special sensor microwave/imager (SSM/I) using a mixed lognormal distribution. Journal of Atmospheric and Oceanic Technology 9: 129--141. Berg W, Olson W, Ferraro R, Goodman SJ, and LaFontaine FJ (1998) An assessment of the first- and second-generation navy operational precipitation retrieval algorithms. Journal of the Atmospheric Sciences 55: 1558--1575. Bergeron T (1935) On the physics of cloud and precipitation. In: Proceedings of the Fifth Assembly of the International Union of Geodesy and Geophysics (IUGG), pp. 156–178. Lisbon, Portugal. Berry EX and Reinhardt RL (1974a) An analysis of cloud drop growth by collection: Part I. Double distributions. Journal of the Atmospheric Sciences 31: 1814--1824. Berry EX and Reinhardt RL (1974b) An analysis of cloud drop growth by collection: Part II. Single initial distributions. Journal of the Atmospheric Sciences 31: 1825--1831. Bluestein HB (2003) Tornadoes. In: Holton JR, Curry JA, and Pyle JA (eds.) Encyclopedia of Atmospheric Sciences, pp. 2290--2297. London: Academic Press. Bowen EG (1950) The formation of rain by coalescence. Australian Journal of Scientific Research A3: 192--213. Bringi VN and Chandrasekar V (2001) Polarimetric Doppler Weather Radar. New York, NY: Cambridge University Press. Brock JR (1972) Condensational growth of atmospheric aerosols. In: Hidy GM (ed.) Aerosols and Atmospheric Chemistry, pp. 149--153. New York, NY: Academic Press.

Precipitation

Browning KA and Ludlam FH (1962) Airflow in convective storms. Quarterly Journal of the Royal Meteorological Society 88: 117--135. Byers HR and Braham RR (1949) The thunderstorm. Final Report of the Thunderstorm Project, US Deptartment of Commerce. Washington, DC: US Government Printing Office. Calenda G, Gorgucci E, Napolitano F, Novella A, and Volpi E (2005) Multifractal analysis of radar rainfall fields over the area of Rome. Advances in Geosciences 2: 293--299. Carvalho LMV, Lavalle´e D, and Jones C (2002) Multifractal properties of evolving convective systems over tropical South America. Geophysical Research Letters 29: doi: 10.1029/2001GL014276. Chaouche K (2001) Approche multifractale de la mode´lisation stochastique en hydrologie. The`se, Ecole Nationale du Ge´nie Rural, des Eaux et des Foreˆts, Centre de Paris, France. http://www.engref.fr/thesechaouche.htm (accessed March 2010). Chaouche K, Hubert P, and Lang G (2002) Graphical characterisation of probability distribution tails. Stochastic Environmental Research and Risk Assessment 16(5): 342--357. Chen SS, Lonfat M, Knaff JA, and Marks FD Jr. (2006) Effects of vertical wind shear and storm motion on tropical cyclone rainfall asymmetries deduced from TRMM. Monthly Weather Review 134: 3190--3208. Chen SS, Price JF, Zhao W, Donelan M, and Walsh EJ (2007) The CBLAST-hurricane program and the next-generation fully coupled atmosphere–wave–ocean models for hurricane research and prediction. Bulletin of the American Meteorological Society 88: 311--317. Cheng M and Brown R (1995) Delineation of precipitation areas by correlation of Meteosat visible and infrared data with radar data. Monthly Weather Review 123: 2743--2757. Cheng M, Brown R, and Collier C (1993) Delineation of precipitation areas using Meteosat infrared and visible data in the region of the United Kingdom. Journal of Applied Meteorology 32: 884--898. Chin EH (1977) Modelling daily precipitation occurrence process with Markov chains. Water Resources Research 13(6): 956--959. Clarke RT (1998) Stochastic Processes for Water Scientists. New York, NY: Wiley. Coles S and Pericchi L (2003) Anticipating catastrophes through extreme value modelling. Applied Statistics 52: 405--416. Coles S, Pericchi LR, and Sisson S (2003) A fully probabilistic approach to extreme rainfall modeling. Journal of Hydrology 273(1–4): 35--50. Cotton WR and Anthes RA (1989) Storm and Cloud Dynamics, International Geophysics Series, vol. 44. San Diego, CA: Academic Press. Cotton WR, Thompson G, and Mielke PW (1994) Realtime mesoscale prediction on workstations. Bulletin of the American Meteorological Society 75: 349--362. Cowpertwait P, Isham V, and Onof C (2007) Point process models of rainfall: Developments for fine-scale structure. Proceedings of the Royal Society A 463(2086): 2569--2587. Cox D and Isham V (1988) A simple spatial–temporal model of rainfall. Proceedings of the Royal Society A 415: 317--328. De Bruin HAR (1980) A stochastic description of wet and dry spells in terms of an effective number of days. Journal of Hydrology 45: 91--99. Deidda R (2000) Rainfall downscaling in a space–time multifractal framework. Water Resources Research 36(7): 1779--1794. Deidda R, Badas MG, and Piga E (2004) Space–time scaling in high-intensity tropical ocean global atmosphere coupled ocean–atmosphere response experiment (TOGACOARE) storms. Water Resources Research 40: W02506.1 (doi:10.1029/ 2003WR002574). Deidda R, Benzi R, and Siccardi F (1999) Multifractal modeling of anomalous scaling laws in rainfall. Water Resources Research 35(6): 1853--1867. Deidda R, Grazia-Badas M, and Piga E (2006) Space–time multifractality of remotely sensed rainfall fields. Journal of Hydrology 322: 2--13 (doi:10.1016/ j.jhydrol.2005.02.036). Dingman SL (1994) Physical Hydrology. Englewood Cliffs, NJ: Prentice Hall. Doviak R and Zrnic D (1993) Doppler Radar and Weather Observations. p. 562. New York, NY: Elsevier. Drake RL (1972a) A general mathematical survey of the coagulation equation. In: Hidy GM and Brock JR (eds.) Topics in Current Aerosol Research (Part 2), pp. 201--384. Oxford: Pergamon. Drake RL (1972b) The scalar transport equation of coalescence theory: Moments and kernels. Journal of the Atmospheric Sciences 29: 537--547. Elachi C (1987) Introduction to the Physics and Techniques of Remote Sensing. Wiley Series in Remote Sensing. Hoboken, NJ: Wiley. Emanuel KA (1986) An air–sea interaction theory for tropical cyclones. Part I: Steady state maintenance. Journal of the Atmospheric Sciences 43: 585--604. Emanuel KA (1989) Polar lows as artic hurricanes. Tellus 41A: 1--17.

73

Entekhabi D, Rodriguez-Iturbe I, and Eagleson PS (1989) Probabilistic representation of the temporal rainfall process by the modified Neyman–Scott rectangular pulses model: Parameter estimation and validation. Water Resources Research 25(2): 295--302. Epstein ES (1969) Stochastic dynamic prediction. Tellus 21: 739--757. Federer B and Waldvogel A (1975) Hail and raindrop size distributions from a Swiss multicell storm. Journal of Applied Meteorology 14: 91--97. Feingold G and Levin Z (1986) The lognormal fit to raindrop spectra from frontal convective clouds in Israel. Journal of Climate and Applied Meteorology 25: 1346--1363. Feller W (1950) An Introduction to Probability Theory and Its Applications. New York, NY: Wiley. Ferraro RR and Marks GF (1995) The development of SSM/I rain-rate retrieval algorithms using ground-based radar measurements. Journal of Atmospheric and Oceanic Technology 12: 755--770. Ferriday JG and Avery SK (1994) Passive microwave remote sensing of rainfall with SSM/I: Algorithm development and implementation. Journal of Applied Meteorology 33: 1587--1596. Feyerherm AM and Bark LD (1964) Statistical methods for persistent precipitation patterns. Journal of Applied Meteorology 4: 320--328. Findeisen W (1938) Die Kolloid Meteorologischen Vorgang der Niederschlags Bildung. Meteorologische Zeitschrift 55: 121--133. Fletcher NH (1962) The Physics of Rainclouds. Cambridge: Cambridge University Press. Foufoula-Georgiou E and Guttorp P (1986) Compatibility of continuous rainfall occurrence models with discrete rainfall observations. Water Resources Research 22(8): 1316--1322. Foufoula-Georgiou E and Lettenmaier DP (1986) Continuous-time versus discrete-time point process models for rainfall occurrence series. Water Resources Research 22(4): 531--542. Fraedrich K and Larnder C (1993) Scaling regimes of composite rainfall time series. Tellus 45A: 289--298. Frank WM (1977) The structure and energetics of the tropical cyclone: I. Storm structure. Monthly Weather Review 105: 1119--1135. Gabriel KR and Neumann J (1962) A Markov chain model for daily rainfall occurrences at Tel Aviv. Quarterly Journal of the Royal Meteorological Society 88: 90--95. Garcı´a-Marı´n AP, Jime´nez-Hornero FJ, and Ayuso JL (2007) Applying multifractality and the self-organized criticality theory to describe the temporal rainfall regimes in Andalusia (southern Spain). Hydrological Processes (doi:10.1002/hyp.6603). Gebremichael M and Krajewski WF (2004) Assessment of the statistical characterization of small-scale rainfall variability from radar: Analysis of TRMM ground validation datasets. Journal of Applied Meteorology 43(8): 1180--1199. Gebremichael M, Over TM, and Krajewski WF (2006) Comparison of the scaling properties of rainfall derived from space- and surface-based radars. Journal of Hydrometeorology 7: 1277--1294. Georgakakos KP, Carsteanu AA, Sturdevant PL, and Cramer JA (1994) Observation and analysis of Midwestern rain rates. Journal of Applied Meteorology 33: 1433--1444. Gillespie DT (1972) The stochastic coalescence model for cloud droplet growth. Journal of the Atmospheric Sciences 29: 1496--1510. Gillespie DT (1975) Three models for the coalescence growth of cloud drops. Journal of the Atmospheric Sciences 32: 600--607. Gray WM (1968) Global view of the origins of tropical disturbances and storms. Monthly Weather Review 96: 669--700. Gray WM (1979) Hurricanes: Their formation, structure and likely role in the tropical circulation. In: Shaw DB (ed.) Meteorology Over the Tropical Oceans. London: Royal Meteorological Society. Grell GA, Dudhia J, and Stauffer DR (1995) A description of the fifth-generation Pennsylvania State/NCAR mesoscale model (MM5), NCAR Tech. Note TN398 þ STR122 pp. (Available from UCAR Communications, P.O. Box 3000, Boulder, CO 80307.) Griffith CG, Woodley WL, Grube PG, Martin DW, Stout J, and Sikdar DN (1978) Rain estimation from geosynchronous satellite imagery – visible and infrared studies. Monthly Weather Review 106: 1153--1171. Grody NC (1991) Classification of snow cover and precipitation using the special sensor microwave imager. Journal of Geophysical Research 96(D4): 7423--7435. Gunn KLS and Marshall JS (1958) The distribution with size of aggregate snowflakes. Journal of Meteorology 15: 452--461. Gunn R and Kinzer GD (1949) The terminal velocity of fall for water drops in stagnant air. Journal of Meteorology 6: 243--248. Gu¨ntner A, Olsson J, Calver A, and Gannon B (2001) Cascade-based disaggregation of continuous rainfall time series: The influence of climate. Hydrology and Earth System Sciences 5: 145--164.

74

Precipitation

Gupta VK and Waymire E (1990) Multiscaling properties of spatial rainfall and river flow distributions. Journal of Geophysical Research 95: 1999--2009. Gyasi-Agyei Y and Willgoose GR (1997) A hybrid model for point rainfall modelling. Water Resources Research 33(7): 1699--1706. Haan CT, Allen DM, and Street JO (1976) A Markov chain model of daily rainfall. Water Resources Research 12(3): 443--449. Hakim GJ (2003) Cyclogenesis. In: Holton JR, Curry JA, and Pyle JA (eds.) Encyclopedia of Atmospheric Sciences, pp. 589--594. London: Academic Press. Hane CE (1986) Extratropical squall lines and rainbands. In: Ray PS (ed.) Mesoscale Meteorology and Forecasting, ch. 1, pp. 359–389. Boston: American Meteorological Society. Harr RD (1982) Fog drip in the Bull Run municipal watershed, Oregon. Water Resources Bulletin 18: 785--789. Harrold TI, Sharma A, and Sheather SJ (2003) A nonparametric model for stochastic generation of daily rainfall occurrence. Water Resources Research 39(10): 1300 (doi: 10.1029/2003WR002182). Harrold TW and Austin PM (1974) The structure of precipitation systems – a review. Journal de Recherches Atmospheriques 8: 41--57. Hawkins HF and Rubsam DT (1968) Hurricane Hilda, 1964. II. Structure and budgets of the hurricane on October 1, 1964. Monthly Weather Review 96: 617--636. Hemond HF and Fechner-Levy EJ (2000) Chemical Fate and Transport in the Environment, 2nd edn. London: Academic Press. Hershfield DM (1961) Estimating the probable maximum precipitation. ASCE: Journal of the Hydraulics Division 87(HY5): 99--106. Hershfield DM (1965) Method for estimating probable maximum precipitation. Journal of the American Water Works Association 57: 965--972. Hershfield DM (1970) A comparison of conditional and unconditional probabilities for wet- and dry-day sequences. Journal of Applied Meteorology 9(5): 825-827. Hinton BB, Olson WS, Martin DW, and Auvine B (1992) A passive microwave algorithm for tropical oceanic rainfall. Journal of Applied Meteorology 31: 1379--1395. Hobbs PV, Bowdle DA, and Radke LF (1985) Particles in the lower troposphere over the high plains of the United States. Part I: Size distributions, elemental compositions and morphologies. Journal of Climate and Applied Meteorology 24: 1344--1356. Holland GJ (1980) An analytic model of the wind and pressure profiles in hurricanes. Monthly Weather Review 108: 1212--1218. Hong S-Y, Dudhia J, and Chen S-H (2004) A revised approach to ice microphysical processes for the bulk parameterization of clouds and precipitation. Monthly Weather Review 132: 103--120. Horton RE (1931) The field, scope and status of the science of hydrology. In: Trans. Am. Geophys. Union, 12th Annu. Mtg., National Research Council, pp. 189–202. Washington, DC. Hosking JRM and Wallis JR (1997) Regional Frequency Analysis – an Approach Based on L-Moments. New York, NY: Cambridge University Press. Houghton HG (1985) Physical Meteorology. Cambridge, MA: MIT Press. Houze RA (1993) Cloud Dynamics. San Diego, CA: Academic Press. Houze RA, Hobbs PV, Herzegh PH, and Parsons DB (1979) Size distributions of precipitation particles in frontal clouds. Journal of the Atmospheric Sciences 36: 156--162. Houze RA, Marks FD Jr., and Black RA (1992) Dual-aircraft investigation of the inner core of hurricane Norbert. Part II: Mesoscale distribution of ice particles. Journal of the Atmospheric Sciences 49: 943--962. Hsu HM, Moncrieff MW, Tung WW, and Liu C (2006) Multiscale temporal variability of warm-season precipitation over North America: Statistical analysis of radar measurements. Journal of the Atmospheric Sciences 63: 2355--2368. Hubert P, Bendjoudi H, Schertzer D, and Lovejoy S (1998) A Multifractal explanation for rainfall intensity–duration–frequency curves. In: Llasat C, Versace P, and Ferrari E (eds.) Heavy Rains and Flash Floods, pp. 21–28, Cosenza, Italy: Natl. Res. Counc., Group for Prev. from Hydrol. Disasters. Hurst HE (1951) Long term storage capacities of reservoirs. Transactions of the American Society of Civil Engineers 116: 776--808. Hutchinson MF (1990) A point rainfall model based on a three-state continuous Markov occurrence process. Journal of Hydrology 114: 125--148. Jorgensen DP (1984a) Mesoscale and convective-scale characterisitcs of mature hurricanes. Part I: General observations by research aircraft. Journal of the Atmospheric Sciences 41(8): 1268--1286. Jorgensen DP (1984b) Mesoscale and convective-scale characterisitcs of mature hurricanes. Part II: Inner core structure of hurricane Allen (1980). Journal of the Atmospheric Sciences 41(8): 1287--1311. Joss J and Gori E (1978) Shapes of raindrop size distributions. Journal of Applied Meteorology 17: 1054--1061.

Joss J and Waldvogel A (1990) Precipitation measurement and hydrology. In: Atlas D (ed.) Radar in Meteorology, Battan Memorial and 40th Anniversary Radar Meteorology Conference, American Meteorological Society, pp. 577–597. Chicago, IL: The University of Chicago Press. Katsanos D, Lagouvardos K, Kotroni V, and Argiriou A (2007) Relationship of lightning activity with microwave brightness temperatures and spaceborne radar reflectivity profiles in the Central and Eastern Mediterranean. Journal of Applied Meteorology 46: 1901--1912. Katz RW (1977a) An application of chain-dependent processes to meteorology. Journal of Applied Probability 14: 598--603. Katz RW (1977b) Precipitation as a chain-dependent process. Journal of Applied Meteorology 16: 671--676. Katz RW and Parlange MB (1998) Overdispersion phenomenon in stochastic modeling of precipitation. Journal of Climate 11: 591--601. Kavvas ML and Delleur JW (1981) A stochastic cluster model of daily rainfall sequences. Water Resources Research 17(4): 1151--1160. Kepert J (2001) The dynamics of boundary layer jets within the tropical cyclone core. Part I: Linear theory. Journal of the Atmospheric Sciences 58: 2469--2484. Kepert J and Wang Y (2001) The dynamics of boundary layer jets within the tropical cyclone core. Part II: Nonlinear enhancement. Journal of the Atmospheric Sciences 58: 2485--2501. Kessler E (1969) On the Distribution and Continuity of Water Substance in Atmospheric Circulation. Meteorological Monographs 10, no. 32, 84pp. Boston, MA: American Meteorological Society. Kidd C, Kniveton D, and Barrett EC (1998) The advantages and disadvantages of statistically derived–empirically calibrated passive microwave algorithms for rainfall estimation. Journal of the Atmospheric Sciences 55: 1576--1582. Kidder SQ and Vonder Haar TH (1995) Satellite Meteorology. London: Academic Press. King PW, Hogg WD, and Arkin PA (1995) The role of visible data in improving satellite rain-rate estimates. Journal of Applied Meteorology 34: 1608--1621. Klemes V (2000) Tall tales about tails of hydrological distributions. Journal of Hydrologic Engineering 5(3): 227--239. Kolmogorov AN (1933) Grundbegrijfe der Wahrscheinlichkeitsrechnung Ergebnisse der Math. (2), Berlin; 2nd English edn. (1956): Foundations of the Theory of Probability, 84pp. New York, NY: Chelsea Publishing Company. Kolmogorov AN (1940) Wienersche Spiralen und einige andere interessante Kurven in Hilbertschen Raum. Doklady Akademii Nauk SSSR 26: 115--118. Kosambi DD (2005) The Culture and Civilization of Ancient India in Historical Outline. Delhi: Vikas Publishing House. Kotroni V and Lagouvardos K (2004) Evaluation of MM5 high-resolution real-time forecasts over the urban area of Athens, Greece. Journal of Applied Meteorology 43: 1666--1678. Kottegoda NT and Horder MA (1980) Daily flow model based on rainfall occurrences using pulses and a transfer function. Journal of Hydrology 47: 215--234. Koutsoyiannis D (1999) A probabilistic view of Hershfield’s method for estimating probable maximum precipitation. Water Resources Research 35(4): 1313--1322. Koutsoyiannis D (2000a) A generalized mathematical framework for stochastic simulation and forecast of hydrologic time series. Water Resources Research 36(6): 1519--1533. Koutsoyiannis D (2000b) Lecture Notes on Hydrometeorology – Part 1, 2nd edn., 157pp. Athens: National Technical University of Athens. http://www.itia.ntua.gr/en/ docinfo/116 (accessed March 2010). Koutsoyiannis D (2001) Coupling stochastic models of different time scales. Water Resources Research 37(2): 379--392. Koutsoyiannis D (2002) The Hurst phenomenon and fractional Gaussian noise made easy. Hydrological Sciences Journal 47(4): 573--595. Koutsoyiannis D (2003a) Climate change, the Hurst phenomenon, and hydrological statistics. Hydrological Sciences Journal 48(1): 3--24. Koutsoyiannis D (2003b) Rainfall disaggregation methods: Theory and applications. In: Piccolo D and Ubertini L (eds.) Proceedings, Workshop on Statistical and Mathematical Methods for Hydrological Analysis, pp. 1--23. Rome: Universita` di Roma ‘‘La Sapienza’’. Koutsoyiannis D (2004a) Statistics of extremes and estimation of extreme rainfall: 1. Theoretical investigation. Hydrological Sciences Journal 49(4): 575--590. Koutsoyiannis D (2004b) Statistics of extremes and estimation of extreme rainfall: 2. Empirical investigation of long rainfall records. Hydrological Sciences Journal 49(4): 591--610. Koutsoyiannis D (2005a) Uncertainty, entropy, scaling and hydrological stochastics: 1. Marginal distributional properties of hydrological processes and state scaling. Hydrological Sciences Journal 50(3): 381--404. Koutsoyiannis D (2005b) Uncertainty, entropy, scaling and hydrological stochastics: 2. Time dependence of hydrological processes and time scaling. Hydrological Sciences Journal 50(3): 405--426.

Precipitation

Koutsoyiannis D (2006a) An entropic-stochastic representation of rainfall intermittency: The origin of clustering and persistence. Water Resources Research 42(1): W01401 (doi:10.1029/2005WR004175). Koutsoyiannis D (2006b) Nonstationarity versus scaling in hydrology. Journal of Hydrology 324: 239--254. Koutsoyiannis D (2007) A critical review of probability of extreme rainfall: Principles and models. In: Ashley R, Garvin S, Pasche E, Vassilopoulos A, and Zevenbergen C (eds.) Advances in Urban Flood Management, pp. 139--166. London: Taylor and Francis. Koutsoyiannis D (2010) A random walk on water. Hydrology and Earth System Sciences 14: 585--601. Koutsoyiannis D, Efstratiadis A, Mamassis N, and Christofides A (2008) On the credibility of climate predictions. Hydrological Sciences Journal 53(4): 671--684. Koutsoyiannis D, Kozonis D, and Manetas A (1998) A mathematical framework for studying rain-fall intensity–duration–frequency relationships. Journal of Hydrology 206(1–2): 118--135. Koutsoyiannis D, Mamassis N, and Tegos A (2007) Logical and illogical exegeses of hydrometeorological phenomena in ancient Greece. Water Science and Technology: Water Supply 7(1): 13--22. Koutsoyiannis D and Montanari A (2007) Statistical analysis of hydroclimatic time series: Uncertainty and insights. Water Resources Research 43(5): W05429 (doi: 10.1029/2006WR005592). Koutsoyiannis D and Onof C (2001) Rainfall disaggregation using adjusting procedures on a Poisson cluster model. Journal of Hydrology 246: 109--122. Koutsoyiannis D, Onof C, and Wheater HS (2003) Multivariate rainfall disaggregation at a fine timescale. Water Resources Research 39(7): 1173 (doi:10.1029/ 2002WR001600). Koutsoyiannis D and Xanthopoulos T (1999) Engineering Hydrology, 3rd edn., p. 418. Athens: National Technical University of Athens (in Greek). Krajewski WF and Smith JA (2002) Radar hydrology: Rainfall estimation. Advances in Water Resources 25(8–12): 1387--1394 (doi: 10.1016/S03091708(02)00062-3). Kummerow C and Giglio L (1994a) A passive microwave technique for estimating rainfall and vertical structure information from space. Part I: Algorithm description. Journal of Applied Meteorology 33: 3--18. Kummerow C and Giglio L (1994b) A passive microwave technique for estimating rainfall and vertical structure information from space. Part II: Applications to SSM/I data. Journal of Applied Meteorology 33: 19--34. Kummerow C, Barnes W, Kozu T, Shiue J, and Simpson J (1998) The tropical rainfall measuring mission (TRMM) sensor package. Journal of Atmospheric and Oceanic Technology 15: 809--817. Kummerow C, Hong Y, Olson WS, et al. (2001) The evolution of the Goddard profiling algorithm (GPROF) for rainfall estimation from passive microwave sensors. Journal of Applied Meteorology 40: 1801--1820. Kummerow CD, Olson WS, and Giglio L (1996) A simplified scheme for obtaining precipitation and vertical hydrometeor profiles from passive microwave sensors. IEEE Transactions on Geoscience and Remote Sensing 34: 1213--1232. Kundu PK and Bell TL (2003) A Stochastic model of space–time variability of mesoscale rainfall: Statistics of spatial averages. Water Resources Research 39(12): 1328 (doi:10.1029/2002WR001802). Lagouvardos K and Kotroni V (2005) Improvement of high resolution weather forecasts through humidity adjustment, based on satellite data. Quarterly Journal of the Royal Meteorological Society 131: 2695--2712. Lagouvardos K, Kotroni V, Koussis A, Feidas C, Buzzi A, and Malguzzi P (2003) The meteorological model BOLAM at the National Observatory of Athens: Assessment of two-year operational use. Journal of Applied Meteorology 42: 1667--1678. Landsea CW (2000) Climate variability of tropical cyclones. In: Pielke R Jr. and Pielke R Sr. (eds.) Storms, vol. 1, New York, NY: Routledge. Langmuir I (1948) The production of rain by a chain reaction in cumulus clouds at temperatures above freezing. Journal of Meteorology 5: 175--192. Langousis A and Koutsoyiannis D (2006) A stochastic methodology for generation of seasonal time series reproducing over-year scaling behavior. Journal of Hydrology 322(1–4): 138--154. Langousis A and Veneziano D (2007) Intensity–duration–frequency curves from scaling representations of rainfall. Water Resources Research 43(2): W02422 (doi: 10.1029/2006WR005245). Langousis A and Veneziano D (2009a) Theoretical model of rainfall in tropical cyclones for the assessment of long-term risk. Journal of Geophysical Research 114: D02106 (doi:10.1029/2008JD010080). Langousis A and Veneziano D (2009b) Long term rainfall risk from tropical cyclones in coastal areas. Water Resources Research 45: W11430 (doi:10.1029/ 2008WR007624).

75

Langousis A, Veneziano D, Furcolo P, and Lepore C (2007) Multifractal rainfall extremes: Theoretical analysis and practical estimation. Chaos Solitons and Fractals 39(3): 1182--1194 (doi:10.1016/j.chaos.2007.06.004). Langousis A, Veneziano D, and Chen S (2008) Boundary layer model for moving tropical cyclones. In: Elsner J and Jagger TH (eds.) Hurricanes and Climate Change, p. 265. New York, NY: Springer. La Seur NE and Hawkins HF (1963) An analysis of hurricane Cleo (1958) based on data from research reconnaissance aircraft. Monthly Weather Review 91: 694--709. Leith CE (1974) Theoretical skill of Monte Carlo forecasts. Monthly Weather Review 102: 409--418. Levizzani V, Amorati R, and Meneguzzo F (2002) A review of satellite-based rainfall estimation methods, MUSIC – multiple sensor precipitation measurements, integration, calibration and flood forecasting, deliverable 6.1, Italy. Lin YL, Farley RD, and Orville HD (1983) Bulk parameterization of the snow field in a cloud model. Journal of Climate and Applied Meteorology 22: 1065--1092. Liu G (2003) Precipitation. In: Holton JR, Curry JA, and Pyle JA (eds.) Encyclopedia of Atmospheric Sciences, pp. 1972--1979. London: Academic Press. Liu G and Curry JA (1992) Retrieval of precipitation from satellite microwave measurement using both emission and scattering. Journal of Geophysical Research 97: 9959--9974. Liu G and Curry JA (1993) Determination of characteristic features of cloud liquid water from satellite microwave measurements. Journal of Geophysical Research 98(D3): 5069--5092. Liu JY and Orville HD (1969) Numerical modeling of precipitation and cloud shadow effects on mountain-induced cumuli. Journal of the Atmospheric Sciences 26: 1283--1298. Locatelli JD and Hobbs PV (1974) Fall speeds and masses of solid precipitation particles. Journal of Geophysical Research 79: 2185--2197. Lonfat M, Marks FD Jr., and Chen SS (2004) Precipitation distribution in tropical cyclones using the tropical rainfall measuring mission (TRMM) microwave imager: A global perspective. Monthly Weather Review 132: 1645--1660. Lonfat M, Rogers R, Marchok T, and Marks FD Jr. (2007) A parametric model for predicting hurricane rainfall. Monthly Weather Review 135: 3086--3097. Long AB (1971) Validity of the finite-difference droplet equation. Journal of the Atmospheric Sciences 28: 210--218. Lorenz EN (1963) Deterministic nonperiodic flow. Journal of the Atmospheric Sciences 20: 130--141. Lovejoy S and Austin GL (1979) The delineation of rain areas from visible and IR satellite data for GATE and mid-latitudes. Atmosphere-Ocean 17(1): 77--92. Lovejoy S and Schertzer D (1995) Multifractals and rain. In: Kundzewicz ZW (ed.) New Uncertainty Concepts in Hydrology and Hydrological Modelling, pp. 61--103. Cambridge: Cambridge University Press. Lull HW (1964) Ecological and silvicultural aspects. In: Chow VT (ed.) Handbook of Applied Hydrology, sec. 6. New York, NY: McGraw-Hill. Lutgens FK and Tarbuck EJ (1992) The Atmosphere: An Introduction to Meteorology, 5th edn.. Englewood Cliffs, NJ: Prentice Hall. Mandelbrot BB and van Ness JW (1968) Fractional Brownian motion, fractional noises and applications. SIAM Review 10: 422--437. Marks FD (1985) Evolution of the structure of precipitation in hurricane Allen (1980). Monthly Weather Review 113: 909--930. Marks FD (2003) Hurricanes. In: Holton JR, Curry JA, and Pyle JA (eds.) Encyclopedia of Atmospheric Sciences, pp. 942--966. London: Academic Press. Marks FD and Houze RA Jr. (1987) Inner-core structure of hurricane Alicia from airborne Doppler radar observations. Journal of the Atmospheric Sciences 44: 1296--1317. Marks FD, Houze RA Jr., and Gamache JF (1992) Dual-aircraft investigation of the inner core of hurricane Norbert. Part I: Kinematic structure. Journal of the Atmospheric Sciences 49: 919--942. Marshall JS, Hitschfeld W, and Gunn KLS (1955) Advances in radar weather. Advances in Geophysics 2: 1--56. Marshall JS and Palmer WM (1948) The distribution of raindrops with size. Journal of Meteorology 5: 165--166. Mason BJ (1971) The Physics of Clouds. Oxford: Clarendon. Mass CF, Ovens D, Westrick K, and Colle B (2002) Does increasing horizontal resolution produce more skillful forecasts? Bulletin of the American Meteorological Society 83: 407--430. Meehl GA, Stocker TF, Collins WD, et al. (2007) Global climate projections. In: Solomon S, Qin D, Manning M, et al. (eds.) Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, pp. 747--845. Cambridge: Cambridge University Press.

76

Precipitation

Menabde M, Harris D, Seed A, Austin G, and Stow D (1997) Multiscaling properties of rainfall and bounded random cascades. Water Resources Research 33(12): 2823--2830. Menabde M and Sivapalan M (2000) Modelling of rainfall time series and extremes using bounded random cascades and levy-stable distributions. Water Resources Research 36(11): 3293--3300. Miller BL (1958) Rainfall rates in Florida hurricanes. Monthly Weather Review 86(7): 258--264. Molinari J, Moore PK, and Idone VP (1999) Convective structure of hurricanes as revealed by lightning locations. Monthly Weather Review 127: 520--534. Molinari J, Moore PK, Idone VP, Henderson RW, and Saljoughy AB (1994) Cloud-toground lightning in hurricane Andrew. Journal of Geophysical Research 99: 16/ 665--16/676. Mora RD, Bouvier C, Neppel L, and Niel H (2005) Approche re´gionale pour l’estimation des distributions ponctuelles des pluies journalie`res dans le Languedoc-Roussillon (France). Hydrological Sciences Journal 50(1): 17--29. National Oceanic and Atmospheric Administration (2009) About the Radar Operations Center. http://www.roc.noaa.gov/WSR88D/About.aspx (accessed March 2010). Neyman J and Scott EL (1952) A theory of the spatial distribution of galaxies. Astrophysical Journal 116: 144--163. Northrop P (1996) Modelling and Statistical Analysis of Spatial–Temporal Rainfall Fields. PhD Thesis, Department of Statistical Science, University College, London. Northrop PJ (1998) A clustered spatial–temporal model of rainfall. Proceedings of the Royal Society A 454: 1875--1888. Nykanen DK and Harris D (2003) Orographic influences on the multiscale statistical properties of precipitation. Journal of Geophysical Research 108(D8): CIP6.1--CIP6.13 (doi:10.1029/2001JD001518). Olson DA, Junker NW, and Korty B (1995) Evaluation of 33 years of quantitative precipitation forecasting at the NMC. Weather and Forecasting 10: 498--511. Olsson J (1995) Limits and characteristics of the multifractal behavior of a highresolution rainfall time series. Nonlinear Processes in Geophysics 2: 23--29. Olsson J (1998) Evaluation of a scaling cascade model for temporal rainfall disaggregation. Hydrology and Earth System Sciences 2: 19--30. Onof C, Chandler RE, Kakou A, Northrop P, Wheater HS, and Isham V (2000) Rainfall modelling using Poisson-cluster processes: A review of developments. Stochastic Environmental Research and Risk Assessment 14: 384--411. Onof C and Wheater HS (1993) Modelling of British rainfall using a random parameter Bartlett–Lewis rectangular pulse model. Journal of Hydrology 149: 67--95. Onof C and Wheater HS (1994) Improvements to the modeling of British rainfall using a modified random parameter Bartlett–Lewis rectangular pulses model. Journal of Hydrology 157: 177--195. Ooyama KV (1982) Conceptual evolution of theory and modeling of the tropical cyclone. Journal of the Meteorological Society of Japan 60: 369--379. Orlanski I (1975) A rational subdivision of scales for atmospheric processes. Bulletin of the American Meteorological Society 56(5): 527--530. O’Sullivan F, Wash CH, Stewart M, and Motell CE (1990) Rain estimation from infrared and visible GOES satellite data. Journal of Applied Meteorology 29: 209--223. Over TM and Gupta VK (1996) A space–time theory of mesoscale rainfal using random cascades. Journal of Geophysical Research 101(D21): 26319--26331. Papalexiou SM and Koutsoyiannis D (2006) A probabilistic approach to the concept of probable maximum precipitation. Advances in Geosciences 7: 51--54. Papalexiou SM and Koutsoyiannis D (2008) Probabilistic description of rainfall intensity at multiple time scales. IHP 2008 Capri Symposium: ‘‘The Role of Hydrology in Water Resources Management’’, UNESCO, International Association of Hydrological Sciences. Capri, Italy. http://www.itia.ntua.gr/en/docinfo/884 (accessed March 2010). Papalexiou SM and Koutsoyiannis D (2009) Ombrian curves: From theoretical consistency to engineering practice. In: 8th IAHS Scientific Assembly/37th IAH Congress. Hyderabad, India, 2009. http://www.itia.ntua.gr/en/docinfo/926 (accessed March 2010). Papoulis A (1991) Probability, Random Variables, and Stochastic Processes, 3rd edn. New York, NY: McGraw-Hill. Pegram GGS (1980) An auto-regressive model for multi-lag Markov chains. Journal of Applied Probability 17: 350--362. Pegram GGS and Seed AW (1998) The feasibility of stochastically modelling the spatial and temporal distribution of rainfields. WRC Report No. 550/1/98 to the Water Research Commission, Pretoria, South Africa. Perica S and Foufoula-Georgiou E (1996a) Linkage of scaling and thermodynamic parameters of rainfall: Results from midlatitude mesoscale convective systems. Journal of Geophysical Research 101(D3): 7431--7448. Perica S and Foufoula-Georgiou E (1996b) Model for multiscale disaggregation of spatial rainfall based on coupling meteorological and scaling descriptions. Journal of Geophysical Research 101(D21): 26347--26361.

Petty GW (1994a) Physical retrievals of over-ocean rain rate from multichannel microwave imaging. Part I: Theoretical characteristics of normalized polarization and scattering indices. Meteorology and Atmospheric Physics 54: 79--99. Petty GW (1994b) Physical retrievals of over-ocean rain rate from multichannel microwave imaging. Part II: Algorithm implementation. Meteorology and Atmospheric Physics 54: 101--121. Petty GW (1995) The status of satellite-based rainfall estimation over land. Remote Sensing of Environment 51: 125--137. Petty GW (2001a) Physical and microwave radiative properties of precipitating clouds. Part I: Principal component analysis of observed multichannel microwave radiances in tropical stratiform rainfall. Journal of Applied Meteorology 40: 2105--2114. Petty GW (2001b) Physical and microwave radiative properties of precipitating clouds. Part II: A parametric 1D rain-cloud model for use in microwave radiative transfer simulations. Journal of Applied Meteorology 40: 2115--2129. Pinsky M and Khain A (2004) Collisions of small drops in a turbulent flow. Part II: Effects of flow accelerations. Journal of the Atmospheric Sciences 61: 1926--1939. Pinsky M, Khain A, and Shapiro M (1999) Collisions of small drops in a turbulent flow. Part I: Collision efficiency. Problem formulation and preliminary results. American Meteorological Society 56: 2585--2600. Pinsky M, Khain A, and Shapiro M (2000) Stochastic effects of cloud droplet hydrodynamic interaction in a turbulent flow. Atmospheric Research 53: 131--169. Poincare´ H (1908) Science et Me´thode. Reproduced in Poincare´ H (1956) Chance. In: Newman JR (ed.) The World of Mathematics. New York, NY: Simon and Schuster. Popper K (1982) Quantum Physics and the Schism in Physics. London: Unwin Hyman. Powell MD (1990) Boundary layer structure and dynamics in outer hurricane rainbands. Part I: Mesoscale rainfall and kinematic structure. Monthly Weather Review 118: 891--917. Probert-Jones JR (1962) The radar equation in meteorology. Quarterly Journal of the Royal Meteorological Society 88: 485--495. Ramage C (1971) Monsoon Meteorology, International Geophysics Series, vol. 15. San Diego, CA: Academic Press. Randall DA, Wood RA, Bony S, et al. (2007) Climate models and their evaluation. In: Solomon S, Qin D, Manning M, et al. (eds.) Climate Change 2007: The Physical Science Basis, Contribution of Working Group I to the Fourth Assessment Report of the IPCC, pp. 589–662. Cambridge, UK and New York, NY: Cambridge University Press. Reisner J, Rasmussen RJ, and Bruintjes RT (1998) Explicit forecasting of supercooled liquid water in winter storms using the MM5 mesoscale model. Quarterly Journal of the Royal Meteorological Society 124B: 1071--1107. Renno N and Ingersoll A (1996) Natural convection as a heat engine: A theory for CAPE. Journal of the Atmospheric Sciences 53(4): 572--585. Richards F and Arkin P (1981) On the relationship between satellite-observed cloud cover and precipitation. Monthly Weather Review 109: 1081--1093. Robert CP (2007) The Bayesian Choice, From Decision-Theoretic Foundations to Computational Implementation. New York, NY: Springer. Robertson D (1974) Monte Carlo simulations of drop growth by accretion. Journal of the Atmospheric Sciences 31: 1344--1350. Rodriguez-Iturbe I, Cox DR, and Isham V (1987) Some models for rainfall based on stochastic point processes. Proceedings of the Royal Society A 410: 269--298. Rodriguez-Iturbe I, Cox DR, and Isham V (1988) A point process model for rainfall: Further developments. Proceedings of the Royal Society A 417: 283--298. Rodriguez-Iturbe I, Gupta VK, and Waymire E (1984) Scale considerations in the modeling of temporal rainfall. Water Resources Research 20(11): 1611--1619. Rogers RR and Yau MK (1996) A Short Course in Cloud Physics, 3rd edn. Woburn, MA: Butterworth-Heinemann. Roldan J and Woolhiser DA (1982) Stochastic daily precipitation models: 1. A comparison of occurrence processes. Water Resources Research 18(5): 1451--1459. Rossi F, Fiorentino M, and Versace P (1984) Two-component extreme value distribution for flood frequency analysis. Water Resources Research 20(7): 847--856. Rotunno R (1986) Tornadoes and tornadogenesis. In: Ray PS (ed.) Mesoscale Meteorology and Forecasting, ch. 18, pp. 414–436. Boston, MA: The American Meteorological Society. Ryan BF (1996) On the global variation of precipitating layer clouds. Bulletin of the American Meteorological Society 77: 53--70. Ryan BT, Blau HH, Thuna PC, Cohen ML, and Roberts GD (1972) Cloud microstructure as determined by optical cloud particle spectrometer. Journal of Applied Meteorology 11: 149--156. Schaefer MG, Barker BL, Taylor GH, and Wallis JR (2006) Regional precipitationfrequency analysis and spatial mapping for 24-hour and 2-hour durations for Wahington State. Geophysical Research Abstracts 8: 10899.

Precipitation

Schertzer D and Lovejoy S (1987) Physical modeling and analysis of rain and clouds by anisotropic scaling of multiplicative processes. Journal of Geophysical Research 92: 9693--9714. Schertzer D and Lovejoy S (1989) Generalized scale invariance and multiplicative processes in the atmosphere. Pure and Applied Geophysics 130: 57--81. Schmitt F, Vannitsem S, and Barbosa A (1998) Modeling of rainfall time series using two-state renewal processes and multifractals. Journal of Geophysical Research 103(D18): 23181--23193. Scorer RS and Wexler H (1963) A Colour Guide to Clouds. Oxford: Pergamon. Scott WT (1968) On the connection between the Telford and kinetic equation approaches to droplet coalescence theory. Journal of the Atmospheric Sciences 25: 871--873. Scott WT (1972) Comments on ‘‘Validity of the finite-difference droplet collection equation’’. Journal of the Atmospheric Sciences 29: 593--594. Seltzer MA, Passarelli RE, and Emanuel KA (1985) The possible role of symmetric instability in the formation of precipitation bands. Journal of the Atmospheric Sciences 42: 2207--2219. Sempere-Torres D, Porra` JM, and Creutin JD (1994) A general formulation for raindrop size distribution. Journal of Applied Meteorology 33: 1494--1502. Sempere-Torres D, Porra` JM, and Creutin JD (1998) Experimental evidence of a general description for raindrop size distribution properties. Journal of Geophysical Research 103(D): 1785--1797. Shapiro LJ (1983) The asymmetric boundary layer flow under a translating hurricane. Journal of the Atmospheric Sciences 40: 1984--1998. Sharma A and O’Neill R (2002) A nonparametric approach for representing interannual dependence in monthly streamflow sequences. Water Resources Research 38(7): 1100 (doi: 10.1029/2001WR000953). Simpson J, Adler RF, and North GR (1988) Proposed tropical rainfall measuring mission (TRMM) satellite. Bulletin of the American Meteorological Society 69: 278--295. Sisson SA, Pericchi LR, and Coles SG (2006) A case for a reassessment of the risks of extreme hydrological hazards in the Caribbean. Stochastic Environmental Research and Risk Assessment 20: 296--306. Skamarock W, Klemp J, Dudhia J, et al. (2005) A Description of the Advanced Research WRF Version 2. Mesoscale and Microscale Meteorology Division, National Center for Atmospheric Research, Boulder, CO, USA. Slingo J (2003) Monsoon. In: Holton JR, Curry JA, and Pyle JA (eds.) Encyclopedia of Atmospheric Sciences, pp. 1365--1370. London: Academic Press. Slinn WGN (1975) Atmospheric aerosol particles in surface-level air. Atmospheric Environment 9: 763--764. Small MJ and Morgan DJ (1986) The relationship between a continuous-time renewal model and a discrete Markov chain model of precipitation occurrence. Water Resources Research 22(10): 1422--1430. Smith JA (1993) Precipitation In: Maidment DA (ed.) Handbook of Applied Hydrology, ch. 3, pp. 3.1–3.47. New York, NY: McGraw-Hill. Smith RK (1968) The surface boundary layer of a hurricane. Tellus 20: 473--484. Smith RK (2000) The role of cumulus convection in hurricanes and its representation in hurricane models. Reviews of Geophysics 38: 465--489. Smith RK (2003) A simple model of hurricane boundary layer. Quarterly Journal of the Royal Meteorological Society 129: 1007--1027. Spencer RW, Goodman HM, and Hood RE (1989) Precipitation retrieval over land and ocean with the SSM/I: Identification and characteristics of the scattering signal. Journal of Atmospheric and Oceanic Technology 6: 254--273. Srikanthan R and McMahon TA (2001) Stochastic generation of annual, monthly and daily climate data: A review. Hydrology and Earth System Sciences 5(4): 653--670. Stephens GL (1994) Remote Sensing of the Lower Atmosphere: An Introduction. New York, NY: Oxford University Press. Stewart SR (2002) Tropical cyclone report. Tropical Storm Allison, 5–17 June 2001, US National Oceanic and Atmospheric Administration. http://www.nhc.noaa.gov/ 2001allison.html (accessed March 2010). Stoelinga MT, Woods CP, Locatelli JD, and Hobbs PV (2005) On the representation of snow in bulk microphysical parameterization schemes. WRF/MM5 Users’ Workshop, June 2005. Stout JS, Martin DW, and Sikdar DN (1979) Estimating GATE rainfall with geosynchronous satellite images. Monthly Weather Review 107: 585--598. Sutcliffe JV (1978) Methods of Flood Estimation, A Guide to Flood Studies Report, Report 49. Wallingford: Institute of Hydrology. Telford JW (1955) A new aspect of coalescence theory. Journal of Meteorology 12: 436--444. Telford JW and Chai SK (1980) A new aspect of condensation theory. Pure and Applied Geophysics 118: 720--742. Telford JW and Wagner PB (1981) Observations of condensation growth determined by entity type mixing. Pure and Applied Geophysics 119: 934--965.

77

Tessier Y, Lovejoy S, Hubert P, Schertzer D, and Pecknold S (1996) Multifractal analysis and modeling of rainfall and river flows and scaling, causal transfer functions. Journal of Geophysical Research 101(D21): 26427--26440. Tessier Y, Lovejoy S, and Schertzer D (1993) Universal multifractals in rain and clouds: Theory and observations. Journal of Applied Meteorology 32: 223--250. Testik FY and Barros AP (2007) Toward elucidating the microstructure of warm rainfall: A survey. Reviews of Geophysics 45: RG2003 (doi:10.1029/2005RG000182). Thompson G, Rasmussen RM, and Manning K (2004) Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part I: Description and sensitivity analysis. Monthly Weather Review 132: 519--542. Todorovic P and Woolhiser DA (1975) A stochastic model of n-day precipitation. Journal of Applied Meteorology 14: 17--24. Todorovic P and Zelenhasic E (1970) A stochastic model for flood analysis. Water Resources Research 6(6): 1641--1648. Tsonis A (1987) Determining rainfall intensity and type from GOES imagery in the midlatitudes. Pure and Applied Geophysics 21: 29--36. Tsonis A and Isaac G (1985) On a new approach for instantaneous rain area delineation in the midlatitudes using GOES data. Journal of Applied Meteorology 24: 1208--1218. Turpeinen OM, Abidi A, and Belhouane W (1987) Determination of rainfall with the ESOC precipitation index. Monthly Weather Review 115: 2699--2706. Twomey S (1964) Statistical effects in the evolution of a distribution of cloud droplets by coalescence. Journal of the Atmospheric Sciences 21: 553--557. Twomey S (1966) Computation of rain formation by coalescence. Journal of the Atmospheric Sciences 23: 405--411. Uijlenhoet R (1999) Parameterization of Rainfall Microstructure for Radar Meteorology and Hydrology. Doctoral Dissertation, Wageningen University, The Netherlands. Uijlenhoet R (2001) Raindrop size distributions and radar reflectivity –rain rate relationships for radar hydrology. Hydrology and Earth System Sciences 5: 615--627. Uijlenhoet R (2008) Precipitation physics and rainfall observation. In: Bierkens MFP, Dolman AJ, and Troch PA (eds.) Climate and the Hydrological Cycle, IAHS Special Publication 8, pp. 59–97. Wallingford: IAHS. Uijlenhoet R and Berne A (2008) Stochastic simulation experiment to assess radar rainfall retrieval uncertainties associated with attenuation and its correction. Hydrology and Earth System Sciences 12: 587--601. Uijlenhoet R and Sempere-Torres D (2006) Measurement and parameterization of rainfall microstructure. Journal of Hydrology 328(1–2): 1--7 (ISSN 0022-1694; doi: 10.1016/j.jhydrol.2005.11.038). Ulbrich C (1983) Natural variations in the analytical form of the raindrop size distribution. Journal of Climate and Applied Meteorology 22(10): 1764--1775. Ulbrich CW and Atlas D (1998) Rainfall microphysics and radar properties: Analysis methods for drop size spectra. Journal of Applied Meteorology 37: 912--923. US Department of the Interior, Bureau of Reclamation (1977) Design of Arch Dams. Denver, CO: US Government Printing Office. US Department of the Interior, Bureau of Reclamation (1987) Design of Small Dams, 3rd edn.. Denver, CO: US Government Printing Office. Velghe T, Troch PA, De Troch FP, and Van de Velde J (1994) Evaluation of clusterbased rectangular pulses point process models for rainfall. Water Resources Research 30(10): 2847--2857. Veneziano D (1999) Basic properties and characterization of stochastically self-similar processes in RD. Fractals 7(1): 59--78. Veneziano D and Furcolo P (2002) Multifractality of rainfall and intensity–duration– frequency curves. Water Resources Research 38(12): 1306--1317. Veneziano D, Furcolo P, and Iacobellis V (2006a) Imperfect scaling of time and space– time rainfall. Journal of Hydrology 322(1–4): 105--119. Veneziano D and Langousis A (2005) The areal reduction factor a multifractal analysis. Water Resources Research 41: W07008 (doi:10.1029/2004WR003765). Veneziano D and Langousis A (2010) Scaling and fractals in hydrology. In: Sivakumar B (ed.) Advances in Data-Based Approaches for Hydrologic Modeling and Forecasting, ch. 4, p. 145. Singapore: World Scientific. Veneziano D, Langousis A, and Furcolo P (2006b) Multifractality and rainfall extremes: A review. Water Resources Research 42: W06D15 (doi:10.1029/ 2005WR004716). Veneziano D, Langousis A, and Lepore C (2009) New asymptotic and pre-asymptotic results on rainfall maxima from multifractal theory. Water Resources Research 45: W11421 (doi:10.1029/2009WR008257). Veneziano D, Lepore C, Langousis A, and Furcolo P (2007) Marginal methods of intensity–duration–frequency estimation in scaling and nonscaling rainfall. Water Resources Research 43: W10418 (doi:10.1029/2007WR006040). Venugopal V, Roux SG, Foufoula-Georgiou E, and Arneodo A (2006) Revisiting multifractality of high resolution temporal rainfall using a wavelet-based formalism. Water Resources Research 42: W06D14 (doi:10.1029/2005WR004489).

2.03 Evaporation in the Global Hydrological Cycle AJ Dolman and JH Gash, VU University Amsterdam, Amsterdam, The Netherlands & 2011 Elsevier B.V. All rights reserved.

2.03.1 2.03.2 2.03.2.1 2.03.2.2 2.03.2.3 2.03.3 2.03.4 2.03.5 References

Introduction General Theory of Evaporation Vegetated Surfaces Bare Soil Open Water and Lakes Regional and Equilibrium Evaporation Trends and Variability in Global Evaporation Summary and Conclusions

2.03.1 Introduction Evaporation is the transfer of moisture from a particular surface to the overlying atmosphere. The physical process of evaporation consists of the exchange of water molecules between a free water surface and the air. The surface can be any among the following nonexhaustive list: a lake, the inside of plant leaves, the water surface adhering to a soil conglomerate, or the surface of a soil or canopy during or just after rain. The evaporation rate is expressed as the quantity of water evaporated per unit area per unit time from a (water) surface under existing atmospheric conditions. This chapter describes the progress in understanding evaporation at the local scale, both from an observational and a conceptual perspective. It then moves on to the global scale, through a discussion of regional scale feedbacks. The main conclusion is that while we have gained considerable understanding in local scale evaporation, the data sets required to study the impact of evaporation on the global water cycle are still lacking. Evaporation at the Earth’s surface is constrained both by the energy available to convert liquid water into vapor and by the capability of the surrounding air to transfer moisture away from the saturated surface. At the surface, these constraints are best expressed through the energy balance equation and the transfer equations of latent and sensible heat:

Rg ð1  zÞ þ Lk  Lm ¼ lE þ H þ G þ dS

ð1Þ

lE ¼ rlKv

qq qz

ð2Þ

H ¼ rcp Kh

qT qz

ð3Þ

b¼

H lE

ð4Þ

with Rg the incoming short-wave radiation (W m2), z the short-wave albedo, Lk and Lm the incoming and outgoing long-wave radiation (W m2), lE the evaporation (or latent heat flux, W m2) (with l the latent heat of vaporization and E the mass flux), H the sensible heat flux (W m2), G the soil heat flux (W m2), dS the change in heat storage in the

79 80 81 82 82 83 83 85 85

biomass and atmosphere below a reference height above the surface (W m2), Kv and Kh the transfer coefficients for water vapor and heat, respectively, q q/q z the vertical gradient in specific humidity (g kg1), and similarly q T/q z is the vertical gradient in temperature (K), r is the density of air (kg m3), and cp the specific heat of air (J kg1 K1). b is the Bowen ratio and a useful indicator of the dryness of the surface because it shows how the available energy is partitioned into sensible and latent heat. Factors influencing the rate of evaporation can easily be determined from the above equations: radiation as the limit of available energy, the gradient in moisture, or the specific humidity deficit and factors that influence the transfer coefficients such as wind speed and roughness of the surface, and, crucially, when dealing with evaporation from leaves (transpiration), water availability (soil moisture). Thus, despite the complexity of the interaction between a partially wet surface and the atmosphere, it may be possible to simplify this interaction and to approximate evaporation by considering only two factors, available energy (radiation) and water availability. Available energy determines the maximum evaporation possible for the given climatic conditions and unlimited water availability, that is, potential evaporation (see also Allen et al., 1998). Water availability can be characterized by the amount of precipitation. Under dry conditions, potential evaporation exceeds precipitation, and the actual evaporation from an area will approach the amount of precipitation received. Conversely, under wet conditions, water availability exceeds potential evaporation and actual evaporation will approach asymptotically potential evaporation. This relationship was first used by Budyko (1974) to set a constraint on global evaporation. Later, Milly and Dunne (2002) suggested that the long-term water balance of a catchment is determined by the interaction of supply (precipitation) and demand (potential evaporation), mediated by soil moisture storage. At longer timescales, the change in soil moisture can be neglected. Figure 1 shows results of Zhang et al. (2001), from a simple model based on the above considerations for grass and forested catchments, using catchment discharge and precipitation data, that allow evaporation to be calculated as the residual of these two. These results suggest that at the scale of catchments such a simple approach works remarkably well, with a linear (1:1) relation between rainfall and evaporation for catchments that receive up to 500 mm yr1, and an

79

80

Evaporation in the Global Hydrological Cycle 1600

Annual evapotranspiration (mm)

Forest Mixed veg. Pasture 1200

800

400

0

0

500

1000

1500

2000

2500

3000

3500

Annual rainf all (mm) Figure 1 Relationship between annual precipitation and evaporation. The dashed line represents the best fit of a theoretical model (Zhang et al., 2001) for grass-only catchments, the solid line the best fit for forests. From Zhang et al. (2001). (Note that we use the word evapotranspiration here in this graph consistent with its origin. This rather loosely defined term refers to the total of (wet canopy and bare soil) evaporation and transpiration. Throughout the text, we prefer to use the word evaporation as the physical term denoting the process of transforming liquid water into its vapor form.)

asymptotic relation (driven by available energy and not by water availability) for catchments receiving more than 500 mm yr1. The difference between forest and grassland is a result of the high rate of evaporation of intercepted rainfall from forest; interception loss creates a considerable additional evaporative loss for forest (see also Chapter 2.04 Interception). Total evaporation from forest saturates at about 1400 mm yr1, while for grasslands this value is around 900 mm yr1. Direct observations of evaporation have been made since the mid-1990s using the eddy-covariance technique in a global network, Fluxnet (see Aubinet et al., 2001). While there are other techniques based on scintillometry available, these have not been used extensively in networks such as Fluxnet. Evaporation measured by micrometeorological techniques usually refers to dry-canopy evaporation only (transpiration) and contains little information on evaporation during wet-canopy conditions (interception). Consequently, total evaporation values such as those given by Law et al. (2002) should be treated with caution, when interception losses are not explicitly treated. With this caveat in mind, the annual evaporation from Fluxnet data for coniferous forests is 397 (731) mm yr1; for mixed evergreen and deciduous forest, 386 (718) mm yr1; for deciduous broadleaf, 512 (769) mm yr1; for grassland, 494 (7104) mm yr1; and for crops, 666 (767) mm yr1. Grassland and crops have higher dry-canopy evaporation than forest, because they are less strongly coupled to the atmosphere and thus show less stomatal control (e.g., Shuttleworth and Calder, 1979). This also gives forest the possibility to survive occasional drought that would kill off annual grassland species. The Fluxnet data can be used to identify the main controls on evaporation. Wilson et al. (2002) showed how the

partitioning of energy, as expressed by the ratio of sensible heat to latent heat, the Bowen ratio (b), can be used to classify different vegetation types in climate space. Figure 2 shows the position of the individual sites with respect to the magnitude of the latent heat (evaporation) and sensible heat fluxes. In contrast to the annual rates discussed earlier, these data refer to the growing season only. Also shown are lines of constant available energy (the sum of latent and sensible heat) and lines of constant b. Moving from the lower left corner of the diagram to the upper right, the available energy increases, whereas moving from the lower right to the upper left, the value of b increases. Low evaporation rates are found in Sitka spruce, tundra, and boreal forest, and high evaporation rates in deciduous forests and agriculture. Most of the forests have high b’s, implying that much of the energy received is transmitted back to the atmosphere as sensible heat. Deciduous forests tend to have lower b’s with higher evaporation rates than coniferous forests. The average b at tundra sites appears to be close to 1.

2.03.2 General Theory of Evaporation Penman (1948) was among the first to achieve the crucial combination of the energy balance equation (1) with the transfer Equations (2) and (3) to derive an expression for actual evaporation from vegetation well supplied with water. Although the vertical gradients in Equations (2) and (3) could be derived from the differences between air and surface values, measurements of surface temperature are difficult and not made routinely. Penman overcame this problem by introducing the slope of the saturated vapor pressure versus temperature curve, D, approximated as a linear function and

Evaporation in the Global Hydrological Cycle 15

81

=3 T1

=2

T2

12

M

Daily sensible heat flux (MJ m−2 d−1)

ed

ite

rra

ne

an

m

S1

9

=1

cli

at

es

Boreal Canadian

S2 N4 R 2 R1 31 Y2 NN

I1

6 Sitka spruce

Q3

 = 0.5

M1 P1 T3

O1

N2

K1 G 2

K 2 F3 G1Y1F1 Q1 H1 A8 L1 F2 Tundra A7A1 M2 I2 D1 L2 A1 A4 X1 A5 B B1 A2 H3 V1V 2 B4 G3 A6 B5 B3 U 3 G2 A3 J1 Z1 B2 Conifers U2 B6 C1 O2

Q2

HW1 2

X2

3

E2

 = 0.25

Agriculture E1

Deciduous forests

0 0

3

6

9

Daily latent heat flux (MJ

m−2

12

15

d−1)

Figure 2 The daily cumulative sensible heat flux vs. the daily cumulative latent heat flux between days 165 and 235 for the Fluxnet sites analyzed by Wilson et al. (2002). The letter and number codes refer to the sites as given by Wilson et al. (2002). Also shown are lines of constant Bowen ratio (dashed lines) and lines of constant total turbulent energy fluxes (solid diagonal lines). Enclosed areas denote subjective delineations between different vegetation types and climates. From Wilson et al. (2002).

evaluated at air temperature:

D¼

e  ðTs Þ  e ðTa Þ Ts  Ta

ð5Þ

where e* (Ts) is the saturated vapor pressure at surface temperature Ts (K), and e* (Ta) is the saturated vapor pressure at air temperature Ta. The evaporation, E, in mm d1 is given by

E¼

DQ þ gEa Dþg

ð6aÞ

where Q is water equivalent of the net radiation (from the lefthand side of Equation (1)), g the psychrometric constant and

Ea ¼ f ðUÞðe ðTa Þ  ea Þ

ð6bÞ

Ea is the aerodynamic or demand term for evaporation, with f(U) a wind function, and (e* (Ta)  ea) the vapor pressure deficit. The original Penman equation contains an empirical wind function that replaces the transfer coefficients in Equations (2) and (3). This wind function was difficult to generalize, and subsequently Thom and Oliver (1977) provided a more

physical basis including considerations of aerodynamic transfer over rough surfaces by introducing an explicit aerodynamic resistance. Penman applied his equation to bare soil evaporation, vegetated or cropped surfaces, and open water bodies. The equation is still widely used to calculate evaporation from well-watered, short vegetation.

2.03.2.1 Vegetated Surfaces Monteith (1965) (see also Gash and Shuttleworth, 2007) introduced the control of vegetation on evaporation by including a canopy scale resistance rs. This resistance represents the restriction on the transfer of water from the collective saturated surfaces inside the plant stomatal cavities to the air outside the leaves. The resulting equation, now carrying both an aerodynamic (to replace the Penman wind function) and a canopy resistance, is known as the Penman–Monteith (PM) equation and is arguably still the most elegant yet advanced resistance model of evaporation used in hydrological practice today (Shuttleworth, 1993). For a mathematically precise and exact definition and brief historical overview of its development, the reader is referred to Raupach (2001). Most commonly, the PM equation is expressed in specific humidity units (rather than vapor pressure deficit as in Equation (6)) with

82

Evaporation in the Global Hydrological Cycle

evaporation expressed in energy units (W m2) and reads as

lE ¼

rcp ðq * ðTa Þ  qa Þ ra   cp rs 1þ Dþ l ra

DðQ *  GÞ þ

ð7Þ

with D the slope of the saturated specific humidity versus temperature curve and Q* the net radiation (W m2). q* (Ta) is the saturated specific humidity of the air (g kg1) at the reference level and qa the specific humidity at the same level. Two important variables appear in this equation that replace the transfer coefficients of the transfer equations of heat and moisture: the aerodynamic and surface resistance, ra and rs (s m1). (Only in the case of a full canopy cover, can the surface resistance be equal to the canopy resistance. For the derivation of the big-leaf version of the PM model as presented here, this is not important. When the canopy cover is not full, and there is bare ground directly in contact with the overlying atmosphere, the surface resistance is not equal to the canopy resistance and approaches the reciprocal sum of the canopy and an assumed soil resistance.) The PM equation assumes that evaporation and sensible heat originate from the same source in the canopy. The main advantage of the PM equation is that the meteorological driving variables, wind speed, specific humidity deficit, and temperature are required only at a single level above the surface, removing the need for the notoriously difficult observation of surface values. The main obstacle for the practical application of this equation is the estimation of the values for aerodynamic and surface resistance. When there is unlimited supply of water, the PM equation can be used to calculate potential evaporation. It can be shown then to collapse to the Penman equation with an aerodynamic resistance rather than a wind function. The PM equation is now the preferred method of estimating crop water requirements as reference crop evaporation (Allen et al., 1998). When the canopy is wet, the surface resistance equals zero and the PM equation can be used to calculate evaporation of intercepted rainfall (see also Chapter 2.04 Interception). Although there is a wide range of empirical evaporation equations of which some are particular cases of the PM equation (see Brutsaert, 1982; Shuttleworth, 1993), the use of the PM equation is widespread because of its clear physical interpretation.

2.03.2.2 Bare Soil Evaporation from bare soil can be a significant component of the water balance, particularly in semi-arid environments (Wallace and Holwill, 1997). Soil evaporation can be described as a two-stage process. The first stage occurs when the available soil moisture is sufficient to meet the atmospheric demand. This occurs immediately after rainfall or irrigation events. Soil evaporation under these conditions equals potential evaporation. Typically, this stage lasts 1–2 days, although in some cases when evaporative demand is low and the soil contains a high amount of clay, this stage may last for up to 5 days. In the second stage, the amount of soil moisture has dropped and soil evaporation is no longer only restricted by evaporative demand but also by availability of moisture. In

these conditions, the change of soil moisture with time can be described as a desorption process with evaporation proportional to the square root of the time since the start of the process:

lEs ¼

1 aðt  t0 Þ1=2 2

ð8Þ

with lEs the soil evaporation, t the time and t0 the time since the start of second stage drying, Ds is a desorptivity (in units of W m2 d1/2 when evaporation is expressed as a heat flux, or in mm d1/2 when evaporation is expressed as water flux). The desorptivity is assumed constant for a particular soil type. It varies from a value of 2.1 for sandy loam with gravel, to a value of 5 mm d1 for a clay loam soil (Kustas et al., 2002). Although the two-stage process describes soil evaporation at diurnal timescales, extension of the theory to (sub) hourly timescales is straightforward (Brutsaert and Chen, 1996; Porte´Agel et al., 2000). The determination of the desorptivity coefficient can be problematic, as can the identification of the switch from stage 1 to 2 (Kustas et al., 2002). The observed dependence of soil evaporation on available soil moisture suggests the feasibility of a resistance approach that incorporates a dependence of soil surface resistance on soil moisture. Mahfouf and Noilhan (1991) review several such formulations. These approaches can be divided into socalled a- and bs-approaches. In the a-approach, the saturated humidity in the soil pore space is adjusted by a factor a that may be related to soil matrix potential and takes into account that, averaged over a certain depth, the evaporation takes place from a nonsaturated surface. In the bs-approach, the humidity in the pore space at the evaporation front is assumed to be saturated, and bs is the ratio of an aerodynamic resistance to the sum of the aerodynamic and soil surface resistance.

2.03.2.3 Open Water and Lakes The key process controlling evaporation from large lakes (or reservoirs) is the absorption of solar energy. This energy is not absorbed at the surface; because water is a semitransparent medium, the solar radiation is absorbed over depth. The rate of change of absorption with depth depends on the turbidity, but may be significant down to several meters below the surface. Solar energy heats the water up during the spring and summer, and this energy is not available for evaporation; but energy released as the water cools in autumn and winter is available and enhances the evaporation. This creates a phase lag between lake evaporation and the annual radiation cycle. While in the tropics this lag will be small, at high latitudes the phase lag may be as much as 5 or 6 months (Blanken et al., 2000) with the rate of change in storage, dS/dt, being the dominant source of energy for evaporation. The energy available for evaporation is given by

A ¼ Q þ

dS þ Aq dt

ð9Þ

where Aq is the rate of net energy advection due to inflow and outflow of water. Net radiation, Q* , is given by Equation (1)

Evaporation in the Global Hydrological Cycle

with long-wave radiation emitted by the lake as

Lm ¼ esT4w

ð10Þ

e is emissivity, s the Stefan–Boltzmann constant, and Tw the surface temperature of the water (K). Q* over the lake can be estimated from measurements over land, but must take account of the different albedo and different surface temperature of the lake. Like the ocean, lake albedo varies strongly with solar elevation (see Finch and Gash, 2002; Finch and Hall, 2005; Payne, 1972). The Penman (1948) equation (Equation (6)) is often used to estimate lake evaporation, as it appears to remove the need for surface temperature measurement; however, this is not the case as water temperature is still needed to calculate the emitted long-wave radiation. Nevertheless, it should give good results if used with the available energy calculated from Equation (9) and measurements made over the lake (see Linacre, 1993). Working in the tropics where the annual cycle in water temperature is small and energy storage can be neglected, Sene et al. (1991) found good agreement between daily estimates made with the Penman equation and eddy-covariance measurements of evaporation. To overcome the lack of water temperature measurements, Finch and Gash (2002) applied a simple numerical, finite difference scheme to calculate a running balance of lake energy storage. A new value of the water temperature required to force energy closure was calculated at each time step. For a well-mixed lake of known depth, the evaporation could then be calculated from land-based, daily meteorological observations of sunshine hours, relative humidity, wind run, and average air temperature. The model gave good agreement with mass-balance measurements of the water loss from a reservoir with no inflow or outflow.

2.03.3 Regional and Equilibrium Evaporation

D Dþ

cp A l

the parameter a. a can be shown to be unity only when the specific humidity deficit in the PM equation is zero, in other words, when advection is negligible. That this is hardly ever the case proves the fact that most empirical values of a for short crops are of the order 1.2–1.3 (e.g., Brutsaert, 1982). For tall crops, Shuttleworth and Calder (1979) showed convincingly that the equilibrium approach is not appropriate because the physiological control of the forest transpiration reduces a below a value of 1 in dry canopy conditions, while in wet canopy conditions large-scale advection and negative sensible heat fluxes (Stewart, 1977) may form an additional supply of energy and force a to be well above the value for short crops. This emphasizes the important point that for tall crops in particular, it is important to estimate dry and wet canopy evaporation separately (see also Figure 1). Thus, at larger scale, atmospheric conditions can override the surface control by exerting a strong feedback on evaporation through the humidity and temperature of the atmospheric boundary layer. A number of concepts have been derived that use the feedback power of the atmosphere to estimate regional-scale evaporation (e.g., Bouchet, 1963; Morton, 1983). Although McNaughton and Jarvis (1991) show that at larger scale the feedback of the atmospheric boundary layer dampens the effects of surface controls – and thus makes precise estimation of the surface resistance less important – the physical basis of the Bouchet and Morton schemes remains doubtful (see de Bruin, 1983; McNaughton and Spriggs, 1989). Nevertheless, de Bruin showed that the feedback of the increasing atmospheric boundary-layer humidity during the day causes the regional surface conductance to vary less than if the feedback were neglected. The relatively large confidence of hydrologists in using potential evaporation formulas probably finds its physical explanation in this feedback.

2.03.4 Trends and Variability in Global Evaporation

Priestley and Taylor (1972) showed that the Bowen ratio would approach a constant value defined by b ¼ s/l when air moves over a moist surface, and gradients of temperature and specific humidity with height are small or become saturated with respect to moisture. Combining this insight with the PM equation, and setting the second term above the nominator to zero as well as the surface or canopy resistance (rs ¼ 0) as would be appropriate for a moist surface, yields the equilibrium evaporation (see also Brutsaert, 1982; Raupach, 2001):

lE ¼

83

ð11Þ

Equilibrium evaporation as defined by these authors refers to the lower limit of evaporation from a moist surface where the specific humidity deficit of the second term in the nominator of the PM equation has become zero as a result of contact of air with a moist surface over a very long fetch. It can easily be shown that the second term in the nominator of the PM equation (Equation (7)) now represents the departure from this equilibrium. Priestley and Taylor (1972) represented this departure from equilibrium evaporation by

Globally, evaporation from the land surface to the atmosphere amounts to 71 103 km3 yr1 (Baumgartner and Reichel, 1975). It is the key return flow in the hydrological cycle from the surface on which the precipitation falls, back to the atmosphere. Evaporation from the oceans is a far larger component of the hydrological cycle at an estimated 428 103 km3 yr1 (Baumgartner and Reichel, 1975). A recent multi-model ensemble of 11 state-of-the-art land surface models (Dirmeyer et al., 2006) estimated annual evaporation over a range of 58 103 to 85 103 km3, indicating the uncertainty in our ability to model evaporation from land. Oki and Kanae (2006) estimated total terrestrial evaporation at 66 103 km3 yr1. Thus, although it is an important component of the hydrological cycle, the exact magnitude and variability, both spatially and temporally, of evaporation from land remains highly uncertain. Figure 3 shows one of the few available estimates of the latitudinal distribution of evaporation in mm yr1 for both ocean and land. This estimate (Baumgartner and Reichel, 1975) is based on the balance between precipitation and runoff on land and a variety of other methods (Peixoto and Oort, 1996). Note the relatively large contribution of Southern

84

Evaporation in the Global Hydrological Cycle

Latittude

0 85 75 65 55 45 35 25 15 5 −5 −15 −25 −35 −45 −55 −65 −75 −85

375

750

1125

1500

Evaporation (mm yr−1)

Figure 3 Global evaporation according to Baumgartner and Reichel (1975) for different 101 latitude bands.

ρRg,E

−1 +1

ρP,E 0

+1

0

−1 Figure 4 Multi-model analysis of controls on yearly evaporation. Correlation between yearly evaporation and global radiation (rRg,E), and precipitation (rP,E), for the period 1986–95. Each color corresponds to a unique combination of rRg,E and rP,E. The gray lines (legend) show the global frequency distribution (Teuling et al., 2009).

Hemisphere latitude bands 20–401 S compared to their northern equivalents. This difference is largely due to the greater area of land in the Northern Hemisphere, which evaporates at a significantly lower rate than the ocean. The largest evaporative flux is found in the humid tropics, mainly as a result of large amounts of precipitation and high solar radiation. Compared to the latitudinal distribution of precipitation, the evaporation is more smoothly distributed with a general tendency of decreasing evaporation when moving poleward. Decreases in radiation, global dimming, have caused a debate about an observed decline in pan measured evaporation (e.g., Roderick and Farquhar, 2002) that would be contrary to expectations for a warming climate. Peterson et al. (1995), using data from a network of pan evaporimeters in the US and the former Soviet Union, found a decrease in pan evaporation between 1950 and 1990. However, Fu et al. (2009) in a more comprehensive analysis suggested that, although many observations across the world indicate a general trend of pan evaporation decreasing over the last 50 years, this trend is not universal. A decrease in evaporation presents a

paradox, as with global warming one would expect an increase due to the larger water holding capacity of the atmosphere through the Clausius–Clapeyron equation. An increase in evaporation would also match an increase in precipitation, although this is regionally very variable. Roderick and Farquhar (2002) explained the paradox by relating the decrease in evaporation to a decrease in solar radiation (global dimming). However, the dimming trend has recently reversed into a brightening trend and thus cannot singularly be held responsible for the decrease in pan evaporation. Teuling et al. (2009) presented an analysis of the major controls on evaporation using an ensemble of land surface models forced off line with meteorological data. They investigate the control of two key drivers on evaporation, incoming solar radiation and soil moisture. Figure 4 shows that Europe, North Africa, and North America are characterized by two evaporation regimes: a humid regime with high correlation with radiation expressed through the correlation coefficient rRg,E, but low correlation with precipitation, P(rP,E); and a more arid regime with high rP,E, but low rRg,E. Because radiation and precipitation tend to be negatively correlated,

Evaporation in the Global Hydrological Cycle

Teuling et al. (2009) concluded that yearly variations in evaporation reflect either variations in Rg or P, but not in both. Central Europe is among the regions with the highest rRg,E correlation, while in more arid regions such as the US Midwest and the Sahara, evaporation correlates only with precipitation. Using data from direct observations of evaporation (Fluxnet; see Baldocchi et al., 2001), Teuling et al. were able to reproduce and validate these modeled patterns quite well. Teuling et al. (2009) made a strong argument for a regional approach to explain some of the evaporation trends. Based on the different sensitivities of the various drivers of evaporation (see Figure 4) and the conclusion of other work that a dimming trend has been reversed, they concluded that scenarios of both decreasing actual evaporation with decreasing pan evaporation in regions with ample supply of water (e.g., central Europe), and of increasing evaporation with decreasing pan evaporation (e.g., the US Midwest) are consistent. Using basin-scale discharge data and precipitation they found that evaporation decreased over Europe during the dimming period and increased later, consistent with the high sensitivity of European evaporation to radiation rather than precipitation. During the dimming period, the positive trend in runoff is induced by reduced evaporation, rather than increased precipitation. After 1983, evaporation derived as the residual of precipitation minus runoff, increased in all central European basins during the brightening phase. These results suggest that evaporation trends follow radiation trends in central Europe. In contrast, in the US Midwest the upward trends in evaporation derived as a catchment residual before 1983 are followed by decreasing trends. These may be explained by trends in precipitation combined with high correlation between solar radiation and evaporation as inferred from Figure 2. Next to the analysis of Teuling et al. (2009) that attribute changes in evaporation to either radiation or water availability (precipitation), recent studies of changes in pan evaporation in Australia attribute most of the reduction in pan evaporation to reduced wind speed (Roderick et al., 2007; Rayner, 2007; McVicar et al., 2008). The cause of such a reduction in regional wind speed is not certain, although wind speed reductions have been widely reported in mid-latitudes in both hemispheres (see also Shuttleworth, 2009).

2.03.5 Summary and Conclusions Evaporation is an important, but regionally a still poorly quantified term in the global water balance. At global scale its determination as a residual of the continental scale water balance hinges on adequate estimation of precipitation and river discharge. Estimates obtained by direct bottom-up modeling vary considerably. This is in some contrast to our understanding of the basic physics of evaporation, that is well known, as for instance is shown by the Penman and PM equations. Application of these equations for use at local scales, for instance irrigation practice, is widespread (e.g., Allen et al., 1998), and to a large extent very successful. The importance of evaporation in the global hydrological cycle critically encompasses two related aspects: its direct role as a term in the water budget, and its potential to impact

85

weather and climate processes, by changing aspects of the surface energy balance and boundary layer. Both these roles depend on the balance between the controlling forces of evaporation, surface moisture and available energy. Where only water availability is limiting and radiation plentiful, such as in the (semi)-arid tropics, evaporation may add moisture to the overlying air that can then tip the balance to produce precipitation. These semi-arid areas have been found (e.g., Koster et al., 2002) to be sensitive to land surface precipitation feedbacks. In these feedbacks the role of evaporation is critical. On the other hand, in areas where both precipitation and radiation are not limiting, large-scale evaporation appears constrained by the available energy. Under these conditions, equations such as the Priestly–Taylor equation that do not explicitly take account of water availability on average perform well. Thus, to be able to predict the effect of, for instance, landuse change on evaporation (and resulting catchment discharge), one would need to determine first which process, if any, is limiting. In cases where water availability is limiting, atmospheric feedbacks may also become important and simple bottom-up estimates of the change in evaporation may be wrong. In cases where neither water nor energy is limiting, to first order a bottom-up estimate based on energy constraints would be appropriate. Changes in the pattern of largescale moisture recycling, such as found in the Amazon (e.g., Meesters et al., 2009), may however be important to estimate changes in the resulting precipitation climate. The debate whether and where evaporation is increasing or decreasing is fundamental to our understanding of the role of evaporation in the global water cycle and climate. Improved, high-quality data sets are needed to provide benchmarks for climate models and to increase our process understanding. Currently, such data sets unfortunately do not exist. It is worth noting that the role of evaporation in climate is not related only to its direct effects on the hydrological cycle. Through the influence evaporation exerts on the partitioning of the energy balance, the effects on climate are also seen in surface temperatures. High evaporation keeps surfaces cool; low evaporation makes them hot. Seneviratne et al. (2006) and Fischer et al. (2007) used regional and global climate modeling to investigate the role of land surface atmosphere feedbacks on temperature in a changing climate. They concluded that soil moisture through its effect in reducing evaporation has a major impact on the variability and mean and maxima of surface temperatures in Europe. Fischer et al. (2007) concluded that land–atmosphere interactions over drought regions account for typically 50–80% of the number of hot days in a Northern Hemisphere summer. This is mainly due to local effects through the limitation of evaporation (and increased sensible heat flux) due to drought conditions. Drought conditions may also have remote effects on areas around or outside the actual drought region, through changes in atmospheric circulation and advection of air masses. These mechanisms can enhance an existing anticyclonic circulation over, or slightly downstream of, a drought anomaly. Evaporation plays a key role in the global water cycle and hence in the global climate system. There have, however, been very few attempts to produce robust estimates of evaporation based on a global approach. Yet such data are urgently needed

86

Evaporation in the Global Hydrological Cycle

to validate and constrain current climate models. Thus, despite that at the practical level, considerable advances have been made in our ability to estimate and observe evaporation at local level, our understanding of evaporation in the global climate system (e.g., Kleidon and Schymanski, 2008) still shows significant gaps. The challenge for the next decade in evaporation research is to fill these gaps.

References Allen RG, Pereira LS, Raes D, and Smith M (1998) Crop Evapotranspiration – Guidelines for Computing Crop Water Requirements, p. 300. Rome: Irrigation and Drainage, FAO. Aubinet M, Chermanne B, Vandenhaute M, Longdoz B, Yernaux M, and Laitat E (2001) Long term carbon dioxide exchange above a mixed forest in the Belgian Ardennes. Agricultural and Forest Meteorology 108: 293--315. Baldocchi D, Falge E, Gu L, et al. (2001) FLUXNET: A new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide, water vapor, and energy flux densities. Bulletin of the American Meteorological Society 82: 2415--2434. Baumgartner A and Reichel E (1975) Die Weltwasserbilanz, 179pp. Mu¨nchen: Oldenbourg Verlag. Blanken PD, Rouse WR, Culf AD, et al. (2000) Eddy covariance measurements of evaporation from Great Slave Lake, Northwest Territories, Canada. Water Resources Research 36: 1069--1077. Bouchet RJ (1963) Evapotranspiration re´elle et potentielle, signification climatique, Int. Assoc. Sci. Hydrol., Proc. Berkeley, Calif. Symp., Publ. 62: 134–142. Brutsaert W (1982) Evaporation into the Atmosphere. Dordrecht: Kluwer. Brutsaert W and Chen D (1996) Diurnal variation of surface fluxes during thorough drying (or severe drought) of natural prairie. Water Resources Research 32: 2013--2019. Budyko MI (1974) Climate and Life, 508p. New York: Academic Press. De Bruin HAR (1983) A model of the Priestley–Taylor parameter, a. Journal of the Applied Meteorology 22: 572--578. Dirmeyer PA, Gao X, Zhao M, Guo Z, Oki T, and Hanasaki N (2006) GSWP-2: Multimodel analysis and implications for our perception of the land surface. Bulletin of the American Meteorological Society 87: 1381--1397. Finch JW and Gash JHC (2002) Application of a simple finite difference model for estimating evaporation from open water. Journal of Hydrology 255: 253--259. Finch JW and Hall RL (2005) Evaporation from lakes. In: Anderson MG (ed.) Encyclopedia of Hydrological Sciences, pp. 635--646. Chichester: Wiley. Fischer EM, Seneviratne S, Luethi M, and Schaer C (2007) Contribution of land– atmosphere coupling to recent European summer heat waves. Geophysical Research Letters 34: L06707 (doi:10.1029/2006GL029068). Fu G, Charles SP, and Yu J (2009) A critical overview of pan evaporation trends over the last 50 years. Climatic Change 97(1–2): 193--214. Gash JHC and Shuttleworth WJ (2007) Evaporation. Wallingford: IAHS Press. Gedney N, Cox PM, Betts RA, Boucher O, Huntingford C, and Stott PA (2006) Detection of a direct carbon dioxide effect in continental river runoff records. Nature 439: 835--838. Kleidon A and Schymanski S (2008) Thermodynamics and optimality of the water budget on land: A review. Geophysical Research Letters 35: L20404. Koster RD, Dirmeyer PA, Guo Z, et al. (2004) Regions of strong coupling between soil moisture and precipitation. Science 305: 1138–1140. Kustas WP, Prueger JH, and Hipps LE (2002) Impact of using different time-averaged inputs for estimating sensible heat flux of riparian vegetation using radiometric surface temperature. Journal of Applied Meteorology 41: 319--332. Law BE, Falge E, Gu L, et al. (2002) Environmental controls over carbon dioxide and water vapor exchange of terrestrial vegetation. Agricultural and Forest Meteorology 113: 97--120. Linacre ET (1993) Data-sparse estimation of lake evaporation, using a simplified Penman equation. Agricultural and Forest Meteorology 64: 237--256. Mahfouf JF and Noilhan J (1991) Comparative study of various formulations of evaporation from bare soil using in situ data. Journal of Applied Meteorology 30: 1354--1365. McNaughton KG and Jarvis PG (1991) Effects of spatial scale on stomatal control of transpiration. Agricultural and Forest Meteorology 54: 279--302. McNaughton KG and Spriggs TW (1989) An evaluation of the Priestley–Taylor equation. In: Black TA, Spittlehouse DL, Novak MD, and Price DT (eds.) Estimation

of Areal Evaporation, IAHS Publication No. 177, pp. 89–104. Wallingford: IAHS Press. McVicar TR, Van Niel TG, Li LT, et al. (2008) Wind speed climatology and trends for Australia, 1975–2006: Capturing the stilling phenomenon and comparison with near-surface reanalysis output. Geophysical Research Letters 35: L20403 (doi:10.1029/2008GL035627) Meesters AGCA, Dolman AJ, and Bruijnzeel LA (2009) Comment on ‘‘Biotic pump of atmospheric moisture as driver of the hydrological cycle on land’’ by AM Makarieva and VG Gorshkov, Hydrol. Earth Syst. Sci., 11, 1013–1033, 2007. Hydrology and Earth System Sciences 13: 1299--1305. Milly PCD and Dunne KA (2002) Macroscale water fluxes, 2, Water and energy supply control of their interannual variability. Water Resources Research 38(10): 1206 (doi:10.1029/2001WR000760) Monteith JL (1965) Evaporation and environment. In: The State and Movement of Water in Living Organisms. Proceedings of the 19th Symposium Society for Experimental Biology, pp. 205--234. Swansea: Cambridge University Press. Morton FI (1983) Operational estimates of areal evapotranspiration and their significance to the science and practice of hydrology. Journal of Hydrology 66: 1–76. Oki T and Kanae S (2006) Global hydrological cycles and world water resources. Science 313: 1068--1072. Payne RE (1972) Albedo of the sea surface. Journal of the Atmospheric Sciences 29: 959--970. Peixoto JP and Oort AH (1996) The climatology of relative humidity in the atmosphere. Journal of Climate 9: 3443--3463. Penman HL (1948) Natural evaporation from open water, bare soil and grass. Proceedings of the Royal Society of London, Series A: Mathematical and Physical Sciences 193: 120--145. Peterson TC, Golubev VS, and Groisman PY (1995) Evaporation losing its strength. Nature 377: 687--688. Piao S, Friedlingstein P, Ciais P, De Noblet-Ducoudre N, Labat D, and Zaehle S (2007) Changes in climate and land use have a larger direct impact than rising CO2 on global river runoff trends. Proceedings of the National Academy of Sciences of the United States of America 104: 15242--15247. Porte´-Agel F, Parlange MB, Cahill AT, and Gruber A (2000) Mixture of time scales in evaporation: Desorption and self-similarity of energy fluxes. Agronomy Journal 92: 832--836. Priestley CHB and Taylor RJ (1972) On the assessment of surface heat flux and evaporation using large-scale parameters. Monthly Weather Review 100: 81--92. Raupach MR (2001) Combination theory and equilibrium evaporation. Quarterly Journal of the Royal Meteorological Society 127: 1149--1181. Rayner DP (2007) Wind run changes are the dominant factor affecting pan evaporation trends in Australia. Journal of Climate 20: 3379--3394. Roderick ML and Farquhar GD (2002) The cause of decreased pan evaporation over the past 50 years. Science 298: 1410--1411. Roderick ML, Rotstayn LD, Farquhar GD, and Hobbins MT (2007) On the attribution of changing pan evaporation. Geophysical Research Letters 34: L17403 (doi:10.1029/ 2007GL031166). Sene KJ, Gash JHC, and McNeil DD (1991) Evaporation from a tropical lake: Comparison of theory with direct measurements. Journal of Hydrology 127: 193--217. Seneviratne SI, Luethi D, Litschi M, and Schaer C (2006) Land–atmosphere coupling and climate change in Europe. Nature 443, doi:10.1038/nature05095. Shuttleworth WJ (1993) Evaporation. In: Maidment DR (ed.) Handbook of Hydrology, pp. 4.1--4.53. New York: McGraw-Hill. Shuttleworth WJ (2009) On the theory relating changes in area-average and pan evaporation. Quarterly Journal of the Royal Meteorological Society 135: 1230--1247. Shuttleworth WJ and Calder IR (1979) Has the Priestley–Taylor equation any relevance to forest evaporation? Journal of Applied Meteorology 18: 639--646. Stewart JB (1977) Evaporation from the wet canopy of a pine forest. Water Resources Research 13: 915–921. Sweers HE (1976) A nomogram to estimate the heat-exchange coefficient at the air– water interface as a function of wind speed and temperature: A critical survey of some literature. Journal of Hydrology 30: 375--401. Teuling AJ, Hirschi M, Ohmura A, et al. (2009) A regional perspective on trends in continental evaporation. Geophysical Research Letters 36(2): L02404. Thom AS and Oliver HR (1977) On Penman’s equation for estimating regional evaporation. Quarterly Journal of the Royal Meteorological Society 103: 345--357. Valentini R, Mateucci G, Dolman AJ, et al. (2000) Respiration as the main determinant of carbon balance in European forests. Nature 404: 861--865.

Evaporation in the Global Hydrological Cycle van der Molen MK, Dolman AJ, Waterloo MJ, and Bruijnzeel LA (2006) Climate is affected more by maritime than by continental land use change: A multiple scale analysis. Global and Planetary Change 54: 128--149. Wallace JS and Holwill CJ (1997) Soil evaporation from tiger-bush in south-west Niger. Journal of Hydrology 188–189: 426--442. Willett KM, Gillett NP, Jones PD, and Thorne PW (2007) Attribution of observed surface humidity changes to human influence. Nature 449, doi:10.1038/ nature06207.

87

Wilson KB, Baldocchi DD, Aubinet M, et al. (2002) Energy partitioning between latent and sensible heat flux during the warm season at FLUXNET sites. Water Resources Research 38(12): 1294 (doi:10.1029/2001WR000989). Zhang L, Dawes WR, and Walker GR (2001) Response of mean annual evapotranspiration to vegetation changes at catchment scale. Water Resources Research 37: 701--708.

2.04 Interception AMJ Gerrits and HHG Savenije, Delft University of Technology, Delft, The Netherlands & 2011 Elsevier B.V. All rights reserved.

2.04.1 2.04.2 2.04.3 2.04.3.1 2.04.3.2 2.04.3.3 2.04.3.4 2.04.3.5 2.04.4 2.04.4.1 2.04.4.2 2.04.5 2.04.5.1 2.04.5.2 2.04.5.3 2.04.5.3.1 2.04.5.3.2 2.04.5.3.3 2.04.6 2.04.7 References

Introduction Importance of Interception Types of Interception Canopy Interception Forest Floor Interception Fog Interception Snow Interception Urban Interception Methods to Measure Interception Canopy Forest Floor Interception Models Conceptual Rutter Model Analytical Gash Model Stochastic Interception Models Poisson distribution Markov chains Gamma probability density function and transfer functions Consequences of Underestimating Interception for Hydrological Modeling and Water Resource Assessment Outlook

Nomenclature b Br c cp D Di E Ei Ei,c E li;c E ti;c Ei,f El Ep H I L m n nr,d nm p

constant in Rutter (1971) model (L1) Bowen ratio (–) canopy coverage (–) specific heat (L M T3 K1) drainage rate from the canopy (L T1) interception threshold (L T1) actual evaporation (L T1) interception evaporation (L T1) interception evaporation from canopy (L T1) evaporation from leaves (without trunk) (L T1) evaporation from the trunk (L T1) interception evaporation from forest floor (L T1) evaporation from lower basin (L T1) potential evaporation (L T1) sensible heat flux (M T3) interception process (L T1) number of elemental surface areas per unit ground (L2) mean number of raindrops striking an element (–) mean number of drops retained per element (–) number of rain days per month (–) days within a month ( ¼30.5) (–) throughfall coefficient (–)

pt P Pg P0g P00g q

r Sc Slc Stc Sf Si Sl Su Tf Ts t z b e k q h

89 90 90 90 91 91 91 92 93 93 94 94 94 95 95 95 96 98 98 99 99

trunk fraction coefficient (–) precipitation (L T1) gross precipitation (L T1) gross precipitation necessary for canopy saturation (L T1) gross precipitation necessary for trunk saturation (L T1) maximum amount of rain drops on element (–); specific humidity (M T1L1) amount of rain drops on element (–) storage of canopy (L) storage of leaves (without trunk) (L) storage of trunk (L) storage of forest floor (L) interception storage (L) storage of the lower basin (L) storage of the upper basin (L) throughfall (L T1) stemflow (L T1) mean volume of raindrops (L3) height (L) scaling factor (L T1) constant in Rutter (1971) model (–) latent heat of vaporization coefficient (L2 T2) density of water (M L3) potential temperature (K)

89

90

Interception

2.04.1 Introduction When it rains the entire surface becomes wet: trees, shrubs, grass, forest floor, footpaths, etc. Also in urban areas, roads and roofs become wet, sometimes forming pools of stagnant water. After rainfall has ceased these surfaces soon become dry again. This process is called ‘interception’. It is the part of the rainfall that is captured by surface storage (i.e., vegetation, roofs, etc.) before it can run off or infiltrate into the soil. The intercepted water generally evaporates during the event and shortly after the rainfall ceased, so that it can repeat its function during the next rainfall event. In the literature, interception is defined in different ways: sometimes as a stock, sometimes as a flux, or, more appropriately, as the entire interception process (Savenije, 2005). If only interception storage (Si [L]) is considered, interception is defined as the amount of rainfall which is temporarily stored on the Earth’s surface. Actually, this is the interception capacity or water-holding capacity. If interception is defined as a flux, then it is the intercepted water which evaporated over a certain time [L T1] during and after the event. When interception is considered as a process (I [L T1]), it is defined as the part of the rainfall flux which is intercepted on the wetted surface after which it is fed back to the atmosphere. The interception process equals the sum of the change of interception storage (Si) and the evaporation from this stock (Ei):

I¼

dSi þ Ei dt

ð1Þ

The timescale of the interception process is in the order of 1 day. After 1 day, it is fair to assume for most climates that the first term on the right-hand side in Equation (1) approaches zero, and I ¼ Ei. Of course, in the case of snow under cold climates, this may take longer. How much of the precipitation is intercepted depends on several factors, which can be divided into three groups:

•

Vegetation characteristics. Large vegetation types, such as trees, have a high aerodynamic roughness, causing high potential evaporation rates. Grasses, crops, or bushes, on the other hand, have a much lower roughness and thus do not have as high potential evaporation rates. The storage capacity also depends on the vegetation type. The shape of the leaves, the thickness, the density (leaf area index), and the configuration of the branches determine how much water can be stored. For example, the capacity of a coniferous or a deciduous tree is different (e.g., Rutter et al., 1975; Baird and Wilby, 1999; Bryant et al., 2005; Toba and Ohta, 2005). Although intuitively one might think that a deciduous tree can hold more water in its bucket-like leaves, a coniferous tree can hold much more water by adhesion. Furthermore, it is also important to take the seasonality into account. Deciduous trees lose their leaves in the dormant season, causing a large reduction in the canopy storage capacity. Vegetation also determines the amount of understorey growth and forest floor. The forest floor of different vegetation types can have significantly

•

•

different interception behavior (e.g., a thick needle layer or a thin leaf litter layer). Rainfall characteristics. Rainfall has a large influence on the interception process. The rainfall frequency is a major determining factor. It makes a big difference if rainfall occurs as one continuous storm or as a sequence of several small events with dry spells in between. Even if the total rainfall depth is the same, the last scenario intercepts much more rainwater, because between the events the storage can be (partly) emptied by evaporation and thus more storage is available. Second, the rainfall intensity is important, although there is no consensus in literature. Horton (1919) and Wang et al. (2007) concluded that the interception capacity is lower at higher intensity because high rainfall intensities cause splashing and shaking of leaves. On the other hand, Aston (1979) and Keim et al. (2006b) noted the opposite: high rainfall intensities coincide with high storage capacities, due to dynamic storage. Evaporative demand. If the potential evaporation (i.e., open water evaporation) is high, the intercepted water can evaporate more easily during and after the event. Wind plays an important role in removing moisture from the surface providing a higher vapor deficit, particularly in the canopy. Moreover, the roughness of the vegetation increases the evaporative power, by causing turbulence which makes it easier to take up the intercepted water. However, wind can also reduce the amount of interception by reducing the storage. Horton (1919), Klaassen et al. (1996), and Ho¨rmann et al. (1996) noted that with increasing wind speed the measured storage capacity is less, due to the fact that the wind shakes the rainwater off the leaves.

Of the above three factors, the rainfall characteristics are most dominant for evaporation from interception. Although both the storage capacity (mainly vegetation characteristic) and the available energy form a constraint to the evaporation flux per event, the number of events is a more important factor. This is confirmed by the sensitivity analysis of Gerrits et al. (2009c).

2.04.2 Importance of Interception Although most surfaces can store only a few millimeters of rainfall, which is often not much in comparison to other stocks in the water balance, interception is generally a significant process. The impact becomes evident at longer timescales. Although interception storage is generally small, the number of times that the storage is filled and depleted can be so large that the interception flux is generally of the same order of magnitude as the transpiration flux. In addition, the interception process smooths the rain intensities, causing more gradual infiltration. Interception redistributes the rainfall as well. Some parts of a field receive less water due to interception, whereas other parts receive more due to funneling of the vegetation (e.g., Germer et al., 2006; Gerrits et al., 2009b). Subsequently, this has an influence on the soil moisture patterns, and this is again important for flood generation (Roberts and Klingeman, 1970). Besides the hydrological effects, there are influences on the nutrient cycle of a forest, and on agricultural applications.

Interception

For example, interception affects the efficiency of insecticides and fertilizers (Aston, 1979). Besides, fire retardants are more effective if they are stored by vegetation. Finally, interception may reduce soil erosion by preventing rain drops to directly hit and erode the soil layer (Walsh and Voigt, 1977), although in the case of canopy interception the opposite can be true due to the formation of larger rain drops with a higher impact on the forest floor.

2.04.3 Types of Interception As already stated in Section 2.04.1, it is possible to define an infinite number of interception types. In principle, every surface that can store water can be considered as an interception type. In this chapter, we focus on the major types, mainly occurring in a natural environment plus some special mechanisms. However, more often than not, it is a combination of mechanisms. For example, in a forest, it is likely that a part of the rainfall is intercepted by the canopy of a tree, while the remaining part can be intercepted by epiphytes on the branches and/or bark, and, finally, the understorey and forest floor intercept the throughfall before infiltration starts.

2.04.3.1 Canopy Interception Canopy interception is the rainwater that is stored on the leaves and branches of a tree which is subsequently evaporated. This interception can be calculated by measuring rainfall above the trees or measured in an open area nearby (gross

Precipitation

Canopy interception

91

rainfall Pg) and subtracting the throughfall (Tf) and stemflow (Ts) (Figure 1):

Ei;c þ

dSc ¼ Pg  Tf  Ts dt

ð2Þ

Many research studies have been carried out on canopy interception. In Table 1 an overview is given. We can see in the table and also in tables in Kittredge (1948), Zinke (1967), and Breuer et al. (2003) that there is a large difference in the canopy interception by deciduous and coniferous trees (e.g., Kittredge, 1948; Bryant et al., 2005; Toba and Ohta, 2005). Not only because deciduous trees lose their leaves, but also because the leaf area of coniferous trees is much larger than of deciduous trees; coniferous trees can store much more water. Furthermore, leaves may swing over when they become too heavy, causing a (sudden) decrease of the storage capacity. However, Herbst et al. (2008) found counterintuitive results, where higher evaporation rates were found in deciduous trees in winter caused by rougher aerodynamics of the bare canopy and deeper penetration of the wind. In most cases, the storage of water on the branches is small; however, in some environments, the branches can be overgrown by epiphytes. Pypker et al. (2006) showed that in a Douglas fir forest the canopy water storage can potentially be increased by 41.3 mm and Ho¨lscher et al. (2004) found that epiphytes can account for 50% of the storage capacity. However, this large increase in storage capacity is not necessarily resulting in high interception values (storage þ evaporation), because the water uptake and release by the epiphytes is delayed. It takes a while to saturate the epiphytes, and already before saturation, runoff generation can take place. Successively, after wetting, the drying of the epiphytes takes much longer than drying of the canopy, causing less storage to be available. Another special type of canopy interception is interception by agricultural crops. In essence, there is no difference between crops and other vegetation types. They both can store water up to a certain threshold and then drain water to the floor as throughfall. However, whereas vegetation has a gradual seasonal pattern (summer vs. winter), crops have a phenological growth cycle (seeding to harvesting) which is therefore more abrupt. Hence, when modeling crop interception the appropriate description of the variation in the storage capacity is important.

2.04.3.2 Forest Floor Interception

Throughfall Stemflow Forest floor interception Infiltration Figure 1 Two major interception types in the natural environment.

Forest floor interception is the part of the throughfall that is temporarily stored in the top layer of the forest floor and successively evaporated within a few hours or days during and after the rainfall event. The forest floor can consist of short vegetation (like grasses, mosses, bushes, and creeping vegetation), litter as described by Hoover and Lunt (1952) as the litter and fermentation (L and F) layer (i.e., leaves, twigs, and small branches), or bare soil. Although the latter seems to have an overlap with soil evaporation, we distinguish them by the fact that soil evaporation refers to the water that is stored in the root zone (Groen and Savenije, 2006).

92

Interception

Table 1 Canopy interception values in literature, with Sc,max the water storage capacity and Ei,c the interception evaporation as percentage of gross precipitation Source

Specie

Location

Sc,max (mm)

Rutter et al. (1975)

Corsian pine (Pinus nigra) Douglas fir (Pseudotsuga menziesii) Norway spruce (Picea abies) Hornbeam (Carpinus betulus) Oak (Quercus robur)

United United United United United

Gash and Morton (1978) Gash et al. (1980)

Scots pine (Pinus sylvestris) Sitka spruce (Picea sitchensis) Scots pine (Pinus sylvestris) Beech (Nothofagus) Acacia auriculiformis

United Kingdom United Kingdom United Kingdom New Zealand Indonesia

1.05 1.2 1.5 1.0 (leafy) 0.65 (leafless) 0.875 (leafy) 0.275 (leafless) 0.8 0.75–1.2 1.02 1.5 (leafy) 1.2 (leafless) 0.5–0.6

Norway spruce (Picea abies) Beech (Asperulo-fagetum) Pinus pinaster Eucalyptus globulus Tamaulipan thornscrub Loblolly (Pinus taeda) & shortleaf pine (Pinus echinata) Longleaf pine (Pinus palustris) Scrub oak (Quercus berberidifolia) White oak (Quercus alba) & shortleaf pine (Pinus echinata) & loblolly pine (Pinus palustris) Hardwood Larc (Larix cajanderi) Red pine (Pinus sylvester) Red pine (Pinus densiflora) Sawtooth oak (Quercus acutissima) Oak (Quercus serrata) Rain forest

France Germany Portugal Portugal Mexico USA (GA)

1.97

34.2 18 10.8 17.1 18.9 22.3

USA (GA) USA (GA) USA (GA)

1.70 1.40 1.58

17.6 17.4 18.6

USA (GA) Siberia Siberia Japan Japan Japan Brazil

0.98

17.7 29 36 13–17 24 18 13–22

Rowe (1983) Bruijnzeel and Wiersum (1987) Viville et al. (1993) Ho¨rmann et al. (1996) Valente et al. (1997) Navar et al. (1999) Bryant et al. (2005)

Toba and Ohta (2005)

Cuartas et al. (2007)

Kingdom Kingdom Kingdom Kingdom Kingdom

1.28 (leafy) 0.84 (leafless) 0.41 0.21

1.0

Ei,c (%) 35 39 48 36 18

27–32 42 35 (leafy), 22 (leafless) 11–18

See also tables in Kittredge (1948), Zinke (1967), and Breuer et al. (2003).

In Table 2 some results are presented of previous work on forest floor interception.

moisture. These instruments suffer from various limitations. An overview of fog collectors can be found in Bruijnzeel et al. (2005).

2.04.3.3 Fog Interception A special type of interception is fog interception or cloud interception. Vegetation can intercept not only rain, but also moisture (in the form of small water droplets) from the air. Fog can occur due to different processes. Bruijnzeel et al. (2005) distinguished nine types: radiation fog, sea fog, stream fog, advection fog, ice fog, coastal fog, valley fog, urban fog, and mountain fog. Fog interception is mainly important in tropical montane environments (table in Bruijnzeel (2005): 6–53% of rainfall), and can also play a significant role in semi-arid regions near the coast (e.g., Hursh and Pereira, 1953; Hutley et al., 1997; Hildebrandt et al., 2007). In both environments, the main problem with fog interception studies is to measure precipitation and throughfall (Equation (2)), which is especially important because fog deposition can be twice as high as normal rainfall. Since conventional rain gauges are not suitable to measure fog deposition, special fog collectors have been developed with often wire meshes to intercept the

2.04.3.4 Snow Interception Snowfall is also intercepted by trees. Especially, coniferous trees can store so much snow, that they collapse under its weight. As an example, Storck et al. (2002) found in a Douglas-fir-dominated forest that up to 60% of the snowfall was intercepted, equaling 40 mm of snow water equivalent (swe). The storage of snow on the canopy is different from rain. For rainfall interception the storage capacity is mainly a function of the leaf surface area, whereas for snow interception the branch strength and canopy shape are more important (Ward and Trimble, 2004). Furthermore, the snow storage is also dependent on the temperature. If snow falls with temperatures close to freezing point, the cohesion of snow is higher causing more snow to be accumulated on the canopy (Ward and Trimble, 2004). Another difference between rainfall interception and snow interception is the way in which interception storage is depleted. Rainfall interception is a real threshold process,

Interception

93

Table 2 Forest floor interception values in literature, with the water storage capacity Sf,max and the interception evaporation Ei,f as percentage of net precipitation (i.e., throughfall) Source

Forest floor type

Location

Haynes (1940) Kittredge (1948) Beard (1956) Helvey (1964) Brechtel (1969)

Kentucky bluegrass (Poa pratensis) Californian grass (Avena, Stipa, Lolium, Bromus) Themeda and Cymbopogon Poplar Scot’s pine Norway spruce Beech Oak Shorea robusta and Mallotus philippensis Pinus roxburghii and Quercus glauca Pinus roxburghii Quercus leucotrichophora and Pinus roxburghii Quercus floribunda and Quercus leucotrichophora Quercus lanuginosa and Quercus floribunda Blue stem Andropogon gerardi Vitman Pine (Pinus sylvestris) Beech (Fagus sylvaticus) Bracken litter (Pteridium aquiliunum) Norway spruce Sitka spruce Beech (Asperulo-Fagetum) Pinus radiata Eucalyptus Douglas fir Peble mulch (5–9 cm) Peble mulch (2–6 cm) Cryptomeria japonica Lithocarpus edulis Grass (Aristida divaricata) Woodchips (Pinus) Poplar leaves (Populus nigra)

? USA (CA) South Africa USA (NC) USA (NY) USA (NY) USA (NY) USA (NY) India India India India India India USA (TX) United Kingdom United Kingdom United Kingdom Scotland Scotland Germany Australia Australia Netherlands China China Japan Japan Mexico Mexico Mexico

Pathak et al. (1985)

Clark (1940) in Thurow et al. (1987) Walsh and Voigt (1977) Pitman (1989) Miller et al. (1990) Thamm and Widmoser (1995) Putuhena and Cordery (1996) Schaap and Bouten (1997) Li et al. (2000) Sato et al. (2004) Guevara-Escobar et al. (2007)

a

Sf,max (mm)

Ei,f (%) 56a 26a 13a 34 21 16 16 11 11.8 7.8 9.6 10.6 11.0 11.3 57–84

0.6–1.7 0.9–2.8 1.67

2.5–3.0 2.78 1.70 0.281 0.526 0.27–1.72 0.67–3.05 2.5 8 2.3

18a 16a 12–28

0.23 mm d1 11.5a 17.4a

% of gross precipitation instead of net precipitation.

whereby throughfall starts when the storage capacity is exceeded. The storage capacity is then emptied by evaporation. Snow, on the other hand, can only be removed from the canopy by three ways: sublimation, mechanical removal (sliding leading to mass release), and melt water drip (Miller, 1966).

Island is mainly caused by the (relatively warm) buildings that block the cold night sky. Furthermore, the thermal properties of a city are different: concrete and asphalt have much higher heat capacities than forests and also the surface radiative properties differ (e.g., albedo and emissivity). The lack of vegetation in urban areas, which reduces cooling by transpiration, also causes a difference in the energy balance.

2.04.3.5 Urban Interception Most hydrological studies focus on natural environments and not on urbanized areas, which is also the case for interception studies. However, recently, with the increasing interest for alternative sources of water for nonpotable domestic use (socalled ‘gray water’), water balance studies on (interception) evaporation in urban areas increased (Grimmond and Oke, 1991; Ragab et al., 2003; Gash et al., 2008; Nakayoshi et al., 2009). The difference between urban and rural interception is not only that the typical storage capacities of buildings, roads, etc., are unknown, but also that the entire energy balance is different in a city. Oke (1982) discovered the so-called ‘Urban Heat Island’, that is, higher temperatures in urban areas compared to the surrounding rural areas. The Urban Heat

2.04.4 Methods to Measure Interception 2.04.4.1 Canopy There exist already many methods to measure canopy interception. The most-often used method is by measuring rainfall above the canopy and subtract throughfall and stemflow (e.g., Helvey and Patric, 1965). However, the problem with this method is that the canopy is not homogeneous, which causes it to be difficult to obtain representative throughfall data. Using multiple rain gauges under the canopy (Helvey and Patric, 1965; Keim et al., 2005; Gerrits et al., 2009b) reduces this problem. Sometimes the collectors are moved to achieve a better representation of throughfall (e.g., Lloyd and Marques, 1988; Tobo´n-Marin et al., 2000; Manfroi et al., 2006; Ziegler

94

Interception

et al., 2009). Another method to avoid the problem with the spatial distribution of the canopy was introduced by Calder and Rosier (1976) and applied by, for example, Shuttleworth et al. (1984), Calder et al. (1986), and Calder (1990). They covered the forest floor with plastic sheets and collected the throughfall. The disadvantage of this method is that for long periods irrigation is required, because otherwise, in the end, the trees will dry out and may even die due to water shortage. The method by Hancock and Crowther (1979) avoided these problems, by making use of the cantilever effect of branches. If leaves on a branch hold water, it becomes more heavy and will bend. By measuring the displacement, it is possible to determine the amount of intercepted water. Huang et al. (2005) refined this method by making use of strain gauges. However, the disadvantages of these methods are that only information about one single branch is obtained and it is quite laborious to measure an entire tree. Edwards (1986), Fritschen and Kinerson (1973), and Storck et al. (2002) made use of weighing lysimeters with trees. Although interception of a whole tree is measured with this method, the big disadvantage of this method is that it is expensive and destructive. Friesen et al. (2008) developed a nondestructive method to measure canopy interception of a whole tree. With mechanical displacement sensors, Friesen et al. (2008) measured the stem compression due to interception water, which is an integration of the whole canopy. However, although this method looks promising, it is still under development. A totally different way of measuring canopy interception of a forest plot is to make use of ray attenuation. Calder and Wright (1986) used the attenuation of gamma rays. They transmitted from a tower gamma-rays through the canopy at different heights and measured the gamma-ray density at a receiving tower. The ratio between transmitted and received gamma-ray density during dry conditions is successively compared to this ratio during a rainfall event. This gives an estimate of the amount of water stored on the canopy over time. Although the method gives interception estimated of an entire forest, the method becomes inaccurate under windy conditions. Furthermore, safety standards inhibits unattended use of this method. Bouten et al. (1991) overcame this problem by making use of microwave attenuation. It appears to be a suitable method to measure canopy wetness, although it is an expensive method. Evaporation can also be measured by flux measurements. By measuring temperature (y) and specific humidity (q) at several heights (z) above the canopy, one can calculate the Bowen ratio (Br), which is the sensible heat flux, H, divided by the latent heat flux (lE):

Br ¼

cp dy=dz H ¼ rlE ldq=dz

ð3Þ

Combined with the energy balance, evaporation can be calculated (Gash and Stewart, 1975). The main difficulty with the Bowen ratio method is to measure the humidity gradient more accurately (Stewart, 1977). Another method is the eddy covariance technique, where the net upward or downward flux is determined by fast-response three-dimentional (3D) wind speed measurements combined with a concentration

measurement. This concentration can be humidity, temperature, or CO2 concentrations (Amiro, 2009).

2.04.4.2 Forest Floor In the literature, little can be found on forest floor interception, although some researchers have tried to quantify the interception amounts. Generally, these methods can be divided into two categories (Helvey and Patric, 1965): 1. lab methods, whereby field samples are taken to the lab and successively the wetting and drying curves are determined by measuring the moisture content and 2. field methods, whereby the forest floor is captured into trays or where sheets are placed underneath the forest floor. An example of the first category is that of Helvey (1964), who performed a drainage experiment on the forest floor after it was saturated. During drainage, the samples were covered, and after drainage had stopped (24 h), the samples were taken to the lab, where the samples were weighed and successively dried until a constant weight was reached. By knowing the oven dry weight of the litter per unit area and the drying curve, the evaporation from interception could be calculated. In this way, they found that about 3% of the annual rainfall evaporated from the litter. Similar work was done by Bernard (1963), Walsh and Voigt (1977), and Sato et al. (2004). However, what they all measured was not the flux, but the storage capacity. Another example of lab experiments was carried out by Putuhena and Cordery (1996). First, field measurements were carried out to determine the spatial variation of the different forest floor types. Second, storage capacities of the different forest floor types were measured in the lab using a rainfall simulator. Finally, the lab experiments were extrapolated to the mapping step. In this way, Putuhena and Cordery (1996) found average storage capacities of 2.8 mm for pine and 1.7 mm for eucalyptus forest floors. Moreover, GuevaraEscobar et al. (2007) made use of a rainfall simulator. Examples of the second category have been, for example, carried out by Pathak et al. (1985), who measured the weight of a sample tray before and after a rainfall event. They found litter interception values of 8–12% of the net precipitation. In addition, here, they measured the storage capacity, rather than the flux. Schaap and Bouten (1997) measured the interception flux by the use of a lysimeter and found that 0.23 mm d1 evaporated from a dense Douglas fir stand in early spring and summer. Also, Brechtel (1969) and Thamm and Widmoser (1995) made use of lysimeters. Brechtel (1969) manually measured the infiltrated water and Thamm and Widmoser (1995) developed an automatic and more sophisticated method, whereby the suction under the forest floor is controlled by a tensiometer. Gerrits et al. (2007) developed a method whereby both the forest floor interception and the infiltrated water are continuously weighed in suspended trays with strain gauges. In Figure 2 the schematic setup is shown. Measurements with sheets were done, for example, by Li et al. (2000), who found that pebble mulch intercepts 17% of the gross precipitation. Miller et al. (1990) found comparable results (16–18%) for a mature coniferous plantation in Scotland.

Interception

95

three parts:

Eint

Precipitaion

1. free throughfall, that is, throughfall, which did not touch the canopy at all (pPg), 2. trunk input (ptPg), and 3. canopy input ((1  p  pt)Pg). Litter Su Geotextile El

Weighing device

Infiltration

The rain that falls on the canopy can drain to the ground (i.e., canopy drainage, D), or evaporate ðEli;c Þ, or it can be stored on the canopy ðSlc Þ:

ð1  p  pt Þ

Z

Pg dt ¼

Z

Ddt þ

Z

Eli;c dt þ

Z

dSlc

ð4Þ

Sl

The rain that falls on the trunk can evaporate from the trunk ðEti;c Þ, or drain in the form of stemflow (Ts), or it can be stored on the trunk ðStc Þ:

Valve

Z Figure 2 Forest floor interception device by Gerrits AMJ, Savenije HHG, Hoffmann L, and Pfister L (2007) New technique to measure forest floor interception – an application in a beech forest in Luxembourg. Hydrology and Earth System Sciences 11: 695–701.

2.04.5 Interception Models In literature, several models have been developed to simulate forest interception. Almost all of these models are concentrated on canopy interception, sometimes including stem interception (Table 3). In principle, these models can be expanded to include forest floor or any surface interception as well. The most often used interception models are the conceptual model of Rutter et al. (1971) (Section 2.04.5.1) and the analytical model of Gash (1979) (Section 2.04.5.2) or revisions of these models. Furthermore, there exist some stochastic models, which will be described in Section 2.04.5.3. In Table 3 an overview and summary of the models are given. A more detailed overview and comparison can be found in Muzylo et al. (2009).

2.04.5.1 Conceptual Rutter Model The conceptual framework of the original Rutter model is depicted in Figure 3. As can be seen the rainfall is divided into Table 3

Characteristics of interception models

Main author

Model type

Rutter Gash C alder De Groen

Conceptual Analytical Stochastic Concept./ stoch. Concept./ stoch.

Keim

Interception element: canopy

stem

x x x x

x x

x

x

Timescale

forest floor

x

rhourly event rhourly monthly 6-hourly

pt

Pg dt ¼

Z

Ts dt þ

Z

Eti;c dt þ

Z

dStc

ð5Þ

with Ei;c ¼ Eli;c þ Eti;c and Sc ¼ Slc þ Stc for the total canopy interception. The evaporation from the wet canopy is calculated with the Penman equation (Penman, 1948). Because the canopy is not always completely wet ðSlc o Slc;max Þ, the actual evaporation rate can be calculated by the fraction of the potential evaporation: Ep Slc =Slc;max . The same concept is applied for the trunks. However, for the determination of the potential evaporation of the trunks, the potential evaporation of the canopy is multiplied with and extra constant E. Stemflow is modeled as a threshold process, whereby no stemflow is generated when Stc o Stc;max , and when the threshold is exceeded stemflow equals the difference between Stc and Stc;max . Canopy drainage is modeled in a similar way; however, when the threshold Slc;max is exceeded, drainage is defined as

h i D ¼ Ds exp bðSlc  Slc;max Þ

ð6Þ

with Ds being the rate of drainage when the canopy is saturated and b [L1] an empirical coefficient. Valente et al. (1997) revised the original Rutter model, to model interception in a more realistic way for sparse canopies. The main drawbacks of the original model were the partitioning of free throughfall and canopy input, and the conceptual error that evaporation from interception can theoretically be higher than potential evaporation (Valente et al., 1997). Therefore, they divided the conceptual model into two areas: a covered area (c) and an uncovered area (1  c). Second, in the revised Rutter model, only water can reach the trunk after it has flowed through the canopy as a part of the canopy drainage. Water which is not drained by the trunk is directly dripping to the ground. The final change was made that evaporation from the saturated canopy is not equal to the potential evaporation, but is reduced by a factor 1  E (0oEo1). The remaining energy ((E)Ep) is then available for evaporating water from the saturated trunk (Figure 4).

96

Interception

Canopy evaporation S lc ⎧ ,S lc< S lc,max E ⎪ E li,c = ⎨ p S lc,max ⎪ ,S lc ≥ S lc,max Ep ⎩

Canopy input ( 1−p−pt )Pg

Scl

Gross rainfall Pg

Free throughfall p Pg

Trunk evaporation S tc ⎧ ,S tc < S tc,max ⎪εEp t t E i,c = ⎨ S c,max ⎪ ,S tc ≥ S tc,max ⎩ εEp

Trunk input p t Pg

Sct

Scl, max

Sct,max

Drainage D = Ds exp[b (Scl- Scl,max]

Throughfall, Tf

Stemflow, Ts

Figure 3 Conceptual framework of the Rutter model. Modified from Valente F, David JS, and Gash JHC (1997) Modelling interception loss for two sparse eucalypt and pine forest in central Portugal using reformulated Rutter and Gash analytical models. Journal of Hydrology 190: 141–162.

2.04.5.2 Analytical Gash Model The original Gash model is conceptually the same as the Rutter model (see Section 2.04.5.1); however, it does not require meteorological data of high temporal resolution (hourly) and requires less computation time. The main assumption of the Gash model is that it is possible to represent the real rainfall pattern by different discrete rainfall events, each consisting of three phases: 1. wetting phase, 2. saturation phase, and 3. drying phase (long enough to dry the entire canopy). Similar to the Rutter model, rainfall is divided into canopy input (1  p  pt), free throughfall (p), and trunk input (pt). The Gash model makes a distinction between storms which are not large enough to saturate the canopy ðPg o P0g : m stormsÞ and storms which are large enough to saturate the canopy ðPg  P0g : n stormsÞ. The amount of gross rainfall necessary to saturate the canopy is P0g (see Table 4). Interception evaporation is then calculated for the canopy and the trunk. Although the original Gash model appears to work fine for several types of forests, it contains some weaknesses for modeling sparse forests, similar to the Rutter model. Hence, Gash et al. (1995) revised their existing model according to the revised Rutter model (Rutter et al., 1975). An overview of the formulas of the revised Gash model can be found in Table 4.

2.04.5.3 Stochastic Interception Models

areas, which all have the same probability to be struck by raindrops. The Poisson probability of an element to be struck by r drops equals

Pr ¼

m r m e r!

ð7Þ

with m the mean number of raindrops striking an element per storm. If an element can hold q raindrops, the mean number of drops per element (n) can be expressed as

n¼

q X

r  Pr þ q  Pðr 4 qÞ

ð8Þ

r¼0

¼qþ

q X

Pr  ðr  qÞ

ð9Þ

r¼0

with P (r4q) the probability of elements being struck by more P than q drops and is equal to 1  qr¼ o Pr. To upscale from elemental area to canopy area, the number of elemental surface areas per unit ground (L) is required and the mean volume of raindrops (u):

Sc ¼ nuL

ð10Þ

Sc;max ¼ quL

ð11Þ

Pg ¼ muL

ð12Þ

Evaporation is then obtained by (with dScdEi,c ¼  1)

2.04.5.3.1 Poisson distribution Calder (1986) developed a stochastic interception model, where he assumes that a tree consists of several elemental

dn dSc dn 1 ¼  ¼ dEi;c dEi;c dSc uL

ð13Þ

Interception Gross rainfall Pg

97

Interception from evaporation: Ei,c = c·( E li,c + E ti,c ) Covered area input Pg

Uncovered area input Pg

Uncovered area (1−c)

Covered area c

S lc

⎧ ⎪ (1−ε)Ep l S lc,max E i,c = ⎨ ⎪ (1 − ε) Ep ⎩

Free throughfall Pg

Trunk evaporation:

Canopy evaporation: ,S lc < S lc,max ,S lc ≥ S lc,max

S tc ⎧ ⎪ εEp t S c,max i,c = ⎨ ⎪ εEp ⎩

Et

,S tc,< S tc,max ,S lc, ≥ S tc,max

Scl Scl, max

Drainage Dc = d(S lc−S lc,max)/dt

Drip D i,c =( 1−pd ) Dc

Trunk input pd Dd

Sct Sct, max

Trunk drainage Dt,c = d(Sct- Sct, max)/dt

Throughfall, Tf (1−c)Pg+c Di,c

Stemflow, Ts c Dt,c

Figure 4 Conceptual framework of the revised Rutter model. Modified from Valente F, David JS, and Gash JHC (1997) Modelling interception loss for two sparse eucalypt and pine forest in central Portugal using reformulated Rutter and Gash analytical models. Journal of Hydrology 190: 141–162.

The Calder model is very simple and describes the threshold behavior of interception very well; however, it is difficult to upscale from drop size scale to forest size scale. This hinders the applicability of the model.

2.04.5.3.2 Markov chains Groen and Savenije (2006) developed a monthly interception model based on a daily interception model and the daily rainfall characteristics. They assumed interception on a daily

98 Table 4

Interception Components of interception of the original Gash (1979) model and the revised Gash et al. (1995) model for sparse canopies Original Gash (1979)

Amount of gross rainfall necessary to saturate the canopy ðP 0g Þ and trunk ðP00g Þ

Revised (sparse canopy) Gash et al. (1995)

"

P 0g

 g Sc;max p P E ¼ g  p ln 1  ð1  p  pt ÞP E

P 00g ¼ Stc;max =pt Evaporation from canopy interception ðE li;c Þ: 1. for m storms ðPg o P 0g Þ 2. for n storms ðP g  P 0g Þ Evaporation from trunk interception ðE ti;c Þ: 1. for q storms ðP g  P 00g Þ 2. for m þ n  q storms ðPg o P 00g Þ

# P 0g

" # g p P Sc;max ð1  EÞE ¼  p c ln 1  g ð1  EÞE P

g Stc;max P 0 P 00g ¼   p pt c þ P g P g  ð1  EÞE

Pg;j  p Pn E 0 ðP g;j  P 0g Þ nð1  p  pt ÞP g þ  P g j¼1

Pm P g;j " j¼1 #  p Pn ð1  EÞE 0 0 c nP g þ j¼1 ðP g;j  P g Þ g P

qStc P pt mþnq Pg;j j¼1

qStc "

ð1  p  pt Þ

Pm

j¼1

c

pt c 1 

#  p Pn ð1  EÞE 0 ðP  P Þ g g;j j¼1 g P

Modified from Valente F, David JS, and Gash JHC (1997). Modelling interception loss for two sparse eucalypt and pine forest in central Portugal using reformulated Rutter and Gash analytical models. Journal of Hydrology 190:141–162.

scale as (Savenije, 1997, 2004)

Markov probabilities to model monthly interception based on daily information.

Ei;d ¼ minðDi;d ; Pg;d Þ

ð14Þ

The probability distribution of rainfall on a rain day can be described as

  Pg;d 1 f i;d ðPg;d Þ ¼ exp  b b

ð15Þ

with b being the scaling factor, equal to the expected rainfall on a rain day, which can be expressed as

b¼

Pg;m Eðnr;d =nm Þ

ð16Þ

with Pg,m being the monthly rainfall and nr,d and nm the number of rain days per month and amount of days per month, respectively. The number of rain days per month can be expressed by the use of Markov properties. Being p01 the Markov properties of the transition from a dry day to a rain day, and p11 the probability of a rain day after a rain day:

nr;d ¼ nm

p01 1  p11 þ p01

ð17Þ

Multiplying Equations (14) and (15) and successively integrating results in monthly evaporation from interception:

Ei;m ¼ Eðnr;d jnm Þ

ZN

Ei;d  f i;d ðPd Þd Pd

ð18Þ

2.04.5.3.3 Gamma probability density function and transfer functions Keim et al. (2004) developed a stochastic model to obtain from 6-hourly rainfall to 6-hourly throughfall for extreme events. They made use of the gamma probability density function (PDF; for 6 hourly rainfall): Pg =y Pa1 Tf g e  100% ¼ Pg GðaÞya

ð20Þ

The parameters a and y can be estimated by dividing the 6-hourly rainfall in ranges and find the best-fit sets. After downscaling the rainfall and throughfall data, rainfall is transferred through the canopy by a linear system convolution to obtain high-resolution throughfall data, which allows one to investigate the effect of intensity smoothing:

Tf ðtÞ ¼

Zt

PðtÞgðt  tÞdt

ð21Þ

0

with the transfer function g(t  t). Keim et al. (2004) found that the transfer function can be best described with the exponential distribution:

0

   Di;d ¼ Pm 1  exp b

ð19Þ

Hence, the model of Groen and Savenije (2006) is a parsimonious model with only one measurable parameter, and

gðtÞ ¼ a e at

ð22Þ

By coupling the stochastic model with the the intensity smoothing transfer function, effects of forest canopies on extreme rainfall events can be investigated.

Interception

2.04.6 Consequences of Underestimating Interception for Hydrological Modeling and Water Resource Assessment Hydrologists often consider precipitation as the start of the hydrological cycle. After a rainfall event, the first separation point in the cycle is on the Earth surface. Part of the rainwater is intercepted by the vegetation or ground surface and the remainder infiltrates into the unsaturated zone or runs off. The part of the rainfall that is intercepted successively evaporates from the temporary storage. This first separation point in the hydrological cycle is not always considered a significant process. This is partly due to the technical difficulties that are inherent to interception measurements (Lundberg et al., 1997; Llorens and Gallart, 2000), but it is also generally considered a minor flux, although previous studies tell us that interception can amount to 10–50% of the precipitation depending on the vegetation type (Klaassen et al., 1998). Even then, these studies mostly refer to canopy interception only. If forest floor interception is taken into account as well, the percentage is substantially higher. Furthermore, it is often stated that interception is particularly not important for the generation of floods. This is not true. Interception strongly influences the antecedent soil moisture conditions, which are very important for the generation of floods (Roberts and Klingeman, 1970). Still, interception is regularly (partly) disregarded in hydrological models, or taken as a fixed percentage of the precipitation. As a result, after model calibration, interception is generally compensated by other processes such as transpiration, soil evaporation, or even recharge (Savenije, 2004). Zhang and Savenije (2005) showed that the hydrograph at the outlet of the Geer basin in Belgium improved significantly when interception was included in a rainfall–runoff model using the representative elementary watershed (REW) approach. Both the Nash–Sutcliffe efficiency and the percentage bias improved. They also showed that, in calibration, the soil moisture storage capacity compensated for the neglect of the interception process. Keim et al. (2006a) investigated the effects of (canopy) interception. They looked at the influence on the subsurface stormflow generation and concluded that interception caused a delay in the onset of subsurface stormflow, lowered and delayed stormflow peaks, and decreased total flow and the runoff ratio. They also found that simply reducing the rainfall by a constant factor did not result in a satisfactory peak flow response. Fenicia et al. (2008) looked at the change in the movement of the Pareto front when stepwise new processes were included in a variable model structure. They concluded that when interception was included and especially, when spatially distributed interception was included, the Pareto front moved significantly to the origin. Hence, their conclusion was that interception is an important process and should therefore be included in hydrological models.

2.04.7 Outlook More than 2000 articles have been published on interception studies (source: Scopus and ISI Web of KnowledgeSM) and still

99

new articles are being published. Most of these articles focus on canopy interception and describe in detail the process for different tree species in different climates, resulting in long reference tables as, for example, presented in Table 1 and by Breuer et al. (2003). Although this information is of high value for modeling purposes, it would have been more logical if these tables had also been available for the other types of interception, such as described in Section 2.04.3. Especially, since generally more than one mechanism occur, these mechanisms interact (see, e.g., between canopy and forest floor interception (Gerrits et al., 2009a). It would really be a way forward, if a broader scope was systematically considered in interception studies. Although a more balanced database on interception values will help, it is not the complete solution for hydrological modeling. Often, experimental results are site and time specific. Therefore, it is difficult to upscale literature values on interception for catchment modeling. This problem may be solved by considering the energy balance. If we would know how the available energy is partitioned over the different fluxes and compartments, we would be able to determine interception evaporation as well. However, this would require intensive field experiments where both the energy fluxes and the evaporation processes are measured simultaneously. Remote sensing could provide the necessary spatial and temporal information on energy partitioning. Through a combination of methods, interception could be more adequately incorporated in hydrological models.

References Amiro B (2009) Measuring boreal forest evapotranspiration using the energy balance residual. Journal of Hydrology 366(1–4): 112--118. Aston A (1979) Rainfall interception by eight small trees. Journal of Hydrology 42 (3–4): 383--396. Baird AJ and Wilby RL (eds.) (1999) Eco-Hydrology-Plants and Water in Terrestrial and Aquatic Environments. London: Routledge. Beard JS (1956) Results of the mountain home rainfall interception and infiltration project on black wattle, 1953–1954. Journal of South African Forestry 27: 72--85. Bernard JM (1963) Forest floor moisture capacity of the New Jersey pine barrens. Ecology 44(3): 574--576. Bouten W, Swart PJF, and De Water E (1991) Microwave transmission, a new tool in forest hydrological research. Journal of Hydrology 124(1–2): 119--130. Brechtel HM (1969) Wald und Abfluss-Methoden zur Erforschung der Bedeutung des Waldes fur das Wasserdargebot. Deutsche Gewasserkundliche Mitteilungen 8: 24--31. Breuer L, Eckhardt K, and Frede H-G (2003) Plant parameter values for models in temperate climates. Ecological Modelling 169(2–3): 237--293. Bruijnzeel LA (2005) Tropical montane cloud forest: A unique hydrological case. In: Bonell M and Bruijnzeel LA (eds.) Forests, Water and People in the Humid Tropics, pp. 462--483. Cambridge: Cambridge University Press. Bruijnzeel LA, Eugster W, and Burkard R (2005) Fog as a hydrologic input. In: Anderson MG (ed.) Encyclopedia of Hydrological Sciences, pp. 559--582. Chichester: Wiley. Bruijnzeel LA and Wiersum KF (1987) Rainfall interception by a young Acacia Auriculiformis (a. cunn) plantation forest in West Java, Indonesia: Application of Gash’s analytical model. Hydrological Processes 1: 309--319. Bryant ML, Bhat S, and Jacobs JM (2005) Measurements and modeling of throughfall variability for five forest communities in the southeastern US. Journal of Hydrology 312: 95--108. Calder IR (1986) A stochastic model of rainfall interception. Journal of Hydrology 89: 65--71. Calder IR (1990) Evaporation in the Uplands. Chichester: Wiley. Calder IR and Rosier PTW (1976) The design of large plastic-sheet net-rainfall gauges. Journal of Hydrology 30(4): 403--405.

100

Interception

Calder IR and Wright IR (1986) Gamma ray attenuation studies of interception from Sitka Spruce: Some evidence for an additional transport mechanism. Water Resources Research 22: 409--417. Calder IR, Wright IR, and Murdiyarso D (1986) A study of evaporation from tropical rain forest–West Java. Journal of Hydrology 89: 13--31. Clark OR (1940) Interception of rainfall by prairie grasses, weeds and certain crop plants. Ecological Monographs 10: 243--277. Cuartas LA, Tomasella J, Nobre AD, Hodnett MG, Waterloo MJ, and Mnera JC (2007) Interception water-partitioning dynamics for a pristine rainforest in Central Amazonia: Marked differences between normal and dry years. Agricultural and Forest Meteorology 145(1–2): 69--83. de Groen MM and Savenije HHG (2006) A monthly interception equation based on the statistical characteristics of daily rainfall. Water Resources Research 42: W12417. Edwards WRN (1986) Precision weighing lysimetry for trees, using a simplified taredbalance design. Tree Physiology 1: 127--144. Fenicia F, Savenije HHG, Matgen P, and Pfister L (2008) Understanding catchment behavior through stepwise model concept improvement. Water Resources Research 44: 1--13. Friesen J, Beek C van, Selker J, Savenije HHG, and Giesen N van de (2008) Tree rainfall interception measured by stem compression. Water Resources Research 44: W00D15. Fritschen LJ, Cox L, and Kinerson R (1973) A 28-meter Douglas-fir in a weighing lysimeter. Forest Science 19: 256--261. Gash JHC (1979) An analytical model of rainfall interception by forests. Quarterly Journal of the Royal Meteorological Society 105: 43--55. Gash JHC, Lloyd CR, and Lauchaud G (1995) Estimation sparse forest rainfall interception with an analytical model. Journal of Hydrology 170: 79--86. Gash JHC and Morton AJ (1978) An application of the rutter model to the estimation of the interception loss from Thetford Forest. Journal of Hydrology 38(1–2): 49--58. Gash JHC, Rosier PTW, and Ragab R (2008) A note on estimating urban roof runoff with a forest evaporation model. Hydrological Processes 22(8): 1230--1233. Gash JHC and Stewart JB (1975) The average surface resistance of a pine forest derived from Bowen ratio measurements. Boundary-Layer Meteorology 8: 453--464. Gash JHC, Wright IR, and Lloyd CR (1980) Comparative estimates of interception loss from three coniferous forests in Great Britain. Journal of Hydrology 48(1–2): 89--105. Germer S, Elsenbeer H, and Moraes JM (2006) Throughfall and temporal trends of rainfall redistribution in an open tropical rainforest, South-Western Amazonia (Rondonia, Brazil). Hydrology and Earth System Sciences 10: 383--393. Gerrits AMJ, Pfister L, and Savenije HHG (2009a) Spatial and temporal variability of canopy and forest floor interception in a beech forest. Hydrological Processes, doi: 10.1002/hyp. 7712, published online, 7 june 2010. Gerrits AMJ, Savenije HHG, Hoffmann L, and Pfister L (2007) New technique to measure forest floor interception–an application in a beech forest in Luxembourg. Hydrology and Earth System Sciences 11: 695--701. Gerrits AMJ, Savenije HHG, and Pfister L (2009b) Canopy and forest floor interception and transpiration measurements in a mountainous beech forest in Luxembourg. IAHS Redbook 326: 18--24. Gerrits AMJ, Savenije HHG, Veling EJM, and Pfister L (2009c) Analytical derivation of the Budyko curve based on rainfall characteristics and a simple evaporation model. Water Resources Research 45: W04403. Grimmond CSB and Oke TR (1991) An evapotranspiration-interception model for urban areas. Water Resources Research 27: 1739--1755. Guevara-Escobar A, Gonzalez-Sosa E, Ramos-Salinas M, and Hernandez-Delgado GD (2007) Experimental analysis of drainage and water storage of litter layers. Hydrology and Earth System Sciences 11(5): 1703--1716. Hancock NH and Crowther JM (1979) A technique for the direct measurement of water storage on a forest canopy. Journal of Hydrology 41: 105--122. Haynes JL (1940) Ground rainfall under vegetation canopy of crops. Journal of the American Society of Agronomy 32: 176--184. Helvey JD (1964) Rainfall interception by hardwood forest litter in the southern Appalachians U.S. Forest Service Research Paper SE, vol. 8, pp. 1--8. Asherille, NC: Department of Agriculture, Forest Science, Southeastern Forest Experiment station. Helvey JD and Patric JH (1965) Canopy and litter interception of rainfall by Hardwoods of Eastern United States. Water Resources Research 1(2): 193--206. Herbst M, Rosier PT, McNeil DD, Harding RJ, and Gowing DJ (2008) Seasonal variability of interception evaporation from the canopy of a mixed deciduous forest. Agricultural and Forest Meteorology 148(11): 1655--1667.

Hildebrandt A, Al Aufi M, Amerjeed M, Shammas M, and Eltahir EAB (2007) Ecohydrology of a seasonal cloud forest in Dhofar: 1. Field experiment. Water Resources Research 43: W10411. Ho¨lscher D, Ko¨hler L, Dijk AIJM van, and Bruijnzeel LAS (2004) The importance of epiphytes to total rainfall interception by a tropical montane rain forest in Costa Rica. Journal of Hydrology 292(1–4): 308--322. Hoover MD and Lunt HA (1952) A key for the classification of forest humus types. Soil Science Society Proceedings 16: 368--371. Ho¨rmann G, Branding A, Clemen T, Herbst M, Hinrichs A, and Thamm F (1996) Calculation and simulation of wind controlled canopy interception of a beech forest in Northern Germany. Agricultural and Forest Meteorology 79(3): 131--148. Horton RE (1919) Rainfall interception. Monthly Weather Review 47(9): 603--623. Huang YS, Chen SS, and Lin TP (2005) Continuous monitoring of water loading of trees and canopy rainfall interception using the strain gauge method. Journal of Hydrology 311: 1--7. Hursh CR and Pereira HC (1953) Field moisture balance in the Shimba Hills, Kenya. East African Agricultural Journal 18: 139--148. Hutley LB, Doley D, Yates DJ, and Boonsaner A (1997) Water balance of an Australian subtropical rainforest at altitude: The ecological and physiological significance of intercepted cloud and fog. Australian Journal of Botany 45: 311--329. Keim R, Skaugset A, Link T, and Iroum A (2004) A stochastic model of throughfall for extreme events. Hydrology and Earth System Sciences 8(1): 23--34. Keim RF, Meerveld HJT van, and McDonnell JJ (2006a) A virtual experiment on the effects of evaporation and intensity smoothing by canopy interception on subsurface stormflow generation. Journal of Hydrology 327: 352--364. Keim RF, Skaugset AE, and Weiler M (2005) Temporal persistence of spatial patterns in throughfall. Journal of Hydrology 314: 263--274. Keim RF, Skaugset AE, and Weiler M (2006b) Storage of water on vergetation under simulated rainfall of varying intensity. Advances in Water Resources 29: 974--986. Kittredge J (ed.) (1948) Forest Influences. New York: McGraw-Hill. Klaassen W, Bosveld F, and de Water E (1998) Water storage and evaporation as constituents of rainfall interception. Journal of Hydrology 212–213: 36--50. Klaassen W, Lankreijer HJM, and Veen AWL (1996) Rainfall interception near a forest edge. Journal of Hydrology 185(1–4): 349--361. Li XY, Gong JD, Gao QZ, and Wei XH (2000) Rainfall interception loss by pebble mulch in the semi arid region of China. Journal of Hydrology 228: 165--173. Llorens P and Gallart F (2000) A simplified method for forest water storage capacity measurement. Journal of Hydrology 240: 131--144. Lloyd CR and Marques ADO (1988) Spatial variability of throughfall and stemflow measurements in amazonian rainforest. Agricultural and Forest Meteorology 42(1): 63--73. Lundberg A, Eriksson M, Halldin S, Kellner E, and Seibert J (1997) New approach to the measurement of interception evaporation. Journal of Atmospheric and Oceanic Technology 14: 1023--1035. Manfroi OJ, Kuraji K, Suzuki M, et al. (2006) Comparison of conventionally observed interception evaporation in a 100-m2 subplot with that estimated in a 4-ha area of the same Bornean Lowland tropical forest. Journal of Hydrology 329(1–2): 329--349. Miller HD (1966) Transport of intercepted snow from trees during snowstorms US Forest Service–Research Paper, vol. 33, pp. 1--30. Berkeley, CA: US department of Agriculture, Forest Service, Pacific Southwest Forest & Range Experiment Station. Miller JD, Anderson HA, Ferrier RC, and Walker TAB (1990) Comparison of the hydrological budgets and detailed hydrological responses in two forested catchments. Forestry 63(3): 251--269. Muzylo A, Llorens P, Valente F, Keizer J, Domingo F, and Gash J (2009) A review of rainfall interception modelling. Journal of Hydrology 370(1–4): 191--206. Nakayoshi M, Moriwaki R, Kawai T, and Kanda M (2009) Experimental study on rainfall interception over an outdoor urban-scale model. Water Resources Research 45: W04415. Navar J, Charles F, and Jurado E (1999) Spatial variations of interception loss components by Tamaulipan thornscrub in Northeastern Mexico. Forest Ecology and Management 24: 231--239. Oke TR (1982) The energetic basis of the urban heat island. Quarterly Journal of the Royal Meteorological Society 108(455): 1--24. Pathak PC, Pandey AN, and Singh JS (1985) Apportionment of rainfall in central Himalayan forests (India). Journal of Hydrology 76: 319--332. Penman HL (1948) Natural evaporation from open water, bare soil and grass. Proceedings of the Royal Society of London 193: 120--146. Pitman JI (1989) Rainfall interception by bracken litter–relationship between biomass, storage and drainage rate. Journal of Hydrology 111: 281--291. Putuhena W and Cordery I (1996) Estimation of interception capacity of the forest floor. Journal of Hydrology 180: 283--299.

Interception Pypker TG, Unsworth MH, and Bond BJ (2006) The role of epiphytes in rainfall interception by forests in the Pacific Northwest. I. Laboratory measurements of water storage. Canadian Journal of Forest Research 36: 808--818. Ragab R, Bromley J, Rosier P, Cooper JD, and Gash JHC (2003) Experimental study of water fluxes in a residential area: 1. Rainfall, roof runoff and evaporation: The effect of slope and aspect. Hydrological Processes 17(12): 2409--2422. Roberts MC and Klingeman PC (1970) The influence of landform and precipitation parameters on flood hydrograph. Journal of Hydrology 11: 393--411. Rowe L (1983) Rainfall interception by an evergreen beech forest, Nelson, New Zealand. Journal of Hydrology 66(1–4): 143--158. Rutter AJ, Kershaw KA, Robins PC, and Morton AJ (1971) A predictive model of rainfall interception in forests. I. Derivation of the model and comparison with observations in a plantation of Corsican pine. Agricultural Meteorology 9: 367--384. Rutter AJ, Morton AJ, and Robins PC (1975) A predictive model of rainfall interception in forests. II. Generalization of the model and comparison with observations in some coniferous and hardwood stands. Journal of Applied Ecology 12: 367--380. Sato Y, Kumagai T, Kume A, Otsuki K, and Ogawa S (2004) Experimental analysis of moisture dynamics of litter layers – the effect of rainfall conditions and leaf shapes. Hydrological Processes 18: 3007--3018. Savenije HHG (1997) Determination of evaporation from a catchment water balance at a monthly time scale. Hydrology and Earth System Sciences 1: 93--100. Savenije HHG (2004) The importance of interception and why we should delete the term evapotranspiration from our vocabulary. Hydrological Processes 18: 1507--1511. Savenije HHG (2005) Interception. In: Lehr JH and Keeley J (eds.) Water Encyclopedia: Surface and Agricultural Water. Hoboken, NJ: Wiley Publishers. Schaap MG and Bouten W (1997) Forest floor evaporation in a dense Douglas fir stand. Journal of Hydrology 193: 97--113. Shuttleworth WJ, Gash JHC, Lloyd CR, Moore CJ, Roberts JM, et al. (1984) Eddy correlation measurements of energy partition for Amazonian forest. Quarterly Journal of the Royal Meteorological Society 110: 1143--1162. Stewart JB (1977) Evaporation from the wet canopy of a pine forest. Water Resources Research 13(6): 915--921.

101

Storck P, Lettenmaier DP, and Bolton SM (2002) Measurement of snow interception and canopy effects on snow accumulation and melt in a mountainous maritime climate, Oregon, United States. Water Resources Research 38: 1223. Thamm F and Widmoser P (1995) Zur hydrologischen Bedeutung der organischen Auflage im Wald: Untersuchungsmethoden und erste Ergebnisse. Zeitschrift fur Pflanzenernahrung und Bodenkunde 158: 287--292. Thurow TL, Blackburn WH, Warren SD, and Taylor CA Jr. (1987) Rainfall interception by midgrass, shortgrass, and live oak mottes. Journal of Range Management 40(5): 455--460. Toba T and Ohta T (2005) An observational study of the factors that influence interception loss in boreal and temperate forests. Journal of Hydrology 313: 208--220. Tobo´n-Marin C, Bouten IW, and Dekker S (2000) Gross rainfall and its partitioning into throughfall, stemflow and evaportation of intercepted water in four forest ecosystems in Western Amazonia. Journal of Hydrology 237: 40--57. Valente F, David JS, and Gash JHC (1997) Modelling interception loss for two sparse eucalypt and pine forest in central Portugal using reformulated Rutter and Gash analytical models. Journal of Hydrology 190: 141--162. Viville D, Biron P, Granier A, Dambrine E, and Probst A (1993) Interception in a mountainous declining spruce stand in the Strengbach catchment (Vosges, France). Journal of Hydrology 144: 273--282. Walsh RPD and Voigt PJ (1977) Vegetation litter: An underestimated variable in hydrology and geomorphology. Journal of Biogeography 4: 253--274. Wang A, Diao Y, Pei T, Jin C, and Zhu J (2007) A semi-theoretical model of canopy rainfall interception for a broad-leaved tree. Hydrological Processes 21(18): 2458--2463. Ward AD and Trimble SW (2004) Environmental hydrology, 2nd edn. Boca Raton, FL: CRC Press. Zhang GP and Savenije HHG (2005) Rainfall-runoff modelling in a catchment with a complex groundwater flow system: Application of the Representative Elementary Watershed (REW) approach. Hydrology and Earth System Sciences 9: 243--261. Ziegler AD, Giambelluca TW, Nullet MA, et al. (2009) Throughfall in an evergreendominated forest stand in Northern Thailand: Comparison of mobile and stationary methods. Agricultural and Forest Meteorology 149(2): 373--384. Zinke PJ (1967) Forest interception studies in the United States. In: Sopper WE and Lull HW (eds.) Forest Hydrology, pp. 137--161. Oxford: Pergamon.

2.05 Infiltration and Unsaturated Zone JW Hopmans, University of California, Davis, CA, USA & 2011 Elsevier B.V. All rights reserved.

2.05.1 Introduction 2.05.2 Soil Properties and Unsaturated Water Flow 2.05.2.1 Soil Water Retention 2.05.2.2 Unsaturated Hydraulic Conductivity 2.05.2.3 Modeling of Unsaturated Water Flow and Transport 2.05.2.4 Infiltration Processes 2.05.3 Infiltration Equations 2.05.3.1 Philip Infiltration Equation 2.05.3.2 Parlange et al. Model 2.05.3.3 Swartzendruber Model 2.05.3.4 Empirical Infiltration Equations 2.05.4 Measurements 2.05.4.1 Infiltration 2.05.4.2 Unsaturated Water Flow 2.05.5 Scaling and Spatial Variability Considerations 2.05.6 Summary and Conclusions Acknowledgments References

2.05.1 Introduction As soils make up the upper part of the unsaturated zone, they are subjected to fluctuations in water and chemical content by infiltration and leaching, water uptake by plant roots, and evaporation from the soil surface. It is the most dynamic region of the subsurface, as changes occur at increasingly smaller time and spatial scales when moving from the groundwater toward the soil surface. Environmental scientists are becoming increasingly aware that soils make up a critically important component of the earth’s biosphere, because of their food production and ecological functions, and the soil’s important role in controlling water quality. For example, prevention or remediation of soil and groundwater contamination starts with proper management of the unsaturated zone. Water entry into the soil by infiltration is among the most important soil hydrological processes, as it controls the partitioning between runoff and soil water storage. Runoff water determines surface water quantity and quality, whereas infiltrated water determines plant available water, evapotranspiration, groundwater recharge, and groundwater quality. Also through exfiltration, infiltrated water affects water quality in waterways and associated riparian zones. Despite its relevance and our reliable physical understanding of infiltration, we have generally many difficulties predicting infiltration at any scale. Mostly, this is so because the infiltration rate is a time-varying parameter of which its magnitude is largely controlled by spatially variable soil properties, in both vertical and horizontal directions of a hydrologic basin. Moreover, infiltration rate and runoff are affected by vegetation cover, as it protects the soil surface from the energy impacts of falling raindrops or intercepting rainfall, serving as temporary water storage. The kinetic energy of rainfall causes soil degradation, leading to soil surface sealing and decreasing infiltration.

103 103 104 105 105 106 108 108 109 109 110 110 110 111 111 112 113 113

Historically, solutions to infiltration problems have been presented by way of analytical solutions or empirically. Analytical solutions provide values of infiltration rate or cumulative infiltration as a function of time, making simplifying assumptions of soil depth variations of water content, before and during infiltration. Instead, we now often use powerful computers to conduct numerical simulations of unsaturated water flow to solve for water content and water fluxes throughout the unsaturated soil domain in a single vertical direction or in multiple spatial dimensions, allowing complex initial and boundary conditions. However, although the modeling of multidimensional unsaturated water flow is extremely useful for many vadose zone applications, it does not necessarily improve the soil surface infiltration rate prediction, in light of the large uncertainty of the soil physical properties and initial and boundary conditions that control infiltration. In contrast, empirical infiltration models serve primarily to fit model parameters to measured infiltration, but have limited power as a predictive tool.

2.05.2 Soil Properties and Unsaturated Water Flow The soil consists of a complex arrangement of mostly connected solid, liquid, and gaseous phases, with the spatial distribution and geometrical arrangement of each phase, and the partitioning of solutes between phases, controlled by physical, chemical, and biological processes. The unsaturated zone is bounded by the soil surface and merges with the groundwater in the capillary fringe. Water in the unsaturated soil matrix is held by capillary and adsorptive forces. Water is a primary factor leading to soil formation from the weathering of parent material such as rock or transported deposits, with additional factors of climate, vegetation, topography, and parent material determining soil physical properties.

103

104

Infiltration and Unsaturated Zone

Defining the soil’s dry bulk density by rb (M L3), soil porosity, e (L3 L3), is defined by

e¼1

rb rs

ð1Þ

with rs being the soil’s particle density (M L3). Equation (1) shows that soil porosity has lower values as bulk soil density is increased such as by compaction. Unsaturated water flow is largely controlled by the physical arrangement of soil particles in relation to the water and air phases within the soil’s pore space, as determined by pore-size distribution and water-filled porosity or volumetric water content, y (L3 water/L3 bulk soil). The volumetric water content y expresses the volume of water present per unit bulk soil as

y¼

wrb rw

ð2Þ

where w is defined as the mass water content (M of water/M dry soil) and we take rw ¼ 1000 kg m3. Alternatively, the soil water content can be described by the degree of saturation S (–) and the equivalent depth of stored water De (L), or

S¼

y e

and

De ¼ yDsoil

ð3Þ

so that y can also be defined by the equivalent depth of water per unit depth of bulk soil, Dsoil (L). The volumetric water content ranges between 0.0 (dry soil) and the saturated water content, ys, which is equal to the porosity if the soil were completely saturated. The degree of saturation varies between 0.0 (completely dry) and 1.0 (all pores completely waterfilled). When considering water flow, the porosity term is replaced by the saturated water content, ys, and both terms in Equation (3) are corrected by subtracting the so-called residual water content, yr (soil water content for which water is considered immobile), so that the effective saturation, Se, is defined as

Se ¼

y  yr ys  yr

weight of water, leading to soil water potential expressed by the equivalent height of a column of water (L). The resulting pressure head equivalent of the combined adsorptive and capillary forces in soils is defined as the matric pressure head, h. When expressed relative to the reference potential of free water, the water potential in unsaturated soils is negative (the soil water potential is less than the water potential of water at atmospheric pressure). Hence, the matric potential decreases or is more negative as the soil water content decreases. In using head units for water potential, the total water potential (H) is defined as the sum of matric potential (h), gravitational potential (z), hydrostatic pressure potential (p), and osmotic potential (p). For most hydrological applications, the contribution of the osmotic potential can be ignored, so that for unsaturated water flow (p ¼ 0) the total soil water potential can be written as

H ¼hþz

ð5Þ

The measurement of the soil water matric potential in situ is difficult and is usually done by tensiometers in the range of matric head values larger (less negative) than  6.0 m. A tensiometer consists of a porous cup, usually ceramic, connected to a water-filled tube (Young and Sisson, 2002). The suction forces of the unsaturated soil draw water from the tensiometer into the soil until the water pressure inside the cup (at pressure smaller than atmospheric pressure) is equal to the pressure equivalent of the soil water matric potential just outside the cup. The water pressure in the tensiometer is usually measured by a vacuum gauge or pressure transducer. Other devices that are used to indirectly measure the soil water matric potential include buried porous units (Scanlon et al., 2002), for which either the electrical resistance or the thermal conductivity is measured in situ, after coming into hydraulic equilibrium with the surrounding soil (h in sensor and soil are equal). Although widely used, these types of sensors require laboratory calibration, before field installation.

2.05.2.1 Soil Water Retention ð4Þ

In addition to the traditional thermogravimetric method to determine soil water content, many other measurement techniques are available, including neutron thermalization, electrical conductivity, dielectric, and heat pulse methods. A recent review on soil moisture measurement methods was presented by Robinson et al. (2008), focusing on measurement constraints between the many available methods across spatial scales. In soils, the driving force for water to flow is the gradient in total water potential. The total potential of bulk soil water can be written as the sum of all possible component potentials, so that the total water potential (ct) is equal to the sum of osmotic, matric, gravitational, and hydrostatic pressure potential. Whereas in physical chemistry the chemical potential of water is usually defined on a molar or mass basis, soil water potential is usually expressed with respect to a unit volume of water, thereby attaining units of pressure (Pa); or per unit

The soil water retention function determines the relation between the volume of water retained by the soil, expressed by y, and the governing soil matric, or suction forces (Dane and Hopmans, 2002). These suction forces are typically expressed by the soil water matric head (strictly negative) or soil suction (strictly positive). These suction forces increase as the size of the water-filled pores decreases, as may occur by drainage, water uptake by plant roots, or soil evaporation. Also known as the soil water release or soil water characteristic function, this soil hydraulic property describes the increase of y and the size of the water-filled pores with an increase in matric potential, as occurs by infiltration. Since the matric forces are controlled by pore-size distribution, specific surface area, and type of physico-chemical interactions at the solid–liquid interfaces, the soil water retention curve is very soil specific and highly nonlinear. It provides an estimate of the soil’s capacity to hold water after irrigation and free drainage (field capacity), minimum soil water content available to the plant (wilting point), and root zone water availability for plants.

Infiltration and Unsaturated Zone

The soil water retention curve exhibits hysteresis, that is, the y value is different for wetting (infiltration) and drying (drainage). By way of the unique relationship between soil water matric head and the radius of curvature of the air–water interface in the soil pores, and using the analogy between capillary tubes and the irregular pores in porous media, a relationship can be derived between soil water matric head (h) and effective pore radius, re, or

rgh ¼

2s cos a re

ð6Þ

where s and a are defined as the surface tension and wetting angle of wetting fluid with soil particle surface (typically values for s and a are 0.072 N m1 and 01, respectively ), r is the density of water, and g is the acceleration due to gravity (9.8 m s2). Because of capillary equation, the effective pore-size distribution can be determined from the soil water retention curve in the region where matric forces dominate. Laboratory and field techniques to measure the soil water retention curve, and functional models to fit the measured soil water retention data, such as the van Genuchten (1980) and Brooks and Corey (1964) models, are described by Kosugi et al. (2002). Alternatively, knowledge of the particle size distribution may provide information on the shape of the soil water retention curve, as presented by Nasta et al. (2009). An example of measured and fitted soil water retention data for two different soils is presented in Figure 1 (Tuli and Hopmans, 2004).

2.05.2.2 Unsaturated Hydraulic Conductivity The relation between the soil’s unsaturated hydraulic conductivity, K, and volumetric water content, y, is the second essential fundamental soil hydraulic property needed to

Soil matric potential head (cm)

describe unsaturated soil water flow. K is a function of the water and soil matrix properties, and controls water infiltration and drainage rates, and is strongly affected by water content and possibly by hysteresis. It is defined by the Darcy– Buckingham equation, which relates the soil water flux density to the total driving force for flow, with K being the proportionality factor. Except for special circumstances, the total driving force for water flow in soils is determined by the sum of the matric and gravitational forces, expressed by the total water potential head gradient, DH/L (L L1), where DH denotes the change in total water potential head over the distance L. For vertical flow, the application of Darcy’s law yields the magnitude of water flux from

 q ¼ KðyÞ

Measured Oso Flaco sand Optimized Oso Flaco sand Measured Columbia sandy loam Optimized Columbia sandy loam

 dh þ1 dz

ð7Þ

where q is the Darcy water flux density (L3 water L2 soil surface T1) and z defines the vertical position (z40, upwards, L). A soil system is usually defined by the bulk soil, without consideration of the size and geometry of the individual flow channels or pores. Therefore, the hydraulic conductivity (K) describes the ability of the bulk soil to transmit water, and is expressed by volume of water flowing per unit area of bulk soil per unit time (L T1). Functional models for unsaturated hydraulic conductivity are based on pore-size distribution, pore geometry, and connectivity, and require integration of soil water retention functions to obtain analytical expressions for the unsaturated hydraulic conductivity. The resulting expressions relate the relative hydraulic conductivity, Kr, defined as the ratio of the unsaturated hydraulic conductivity, K, and the saturated hydraulic conductivity, Ks, to the effective saturation, Se, and can be written in the following generalized form (Kosugi et al., 2002):

2Z

10 000

105

Se

3g

jhj Z dSe 7 6 7 6 Kr ðSe Þ ¼ Sle 6 Z0 S 7 5 4 Z jhj dSe

ð8Þ

0

1000

100

10

1 0

0.2

0.4

0.6

0.8

Sew Figure 1 Measured (symbols) and fitted (lines) soil water retention data. From Tuli AM and JW Hopmans (2004) Effect of degree of saturation on transport coefficients in disturbed soils. European Journal of Soil Science 55: 147–164.

where l and Z are parameters related to the tortuosity and connectivity of the soil pores, and the value of the parameter g is determined by the method of evaluating the effective pore radii. For values of l ¼ 0.5, Z ¼ 1.0, and g ¼ 2.0, Equation (8) reduces to the so-called Mualem (1976) model, that is routinely combined with the van Genuchten (1980) soil water retention model to yield a closed-form expression for the unsaturated hydraulic conductivity function. The moisture dependency is highly nonlinear, with a change in K of five or more orders of magnitude across field-representative changes in unsaturated soil water content. Methods to measure the saturation dependency of the hydraulic conductivity are involved and time consuming. A variety of methods are described in Dane and Topp (2002) and Dirksen (2001). Measurement errors are generally large due to (1) the difficulty of flow measurements in the low water content range and (2) the dominant effect of large pores (macropores), cracks, and fissures in the high water content range. An example of the unsaturated hydraulic conductivity for water, relative to its

Infiltration and Unsaturated Zone 1.0

1.0

0.8

0.8

0.6

0.6 K ra

K rw

106

0.4

0.4

0.2

0.2

0 (a)

0 0

0.2

0.4

0.6

0.8

1.0

(b)

0

0.2

0.4

0.6

0.8

1.0

Volumetric water content Oso Flaco fine sand

Columbia sandy loam

Measured Krw

Measured Krw

Measured Kra

Measured Kra

Measured Drg

Measured Drg

Measured ECra

Measured ECra

Figure 2 (a) Measured relative hydraulic conductivity for water (Krw) and (b) air conductivity (Kra) as a function of degree of water (Sew) and air (Sea) saturation. From Tuli AM and JW Hopmans (2004) Effect of degree of saturation on transport coefficients in disturbed soils. European Journal of Soil Science 55: 147–164.

saturated values (Krw), is presented in Figure 2(a), for the same two soils as in Figure 1. We note that ys in the vadose zone is typically about 85% of the porosity, so that a saturated soil (e.g., as the result of ponded infiltration) is really a satiated soil due to entrapped air, with a saturated hydraulic conductivity that is significantly smaller than the true Ks. The unsaturated hydraulic conductivity is related to the intrinsic soil permeability, k (L2), by

K¼

rgk m

ð9Þ

where m denotes the dynamic viscosity of water (F T L2). The usage of permeability instead of conductivity allows application of the flow equation to liquids other than water with different density and viscosity values. In addition to unsaturated hydraulic conductivity, Figure 2 also includes data for the saturation dependency of the relative air conductivity (Kra), as might be important for water infiltration in soil, when the soil gas phase is trapped and increasing in pressure, so that water infiltration is partly controlled by soil air permeability (Latifi et al., 1994).

2.05.2.3 Modeling of Unsaturated Water Flow and Transport Numerous studies have been published addressing different issues in the numerical modeling of unsaturated water flow using the Richards’ equation. In short, the dynamic water flow equation is a combination of the Darcy expression and a mass balance formulation. Using various solution algorithms, the soil region of interest is discretized in finite-size elements, i, that can be one, two, or three dimensional, to solve for temporal changes in h, y, or water flux, q, for each element or voxel i at any time t.

Most multidimensional soil water flow models use a finiteelement, Picard time-iterative numerical scheme (Sˇimunek et al., 2008) to solve the Richards equation. For isotropic conditions and one-dimensional vertical flow, the general water flow equation simplifies to

   qy q qh ¼ KðhÞ þ 1  Sðz; tÞ qt qz qz

ð10Þ

where S (L3 L3 T1) is the sink term, accounting for root water uptake. Boundary and initial conditions must be included to allow for specified soil water potentials or fluxes at all boundaries of the soil domain. Richards’ equation is a highly nonlinear partial differential equation, and is therefore extremely difficult to solve numerically because of the largely nonlinear dependencies of both water content and unsaturated hydraulic conductivity on the soil water matric head. Both the soil water retention and unsaturated hydraulic conductivity relationships must be known a priori to solve the unsaturated water flow equation. Specifically, it will need the slope of the soil water retention curve, or water capacity C(h), defined as CðhÞ ¼ dy=dh. As dissolved solutes move through the soils with the water, various physical, chemical, and biological soil properties control their fate. In addition to diffusion and dispersion, fate and transport of chemicals in the subsurface are influenced by sorption to the solid phase and biological transformations. Both diffusion and dispersion of the transported chemical are a function of pore-size distribution and water content. Mechanical or hydrodynamic dispersion is the result of water mixing within and between pores as a result of variations in pore water velocity. Increasing dispersivity values cause greater spreading of the chemical, thereby decreasing peak

Infiltration and Unsaturated Zone

2.05.2.4 Infiltration Processes For one-dimensional infiltration, the infiltration rate (L T1), i(t), can be defined by Equation (7) at the soil surface (subscript surf), or

 iðtÞ ¼ KðyÞ

 qh þ1 qz surf

ð11aÞ

Cumulative infiltration I(t), expressed as volume of water per unit soil surface area (L), is defined by

IðtÞ ¼

Z

t

iðtÞdt

ð11bÞ

0

Analytical solutions of infiltration generally assume that the wetted soil profile is homogeneous in texture with uniform initial water content. They also make distinction between ponded (h4 0 or p) and nonponded soil surface (unsaturated, ho0) infiltration. The infiltration capacity of the soil is defined by ic(t), the maximum rate at which a soil can absorb water for ponded soil surface conditions. Its maximal value is at time zero, and decreases with time to its minimum value approaching the soil’s saturated hydraulic conductivity, Ks, as the total water potential gradient decreases, and tends to unity, with the downward moving wetting front. As defined by Equation (11b), the soil’s cumulative infiltration capacity, Ic(t), is defined by the area under the capacity curve. It represents the maximum amount of water that the soil can absorb at any time. Typically, at the onset of infiltration (t ¼ 0), the rainfall rate, r(t), will be lower than ic(t), so that the infiltration rate is equal to the rainfall rate (i.e., r(t)oic(t) for hsurfo0). If at any point in time, the rainfall rate becomes larger than the infiltration capacity, ponding will occur (hsurf40), resulting in runoff. The time at which ponding occurs is defined as tp (time to ponding). Thus, the actual infiltration rate will depend on the rainfall rate and its temporal changes. This makes prediction of infiltration and runoff much more difficult for realistic time-variable rainfall patterns. Therefore, infiltration rate prediction is often described as a function of the cumulative infiltration, I, or i(I), independent of the time domain, and with i(I) curves that are independent of rainfall rate (Skaggs, 1982). An example of such a

time-invariant approach is the IDA or infiltrability-depth approximation (Smith et al., 2002). The main IDA assumption is that time periods between small rainfall events are sufficiently small so that soil water redistribution and evaporation between events do not affect infiltration rate. IDA implies that the infiltration rate at any given time depends only on the cumulative infiltration volume, regardless of the previous rainfall history. Following this approach, tp is defined as the time during a storm event when I becomes equal to Ic(tp), or

R¼

Z

tp

rðtÞdt ¼ Ic ðtp Þ

t¼0

whereas i(t) ¼ r(t) for totp. The time invariance of i(I) holds true also when a layered/sealed soil profile is considered (Mualem and Assouline, 1989). For illustration purposes, we present a hypothetical storm event with time-varying r(t) in Figure 3 (from Hopmans et al., 2007) in combination with an assumed soil-specific infiltration capacity curve, ic(t). At what time will ponding occur? It will not be at t ¼ 7, when r(t) exceeds ic for the first time. In order to approximate tp, we plot both Ic and R for the storm in Figure 4(a), as a function of time and determine tp as the time at which both curves intersect (tp ¼ 13, for R ¼ Ic ¼ 110), since at that time, the cumulative infiltration of the storm is identical to the soil’s infiltration capacity. The final corresponding i(I) for this soil and storm event is presented in Figure 4(b), showing that the soil infiltration rate is equal to r(t) until I ¼ R(t) ¼ Ic(tp) ¼ 110, after which the infiltration rate is soilcontrolled and determined by Ic(t). More accurate approximations to the time-invariant approach can be found in Sivapalan and Milly (1989) and Brutsaert (2005), using the time compression or time condensation approximation that more accurately estimates infiltration prior to surface ponding. In addition to whether the soil is ponded or not, solutions of infiltration distinguish between cases with and without gravity effects, as different analytical solutions apply. As Equation (11a) shows, infiltration rate i(t) is determined by both the soil water matric potential gradient, dh=dz, and gravity. However, at the early stage of infiltration into a relatively dry soil, infiltration rate is dominated by the matric potential gradient so that the gravity effects on infiltration can 15 Infiltration rate, i

concentration. Sorbed chemicals move through the vadose zone slower than noninteracting chemicals, and the degree of sorption will largely depend on mineral type, specific surface area of the solid phase, and organic matter fraction. In addition, biogeochemical processes and radioactive decay affect contaminant concentration, such as by cation exchange, mineral precipitation and dissolution, complexation, oxidation–reduction reactions, and by microbial biodegradation and transformations. However, all these mechanisms depend on soil environmental conditions, such as temperature, pH, water saturation, and redox status, and their soil spatial variations. The solute transport equation is generally referred to as the convection–dispersion equation (CDE), and includes the relevant transport mechanisms to simulate and predict temporal changes in soil solute concentration within the simulation domain (Sˇimunek et al., 2008).

107

r (t ) 10

5 ic 1 0

5

10

15

20

Time, t Figure 3 Hypothetical rainfall event, r(t), and soil infiltration capacity, ic(t). The rainfall event starts at t ¼ 0. From Hopmans JW, Assouline S, and Parlange J-Y (2007) Soil infiltration. In: Delleur JW (ed.) The Handbook of Groundwater Engineering, pp. 7.1–7.18. Boca Raton, FL: CRC Press.

108

Infiltration and Unsaturated Zone

Cumulative infiltration

160

Rainfall R (t ) Ic (t )

120 80 40

tp

0 0

5

10

20

Time

(a)

Infiltration rate, i

15

16 14 12 10 8 6 4 2 0

t p = 13

 I ¼ Ic ¼ K1 t þ ðhsurf  hf ÞDy ln 1 þ

ic (I )

0

50

100

150

Cumulative infiltration, I

(b)

Figure 4 (a) Cumulative infiltration corresponding with infiltration capacity, Ic(t), and cumulative rainfall, R(t) and (b) actual infiltration rate vs. I. Ponding starts only after tp ¼ 13, or I ¼ 110. From Hopmans et al. (2007).

be ignored. Gravity becomes important in the later stages of infiltration, when the wetting front has moved further down. For gravity-free drainage, a simple analytical solution can be found, after transforming Equation (11a) into a y-based form by defining the diffusivity DðyÞ ¼ KðyÞdh=dy, so that

iðtÞ ¼ DðyÞ

  qy q z surf

ð12Þ

Using the Boltzmann transformation for a constant head boundary condition (Bruce and Klute, 1956), and defining the scaling variable j ¼ z=t 1=2, combination of Equations (10) without gravity and sink term and (12) resulted in a unique solution of y as a function of j, from which the wetting profile can be computed for any time t (Kirkham and Powers, 1972). Defining y1 and y0 as the surface water content during infiltration and the initial uniform profile water content, respectively, cumulative infiltration, I, is computed from

I¼

Z

y1

z dy ¼ t 1=2

y0

Z

y1

j dy

ð13aÞ

y0

and results in the simple infiltration equation I ¼ St1/2, where the sorptivity S (L T1/2) is defined as

Sðy1 Þ ¼

Z

y1

j dy

horizontal infiltration, I is a linear function of t1/2, with S being defined as the slope of this line. Hence, for saturated soil conditions where y1 ¼ ys, the infiltration capacity is computed from ic(t) ¼ 12St1/2. Incidentally, this also leads to Ic ¼ S2/2ic. A relatively simple analytical solution without and with gravity effects was suggested by Green and Ampt (1911) for a ponded soil surface, with ysurf ¼ y1. The assumptions are that the wetting front can be approximated as a step function with a constant effective water potential, hf, at the wetting front, a wetting zone hydraulic conductivity of K(y1) ¼ K1 ¼ Ks, and a constant soil water profile of Dy ¼ y1  y0 . Using this so-called delta-function assumption of a D(y) with a Dirac-delta function form, both solutions for horizontal and vertical infiltration can be relatively easily obtained (Jury et al., 1991; Haverkamp et al., 2007). Assuming that K0 at the initial water content, y0, is negligible, the Green and Ampt (GA) solution of vertical infiltration for ponded conditions is (h ¼ hsurf40):

ð13bÞ

y0

Equation (13a) states that for gravity-free infiltration during the early times of vertical infiltration, and at all times for

ðhsurf

 I  hf ÞDy

ð14Þ

which can be solved iteratively for I. This simple, yet physically based, solution appears to work best for dry coarse-textured soils. A theoretical expression for the wetting front potential head,R hf, was defined by Mein and Farrell (1974), to yield that hf ¼ h00 Kr ðhÞdh; where the relative conductivity Kr ¼ K(h)/Ks. The so-called S-form of the GA equation can be obtained by comparing the gravity-free solution of GA with the Boltzmann solution, to yield S20 ¼ 2K1 hf Dy:

S21 ðy1 Þ ¼ 2K1 Dyðhsurf  hf Þ ¼ S20 þ 2K1 hsurf Dy

ð15aÞ

so that

  S20 I I ¼ K1 t þ hsurf Dy þ ln 1 þ hsurf Dy þ S20 =2K1 2K1

ð15bÞ

In reality, the wetting front is not a step function, but will consist of a time-dependent transition zone where water content changes from y1 to y0. The shape of this transition zone will be a function of time and is controlled by soil type. The step function assumption is better for uniform coarsetextured soils that have a Dirac-like D(y), for which there is a sharp decline in K with a decrease in water content near saturation. The wetting front is generally much more diffuse for finer-textured soils that have a wide pore-size distribution. By now, it must be clear that infiltration and its temporal changes are a function of many different soil factors. In addition to rainfall intensity and duration and the soil physical factors, such as soil water retention and hydraulic conductivity, infiltration is controlled by the initial water content, surface sealing and crusting, soil layering, and the ionic composition of the infiltrated water (Kutilek and Nielsen, 1994; Assouline, 2004). For example, Vandervaere et al. (1998) applied the GA model to sealed soil profiles, by assuming that the wetting front potential decreases suddenly as it leaves the seal and enters the soil. This results in a discontinuous drop in the infiltration rate. Many relatively simple infiltration equations have been proposed and are successfully used to

Infiltration and Unsaturated Zone

characterize infiltration. This has been achieved despite that these equations apply for homogeneous soils only, in theory.

2.05.3 Infiltration Equations In addition to the solutions in Section 2.05.2, other physically based analytical solutions have been presented, using different assumptions allowing for a closed-form solution. These can potentially be used to predict infiltration from known soil hydraulic properties of homogeneous soils. However, in practice, this is difficult as soil physical characteristics near the soil surface are time dependent because of soil structural changes and their high spatial variability. Alternatively, various empirical infiltration models have been proposed that are very useful for describing measured infiltration data. A parameter sensitivity analysis of many of the presented infiltration models, analyzing the effects of measurement error, was given by Clausnitzer et al. (1998). This section presents the most frequently used infiltration models in both categories.

2.05.3.1 Philip Infiltration Equation Philip (1957a) presented an analytical infinite-series solution to the water-content-based form of Richards’ equation for the case of vertical infiltration:

  qy q qy ¼ DðyÞ þ KðyÞ qt qz qz

ð16Þ

For the boundary condition of hsurf ¼ 0 and y1 ¼ ys , the Philip (1957a) solution converged to the true solution for small and intermediate times, but failed for large times. In this case, an alternative solution was presented (Philip, 1957b). With additional assumptions regarding the physical nature of soil water properties, Philip (1987) proposed joining solutions that are applicable for all times. Philip (1957c) introduced a truncation of the small-time series solution that is a simple two-parameter model equation (PH model):

Ic ¼ At þ St 1=2

ð17aÞ

which should be accurate for all but very large t, and suitable for applied hydrological studies. The sorptivity S depends on several soil physical properties, including initial water content y0, and the hydraulic conductivity and soil water retention functions. S is equal to the expression defined in Equation (13b). Philip (1969) showed that A may take values between 0.38Ks and 0.66Ks. The physical interpretation of A is not straightforward; however, for long times when gravity is dominant and hsurf ¼ 0, one would expect A to be equal to Ks. Differentiation of Equation (17a) yields the infiltration rate, or

ic ¼ 1=2St 0:5 þ A

ð17bÞ

Using (17b) to express t as a function of ic and substituting in Equation (17a) yields I(i), or

I¼

S 2 ði  A=2Þ 2ði  AÞ 2

ð17cÞ

109

For positive pressure heads (hsurf), the correction of Equation (15a) to S can be applied. In many cases, values of S and A are obtained from curve fitting. We note that for gravity-free flow, the pH solution without the gravity term corresponds with the Boltzmann solution for horizontal flow in Equation (13).

2.05.3.2 Parlange et al. Model Parlange et al. (1982) proposed the following universal model (Parlange et al., model, PA model):

  3 2dK1 I exp þ d  1 7 62K1 S2 S2 7 6 t¼ 2 I  ln 5 4 2 d 2K1 ð1  dÞ S 2

ð18aÞ

assuming that K0 is small so that the DK in Parlange et al. (1982) is equal to K1. The value of the parameter d can be chosen to approach various closed-form solutions. For example, Equation (18a) reduces to the GA solution for d equal to zero. Its value is a function of K(y), and is defined by (Parlange et al., 1985):

d¼

1 ys  y0

Z

ys y0

Ks  KðyÞ dy Ks

ð18bÞ

An approximate value of d ¼ 0.85 was suggested by Parlange et al. (1982) for a range of soil types. After taking the time derivative of I, the following i(I)-relationship can be derived (Espinoza, 1999):

  1 2IdK1 i ¼ K1 þ dK1 1  exp S2

ð18cÞ

Because Equation (18) is based on integration of the watercontent-based form of Richards’ equation, its theoretical scope is limited to nonponded conditions. A generalization of Equation (18) to include ponded conditions without affecting the value of S was introduced by Parlange et al. (1985). Haverkamp et al. (1990) presented a modification of their model to include upward water flow by capillary rise. The resulting infiltration model contained six physical parameters, in addition to the interpolation parameter d (Haverkamp et al., 1990). Both the PA and the Haverkamp et al. (1990) model require an iterative procedure to predict I(t). Barry et al. (1995) presented an explicit approximation to the Haverkamp et al. (1990) model, retaining all six physical parameters (BA model):

S 2 þ 2K1 hsurf Dy I ¼ K1 t þ 2DK    6ð2t  Þ 0:5 2t     t þ 1  g  exp 6 þ ð2t  Þ 0:5 3     g 2t 8 2:5 exp  ½1  ð1  gÞ t  þ 3 1 þ t   t þð2g þ t  Þln 1 þ g

ð19aÞ

110

Infiltration and Unsaturated Zone

where

t ¼

Another simple empirical infiltration equation is the Kostiakov (1932) model (KO):

2tðDKÞ 2 ; S 2 þ 2K1 hsurf Dy

g¼

2K1 ðhsurf þ ha ÞDy S2 þ 2K1 hsurf Dy

ð19bÞ

and ha denotes the absolute value of the soil water pressure head at which the air phase becomes discontinuous upon wetting. By defining

B1 ¼ ðhsurf þ ha ÞDy and B2 ¼

2 S 2 þ 2K1 hsurf Dy

ð19cÞ

Equation (19a) can be expressed by only four fitting parameters K0, K1, B1, and B2. The Clausnitzer et al. (1998) study concluded that both the PA and BA models described infiltration equally well; however, the BA model, while most advanced, was not as well suited to serve as a fitting model due to nonuniqueness problems caused by the larger number of fitting parameters.

i ¼ at b

ð22Þ

Clearly, this equation will not fit infiltration data at long times, as it predicts zero infiltration rate as t-N. The value of a should be equal to the infiltration rate at t ¼ 1, and 0obo1. Mezencev (1948) proposed another infiltration model, and modified the KO model by including a linear term with a coefficient b1, so that b1-K1 for t-N provided 0ob3o1 and b240 (ME model):

I ¼ b1 t þ

b2 tð1  b3 Þ 1  b3

ð23Þ

Other models include the Soil Conservation Service (1972) method and the Holtan solution (Kutilek and Nielsen, 1994; Espinoza, 1999).

2.05.3.3 Swartzendruber Model Swartzendruber (1987) proposed an alternative series solution that is applicable and exact for all infiltration times, and also allows for surface ponding. Its starting point is similar to the GA approach; however, its derivation does not require a step function for the wetted soil profile. Its simplified form is a three-parameter infiltration equation (SW model):

I ¼ K1 t þ

S 1  expðA0 t 1=2 Þ A0

ð20Þ

where A0 is a fitting parameter of which its value depends on the surface water content, y1. As A0-0, it reduces to a form of the Philip (1957b) model with K1 as the coefficient of the linear term, and for which dI/dt approaches K1 as t-N. As for the GA model, the S-term can be corrected using Equation (15a) to account for ponded conditions.

2.05.3.4 Empirical Infiltration Equations For most of these types of infiltration equations, the fitting parameters do not have a physical meaning and are evaluated by fitting to experimental data only. However, in many cases, the specific form of the infiltration equation is physically intuitive. For example, the empirical infiltration equation by Horton (1940) is one the most widely used empirical infiltration equations. It considers infiltration as a natural exhaustion process, during which infiltration rate decreases exponentially with time from a finite initial value, ic|t ¼ 0 ¼ (a1 þ a2), to a final value, a1 ¼ K1. Accordingly, cumulative infiltration I (L) is predicted as a function of time t (HO model):

I ¼ a1 t þ

a2 ½1  expða3 tÞ a3

ð21Þ

with the soil parameter a340, representing the decay of infiltration rate with time. In Equation (21), a1 can be associated with the hydraulic conductivity (LT1) of the wetted soil portion, K1, for t-N.

2.05.4 Measurements 2.05.4.1 Infiltration Infiltration measurements can serve various purposes. In addition to characterizing infiltration, for example, to compare infiltration between different soil types, or to quantify macropore flow, it is often measured to estimate the relevant soil hydraulic parameters from the fitting of the infiltration data to a specific physically based infiltration model. This is generally known as inverse modeling. Infiltration is generally measured using one of three different methods: a sprinkler method, a ring infiltration method, or a permeameter method. The sprinkler method is mostly applied to determine time of ponding for different water application rates, whereas the ring infiltrometer method is used when the infiltration capacity is needed. The permeameter method provides a way to measure infiltration across a small range of h-values p0. A general review of all three methods was recently presented by Smettem and Smith (Smith et al., 2002), whereas a comparison of different infiltration devices using seven criteria was presented by Clothier (2001). Rainfall sprinklers or rainfall simulators are also sprinkler infiltrometers, but they are typically used to study runoff and soil erosion (e.g., Morin et al., 1967). They mimic the rainfall characteristics (e.g., kinetic energy) of natural storms, specifically the rainfall rate, rainfall droplet size distribution, and drop velocity. Most of these devices measure infiltration by subtracting runoff from applied water. Using a range of water application rates, infiltration measurements can be used to determine the i(I) curve for a specific soil type, with specific soil hydraulic properties such as Ks or S. Various design parameters for many developed rainfall simulators, specifically nozzle systems, were presented by Peterson and Bubenzer (1986). A portable and inexpensive simulator for infiltration measurements along hillslopes was developed by Battany and Grismer (2000). This low-pressure system used a hypodermic syringe needle system to form uniform droplets at rainfall intensities ranging from 20 to 90 mm h1.

Infiltration and Unsaturated Zone

Ring infiltrometers have historically been used to characterize soil infiltration by determining the infiltration capacity, ic. A ring is carefully inserted in the soil so that water can be ponded over a known area. Since a constant head is required, a constant water level is maintained either by manually adding water and using a measuring stick to maintain a constant depth of ponded water, by using a Mariotte system, or by a valve connected to a float that closes at a predetermined water level. Measurements are usually continued until the infiltration rate is essentially constant. Water seepage around the infiltrometer is prevented by compaction of the soil around and outside of the infiltrometer. Multidimensional water flow under the ring is minimized by pushing the ring deeper into the soil, or by including an outer buffer ring. In the latter case, the soil between the two concentric rings is ponded at the same depth as the inner ring, to minimize lateral flow directed radially outward. The deviation from the assumed one-dimensionality depends on ring insertion depth, ring diameter, measurement time and soil properties such as its hydraulic conductivity, and the presence of restricting soil layers. A sensitivity analysis on diverging flow of infiltrometers was presented by Bouwer (1986) and Wu et al. (1997). Permeameters are generally smaller than infiltrometers and allow easy control of the soil water pressure head at the soil surface. Generally, multidimensionality of flow must be taken into account, using Wooding’s (1968) equation for steady flow (QN, L3 T1) from a shallow, circular surface pond of free water, or

 QN ¼ Ks

pr20

4r0 þ a

 ð24aÞ

The first and second terms in parentheses denote the gravitational and capillary components of infiltration and a denotes the parameter in Gardner’s (1958) unsaturated hydraulic conductivity function:

KðhÞ ¼ Ks expðahÞ

ð24bÞ

In this model of the so-called Gardner soil, the macroscopic capillary length, lc, is equivalent to 1/a. The basic analysis for most permeameter methods relies on Wooding’s solution. An extensive review of the use of permeameters was presented by Clothier (2001), including the tension infiltrometers and disk permeameters, by which the soil water pressure at water entry is controlled by a bubble tower. Their use is relatively simple, and based on analytical solutions of steady-state water flow. The permeameter method is economical in water use and portable. The soil hydraulic properties (S and K), in an inverse way, can be inferred from measurements using (1) both shortand long-time observations, (2) disks with various radii, or (3) using multiple water pressure heads. Transient solutions of infiltration may be preferable, as it allows analysis of shorter infiltration times, so that the method is faster and likely will better satisfy the homogeneous soil assumption. Differences between one- and three-dimensional solutions for transient infiltration were analyzed by Haverkamp et al. (1994), Vandervaere et al. (2000), and Smith et al. (2002) from multidimensional numerical modeling analysis. These effects were reported to be small if gravity effects were included.

111

Nowadays, permeameters are most often applied to estimate the soil’s hydraulic characteristics in an inverse way, by fitting infiltration data to analytical solutions. In many cases, auxiliary water content or matric potential data are required to yield unique solutions.

2.05.4.2 Unsaturated Water Flow Whereas infiltration measures are typically conducted along the soil surface only, measurement of unsaturated water flow requires installation of instruments and sensors below ground, thereby largely complicating measurement procedures and analysis. The simplest expression for unsaturated water flow estimation is the Darcy equation (7), but still requires the measurement of soil water content (y) or soil water matric potential (h) at various soil depths, and knowledge of the unsaturated hydraulic function, K(y), as expressed by Equation (8). Installation of soil moisture or potential sensors requires extreme care, because of issues of soil disturbance, inadequate soil sensor contact, and inherent soil heterogeneities. In addition, it is not always straightforward to determine installation depth of sensors, as it will depend on a priori knowledge of soil horizon differentiation. Inherently problematic is the fact that no soil water flux meters are available to accurately measure the unsaturated soil water flux q in Equation (7). A review by Gee et al. (2003) provides possible direct and indirect methods, but none of them are adequate because of problems with divergence of water flow near the flux measurement device. Recently, the heat pulse probe was developed (Kamai et al., 2008) for indirect measurement of soil water flux, but is limited to fluxes of 6 mm d1 or higher. Finally, very few routine measurements are available to determine the K(y) relationship. In fact, the lack of the unsaturated conductivity information is the most limiting factor of in situ application of the Darcy equation. Most promising is the application of inverse modeling for parameter estimation of the soil hydraulic functions, using both laboratory and field techniques (Hopmans et al., 2002b), which can be used in conjunction with in situ water content and soil water potential measurements to estimate temporal changes in depth distribution of soil water flux. Selected steady-state solutions are provided in Jury et al. (1991), but are only of limited use for real field conditions since soil water content and matric potential values change continuously. Most realistically, one must apply the transient unsaturated water flow (Equation (10)) that arises from combination of the Darcy equation with mass conservation. However, its solution also requires a priori knowledge of the soil water capacity, C, as determined from the slope of the soil water retention curve, and time measurements of y and h, at the various soil horizon interfaces and at the boundaries of the soil domain of interest, including at the soil profile bottom. Although certainly possible, relatively few of such field experiments are conducted routinely because they are time consuming and wrought with complications. However, in combination with inverse modeling, such field experiments can provide a wealth of information, including plant root water uptake dynamics, plant transpiration, and drainage rates (Vrugt et al., 2001). Therefore, large lysimeters with selected

112

Infiltration and Unsaturated Zone

water content and soil water potential measurements may be very useful.

2.05.5 Scaling and Spatial Variability Considerations Soil hydrologists need to apply locally measured soil physical data to characterize flow and transport processes at large-scale heterogeneous vadose zones. For example, prediction of soil water dynamics, such as infiltration at the field scale, is usually derived from the measurement of soil hydraulic properties from laboratory cores, as collected from a limited number of sampling sites across large spatial extents. Soil parameters obtained from these small-scale measurements are subsequently included in numerical models with a grid or element size many times larger, with the numerical results extrapolated to predict large-scale flow and transport behavior. Because of the typical nonlinearity of soil physical properties, their use across spatial scales is inherently problematic. Specifically, the averaging of processes determined from discrete small-scale samples may not describe the true soil behavior involving larger spatial structures. Moreover, the dominant physical flow processes may vary between spatial scales. Considering that soil physical, chemical, and biological measurements are typically conducted for small measurement volumes and that the natural variability of soils is enormous, the main question asked is how small-scale measurements can provide information about large-scale flow and transport behavior. In their treatise of scale issues of vadose zone modeling, Hopmans et al. (2002a) offer a conceptual solution, considering the control of small-scale processes on larger-scale flow behavior. Hence, vadose zone properties are nonunique and scale dependent, resulting in effective properties that vary across spatial scales and merely serve as calibration parameters in simulation models. Therefore, their accurate prediction in heterogeneous materials can only be accomplished using scale-appropriate measurements, including those that measure at the landscape scale. In addition, infiltration measurements are typically conducted at measurement scales in the range of 0.2–1.0 m. This is relevant for irrigation purposes, especially for micro-irrigation applications. Yet, infiltration information is often needed for much larger spatial scales, at the pedon scale, hillslope scale, and watershed scale. Very little work has been done relating infiltration process to measurement or support scale. Exceptions are the studies by Sisson and Wierenga (1981) and Haws et al. (2004), who measured steady-state infiltration at three spatial scales, ranging from 5 to 127- cm-diameter infiltrometer rings. Their results showed that much of the larger-scale infiltration occurs through smaller-scale regions, and that the spatial variability of infiltration decreased as the measurement scale increased. Thus, in general, we find that the process of infiltration might vary with spatial scale, and that larger spatial scales are required to estimate representative infiltration characteristics across a typical landscape. Many field studies have dealt with the significant areal heterogeneity of soil hydraulic properties, and particularly that of the saturated hydraulic conductivity, Ks (Nielsen et al., 1973). The heterogeneity in Ks is recognized to have a major effect on unsaturated flow, leading to significant variation in

local infiltration. In general, accounting for areal heterogeneity leads to shorter ponding times and to a more gradual decrease of the infiltration flux with time (Smith and Hebbert, 1979; Sivapalan and Wood, 1986). To characterize spatial variable infiltration rates, Sharma et al. (1980) measured infiltration with a double-ring infiltrometer at 26 sites in a 9.6-ha watershed. The infiltration data were fitted to the PH infiltration Equation (17a), and fitting parameters S and A were scaled to express their spatial variability and to describe the ensembleaverage or composite infiltration curve of the watershed. A simpler but similar scaling technique for infiltration data was presented by Hopmans (1989), who measured transient infiltration at 50 sites along a 100-m transect. Data were fitted to both the PH and a modified KO model that includes an additional constant c as a second term in Equation (23). This paper showed that spatial variability of infiltration can be easily described by the probability density function of a single scaling parameter, to be used for applications in Monte Carlo simulation of watershed hydrology, as suggested for the first time by Peck et al. (1977). For application at the field scale, the so-called one-point method was presented by Shepard et al. (1993) to estimate furrow-average infiltration parameters of PH Equation (17a), across a furrow-irrigated agricultural field. They used the volume-balance principle from furrow advance time across the field, water inflow rate, and flow area measurements. For modeling surface hydrology, by subtracting the infiltration rate, i(t), from the rainfall rate, r(t), it is possible to estimate spatial and temporal distributions of rainfall excess or runoff. The influence of spatial heterogeneity in rainfall and soil variability on runoff production was studied by Sivapalan and Wood (1986) from an analytical solution of infiltration and making use of the IDA approximation. Statistical characteristics of ponding time and infiltration rate were presented for two cases, one with a spatially variable soil with a lognormal Ks distribution and uniform rainfall, and the other for a homogeneous soil with spatially variable rainfall. Among the various results, this study concluded that the ensemble infiltration approach is biased for spatially variable soils. Their results also showed that the cumulative distribution of ponding times or proportion of ponded area is an excellent way of analyzing mean areal infiltration. Moreover, the spatial correction of infiltration rate is time dependent and varies depending on the correlation lengths of rainfall and soil Ks. This study neglected the effects of surface water run-on, as caused by accumulated water upstream, running on to neighboring areas, thereby contributing locally to infiltration. A quantitative analysis of soil variability effects on watershed hydraulic response that included surface water interactions, such as run-on, was presented by Smith and Hebbert (1979), through analysis of the effects of deterministic changes of infiltration properties in the direction of surface water flow, using a kinematic watershed model. In a subsequent study by Woolhiser et al. (1996), it was clearly demonstrated that runoff hydrographs along a hillslope are significantly affected by spatial trends in the soil’s saturated hydraulic conductivity. We expect that important new information can be collected by linking this interactive modeling approach with remote sensing and geographical information system (GIS) tools. A detailed analysis and review of the control of spatially

Infiltration and Unsaturated Zone

variable hydrologic properties on overland flow are presented by Govindaraju et al. (2007). Yet another concern regarding nonideal infiltration, causing spatially variable infiltration at small spatial scales, comes from the presence of water-repellent or hydrophobic soils. Since the 1980s much new research and findings have been presented, improving the understanding of the underlying physical processes and its relevance to soil water flow and water infiltration (DeBano, 2000; Wang et al., 2000). Infiltration may be controlled by soil surface crust-forming dynamics, which is another complex phenomenon dominated by a wide variety of factors involving soil properties, rainfall characteristics, and local water flow conditions. Two types of rainfall-induced soil seals can be identified: (1) structural seals that are directly related to rainfall through the impact of raindrops and sudden wetting and (2) depositional or sedimentary seals that are indirectly related to rainfall as it results from the settling of fine particles carried in suspension by runoff in soil depressions. A recent review on concepts and modeling of rainfall-induced soil surface sealing was presented by Assouline (2004).

2.05.6 Summary and Conclusions Although important and seemingly simple, infiltration is a complicated process that is a function of many different soil properties, rainfall, land use, and vegetation characteristics. In addition to rainfall intensity and duration as well as the soil physical factors, such as soil water retention and hydraulic conductivity, infiltration is controlled by the initial water content, surface sealing and crusting, hydrophobicity, soil layering, and the ionic composition of the infiltrated water. Many relatively simple infiltration equations have been proposed historically, and are successfully used to characterize infiltration. Other physically based analytical solutions have been presented that can potentially be used to predict infiltration. However, in practice, this is difficult as soil physical characteristics near the soil surface show naturally high soil spatial variability and are often time dependent because of soil structural changes. Alternatively, infiltration is often measured to estimate the relevant soil hydraulic parameters from the fitting of the infiltration data to a specific infiltration model by inverse modeling, such as by using permeameters. Whereas most infiltration measurement techniques and infiltration models apply to relatively small spatial scales, infiltration information is often needed at the watershed and hillslope scales. Yet, it has been shown that much of the largerscale infiltration occurs through smaller-scale regions, for example, because infiltration is largely controlled by spatial variations of the soil’s physical characteristics at the land surface, vegetation cover, and topography. In general, we expect that the process of infiltration varies with spatial scale, and that measurements at larger spatial scales are needed to estimate representative infiltration characteristics across hillslope and larger spatial scales. For that purpose, improved solutions to infiltration across scales from the field to basin scale are needed, such as may become available using rapidly developing techniques including remote sensing, GIS, and new measurement devices.

113

Acknowledgments This chapter is partly based on the paper by Hopmans et al. (2007), and includes edited sections of that paper. The author acknowledges the significant input received by Drs. J.-Y Parlange and S. Assouline in writing the 2007 paper.

References Assouline S (2004) Rainfall-induced soil surface sealing: A critical review of observations, conceptual models and solutions. Vadose Zone Journal 3: 570--591. Barry DA, Parlange J-Y, Haverkamp R, and Ross PJ (1995) Infiltration under ponded conditions: 4. An explicit predictive infiltration formula. Soil Science 160: 8--17. Battany MC and Grismer ME (2000) Development of a portable field rainfall simulator for use in hillside vineyard runoff and erosion studies. Hydrological Processes 14: 1119--1129. Bouwer H (1986). Intake rate: Cylinder infiltrometer, In: Klute A, (ed.) Methods of Soil Analysis, Part 1. Number 9 in the Series Agronomy, pp. 825–844. Madison, WI: American Society of Agronomy. Brooks RH and Corey AT (1964) Hydraulic Properties of Porous Media, Hydrology Paper No. 3. Fort Collins, CO: Colorado State University. Bruce RR and Klute A (1956) The measurement of soil moisture diffusivity. Soil Science Society American Proceedings 20: 458--462. Brutsaert W (2005) Hydrology – An Introduction. New York, NY: Cambridge University Press. Clausnitzer V, Hopmans JW, and Starr JL (1998) Parameter uncertainty analysis of common infiltration models. Soil Science Society of America Journal 62: 1477--1487. Clothier BE (2001) Infiltration. In: Smith KA and Mullins CE (eds.) Soil and Environmental Analysis, Physical Methods, 2nd edn., Revised and Expanded, pp. 239–280. New York: Dekker. Dane JH and Hopmans JW (2002) Soil water retention and storage – introduction. In: Dane JH and Topp GC (eds.) Methods of Soil Analysis. Part 4. Physical Methods, pp. 671--674. Madison, WI: Soil Science Society of America. Dane JH and Topp GC (eds.) (2002) Methods of Soil Analysis. Part 4. Physical Methods, vol. 5, Madison, WI: Soil Science Society of America. DeBano LF (2000) Water repellency in soils: A historical overview. Journal of Hydrology 231: 4--32. Dirksen C (2001) Hydraulic conductivity. In: Smith KA and Mullins CE (eds.) Soil and Environmental Analysis, pp. 141--238. New York: Dekker. Espinoza RD (1999) Infiltration. In: Delleur JW (ed.) The Handbook of Groundwater Engineering, pp. 7.1--7.18. Boca Raton, FL: CRC Press. Gardner WR (1958) Some steady state solutions of unsaturated moisture flow equations with application to evaporation from a water table. Soil Science 85: 228--232. Gee GW, Zhang F, and Ward AL (2003) A modified vadose zone fluxmeter with solution collection capability. Vadose Zone Journal 2: 627--632. Govindaraju RS, Nahar N, Corradini C, and Morbidelli R (2007) Infiltration and run-on under spatially-variable hydrologic properties. In: Delleur JW (ed.) The Handbook of Groundwater Engineering, pp. 8.1--8.15. Boca Raton, FL: CRC Press. Green WA and Ampt GA (1911) Studies on soils physics: 1. The flow of air and water through soils. Journal of Agricultural Science 4: 1--24. Haverkamp R, Debionne S, Viallet P, Angulo-Jaramillo R, and de Condappa D (2007) Soil properties and moisture movement in the unsaturated zone. In: Delleur JW (ed.) The Handbook of Groundwater Engineering, pp. 6.1--6.59. Boca Raton, FL: CRC Press. Haverkamp R, Parlange J-Y, Starr JL, Schmitz G, and Fuentes C (1990) Infiltration under ponded conditions: 3. A predictive equation based on physical parameters. Soil Science 149: 292--300. Haverkamp R, Ross PJ, Smettem KRJ, and Parlange J-Y (1994) Three-dimensional analysis of infiltration from the disc infiltrometer. 2. Physically based infiltration equation. Water Resources Research 30: 2931--2935. Haws NW, Boast CW, Rao PSC, Kladivko EJ, and Franzmeier DP (2004) Spatial variability and measurement scale of infiltration rate on an agricultural landscape. Soil Science Society of America Journal 68: 1818--1826. Hopmans JW (1989) Stochastic description of field-measured infiltration data. Transactions of the American Society of Agricultural Engineers 32: 1987-1993.

114

Infiltration and Unsaturated Zone

Hopmans JW, Assouline S, and Parlange J-Y (2007) Soil infiltration. In: Delleur JW (ed.) The Handbook of Groundwater Engineering, pp. 7.1--7.18. Boca Raton, FL: CRC Press. Hopmans JW, Nielsen DR, and Bristow KL (2002a) How useful are small-scale soil hydraulic property measurements for large-scale vadose zone modeling. In: Smiles D, Raats PAC, and Warrick A (eds.) Heat and Mass Transfer in the Natural Environment, the Philip Volume. Geophysical Monograph Series No. 129, pp. 247–258. Washington, DC: American Geophysical Union. Hopmans JW, Sˇimunek J, Romano N, and Durner W (2002b) Inverse methods. In: Dane JH and Topp GC (eds.) Methods of Soil Analysis. Part 4. Physical Methods, pp. 963--1008. Madison, WI: Soil Science Society of America. Horton RE (1940) An approach towards a physical interpretation of infiltration capacity. Soil Science Society American Proceedings 5: 399--417. Jury WA, Gardner WR, and Gardner WH (1991) Soil Physics. New York: Wiley. Kamai T, Tuli A, Kluitenberg GJ, and Hopmans JW (2008) Soil water flux density measurements near 1 cm/day using an improved heat pulse probe. Water Resources Research 44: doi: 10.1029/2008WR007036. Kirkham D and Powers WL (1972) Advanced Soil Physics. New York: Wiley. Kostiakov AN (1932) On the dynamics of the coefficient of water percolation in soils and on the necessity of studying it from a dynamic point of view for purposes of amelioration. In: Transactions of the Sixth Commission of the International Society of Soil Science A, pp. 17–21. Kosugi K, Hopmans JW, and Dane JH (2002) Water retention and storage – parametric models. In: Dane JH and Topp GC (eds.) Methods of Soil Analysis. Part 4. Physical Methods, pp. 739--758. Madison, WI: Soil Science Society of America. Kutilek M and Nielsen DR (1994) Soil Hydrology. GeoEcology Textbook. CremlingenDestedt. Germany: Catena Verlag. Latifi H, Prasad SN, and Helweg OJ (1994) Air entrapment and water infiltration in two-layered soil column. Journal of Irrigation and Drainage Engineering 120: 871--891. Mein RG and Farrell DA (1974) Determination of wetting front suction in the Green– Ampt equation. Soil Science Society of America Proceedings 38: 872--876. Mezencev VJ (1948) Theory of formation of the surface runoff (Russian). Meteorologia i Gidrologia 3: 33--40. Morin J, Goldberg D, and Seginer I (1967) A rainfall simulator with a rotating disc. Transactions of the American Society of Agricultural Engineers 10: 74--77. Mualem Y (1976) A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resources Research 12: 513--522. Mualem Y and Assouline S (1989) Modeling soil seal as a non-uniform Layer. Water Resources Research 25: 2101--2108. Nasta P, Kamai T, Chirico GB, Hopmans JW, and Romano N (2009) Scaling soil water retention functions using particle-size distribution. Journal of Hydrology 374: 223–234. Nielsen DR, Biggar JB, and Ehr KT (1973) Spatial variability of field measured soil water properties. Hilgardia 42: 215--260. Parlange J-Y, Haverkamp R, and Touma J (1985) Infiltration under ponded conditions: 1. Optimal analytical solution and comparison with experimental observations. Soil Science 139: 305--311. Parlange J-Y, Lisle I, Braddock RD, and Smith RE (1982) The three-parameter infiltration equation. Soil Science 133: 337--341. Peck AJ, Luxmoore RJ, and Stolzy JL (1977) Effects of spatial variability of soil hydraulic properties in water budget modeling. Water Resources Research 13: 348--354. Peterson AE and Bubenzer GD (1986). Intake rate: Sprinkler infiltrometer, In: Klute A, (ed.) Methods of Soil Analysis, Part 1. Number 9 in the series Agronomy, pp. 45–870. Madison, WI: American Society of Agronomy. Philip JR (1957a) The theory of infiltration: 1. The infiltration equation and its solution. Soil Science 83: 345--357. Philip JR (1957b) The theory of infiltration: 2. The profile at infinity. Soil Science 83: 435--448. Philip JR (1957c) The theory of infiltration: 4. Sorptivity and algebraic infiltration equations. Soil Science 84: 257--264. Philip JR (1969) Theory of infiltration. In: Chow VT (ed.) Advances in Hydroscience, vol. 5, pp. 215--296. New York, NY: Academic Press. Philip JR (1987) The infiltration joining problem. Water Resources Research 12: 2239--2245. Robinson DA, Campbell CS, Hopmans JW, et al. (2008) Soil moisture measurement for ecological and hydrological watershed-scale observatories: A review. Vadose Zone Journal 7: 358--389. Scanlon BR, Andraski BJ, and Bilskie J (2002) Miscellaneous methods for measuring matric or water potential. In: Dane JH and Topp GC (eds.) Methods of Soil Analysis. Part 4. Physical Methods, pp. 643--670. Madison, WI: Soil Science Society of America.

Sharma ML, Gander GA, and Hunt CG (1980) Spatial variability of infiltration in a watershed. Journal of Hydrology 45: 101--122. Shepard JS, Wallender WW, and Hopmans JW (1993) One-point method for estimating furrow infiltration. Transactions of American Society of Agricultural Engineers 36: 395--404. Sˇimunek J, Van Genuchten MTh, and Sejna M (2008) Development and applications of the HYDRUS and STANMOD software packages and related codes. Vadose Zone Journal 7: 587--600. Sisson JB and Wierenga PJ (1981) Spatial variability of steady-state infiltration rates as a stochastic process. Soil Science Society of America Journal 45: 699--704. Sivapalan M and Milly PCD (1989) On the relationship between the time condensation approximation and the flux-concentration relation. Journal of Hydrology 105: 357--367. Sivapalan M and Wood EF (1986) Spatial heterogeneity and scale in the infiltration response of catchments. In: Gupta VK, Rodriguez-Iturbe I, and Wood EF (eds.) Scale Problems in Hydrology, pp. 81--106. Hingham, MA: Reidel. Skaggs RW (1982) Infiltration, In: Haan CT, Johnson HP, and Brakensiek DL, (eds.) Hydrologic Modeling of Small Watersheds, ASAE Monograph No. 5, 121–166. St. Joseph, MI: ASAE. Smith RE and Hebbert RHB (1979) A Monte-Carlo analysis of the hydrologic effects of spatial variability of infiltration. Water Resources Research 15: 419--429. Smith RE, Smettem KRJ, Broadbridge P, and Woolhiser DA (2002) Infiltration Theory for Hydrologic Applications. Water Resources Monograph 15, Washington, DC: American Geophysical Union. Soil Conservation Service (1972) Estimation of direct runoff from storm rainfall National Engineering Handbook, Section 4: Hydrology, pp. 10.1--10.24. Washington, DC: USDA. Swartzendruber D (1987) A quasi-solution of Richards’ equation for the downward infiltration of water into soil. Water Resources Research 23: 809--817. Tuli AM and Hopmans JW (2004) Effect of degree of saturation on transport coefficients in disturbed soils. European Journal of Soil Science 55: 147--164. Vandervaere J-P, Vauclin M, and Elrick DE (2000) Transient flow from tension infiltrometers: I. The two-parameter equation. Soil Science Society of America Journal 64: 1263--1272. Vandervaere J-P, Vauclin M, Haverkamp R, Peugeot C, Thony J-L, and Gilfedder M (1998) Prediction of crust-induced surface runoff with disc infiltrometer data. Soil Science 163: 9--21. Van Genuchten MTh (1980) A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal 44: 892--898. Vrugt JA, Hopmans JW, and Sˇimunek J (2001) Calibration of a two-dimensional root water uptake model. Soil Science Society of America Journal 65: 1027-1037. Wang Z, Wu QJ, Wu L, Ritsema CJ, Dekker LW, and Feyen J (2000) Effects of soil water repellency on infiltration rate and flow instability. Journal of Hydrology 231: 265--276. Wooding RA (1968) Steady infiltration from a shallow circular pond. Water Resources Research 4: 1259--1273. Woolhiser DA, Smith RE, and Giraldez J-V (1996) Effects of spatial variability of saturated hydraulic conductivity on Hortonian overland flow. Water Resources Research 32: 671--678. Wu L, Pan L, Robertson MJ, and Shouse PJ (1997) Numerical evaluation of ring-infiltrometers under various soil conditions. Soil Science 162: 771--777. Young MH and Sisson JB (2002) Tensiometry. In: Dane JH and Topp GC (eds.) Methods of Soil Analysis, Part 4: Physical Methods, pp. 575--606. Madison, WI: Soil Science Society of America.

Relevant Websites http://www.decagon.com Decagon Devices, Mini-Disk Infiltrometer. http://hopmans.lawr.ucdavis.edu Jan W. Hopmans, Vadose Zone Hydrology. http://www.pc-progress.com PC-Progress: Engineering Software Developer; HYDRUS 2D/3 D for Windows, Version 1.xx. http://ag.arizona.edu/sssa-s1 SSSA Soil Physics Division S-1. http://en.wikipedia.org Wikipedia, Infiltration (Hydrology).

2.06 Mechanics of Groundwater Flow M Bakker, Delft University of Technology, Delft, The Netherlands EI Anderson, WHPA, Bloomington, IN, USA & 2011 Elsevier B.V. All rights reserved.

2.06.1 Introduction 2.06.2 Brief History 2.06.3 Hydraulic Head 2.06.4 Darcy’s Law 2.06.5 Steady Conservation of Mass 2.06.6 Flow Types 2.06.6.1 Spatial Dimension 2.06.6.2 Time Dependence 2.06.6.3 Geologic Setting 2.06.7 The Dupuit Approximation 2.06.8 Potential Flow and the Discharge Vector 2.06.9 One-Dimensional Flow 2.06.9.1 Confined Flow between Two Rivers 2.06.9.2 Combined Flow between Two Rivers 2.06.9.3 Unconfined Flow in a River Valley 2.06.10 One-Dimensional Radial Flow 2.06.10.1 Flow to a Well at the Center of a Circular Island without Recharge 2.06.10.2 Recharge on a Circular Island 2.06.10.3 Well at the Center of a Circular Island with Recharge 2.06.11 The Principle of Superposition 2.06.11.1 A Well in Uniform Flow 2.06.11.2 The Method of Images 2.06.11.3 Flow to a Pumping Well in an Alluvial Valley 2.06.12 The Stream Function and the Complex Potential 2.06.12.1 Evaluation of the Capture Zone Envelope Using the Complex Potential 2.06.13 Transient Flow 2.06.13.1 One-Dimensional Periodic Flow 2.06.13.2 Transient Wells 2.06.13.3 Convolution 2.06.14 Computer Models 2.06.15 Discussion Acknowledgments References

2.06.1 Introduction Groundwater is the most important resource of freshwater on earth. It moves very slowly through the top part of the earth’s crust from areas of recharge (often originating from rainfall) to discharge in springs, wells, rivers, lakes, and oceans. The baseflow of rivers, the flow between rainfall or snowmelt events, is caused predominantly by inflow from groundwater. In many parts of the world, groundwater is the only source for drinking water or irrigation. Groundwater resources are threatened by over-exploitation and contamination. Major problems include a rapid decline of the groundwater table caused by pumping of groundwater for irrigated agriculture, salinization of groundwater resources due to heavy pumping in coastal areas, and contamination of groundwater by leakage of toxic chemicals.

115 115 115 116 117 118 118 118 118 118 119 120 120 120 121 121 121 122 122 122 123 123 124 126 127 128 129 129 130 131 133 133 133

Accurate tools are needed to predict whether the current and proposed uses of groundwater resources are sustainable and safe. The field of groundwater flow, also called hydrogeology, is large and only the basic physical principles of groundwater flow through porous media are discussed in this chapter. Detailed textbooks include Verruijt (1970), Bear (1972), Strack (1989), and Fitts (2002). This chapter focuses on groundwater flow through porous materials such as sand, silt, or clay. Significant amounts of groundwater may flow through fractured rock formations. The concepts outlined in this chapter apply to such formations when the fractured rock may be represented by an equivalent porous medium. Compared to other areas of hydrology, the governing equations for groundwater flow are relatively well known. Exact solutions can be obtained for many important flow

115

116

Mechanics of Groundwater Flow

systems. These exact solutions provide important insights into the flow of groundwater and groundwater interactions with the accessible environment. This chapter begins with a description of the governing equations. Flow principles are explained through discussion of a set of steady and transient flow problems. This chapter concludes with a brief discussion of available modeling tools for solving more complicated problems.

2.06.2 Brief History The foundation of the quantitative description of groundwater flow was laid by Henry Darcy and Jules Dupuit. Darcy (1856) peformed column experiments that led to what is now called Darcy’s law for groundwater flow. Jules Dupuit was a classmate of Darcy and his replacement as director of Water and Bridges in Paris in the 1850s. Dupuit (1863) recognized that in many cases the vertical variation of the horizontal components of flow may be neglected, in essence reducing the mathematical description of groundwater flow by one dimension; one of his examples was a formula for flow to a well. At the end of the nineteenth century, Forchheimer (1886) combined Darcy’s law with the continuity equation to show that steady groundwater flow through piecewise homogeneous aquifers is governed by Laplace’s equation. This opened the door to many existing solutions that were derived for other problems governed by the same equation. Equations that describe transient groundwater flow take into account that the aquifer can store water. Phreatic storage, storage through movement of the groundwater table, was included by Boussinesq (1904). The process of elastic storage was conceptualized by Meinzer (1928), Theis (1935), and Jacob (1940), and led to the definition of storativity.

2.06.3 Hydraulic Head The mechanical energy per unit weight in an incompressible fluid is given at a point by the following sum:

p V2 þZþ 2g g

ð1Þ

H A

z L x

Figure 1 Measurement of the hydraulic head at a point in an aquifer using a piezometer. Saturated zone is darker gray and is bounded on top by the groundwater table.

The hydraulic head provides a good estimate of the available energy per unit weight at a point in a groundwater flow field, and is fairly easy to measure. Figure 1 shows a piezometer set into the saturated portion of an aquifer. A piezometer is a hollow tube, open to the aquifer only at the bottom (point A in the figure). The hydraulic head at point A is the sum of the pressure head and elevation head. The fluid that rises into the piezometer is hydrostatic and the pressure at point A is

pA ¼ gH

ð3Þ

or the pressure head at point A is H. The elevation head at point A, measured with respect to the datum shown in the figure, is L. The hydraulic head at point A is given by L þ H and is equal to the height above the datum to which water rises in the piezometer. The hydraulic head may be measured at a point in a groundwater flow field once a piezometer has been placed. Hydraulic head data at surface water features such as lakes and streams, which form the natural boundaries of many aquifers, are often available as time series. Head data are the most abundant information a groundwater engineer has at his disposal. The importance of the hydraulic head in groundwater calculations will become clear in the remainder of this chapter.

where p [ML1T2] is the pressure, g [MT2L2] is the specific weight of the fluid, Z [L] is the elevation above a fixed datum, V [LT1] is the speed of the fluid, and g [LT2] is the acceleration due to gravity. The first term, referred to as pressure head, reflects the pressure energy of the fluid. The second term reflects the potential energy of the system and is called the elevation head. The third term, known as the velocity head, reflects the kinetic energy of the fluid. In groundwater applications, typical fluid speeds within the porous medium are so small that the velocity head is negligible. The combination of pressure head and elevation head is known as the hydraulic head, also called piezometric head or simply head. The dimension of head h is length [L]:

In 1856, Henry Darcy performed experiments from which he concluded that the flow of groundwater is proportional to the head gradient (Darcy, 1856). The general setup of the experiment is simple. A cylinder is filled with aquifer material. The ends of the cylinder are attached to two reservoirs with different levels (Figure 2). Water flows through the aquifer material from the higher reservoir to the lower reservoir. In this fashion, Darcy showed that the discharge Q [L3T1] through the soil column is proportional to the head difference h1  h2 [L] and the cross-sectional area A [L2] of the column, and inversely proportional to the length of the soil column L [L]:

p h¼ þZ g

h1  h2 Q ¼ kA L

ð2Þ

2.06.4 Darcy’s Law

ð4Þ

Mechanics of Groundwater Flow

interconnectedness of the pores of the aquifer. For example, the hydraulic conductivity is smaller for oil than for water when flowing through the same aquifer, as oil is more viscous than water. Similarly, warmer water results in a larger hydraulic conductivity than colder water, as the viscosity of water decreases with temperature. The hydraulic conductivity may be written as

h1 − h2

k¼

L

Figure 2 The experiment of Darcy. Darcy used a vertical column, but the flow is independent of the angle of the soil column.

The proportionality constant is k [L/T] and is called the hydraulic conductivity. Equation (4) is adequate to describe flow through a soil column, but not flow through an aquifer. Flow through an aquifer is expressed in terms of the specific discharge vector ~ q½LT 1 , the discharge per unit area of aquifer normal to the direction of flow:

~ q ¼ krh

ð5Þ

The components of ~ q in the Cartesian x, y, and z directions may be written as

qh qx ¼ k ; qx

qh qy ¼ k ; qy

qh qz ¼ k qz

ð6Þ

Equation (5) is known as Darcy’s law, although it is an empirical formula relating the head gradient to the specific discharge vector. Note the equivalence between Darcy’s law and other physical laws such as Fourier’s law for heat flux, Ohm’s law for current density, and Fick’s law for diffusive flux. The hydraulic conductivity of aquifers may be anisotropic. Sedimentary aquifers often consist of a sequence of thin layers of slightly coarser or slightly finer material. In such aquifers, the average vertical hydraulic conductivity is smaller than the average horizontal hydraulic conductivity. Hence, the hydraulic conductivity is anisotropic and is written as a tensor K so that Darcy’s law becomes

~ q ¼ Krh

ð7Þ

When the principal directions of the hydraulic conductivity tensor coincide with the horizontal and vertical directions of sedimentary aquifers, the Cartesian components of Darcy’s law become

qx ¼ kh

qh ; qx

qy ¼ kh

qh ; qy

qz ¼ kv

117

qh qz

ð8Þ

where kh is the horizontal hydraulic conductivity and kv the vertical hydraulic conductivity. The hydraulic conductivity is a function of the fluid that flows through the aquifer and of the shapes, sizes, and

krg m

ð9Þ

where k [L2] is called the intrinsic permeability of the aquifer and is a characteristic of the pore-size distribution and tortuosity of the porous medium, r [ML3] is the density of the fluid, g [LT2] is the acceleration of gravity, and m [MLT1] is the dynamic viscosity. In this way, the property of the porous material (k) is separated from the properties of the fluid (r and m). Variations of r and m may play a role in coastal aquifers because of changes in salinity, or in cases where the temperature of the groundwater varies significantly, for example, in river bank filtration projects or systems for aquifer thermal energy storage. The hydraulic conductivity of an aquifer may be measured with a Darcy experiment. This requires, however, that undisturbed and representative samples are taken from an aquifer. As this is a difficult, if not impossible, task, field measurements are more likely to give accurate results. Representative values for the hydraulic conductivity are given in Table 1 for water flowing through different aquifer materials. The hydraulic conductivity generally varies spatially throughout an aquifer. The heterogeneity is rarely known well. The head and flow in the aquifer may often be simulated accurately by treating the aquifer properties as piecewise homogeneous. Travel times are, however, strongly affected by heterogeneity of the aquifer (e.g., Moore and Doherty, 2006). The specific discharge vector has the same units as a velocity and is sometimes called the Darcy flux or the Darcy velocity. It is important to note, however, that a water particle that flows through the aquifer does not flow with an average velocity equal to the specific discharge. The specific discharge is the discharge through a unit area of aquifer. Only part of this unit area consists of pores while the larger part consists of solid particles. The ratio of the volume of pores to the volume of aquifer is called the porosity n. Water can only flow through the pores, so that the average velocity vector ~ u may be obtained Table 1 Representative values of hydraulic conductivity for various aquifer materials Material

k (m d1)

Clay Sandy clays Peat Silt Very fine sands Fine sands Coarse sands Sands with gravel Gravels

o0.0001 0.0001–0.001 0.0001–0.01 0.001–0.01 0.1–1 1–10 10–100 100–1000 41000

Modified from Verruijt A (1970) Theory of Groundwater Flow. New York: MacMillan.

118

Mechanics of Groundwater Flow

2.06.6 Flow Types

from the specific discharge as

~ u ¼~ q=n

ð10Þ

This is called an average velocity. The velocity through a larger pore, or through the center of a pore is likely to be larger than the velocity through a smaller pore or along the edge of a soil particle. The velocity of groundwater is generally very small. The head may drop 1 or 2 m every 1000 m. A gradient of 0.002 in a sand with a hydraulic conductivity of 10 m d1 and a porosity of 0.2 gives an average velocity of only 0.1 m d1. Larger velocities occur in very specific cases only, such as near pumping wells.

2.06.5 Steady Conservation of Mass Darcy’s law provides three scalar equations for the four unknowns: qx, qy, qz, and h. The fourth equation for solving the system is obtained from conservation of mass. A derivation of the differential statement of conservation of mass for a flowing fluid is given in any standard fluid mechanics text (i.e., Munson et al., 2002). The result states that the divergence of the mass flow rate and the rate of accumulation of fluid mass are in balance at every point in the flow field:

Groundwater flow may be classified according to the spatial dimensions of the flow field, the dependence of the flow on time, and the aquifer setting in which the flow occurs. The focus in this chapter is on one- and two-dimensional, steady and transient flow in single aquifer systems with isotropic and homogeneous properties. These flow types and others are described in the following.

2.06.6.1 Spatial Dimension Flow in an aquifer may be one, two, or three dimensional depending on the boundary conditions associated with the flow. Most aquifers are relatively thin in comparison to their areal extent. In these settings, which are referred to as shallow aquifers, one- and two-dimensional analyses are often adequate. In shallow aquifers the vertical variations in the hydraulic head are negligibly small when compared to horizontal variations in head. For problems where three-dimensional flow is important, near local features such as partially penetrating or horizontal wells, or near partially penetrating streams, the effects of concentrated vertical flow can be incorporated approximately into two-dimensional models.

2.06.6.2 Time Dependence

qr r  ðr~ uÞ þ ¼0 qt

ð11Þ

where ~ u is the fluild velocity and r the fluid density. By analogy, a statement for conservation of mass made for groundwater flowing through a porous material of porosity n is

r  ðr~ qÞ þ

q ðrnÞ ¼0 qt

ð12Þ

The fluid density is multiplied by the porosity in the second term as the fluid mass occurs only in the pore spaces. If the porous media is rigid (qn/qt ¼ 0) and the fluid density is constant in time, or if the flow is steady, (12) reduces to

Groundwater flow is either steady or transient. In steady flow, there are no changes in flow or hydraulic head in time. Analyses of steady flow are used to reflect long-term, average conditions in an aquifer, for example, the dewatering of an aquifer for a large construction project, or delineation of wellhead protection areas for municipal water supply wells. Transient flow occurs when aquifer boundary conditions change in time, for example, changing aquifer recharge, changing river levels, and varying pumping rates of wells. A specific application of a transient flow analysis is the evaluation of aquifer properties by field tests, such as pumping tests, when it is not practical to run the test until steady conditions are reached.

2.06.6.3 Geologic Setting r  ðr~ qÞ ¼ 0

ð13Þ

If, in addition, the fluid density is constant in space the simplest form of conservation of mass emerges,

r ~ q¼0

ð14Þ

Equation (14) is also known as the continuity of flow equation; when the density is constant, conservation of mass is equivalent to continuity of flow. Conservation of mass (14) may be combined with Dary’s law (5) to obtain a single differential equation governing three-dimensional groundwater flow through a homogeneous aquifer:

r 2h ¼ 0 This result was first obtained by Forcheimer (1886).

ð15Þ

The geologic setting of an aquifer may be used to further define the flow type as confined, unconfined, combined, or multiaquifer flow. The subsoil may be divided in more permeable and less permeable layers. The permeable layers may transmit significant amounts of water. They are called aquifers and can be used as the source for drinking water or irrigation. The less permeable layers transmit little or no water and cannot be used for water supply; they are commonly called aquicludes, confining layers, aquitards, or leaky layers. An aquifer is confined when it is bounded on the top and bottom by impermeable layers, or layers with significantly lower permeability than the aquifer. In contrast, an unconfined aquifer is not bounded on top by an impermeable layer. Flow in an aquifer is called confined when the head in the aquifer is above the impermeable top of the aquifer (Figure 3(a)). For unconfined flow, the saturated part of the aquifer is bounded on top by the groundwater table, also called the

Mechanics of Groundwater Flow

119

Confining layer

Phreatic surface Confined flow

Unconfined flow

Impermeable base (a)

Areal recharge

Phreatic surface

(b)

Impermeable base

Figure 3 Definition of aquifer types and flow types: (a) combined confined and unconfined flow and (b) unconfined flow with recharge.

phreatic surface (Figure 3(b)). The concept of a groundwater table seems simple: when one digs a deep enough hole, it will fill up with water to the level of the groundwater table. Upon closer examination, the concept is less clear, however. When digging down, the soil gets wetter and wetter until the groundwater table is reached. In the section from the surface to the groundwater table, the pores of the soil are filled with both water and air; this section is called the unsaturated zone and is described in detail in Chapter 2.05 Infiltration and Unsaturated Zone. The saturated zone starts at the phreatic surface, which is defined as the depth where the pressure in the water is equal to atmospheric. The phreatic surface is curved when there is flow in the aquifer. The surface goes down in the direction of flow; thus, the velocity of a water particle always has a downward component; in most cases, this component is relatively small. In unconfined flow, the saturated thickness varies with the elevation of the water table. Flow in a confined aquifer becomes unconfined when the head falls below the impermeable top of the aquifer. In a confined aquifer, the flow may consist of both regions where the head is above the confining layer and regions where the head is below the confining layer. This is referred to as combined confined and unconfined flow, or simply combined flow. Combined flow in a confined aquifer is illustrated in Figure 3(a). Often, aquifers are stratified with alternating layers of relatively permeable material separated by layers of less permeable materials. The flow in these systems may move from one aquifer through a leaky layer to another aquifer, and is referred to as multiaquifer flow.

2.06.7 The Dupuit Approximation The basic idea behind the Dupuit approximation (also called the Dupuit–Forchheimer approximation) is to approximate groundwater flow in an aquifer as two-dimensional flow in a horizontal plane. The approximation allows many problems to be solved in simple form that otherwise could not be solved. Conditions of the Dupuit approximation are commonly stated as (e.g., Bear, 1972): 1. the flow is horizontal (qz ¼ 0); 2. the hydraulic head is constant in the vertical (h ¼ h(x,y)); and 3. the hydraulic gradient is equal to the slope of the water table. There are various interpretations of the physical meaning of the Dupuit approximation. Bear (1972) shows that the head predicted with a Dupuit model in a single aquifer represents the average head over the depth of the aquifer. PolubarinovaKochina (1962) shows that Dupuit models are exact for anisotropic aquifers with infinite vertical hydraulic conductivity. This idea was explored further by Kirkham (1967). Strack (1984) showed that conditions 2 and 3 listed above are consequences of neglecting the resistance to vertical flow in an aquifer, and that condition 1 is unnecessary. Strack’s interpretation allows for the calculation of nonzero vertical flow components (qz) and three-dimensional pathlines in two-dimensional Dupuit models of single- and multiaquifer flow. Also, this interpretation clearly identifies where errors may be

120

Mechanics of Groundwater Flow

introduced by making the Dupuit approximation. Strack’s interpretation is adopted here and the Dupuit approximation is defined as neglecting the resistance to vertical flow in an aquifer. As stated, the major advantage of the Dupuit approximation is a two-dimensional head field (h ¼ h(x,y)) with two-dimensional horizontal flow components (qx ¼ qx(x,y), and qy ¼ qy(x,y)), while the vertical flow remains a function of all three coordinates (qz = qz(x,y,z)).

2.06.8 Potential Flow and the Discharge Vector ~ 2 =T is defined as the depth-inteThe discharge vector Q½L grated specific discharge vector. The x-component of the discharge vector is obtained as

Qx ¼

ZZt

qx ðx; y; zÞdz

ð16Þ

Zb

where Zb and Zt are the bottom and top elevations of the saturated portion of the aquifer, respectively. Upon making the Dupuit approximation in a shallow aquifer (16) becomes

Qx ¼ qx ðx; yÞ

ZZt

dz ¼ qx ðZt  Zb Þ

ð17Þ

Zb

The term within parentheses is the saturated thickness of the aquifer. For confined flow, the saturated thickness equals the aquifer thickness H. For unconfined flow, it is equal to h  Zb. In this chapter, the datum for h is chosen at the bottom of the aquifer (Zb ¼ 0), so that the saturated aquifer thickness is h. Substituting Darcy’s law into (17) and applying the appropriate saturated thicknesses gives

8 qh > > > < qx H ¼ kH q x ; confined flow Qx ¼ > > qh > : qx h ¼ kh ; unconfined flow qx

ð18Þ

flow, the product kH in (20) is referred to as the transmissivity T[L2/T] of the aquifer. Groundwater flow may be written as potential flow when the base of the aquifer is horizontal and the aquifer properties are piecewise constant. Writing groundwater flow as potential flow simplifies the formulation of confined and unconfined flow and allows the use of the many potential flow solutions that exist in other fields. The definition of potential flow is that the flow is equal to the gradient of the potential. For groundwater flow, the definition is modified by adding a minus sign (19). The discharge potential has no useful physical meaning, but is merely a convenient quantity in mathematical modeling. Using the discharge potential, Darcy’s law has been rewritten in terms of the discharge vector (19). The differential statement of conservation of mass may also be written in terms of the discharge vector. This may be done either by writing a flow balance on an elementary volume of aquifer (e.g., Strack, 1989), or by integrating the continuity equation over the depth of the aquifer (e.g., Bear, 1972). The result is

~¼N rQ

ð21Þ

where N [L/T] is the steady areal recharge rate, or the rate at which water infiltrates through the unsaturated zone into the saturated portion of the aquifer. If the aquifer is confined, or there is no recharge to the aquifer, N ¼ 0. Combining (19) and (21) results in Poisson’s equation

r 2 F ¼ N

ð22Þ

where the Laplacian is now understood to mean differentiation in the horizontal plane only. Confined and unconfined flow are handled in the same way in terms of the discharge vector. Boundary conditions are written in terms of F using ~ or a combination of F and (20), in terms of components of Q, ~ The resulting boundary-value problem is solved for F and Q. the results translated to heads, using the inverse of (20).

2.06.9 One-Dimensional Flow The y-component of the discharge vector is obtained in a similar manner. Equation (18) suggests the existence of a discharge potential, F [L3/T], from which the discharge vector may be calculated:

~ ¼ rF Q

ð19Þ

The following function satisfies both (19) and (18) and is the discharge potential (e.g., Strack, 1989):

( F¼

kHh  12kH 2 ; confined flow 1 2 unconfined flow 2kh ;

ð20Þ

When hZH, flow is confined, otherwise (or in the absence of a confining layer) it is unconfined. Equation (20) represents a single potential for combined flow; the potential is continuous across the interface where flow changes from confined to unconfined and the head in the aquifer is equal to the aquifer thickness (h ¼ H). For confined

The governing differential equation for steady one-dimensional, confined, unconfined, or combined flow in a shallow aquifer is (see (22))

d 2F ¼ N dx 2

ð23Þ

where F is related to hydraulic head, h, by (20). When the recharge rate N is constant, the general solution to this differential equation is

F ¼ 12Nx 2 þ Ax þ B

ð24Þ

where A and B are constants that must be evaluated from boundary conditions. If the flow is confined, or there is no areal recharge, N ¼ 0. Three examples of one-dimensional flow that demonstrate various boundary conditions to evaluate the constants are presented in the following.

Mechanics of Groundwater Flow 2.06.9.1 Confined Flow between Two Rivers

Application of the boundary conditions to the general solution with N ¼ 0 results in the following discharge potential:

The simplest case of confined flow is one-dimensional flow between two fixed-head boundaries, for example, two fully penetrating rivers as illustrated in Figure 4(a). When the aquifer is homogeneous, the solution shows that the head varies linearly between the head on the left and the head on the right. The head at the river to the left is equal to hL (hLZH), and the head at the river to the right is hR (hRZH). The boundary conditions must be written in terms of the discharge potential. From (20),

F¼

  FL  FR x þ FL L

ð26Þ

The discharge vector is obtained by differentiating (19):

Qx ¼ 

dF FL  FR ¼ L dx

ð27Þ

where Qx is constant throughout the aquifer. Alternatively, the discharge potential may be written as

F ¼ Qx x þ FL

FL ¼ Fðx ¼ 0Þ ¼ kHhL  12kH2 FR ¼ Fðx ¼ LÞ ¼ kHhR  12kH2

121

ð28Þ

ð25Þ 2.06.9.2 Combined Flow between Two Rivers If the head at the river on the right is below the aquifer confining unit (hRoH), the flow in the aquifer will be combined flow. In this case, FR is computed as (see (20))

FR ¼ 12 kh2R hL

z

H

k

The head at the left remains above the confining unit, and therefore FL is defined as before (25). The solution, written in terms of the discharge potential, (26), is still valid, as well as the expression for Qx (27). The location where the flow changes from confined to unconfined flow, as shown in Figure 4(b), is found by setting the discharge potential equal to kH2/2 and solving for the x-coordinate

hR

x L (a)

xc ¼ hL

FL  kH2 =2 Qx

ð30Þ

k hR

(b)

ð29Þ

2.06.9.3 Unconfined Flow in a River Valley Consider unconfined flow in a buried bedrock valley to a river of constant head, illustrated in Figure 5. There is no confining unit on the aquifer. The discharge potential is related to head by (20). The right aquifer boundary (x ¼ L) is the impermeable valley wall and the left boundary (x ¼ 0) is the river. The

xc

Figure 4 One-dimensional flow between two rivers: (a) confined flow and (b) combined confined and unconfined flow.

N

hL

k

z x

L Figure 5 One-dimensional unconfined flow in a river valley with recharge.

122

Mechanics of Groundwater Flow

corresponding boundary conditions are

Qx ðx ¼ LÞ ¼ 0

ð31Þ

Fðx ¼ 0Þ ¼ 12kh2L ¼ FL

ð32Þ

Application of the boundary conditions (31) and (32) to the general solution (24) results in

x  F ¼ Nx  L þ FL 2

ð33Þ

Radial flow in an aquifer is another case of one-dimensional flow. The governing equation for one-dimensional, radial, potential flow, written in radial coordinates r, is

ð34Þ

ð35Þ

where A and B are constants to be evaluated from boundary conditions. The discharge vector is obtained, as before, as minus the gradient of the discharge potential. In polar coordinates, the radial component of the discharge vector is Qr ¼ dF=dr. Radial flow problems and solutions are important in groundwater engineering because they represent the local flow field around pumping wells. Solutions to three example problems of radial flow are provided below.

2.06.10.1 Flow to a Well at the Center of a Circular Island without Recharge The discharge potential that is the solution to this problem is referred to as F1. The following boundary conditions fix the head at the perimeter of the island (r ¼ R) and the perimeter of the well (r ¼ rw), respectively:

F1 ðr ¼ RÞ ¼ FR

Q lnðr=RÞ þ FR 2p

ð36Þ

dF Q ¼ dr 2pr

ð37Þ

Application of the boundary conditions to the general solution, using N ¼ 0, results in the following discharge potential:

F1 ¼

FR  Fw lnðr=RÞ þ FR lnðR=rw Þ

ð38Þ

The discharge rate of the well, Q [L3/T], may be obtained by evaluating the discharge vector Qr(r ¼ rw) and multiplying by the perimeter of the well:

FR  Fw Q ¼ 2prw Qr ðr ¼ rw Þ ¼ 2p lnðR=rw Þ

ð39Þ

ð41Þ

In many two-dimensional problems or problems with recharge, the condition at the well is approximated as

Q ¼ lim 2prw Qr ðr ¼ rw Þ

ð42Þ

Condition (42) produces accurate results as the radius of the well is often much smaller than the horizontal scale of the groundwater problem being considered. However, only in simple radial flow cases are conditions (42) and (39) equivalent.

2.06.10.2 Recharge on a Circular Island The solution for recharge on a circular island is referred to as F2. Once again, the head is fixed at the perimeter of the island

F2 ðr ¼ RÞ ¼ FR

ð43Þ

By considering symmetry, the second boundary condition may be written as

Qr ðr ¼ 0Þ ¼ 0

ð44Þ

Application of the boundary conditions (43) and (44) to the general solution (35) results in the following discharge potential:

  F2 ¼ 14N r 2  R 2 þ FR

ð45Þ

Note that, by continuity of flow, the total groundwater discharge at the perimeter of the island is

Qr ðr ¼ RÞ2pR ¼ NpR2 F1 ðr ¼ rw Þ ¼ Fw

ð40Þ

Equation (40) is known as the Thiem equation. The radial component of the discharge vector for a steady well is

rw -0

The general solution to this differential equation is

F ¼ 14Nr 2 þ Alnr þ B

F1 ¼

Qr ¼ 

2.06.10 One-Dimensional Radial Flow

d 2 F 1dF r 2F ¼ þ ¼ N dr 2 r dr

Equation (39) is useful to compute the discharge for a desired head at the well. In practice, it is more common to know the discharge of a well. If the discharge of the well is known, the solution may be written as (combine (38) and (39))

ð46Þ

which may also be derived by taking the derivative of (45).

2.06.10.3 Well at the Center of a Circular Island with Recharge This problem contains both the features of the first two examples: recharge and a pumping well. Here the boundary conditions are specified as

F3 ðr ¼ RÞ ¼ FR

ð47Þ

Q ¼ lim 2prw Qr ðr ¼ rw Þ

ð48Þ

rw -0

Mechanics of Groundwater Flow

Application of the boundary conditions to the general solution yields the following discharge potential:

 Q 1  F3 ¼  N r 2  R 2 þ lnðr=RÞ þ FR 4 2p

ð49Þ

The head as a function of radial distance from the well is shown in Figure 6. The solid line represents the case for which the well pumps half the total areal recharge entering the aquifer, and the dashed line represents the case for which the well pumps exactly all the recharge on the island. Note that for the latter case, the flow at the perimeter of the island is zero and thus the phreatic surface is horizontal there; furthermore, the drawdown at the well is much larger than for the former case. It is emphasized that it is easy to create a case for which the well cannot pump all the infiltrated water. Theoretically, the maximum discharge is reached when the water level at the well is at the bottom of the aquifer; the practical limit is much less, of course. When the specified discharge in formula (49) is not possible, the potential at the well will be negative, and thus a head cannot be computed with the inverse of formula (20) for unconfined flow.

2.06.11 The Principle of Superposition Comparison of the solutions to the three example problems above reveals that the third solution is the sum of the first two solutions with the additive constant modified. Addition of multiple solutions to obtain another solution is an example of the principle of superposition, which is applicable to all linear differential equations, including the equations of Laplace and Poisson. The sum of the potentials of the first two problems in the previous section satisfies the differential equation of the third problem:

Similarly, the boundary condition at the well (48) is satisfied by the sum of the two potentials. Finally, the value of the sum of the two potentials at r ¼ R is a constant

F3 ðr ¼ RÞ ¼ F1 ðr ¼ RÞ þ F2 ðr ¼ RÞ ¼ 2FR

ð51Þ

This is not the value specified in the boundary condition (47), but this is easily corrected by modification of the additive constant. In this example of superposition, two radial solutions are added such that the resulting solution is also radial, one-dimensional flow. In general, however, superposition of radial flow solutions results in two-dimensional flow. As an example, consider two pumping wells of strengths Q1 and Q2 at locations (x1, y1) and (x2, y2) in an infinite aquifer. By superposition, the discharge potential is

F¼

Q1 Q2 lnr1 þ lnr2 þ A 2p 2p

ð52Þ

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi where r1 ¼ ðx  x1 Þ2 þðy  y1 Þ2 ; r2 ¼ ðx  x2 Þ2 þðy  y2 Þ2, and A is a constant that may be determined from a reference point of known potential. The resulting contours of head are presented in Figure 7 with dotted lines and indicate that the superposition of the two one-dimensional solutions results in a truly two-dimensional flow field. The solid lines in Figure 7 are streamlines.

2.06.11.1 A Well in Uniform Flow The case of confined flow to a well in an otherwise uniform flow field is another example of the principle of superposition. The potential for a uniform flow field with a head gradient G in the positive x-direction is given by

F ¼ TGx

ð53Þ

ð50Þ

Head

r 2 F3 ¼ r2 ðF1 þ F2 Þ ¼ r2 F1 þ r2 F2 ¼ 0  N ¼ N

123

R

R/2

0

R/2

R

Radial distance Figure 6 Head as function of radial distance for well at center of circular island with recharge: half the recharge is pumped by the well (solid), and all the recharge is pumped by the well (dashed).

Figure 7 Flow net for two wells with Q1 (left) larger than Q2: head contours (dotted) and streamlines (solid).

124

Mechanics of Groundwater Flow

such that Qx ¼ TG, where T is the transmissivity. The potential for a well in uniform flow is obtained through superposition of the potential for uniform flow (53) and a steady well located at the origin (40) plus an arbitrary constant F0:

F ¼ TGx þ

Q lnðr Þ þ F0 2p

ð54Þ

An example of head contours obtained from this solution is shown in Figure 8. The heavy line in Figure 8 is part of two streamlines that separate the groundwater flowing to the well from the groundwater that flows past the well. This dividing streamline forms the capture zone envelope of the well. It is important to protect drinking water wells from contamination and many countries have guidelines for the delineation and protection of capture zones for water supply wells. Guidelines commonly require protection of the zone of groundwater around the well that will be captured by the well within a certain period of time, for example, 5 years or 20 years. These capture zones and capture zones for other time periods all lie within the capture zone envelope. The dashed lines in Figure 8 represent the 5- and 20- year capture zones for this case. In most cases, capture zones are actually threedimensional parts of the aquifer, but they are commonly approximated as two-dimensional zones on a map. The width W of the capture zone envelope far upstream of the well in Figure 8 may be computed from continuity as

W ¼ Q=ðGT Þ

ð55Þ

At the well (x ¼ 0), the width of the capture zone is reduced to W/2. Special attention is paid to the point on the capture zone envelope farthest downstream of the well. This is a stagnation point, as the discharge vector is, theoretically, zero there. At the stagnation point, the effect of the well is exactly balanced by the hydraulic gradient of the uniform flow. The capture zone boundaries for large times approach the stagnation point, but only the boundary of the capture zone envelope passes through it. For this simple problem, the capture zones for any time period may be evaluated analytically (Bear and Jacobs, 1965). It is more common, however, to evaluate the capture zone boundaries by particle tracking methods (e.g., Strack, 1989; Bakker and Strack, 1996).

2.06.11.2 The Method of Images The method of images is an application of the superposition principle. Wells or other singularities are placed outside of the problem domain using symmetry to satisfy conditions specified along a boundary. For example, if the two wells in (52) have the same discharge (Q1 ¼ Q2 ¼ Q), and are placed symmetrically about the y-axis (x2 ¼  x1, y1 ¼ y2 ¼ 0), the discharge potential becomes

F¼

Q lnðr1 r2 Þ þ A 2p

ð56Þ

Figure 8 Head contours for a well in uniform flow (dotted). Capture zone envelope (heavy solid line), 5-year capture zone (small dashed contour), 20-year capture zone (large dashed contour).

Mechanics of Groundwater Flow

(a)

125

(b)

Figure 9 The method of images. Equipotentials (dotted) and streamlines (solid) for a well pumping near (a) an impermeable boundary and (b) a boundary of constant potential. The dots to the right of the flow field indicate the locations of the image wells.

Investigation of the behavior of this solution along the line passing midway between the wells (x ¼ 0) shows that the xcomponent of the discharge vector is zero. This potential is the solution to the problem of a well pumping next to an infinitely long impermeable boundary in a semi-infinite aquifer. As the problem domain lies to the left of the impermeable line, the well operating at ( þ x1, 0) is referred to as the image well. Contours of the discharge potential are shown in Figure 9(a). Another solution is obtained when the image well at ( þ x1, 0) is given the opposite discharge of the pumping well:

F¼

Q r1 ln þ A 2p r2

ð57Þ

Investigation of the behavior of this discharge potential shows that the potential is constant and equal to A along the line x ¼ 0. This discharge potential is the solution to the problem of a well pumping near a large lake or fully penetrating stream of constant potential A whose boundary lies along x ¼ 0. Again, the image wells lie outside the problem domain. Contours of the discharge potential are shown in Figure 9(b). Superposition and the method of images are two of the primary tools available to hydrologists and engineers for developing analytical solutions to steady and transient groundwater flow problems. Many analytical solutions to problems with wells and equipotential and/or impermeable boundaries may be obtained by the method of images. The method is also applicable to heterogeneity boundaries (Maxwell, 1873; Muskat, 1933) and leaky (Cauchy-type) boundaries (Keller, 1953; Anderson, 2000). The solution to a more complex and practical problem of groundwater flow is developed below.

2.06.11.3 Flow to a Pumping Well in an Alluvial Valley The problem of groundwater flowing in an alluvial aquifer in a bedrock valley is considered. The aquifer is unconfined and

receives areal recharge at a rate N; the governing differential equation is Poisson’s equation (22). The aquifer is bounded below by impermeable bedrock, and to the right by the bedrock wall of the buried valley as illustrated in Figure 10. The condition specified at the valley wall is

Qx ðx ¼ LÞ ¼ 0

ð58Þ

To the left the aquifer is bounded by a flowing river; the sloping head at the river is approximated with the condition

Fð0; yÞ ¼ Ay þ F0

ð59Þ

where F0 is the potential of the river at x ¼ 0 and A is approximately the slope of the water surface of the river. The condition at the pumping well is

Q ¼ lim 2prQr r-0

ð60Þ

where r2 ¼ (x  xw)2 þ (y  yw)2, and (xw,yw) are the coordinates of the pumping well. A solution is obtained by considering three simpler problems, each representing a particular feature of the whole problem, and applying superposition to the results. First, the effects of the recharge are considered. The problem of onedimensional flow along the x-axis from the valley wall (x ¼ L) to a boundary of zero constant potential at the river (F1(0, y) ¼ 0) was solved previously. Substitution of 0 for FL in (33) gives

x  F1 ¼ Nx  L 2

ð61Þ

Second, the effects of the well are considered. Laplace’s equation is solved subject to (58), (60), and F2(0, y) ¼ 0. The solution to this problem is obtained by repetitive use of the method of images about the two aquifer boundaries, using the elementary solutions (57) and (56), which results in an

126

Mechanics of Groundwater Flow

x

(xw ,yw )

Valley wall

Flowing river

y

Q

z x

L Figure 10 Definition sketch: flow to a well in an alluvial valley.

infinite sum of image wells:

F2 ¼

Q ðx  xw Þ2 þðy  yw Þ2 ln 4p ðx þ xw Þ2 þðy  yw Þ2 N QX þ ð1Þn ln 4p n¼1

(

"(

ðx  xw  2nLÞ2 þðy  yw Þ2 ðx þ xw  2nLÞ2 þðy  yw Þ2

ðx  xw þ 2nLÞ2 þðy  yw Þ2 ðx þ xw þ 2nLÞ2 þðy  yw Þ2

)#

)

ð62Þ

Third, the effect of the sloping stream is included. This solution satisfies Laplace’s equation and represents one-dimensional flow along the y-axis:

F3 ¼ Ay þ F0

ð63Þ

As this solution produces no flow in the x-direction, condition (58) is satisfied. By comparison of (63) and (59), it is seen that this boundary condition is also satisfied. The full solution is the sum of the three potentials (61), (62), and (63):

F ¼ F1 þ F2 þ F3

ð64Þ

A careful check of the solution shows that the correct differential equation is satisfied (Poisson’s equation), and that the boundary conditions (58) through (60) are satisfied exactly. Finally, as the flow is unconfined, the discharge potential is related to the head through equation (20). A threedimensional depiction of the groundwater table is shown for this example in Figure 11. The distance between the river and valley wall is L ¼ 500 m, and the distance between the well and the river is 100 m. The bottom of the aquifer is at z ¼ 0 m. In addition, contours are shown on the bottom of the figure.

2.06.12 The Stream Function and the Complex Potential Head contours have been defined previously as curves of constant head which allow us to visualize the variation of mechanical energy within a flow field. In homogeneous aquifers, head contours are equal to potential contours, called equipotentials. Streamlines allow for the visualization of the average paths of groundwater flow. Streamlines are defined as lines that are everywhere tangent to the discharge vector. Using this definition, a differential equation may be written for a

Mechanics of Groundwater Flow

127

40.5

40.0 Vall ey

wall

39.5

39.0 200

400 100 Riv er

300 0

200

−100 −200

100

Figure 11 A three-dimensional depiction of the groundwater table for a well pumping in an alluvial valley as shown in Figure 10. Contours of the head are shown on the bottom of the figure.

streamline

Qy cosa dy=ds ¼ ¼ Qx sina dx=ds

ð65Þ

where s is the position along the streamline and a is the angle between the s and x axes. Substituting in Darcy’s law the equation becomes



q F dx q F dy þ ¼0 q y ds q x ds

ð66Þ

By Darcy’s law, the discharge vector is everywhere normal to equipotentials. Examples are provided in Figure 9 which shows streamlines and equipotentials for flow to a well near an impermeable boundary and near an equipotential boundary. Note that the two sets of lines cross everywhere at right angles. Equipotentials and streamlines may both be drawn for problems of steady groundwater flow governed by the equations of Laplace and Poisson. However, in the case of Laplace’s equation, a special function exists – the stream function, or C(x, y) – whose contours represent streamlines. The stream function exists for a steady, two-dimensional, divergence-free groundwater flow. As the value of the stream function does not change along a streamline,

dC q C dx q C dy ¼ þ ¼0 q x ds q y ds ds

ð67Þ

By comparing (66) with (67), the following relationships are obtained between derivatives of the discharge potential and the stream function:

qF qC ¼ qx qy

ð68Þ

qF qC ¼ qy qx

ð69Þ

Equations (68) and (69) are known as the Cauchy–Riemann equations (e.g., Strack, 1989). It may also be shown that the stream function is single-valued, and harmonic (r2C ¼ 0). These properties of the stream function indicate that it is the harmonic conjugate of the discharge potential. Given a discharge potential that satisfies Laplace’s equation, the corresponding stream function may be evaluated from (68) and (69). The properties of the discharge potential and the stream function suggest the use of complex variables to solve groundwater flow problems governed by Laplace’s equation. There are many texts on complex variables including their use in solving groundwater flow problems (e.g., PolubarinovaKochina, 1962; Verruijt, 1970; Bear, 1972; Strack, 1989). The topic is only briefly discussed here. For groundwater problems governed by Laplace’s equation, a complex potential O exists which is an analytic function of the complex coordinate z ¼ x þ iy. The real part of the complex potential is the discharge potential and the imaginary part is

128

Mechanics of Groundwater Flow

2.06.13 Transient Flow

the stream function:

OðzÞ ¼ Fðx; yÞ þ iCðx; yÞ

ð70Þ

The negative derivative of the complex potential is the complex discharge

WðzÞ ¼ 

dO ¼ Qx  iQy dz

ð71Þ

Introduction of complex variables allows for the use of more sophisticated tools, including conformal mapping, to solve many groundwater flow problems. In particular, using complex variables allows for the simultaneous solution of the discharge potential and the stream function. An example demonstrating the utility of the stream function and the complex potential is presented in the following.

2.06.12.1 Evaluation of the Capture Zone Envelope Using the Complex Potential The complex potential for a well in an otherwise uniform flow field is

O ¼ TGz þ

Q lnz þ F0 2p

ð72Þ

Separation into real and imaginary parts shows that the discharge potential is the same as obtained previously (54):

F ¼ TGx þ

Q lnr þ F0 2p

ð73Þ

Q y 2p

ð74Þ

C ¼ TGy þ

where (r, y) are polar coordinates. The location of the stagnation point, zs, is evaluated as

Wðz ¼ zs Þ ¼ 0

ð75Þ

which gives

zs ¼

Q 2pTG

ð76Þ

The value of the complex potential at the stagnation point is

Oðz ¼ zs Þ ¼ 

 

Q Q 1  ln þ F0 2p 2pTG

ð77Þ

which is a purely real number. Therefore, the value of the stream function at the stagnation point is zero. The contour C ¼ 0 ¼ Cs defines the capture zone envelope:

TGy þ

Q y¼0 2p

ð78Þ

The equation for the capture zone envelope is obtained in polar coordinates using y ¼ r sin y and solving (78) for r:

r¼

Q y 2p TGsiny

ð79Þ

In the previous sections, steady-state flow was treated: the head was only a function of the spatial coordinates. In reality, the head is often also a function of time. When the head increases, more water is stored in the aquifer, and when the head decreases, less water is stored in the aquifer. For steady flow, continuity of flow states that the divergence of the discharge vector (21) is equal to the areal recharge rate N. When groundwater flow is transient, the divergence of the discharge vector is equal to the areal recharge plus the decrease in storage of water in the aquifer. The physics of the storage process is different for unconfined aquifers than for confined aquifers, but with suitable approximations, both lead to the same governing differential equation. The derivation of the governing equation for transient flow from the general statement of conservation of mass includes many approximations which are not discussed here. Rigorous derivations stating all necessary approximations are provided by Verruijt (1969) and Brutsaert (2005). First, consider a column of an unconfined aquifer with constant surface area A. When the head in the column is increased by an amount dh (i.e., the phreatic surface is raised dh), the volume of water in the column increases by an amount

dV ¼ SdhA

ð80Þ

were S [–] is the storativity of the unconfined aquifer. When the aquifer material above the phreatic surface is dry, the storativity of the unconfined aquifer is equal to the porosity. In practice, the storativity is always smaller than the porosity, as there is water present in the pores above the phreatic surface. The storativity of an unconfined aquifer is also called the specific yield. Next, consider a column of a confined aquifer with constant surface area A. When the head is now increased by dh, the volume of water still increases by an amount dV (80), but the storage coefficient is much smaller. Additional water can only be stored in the column through compression of the water and expansion of the aquifer. For most unconsolidated aquifers, the ability of the aquifer to expand is significantly larger than the ability of the water to compress, so that the compression of the water may be neglected. The storage coefficient of a confined aquifer is a function of the aquifer thickness: an aquifer of the same material but twice the thickness has a storage coefficient that is twice as large. The storage coefficient of a confined aquifer may be written as

S ¼ Ss H

ð81Þ

where Ss [L1] is the specific storage of the aquifer. Typical values for the specific storage of sand are between Ss ¼ 103 m1 and Ss ¼ 105 m1. Inclusion of the storage term in the divergence of the discharge vector (21) gives

~ ¼ S rQ

qh þN qt

ð82Þ

Mechanics of Groundwater Flow

1.0

where the areal recharge N may now vary with time. Using the potential for confined flow, this equation may be converted to

1 qF N D qt

0.8

ð83Þ

where the aquifer diffusivity D is defined as

D ¼ T=S

ð84Þ

and T is the transmissivity. The governing differential equation reduces to the diffusion equation when the areal recharge equals zero:

r 2F ¼

1 qF D qt

S qF N kh q t

ð86Þ

This nonlinear differential equation for transient unconfined flow is called the Boussinesq equation (Boussinesq, 1904). A common way to linearize the equation is to replace the head h in front of the time derivative on the right-hand side by an average head h (Strack, 1989), so that the diffusivity of an unconfined aquifer becomes D ¼ S=ðkhÞ: Note that after linearization, unconfined flow is also described by the diffusion equation (in absence of areal infiltration). Another way to linearize the differential equation for transient unconfined flow is to use the differential equation for transient confined flow, to approximate the transmissivity by T E kh and to use the storage coefficient for unconfined flow. The latter approach is used in this chapter. The solution of combined transient confined and transient unconfined flow is not as easy as it was for steady flow, because the storage coefficients differ between confined and unconfined flow. Exact solutions for transient groundwater flow are, not surprisingly, more difficult to obtain than those for steady flow. Common mathematical approaches include separation of variables, Fourier series, and Laplace or other transforms (e.g., Bruggeman, 1999). In this chapter solutions are presented, without derivation, for one-dimensional flow. These solutions are valid for both confined and for unconfined flow as long as the linearization of the differential equation for unconfined flow is reasonable.

2.06.13.1 One-Dimensional Periodic Flow Consider one-dimensional transient flow where the boundary condition varies periodically through time. The aquifer is semi-infinite and is bounded by open water at x ¼ 0; there is no areal infiltration and no flow at infinity. The water table at the boundary varies sinusoidally:

hð0; tÞ ¼ h0 þ Acosð2pt=tÞ

0.6

0.4

0.2

0.0

ð85Þ

The diffusion equation governs the transient behavior of many other physical processes. Using the potential for unconfined flow, the continuity equation (82) may be written as

r 2F ¼

Damping

r 2F ¼

129

ð87Þ

0

2

4

6

8

10

x/ or r/ Figure 12 Damping of the amplitude with distance for one-dimensional periodic flow. Damping of the head when head at boundary varies as (87) (solid) and damping of the radial flow when discharge of a well varies as (90) (dashed).

where A is the amplitude of the fluctuation and t is the time period of the fluctuation. The sinusoidal fluctuation in the surface water (87) may be caused, for example, by tides, by the periodic operation of hydroelectric dams, or by seasonal fluctuations of the surface water level. Solutions to problems of periodic flow may be obtained by separtation of variables. The solution to this problem is

pffi F ¼ Th0 þ AT< expðx i=l þ 2pit=tÞ

ð88Þ

where < stands for taking the real part of the complex function, and l is a characteristic length defined as

l¼

pffiffiffiffiffiffiffiffiffiffiffiffiffiffi tD=2p

ð89Þ

The amplitude A dampens away from the open water as pffiffiffi exp½x=ðl 2Þ; and is shown in Figure 12. At a distance of 3l, the amplitude has damped to less than 5% of the amplitude at x ¼ 0, and at a distance of 6l, the amplitude has damped to less than 0.25% of the amplitude at x ¼ 0. This result may be used as a rule of thumb to assess whether fluctuations in surface water levels need to be taken into account when considering the head and flow in an aquifer. If the area of interest is farther away from a surface water body than 6l, periodic fluctuations of the surface water level with a period of t may be neglected. Note that l is a function of the period t: the longer the period t, the larger the characteristic length l. Fluctuations with different periods and amplitudes may be superimposed in time. An arbitrary fluctuation of the water level may be approximated by a Fourier series. A similar analysis may be carried out for a well with an average discharge of Q0 and a sinusoidal discharge with an amplitude of Q0:

QðtÞ ¼ Q0 þ Q0 cosð2pt=tÞ

ð90Þ

At a certain distance from the well, the sinusoidal fluctuation of the discharge is unnoticeable and it seems that the well pumps with a steady discharge Q0. This distance depends

130

Mechanics of Groundwater Flow

again on the characteristic length l (89). The discharge vector for a well with a constant discharge Q0 is given by Equation (41). The relative difference between the radial flow caused by the well with sinusoidal discharge (90) and the flow caused by a well with constant discharge Q0 is 4.6% at a distance of 6l, reducing to 0.3% at 10l (see Figure 12). Hence, a well with a periodic discharge (90) varying between 0 and 2Q0 may be represented by a well with steady discharge Q0 beyond a distance of 10l from the well.

water source closer than infinity, and if that source is included in the solution, the transient solution will approach a steady solution for large time. For example, consider a well at (  x1, y1) near a large lake with a constant potential A along y ¼ 0; the steady solution was obtained with the method of images and is given in Equation (57). A transient solution may also be obtained with the method of images as

F¼

   

Q Sr21 Sr22  E1 þA E1 4Tðt  t0 Þ 4Tðt  t0 Þ 4p

ð96Þ

2.06.13.2 Transient Wells In Equation (40) the solution was presented for steady flow to a well with discharge Q. Here, the transient equivalent is discussed. At time t ¼ t0 the head in the aquifer is constant and equal to h0 everywhere and a well starts pumping with discharge Q. The head h0, and thus the corresponding potential F0, at infinity remains constant throughout time:

FðN; tÞ ¼ F0

ð91Þ

This problem may be solved as a similarity solution or by Laplace transforms. The potential as a function of time and the radial distance from the well is



F ¼ F0 þ

Q Sr 2 E1 4p 4Tðt  t0 Þ

 ;

t  t0

ð92Þ

where E1 is the exponential integral defined as

E1 ðuÞ ¼

ZN

expðsÞ ds s

ð93Þ

u

Solution (92) is known as the Theis solution (Theis, 1935). The head is a function of only one dimensionless parameter, u

u¼

Sr 2 4Tðt  t0 Þ

ð94Þ

Hence, if a certain drawdown h0  h(r1, t1) is reached at a distance r1 at time t1, the same drawdown is reached at a distance 2r1 at time 4t1. A common approximation for E1 is the series

E1 ðuÞ ¼ g  ln u 

N X ðuÞ n n¼1

nðn!Þ

ð95Þ

where g ¼ 0.5772y is Euler’s constant. The infinite series in (95) converges quickly (when uo1), so that in practice only a small number of terms needs to be used. One might expect that if the well is pumped for a longenough period of time, the head will approach a steady-state position. This is not the case: the Theis solution (92) does not approach the Thiem solution (40) for large time. For the Thiem solution, the head approaches infinity when r approaches infinity, because the source of water for the Thiem solution lies at infinity. The Theis solution approaches h0 when r approaches infinity according to (91), and all the pumped water comes from storage. In reality, there is always a

When time approaches infinity, u approaches zero, and E1 may be represented with the first two terms of (95). Substitution of these terms for E1 in (96) leads to the steady solution (57). Even though the head of the Theis solution by itself does not approach the steady–state head of the Thiem solution, the discharge vector does approach the steady-solution. The radial flow Qr of the Theis solution may be obtained through differentiation of (92) to give

Qr ¼ 

Q expðuÞ 2pr

ð97Þ

It is seen from this equation that when time approaches infinity, and u approaches zero, Qr approaches the steady discharge vector (41). The consequence is that head gradients in the Theis solution approach the steady head gradients obtained with the Thiem solution, even though the head values themselves do not. The Theis solution is very useful to determine aquifer parameters from a pumping test. During a pumping test, a well is turned on and the drawdown is measured in a nearby observation well. The Theis solution may be fit to observed head data to determine the transmissivity T and the storage coefficient S in the neighborhood of the well. Tansient solutions may be superimposed in time as well as in space. For example, consider a well with a discharge Q operating from t ¼ t0 to t ¼ t1 and with zero discharge after t1. For the period t 4 t1, the potential may be represented by two Theis wells, one with a discharge Q starting at t ¼ t0 and one with a discharge –Q starting at t ¼ t1:

   

Q Sr 2 Sr 2 F¼ E1  E1 ; 4p 4Tðt  t0 Þ 4Tðt  t1 Þ

t  t1

ð98Þ

This is called a pulse solution, where the pulse lasts from t0 until t1.

2.06.13.3 Convolution In the last example of the previous section, a solution was presented for a well that pumped with a discharge Q from t0 to t1. When the pumping period is 1 day, this solution may be used to compute the head variation caused by a well for which daily discharge records are available. This requires the repeated superposition through time of solution (98), called convolution. This solution approach is an example of a standard technique to solve differential equations. More formally, the approach is based on the determination of the solution for a unit impulse, in this case a discharge of unit volume over a short period, theoretically an infinitely short period. The

Mechanics of Groundwater Flow

131

defined as

Fj ðx; y; tÞ ¼ Yðx; y; t; tj Þ  Yðx; y; t; tjþ1 Þ;

t  tjþ1

ð103Þ

Response

where tj ¼ jDt. An example of a pulse response is given in Figure 13. Consider the case for which the applied stress is known over periods of equal length Dt, tn is defined as tn ¼ nDt, and Qn is the stress from t ¼ tn until t ¼ tnþ1. The potential at time tn may be computed with the convolution sum:

Fðx; y; tn Þ ¼

n1 X

Qj Fj ðx; y; tn Þ;

n1

ð104Þ

j¼0

t0

t0 + Δt

Time

Figure 13 Examples at one specific point for an impulse response (solid), step response (dashed), and pulse response (dash-dotted) for a well near a long straight river; the pulse response is identical to the step response for the period of the pulse Dt.

response due to a unit impulse is called the impulse response function (e.g., Figure 13). For a well, the impulse response function y of the potential is

yðr; tÞ ¼

1 expðuÞ 4p t

ð99Þ

The potential for a time-varying discharge Q(t) is obtained with the convolution integral (Duhamel’s principle):

Fðr; tÞ ¼

Zt

Qðt 0 Þyðr; t  t 0 Þdt 0

ð100Þ

N

The Theis equation (92) may be obtained with the convolution integral by specifying the discharge as Q(t0 ) ¼ Q for t0 Z t0 in (99), and using that

dt 0 du ¼ t  t 0 uðr; t  t 0 Þ

ð101Þ

The Theis equation is an example of a step response (e.g., Figure 13): at time t0, the discharge changes from 0 to Q. In general, the unit step response Y is obtained from the impulse response through integration

Yðx; y; t; t0 Þ ¼

Zt

yðx; y; t  t 0 Þdt 0

ð102Þ

t0

where the step occurs at time t0. For practical application, the convolution integral is often written as a sum of pulse response functions. An example of a pulse response was given by the last example in the previous section, where a well was pumped at a constant discharge for a finite period. In general, the pulse response Fj for a pulse of length Dt starting at t ¼ tj is

The convolution approach assumes that the system is linear. Nonlinear behavior may occur, for example, in the summer time when pumping is at its peak and there is little rainfall. During such periods, ditches or streams may go dry, which means that the hydrological system and thus the impulse response function change. In such cases it is not possible to simulate the head variation with a straightforward convolution. In practice, when a system is sufficiently linear, the convolution approach works very well. The pulse response is different for different stresses (areal recharge, pumping, lake-level changes) and needs to take into account all nearby boundary conditions. Once the pulse response for a stress is known at a point, the head variation may be simulated using the convolution approach. Consider again the problem of unconfined flow in a buried bedrock valley to a river of constant head, as illustrated in Figure 5. The solution for a constant recharge rate is given in Equation (33). Instead of a constant recharge rate, the recharge now varies daily as shown in Figure 14(a) for a period of 7 years. Note that the recharge is negative for days without rainfall due to evaporation. The step response for this problem may be obtained with the Laplace transform technique and is given in Bruggeman (1999, Eq. 133.16). Alternatively, the pulse response may be obtained with a computer model; computer models are discussed in the next section. Convolution of the recharge with the step response gives the head variation. The head variation at the valley wall (x ¼ L) and near the river at x ¼ 0.1L are shown in Figure 14(b). Note that the total head variation at the valley wall (B1.8 m) is much larger than near the head boundary (B0.4 m). The head variation at the valley wall has a long memory: the head value depends on the recharge that fell almost 2 years ago. In other words, the pulse response approaches zero after approximately 2 years. The head variation is not shown for the first 2 years in Figure 14(b) as it would require recharge information prior to the record shown in Figure 14(a). When heads are measured, they always show the effect of recharge, as shown in Figure 14. Most head measurements also show the effect of barometric variations and earth tides. The latter are often undesirable and need to be removed; a computer program to remove these variations is called BETCO, which is available for download from the Internet. Time series of head observations always show the effect of the different time-varying stresses that act on the groundwater system. A stochastic approach called ‘time series analysis’ may

132

Mechanics of Groundwater Flow 35

Recharge (mm)

30 25 20 15 10 5 0 −5 2001

2002

2003

(a)

2004 2005 2006 Time, beginning of the year

2007

2008

2004 2005 2006 Time, beginning of the year

2007

2008

2.0

Head (m)

1.5 1.0 x=L

0.5

x = 0.1L

0.0 −0.5 2001 (b)

2002

2003

Figure 14 Daily recharge rate (a) and head variation (b) at valley wall (x ¼ L, large fluctuation), and near specified-head boundary (x ¼ 0.1L, small fluctuation) for the system shown in Figure 10.

be used to unravel the series and compute the head variations due to the individual stresses. The traditional method for time series analysis is the Box–Jenkins method (Box and Jenkins, 1970). Recently, the PIRFICT method was developed for time series analysis of hydrological data (Von Asmuth et al., 2008). The PIRFICT method uses predefined, parameterized shapes for the impulse response functions and allows for irregular time series, and time series with missing data.

2.06.14 Computer Models The relatively simple solutions presented in this chapter may be used to solve real problems. They may be used for first estimates, to verify more complicated models, and to gain insight in the flow problem. In many cases, however, the setting is more complicated than, for example, a well near a long, straight lake boundary. To obtain solutions for more complicated problems, general solution approaches for the governing differential equations are implemented in computer programs. These computer programs may be applied to simulate groundwater flow in domains with more complicated boundary shapes, with a variety of boundary conditions, as well as flow in aquifers that are not homogeneous. The resulting computer models remain an approximation of reality and the modeler must decide what details to put into the model based on the purpose of the model. Most existing computer programs for modeling groundwater flow are based on one of three methods to solve the mathematical problem: the analytic element method, the finite difference method, or

the finite element method. Characteristics of these three methods are discussed here briefly and some references to free software are given. The analytic element method is based on the superposition of analytic functions (Strack, 1989; Haitjema, 1995; Fitts, 2002). In this respect, it is an extension of many of the solutions presented in this chapter. Each analytic function represents a hydrogeologic feature in the aquifer, such as a well, the section of a stream, or the boundary between two geologic formations. Each analytic element has at least one free parameter. The free parameter of an element may be specified or a condition may be specified so that the computer program can compute the value of the free parameter. For example, the free parameter of a well is its discharge. The discharge may be specified, or the modeler may specify the desired head (or drawdown) at the well, and the computer program computes the corresponding discharge (often simultaneously with the other free parameters in the model). An advantage of the analytic element method is that the head and flow can be computed analytically at any point in the aquifer. The model extends, theoretically, to infinity, but has no practical significance beyond the area where sufficient analytic elements are defined to simulate the flow. Many analytic elements exist, including wells, stream segments with and without leaky bottoms, boundaries between zones with different aquifer properties, leaky and impermeable walls, areal recharge, and lakes. Most currently available analytic element programs are restricted to steady flow in piece-wise homogeneous aquifer systems. Analytic element approaches have been developed for

Mechanics of Groundwater Flow

transient flow through piecewise homogeneous aquifers (e.g., Bakker, 2004; Kuhlman and Neuman, 2009) and for flow through aquifers with continuously varying properties (e.g., Craig, 2009). Analytic elements are ideally suited for implementation in an object-oriented computer code; a simple design is presented in Bakker and Kelson (2009). Several analytic element programs are available for modeling steady flow. Single aquifer codes include WhAEM, which contains a graphical interface, and Split. An approach for steady multiaquifer flow was developed by Bakker and Strack (2003) and is implemented in the program TimML. A graphical user interface for both Split and TimML is VisualAEM. Commercial programs are available as well, but are not discussed here. The most popular computer program for modeling groundwater flow is MODFLOW (Harbaugh et al., 2000), which is based on the finite difference method and is available by the download form the internet. MODFLOW model domains are discretized in a grid of rectangles, called cells. Heads are computed at cell centers and flows are computed between cell centers. Hydrogeologic features need to be simplified to fit the chosen grid. The governing differential equation is solved approximately by replacing the derivatives in the differential equation by difference equations. Transient solutions are obtained by stepping through time. The cell size and time step need to be chosen sufficiently small to obtain an accurate solution. A condition must be specified along the entire model boundary. This is in contrast to analytic element models, which do not have a formal boundary. The model boundary should be chosen, where possible, along real hydrogeologic boundaries, such as rivers or impermeable rock outcrops. This is never possible everywhere, however, especially in deeper aquifers. Hence, along parts of the boundary artificial boundary conditions need to be specified based on the modeler’s expert knowledge. The finite difference method is relatively easy to implement in computer codes and allows for continuously varying aquifer properties. Many specialized packages exist for MODFLOW to model a variety of features (e.g., wells, lakes, and drains) and flow types (e.g., unsaturated flow). Seawater intrusion may be simulated with the SWI package or SEAWAT. Creation of input files for MODFLOW is cumbersome. Many powerful graphical interfaces are available commercially. A free graphical interface is version 5 of PMWIN. Python scripts to run MODFLOW are available by the download from the Internet. The finite element method also requires a discretization of the model area, although the common choice is a discretization in triangles. Heads are computed at the corners of the triangles (called nodes) and flows are computed between them. The mathematics behind the finite element method is based on a minimization principle, which is more complex than the finite difference method. Grids of triangles are more flexible than grids of rectangles as it is much easier to represent shapes of hydrogeologic features with triangles, and small triangles can be used where needed. Sophisticated grid builders are available to construct complicated grids of triangles. Other practical advantages and disadvantages of the finite element method are similar to the finite difference method. It is more complicated to implement the finite element method in a computer code than the finite difference method. A commonly used free finite element code is SUTRA, which can also be used to model unsaturated flow and variable density flow.

133

2.06.15 Discussion The basic principles of groundwater flow were discussed in this chapter. Essential principles, such as Darcy’s law, the Dupuit approximation, and the derivation of Laplace’s equation, originate from the nineteenth century. Many analytic solutions were developed in the twentieth century, while numerical methods and computer models became important in the last few decades. The size of numerical models has grown over the years with the available computational power. Evaluation in other fields shows that the size of modeling grids follows Moore’s law and doubles approximately every 1.5 years, as does the computational power (Voller and Porte´ Agel, 2003). Analytic solutions play an important role in the evaluation of numerical model results. Analytic formulas may be used to assess whether the results of a numerical model are approximately correct. Such a comparison often shows that application of a simple analytic model provides a very reasonable solution. A number of discussions have been published on the future of hydrogeology in the first decade of the twenty-first century (e.g., Voss, 2005; Miller and Gray, 2008). Progress is being made in the modeling of heterogeneous domains using advanced calibration tools and assessment of the predictive ability of groundwater models (e.g., Doherty, 2008) and stochastic modeling (e.g., Zhang and Zhang, 2004). Another active area of research is the linkage of groundwater models with unsaturated zone models, surface water models, and ultimately atmospheric models. Such linkages run into serious issues of differences in temporal and spatial scales that have yet to be resolved satisfactorily. As in other areas of hydrology, some groundwater models try to include details because they exist, not because they matter (Haitjema, 1995). The inability to capture the full complexity of systems and processes makes the search for accurate simplifications a continuing endeavor.

Acknowledgments We gratefully acknowledge Tanja Euser for creating Figures 1– 5 and 32.10.

References Anderson EI (2000) The method of images for leaky boundaries. Advances in Water Resources 23: 461--474. Bakker M (2004) Transient analytic elements for periodic Dupuit–Forchheimer flow. Advances in Water Resources 27(1): 3--12. Bakker M and Kelson VA (2009) Writing analytic element programs in Python. Ground Water 47(6): 828--834. Bakker M and Strack ODL (1996) Capture zone delineation in two-dimensional groundwater flow models. Water Resources Research 32(5): 1309--1315. Bakker M and Strack ODL (2003) Analytic elements for multiaquifer flow. Journal of Hydrology 271(1–4): 119--129. Bear J (1972) Dynamics of Fluids in Porous Media. New York: Dover Publications. Bear J and Jacobs M (1965) On the movement of water bodies injected into aquifers. Journal of Hydrology 3: 37--57. Boussinesq J (1904) Recherches the´oriques sur le coulement des nappes d’eau infiltre´es dans le sol. Journal de Mathe´matiques Pures et Applique´es 10: 5--78. Box GEP and Jenkins GM (1970) Time Series Analysis, Forecasting and Control. San Francisco, CA: Holden-Day.

134

Mechanics of Groundwater Flow

Bruggeman GA (1999) Analytical Solutions of Geohydrological Problems, Developments in Water Science, 46, Amsterdam: Elsevier. Brutsaert W (2005) Hydrology – An Introduction. New York: Cambridge University Press. Craig JR (2009) Analytic elements for flow in harmonically heterogeneous aquifers. Water Resources Research 45: W06422 (doi:10.1029/2009WR007800). Darcy H (1856) Les Fourntaines Publiques de la Vlle de Dijon. Paris: Dalmont. Doherty J (2008) Manual and Addendum for PEST: Model Independent Parameter Estimation. Brisbane, Australia: Watermark Numerical Computing. Dupuit J (1863) E´tudes The´oriques et Practiques sur le Mouvement des Eaux dans les Canaux Decouverts et a´ Travers les Terrains Perme´ables, 2nd ed. Dunod: Paris. Fitts CR (2002) Groundwater Science. New York: Academic Press. Forchheimer P (1886) Ueber die Ergiebigkeit von Brunnenanlagen und Sickerschlitzen. Z. Architekt. Ing. Ver. Hannover 32: 539--563. Haitjema HM (1995) Analytic Element Modeling of Groundwater Flow. San Diego, CA: Academic Press. Harbaugh AW, Banta ER, Hill MC, and McDonald MG (2000) MODFLOW-2000, the US Geological Survey modular ground-water model-user guide to modularization concepts and the ground-water flow process. USGS Open-File Report 00–92. Jacob CE (1940) The flow of water in an elastic artesian aquifer. Transactions, American Geophysical Union 21: 574--586. Keller JB (1953) The scope of the image method. Communications on Pure and Applied Mathematics VI: 505--512. Kirkham D (1967) Explanation of paradoxes in Dupuit–Forchheimer seepage theory. Water Resources Research 3(2): 609--622. Kuhlman KL and Neuman SP (2009) Laplace-transform analytic-element method for transient porous-media flow. Journal of Engineering Mathematics 64(2). 113–13. Maxwell JC (1873) A Treatise on Electricity and Magnetism, vol. 1, Oxford: Clarendon Press. Meinzer OE (1928) Compressibility and elasticity of artesian aquifers. Economic Geology 23: 263--291. Miller CT and Gray WG (2008) Hydrogeological research, education, and practice: A path to future contributions. Journal of Hydrologic Engineering 13(7): 7--12. Moore C and Doherty J (2006) The cost of uniqueness in groundwater model calibration. Advances in Water Resources 29: 605623. Munson BR, Young DF, and Okiishi TH (2002) Fundamentals of Fluid Mechanics, 4th edn. New York: John Wiley and Sons. Muskat M (1933) Potential distribution about an electrode on the surface of the Earth. Physics 4(4): 129--147. Polubarinova-Kochina PY (1962) Theory of Groundwater Movement. De Wiest JMR. (trans.). Princeton, NJ: Princeton University Press.

Strack ODL (1984) Three-dimensional streamlines in Dupuit-Forchheimer models. Water Resources Research 20(7): 812--822. Strack ODL (1989) Groundwater Mechanics. Englewood Cliffs, NJ: Prentice Hall. Theis CV (1935) The relation between the lowering of the piezometric surface and the rate and duration of discharge of a well using ground-water storage. Transactions, American Geophysical Union 16: 519524. Verruijt A (1969) Elastic stage of aquifers. In: de Wiest RJM (ed.) Flow Through Porous Media, pp. 331--376. New York: Academic press. Verruijt A (1970) Theory of Groundwater Flow. New York: MacMillan. Voller VR and Porte´ Agel (2003)) Moore’s law and numerical modeling. Journal of Computational Physics 172(2): 698--703. Von Asmuth JR, Maas K, Bakker M, and Petersen J (2008) Modeling time series of ground water head fluctuations subjected to multiple stresses. Ground Water 46(1): 30--40. Voss CI (2005). The future of hydrogeology. Hydrogeology Journal 13(1): 1–6. Zhang Y-K and Zhang D (2004) Forum: The state of stochastic hydrology. Journal of Stochastic Environmental Research and Risk Assessment 18(4): 265.

Relevant Websites http://www.civil.uwaterloo.ca Civil and Environmental Engineering, University of Waterloo; James R. Craig, Visual AEM. http://www,epa.gov EPA: United States Environmental Protection Agency; Ecosystems Research Division, WLAEM2000. http://code.google.com Google.Wigaem; timml; flopy. http://www.groundwater.buffalo.edu Groundwater Research Group, UB Groundwater Group Software. http://www.hydrology.uga.edu Hydrology@University of Georgia; BETCO: Barometric and Earth tide Correction. http://bakkerhydro.org Mark Bakker, SWI package. http://www.pmwin.net PMWiN.NET by Wen-Hsing Chiang, PMWIN Version 5.3. http://water.usgs.gov USGS: U.S. Geological Survey. SEAWAT; SUTRA Version 2.1; MODFLOW-2000 version 1.18.01.

2.07 The Hydrodynamics and Morphodynamics of Rivers N Wright, University of Leeds, Leeds, UK A Crosato, UNESCO-IHE, Delft, The Netherlands & 2011 Elsevier B.V. All rights reserved.

2.07.1 2.07.2 2.07.2.1 2.07.2.1.1 2.07.2.1.2 2.07.2.1.3 2.07.2.2 2.07.2.2.1 2.07.2.3 2.07.2.4 2.07.2.5 2.07.2.6 2.07.2.7 2.07.2.7.1 2.07.2.7.2 2.07.2.8 2.07.2.9 2.07.2.10 2.07.2.10.1 2.07.2.10.2 References

Early History of Hydrodynamics and Morphodynamics in Rivers and Channels State of the Art in Hydrodynamics and Morphodynamics Fluid Flow Mass Momentum Energy Numerical Solution Boundary conditions Depth and Process Scales Cross-Section Scale River Reach Scale Spatial Scales in River Morphodynamics Geomorphological Forms in Alluvial River Beds Ripples and dunes Bars River Planimetric Changes Bed Resistance and Vegetation Discussion of Current Research and Future Directions Incremental changes Step changes

2.07.1 Early History of Hydrodynamics and Morphodynamics in Rivers and Channels The study of flow in open channels and their shape is inextricably linked to the study of fluid dynamics more generally, and hydrodynamics can perhaps be best defined as the application of the theory of fluid dynamics to flows in open channels. Early work on the general properties of fluids was carried out by the ancient Greeks. They studied many fluid phenomena, and the work of Archimedes on hydrostatics is well known. However, it was the Romans who demonstrated a more practical knowledge of fluid flow and open-channel flow in particular. They constructed advanced water-supply systems including aqueducts and water wheels. Archaeological evidence confirms their use of sophisticated siphon systems that required advanced techniques to seal the pipes in order to maintain the necessary pressures and this is likely to have required an understanding of pressure and fluid potential energy. Unfortunately, there is no documentary evidence of the knowledge that they had, as it was a practical skill. In Islamic civilizations around the ninth century, engineers and physicists studied fluid flow and made use of hydraulics through water wheels in order to process grain and carry out other mechanical tasks. They also engineered channels for irrigation and developed the systems of qanats for irrigation. Chinese engineers also harnessed energy by using water wheels to power furnaces.

135 137 138 138 138 138 139 139 139 140 141 141 143 143 145 146 148 151 152 152 152

Despite its widespread use and study the theory of open channel flow did not advance, and by the beginning of the nineteenth century the study of flow in pipes was probably more advanced, particularly in its mathematical description. This reflects the intrinsic difficulty of open-channel flow that is often not fully appreciated by a cursory examination. Under more detailed examination, it becomes clear that we do not know a priori what the depth will be in a channel as opposed to full pipe flow where the cross-sectional area is known: that is, the relationship between depth (m), discharge (m3 s1), and cross-sectional geometry cannot be expressed in a simple formula. In essence, this is the fundamental question to be answered by both theoreticians and practitioners. The situation is further complicated by the high variation in bed and bank material. Due to this complexity, early studies were empirical. The first step to a more mathematics- and physics-based approach had been taken by Leonardo da Vinci (1452–1519). His book entitled Del moto e misura dell’acqua (Water Motion and Measurement), written in around 1500 and published in 1649 after his death, is a treatise of nine individual books, of which the first four deal with open-channel flows (Graf, 1984). In this book, da Vinci made an early attempt to formulate the law of continuity linking the water flow to channel width, depth, slope, and roughness. Nevertheless, the founder of river hydraulics has been traditionally viewed as Benedetto Castelli (1577–1644), a pupil of Galileo Galilei, who wrote

135

136

The Hydrodynamics and Morphodynamics of Rivers

the book entitled Della misura delle acque correnti (Measurement of Water Flows) (1628), in which he explained the law of continuity in more precise terms. It is perhaps worth noting that Castelli was also engaged by the Pope as a consultant on the management of rivers in the Papal States, reflecting the combination of theoretical and practical approaches. Sir Isaac Newton (1642–1727) discussed fluid statics and dynamics at length in his Principia Mathematica (1687) (Anderson, 2005). He proposed his law of viscosity stating that shear stress was proportion to the velocity gradient with the constant of proportionality being the viscosity. Newton’s work informed later studies and Prandtl used the shear stress relationship to create an analogy for turbulent flow. In the eighteenth century, work on the fundamental mathematical description of fluid mechanics was advanced by Daniel Bernoulli (1700–82), Jen le Rond d’Alembert (1717–83), and Leonhard Euler (1707–83). The latter used momentum and mass conservation to derive the Euler equations for fluid flow and these were not surpassed until the Navier–Stokes equations were derived with their treatment of viscous shear stress. These were derived independently by Claude-Louis Navier in 1822 and George Stokes in 1845 (Anderson, 2005). The Navier–Stokes equations were of a general nature. In terms of open-channel flow, it was realized that the key parameters were discharge (m3 s1), depth (m), cross-section geometry, longitudinal bed slope, and the nature of the bed and banks. The cross-section geometry is clearly infinitely variable and difficult to encapsulate in a formula and so key geometric properties were chosen to represent it. These are wet area (A (m2)) and wetted perimeter (P (m)), and these are often used to derive the hydraulic radius, R ( ¼ A/P (m)). Based on this theory, Che´zy (1717–98) developed his theory of open-channel flow as balance of the frictional and gravitational force. He proposed the formula

pffiffiffiffiffiffi V ¼ C RS

ð1Þ

where C is the Che´zy coefficient (m1/2 s1), R the hydraulic radius (m), and S the longitudinal bed slope (m m1). Although C is often assumed to be constant for a given channel, it has dimensions and does vary with the water depth. Later Manning proposed an alternative formula based on his measurements and this has been widely adopted in the English-speaking world:

1 V ¼ R 2=3 S 1=2 n

ð2Þ

where n is Manning’s coefficient. Again, this is dimensional (m1/3 s1) and varies with water depth. The formulations by Che´zy and Manning are valid for flows that are steady state and uniform. These assumptions clearly do not apply in many cases, particularly in natural rivers. In a treatise published in 1828, Be´langer put forward an equation for a backwater in steady, one-dimensional (1-D) gradually varied flows, that is, flows with constant discharge, but gradually varied depth (Chanson, 2009). This equation can be used to qualitatively assess the flow profile in a section of a river and further allows for the analysis of the profile across a series of different reaches with different characteristics

(Chanson, 1999). It still uses Che´zy or Manning to calculate a friction slope, but it must be borne in mind that this takes these equations beyond their validity. A full solution of the backwater equation is not possible with a closed or continuous solution, but it is possible to use discrete, stepping methods to calculate solutions as a set of points moving away from a control section. This is one of the early examples of numerical solution. Be´langer used the direct step method to calculate the longitudinal distance taken for a given depth change, and other methods such as the standard step method, Euler method, and predictor–corrector methods have subsequently been developed. Be´langer also recognized the importance of the Froude number, which is the ratio of momentum to gravitational effects in an open channel and which governs whether information can flow upstream, in a similar way to its analogy, the Mach number, in compressible gas dynamics. Be´langer also identified that there were singular points in the solution of the backwater equations where the flow was critical and where the Froude number has the value of 1. The ability to calculate gradually varied flow allowed for the calculation of water profiles between control points and critical points, but it is not applicable at the control points themselves. These control points include structures such as weirs, sluices, and bridges which were increasingly being used in the nineteenth century as a result of the industrial development in Europe. Be´langer paid much attention to the phenomenon of the hydraulic jump. This is observed when the water flow changes from a shallow, fast flow with a Froude number greater than 1 to a flow that is deep and slow with a Froude number less than 1. This transition cannot occur smoothly and is therefore highly turbulent and complex. Be´langer used the momentum concept to derive an equation relating the depths upstream and downstream of the jump (the conjugate depths). After a first attempt, he presented his complete theory in 1841 (Chanson, 2009) and the equation bearing his name is still in use today. Be´langer also went on to examine other control structures such as the broad-crested weir. This formed the basis of the study of rapidly varied flows using the concept of specific energy to obtain insight into the phenomena. Further progression in 1-D open-channel flow led to the development of the full shallow water equations by Barre´ de Saint Venant (1871) but these are discussed in the next section in view of their continued widespread use in modern river modeling software. The next major development of relevance to open-channel flow came in the more general field of boundary layer theory. The boundary where the main flow in a channel meets the bed and banks is of crucial importance particularly in steady flows where there is a balance between gravity and the friction generated at the interface. The contribution of Ludwig Prandtl (1875–1953) to fluid dynamics was significant and comprehensive (Anderson, 2005), but the most significant contribution was to identify the concept of the boundary layer. He postulated that the flow at a surface was zero and that the effect of friction was experienced in a narrow layer adjacent to the surface: away from this boundary layer, the flow was inviscid and could be studied with simpler techniques such as those of Euler. Prandtl then used his theory to derive

The Hydrodynamics and Morphodynamics of Rivers

equations for the velocity profile and consequent shear stresses in the boundary layer. These concepts are particularly relevant to open-channel flows as they demonstrate that the friction effects are confined to a narrow region adjacent to the bed and banks; they also provide a theoretical framework for studying these. Nikuradse used these concepts to study the effect of roughness in pipes and this led to his seminal work that produced the concept of sand grain roughness in pipes. He used the latter to derive friction factors for pipes and much of this theory was later transferred to the study of resistance due to friction in open channels. In the above, we can see that there has been a move from empiricism to a more physical and mathematical basis for the equations used in open-channel flow. However, a completely nonempirical formulation is still not available and is arguably impossible to achieve. This distinction should always be borne in mind and it is vital to remember that although we can find accurate solutions to the equations, these solutions represent models of reality and whoever is conducting the analysis must also use their knowledge and judgment in drawing conclusions. So far, this brief history has focused on hydrodynamics, but in addition to the movement of water, an understanding of rivers needs a sound understanding of the movement of sediment and changes in the shape and location of the river channel. The balance between entrainment and deposition of sediment by water flow is the fundamental process governing the geomorphological changes of alluvial rivers at all spatial and temporal scales. The water flow over a mobile bed generates spatial and temporal variations of the sediment transport capacity, causing either net entrainment or net deposition of sediment. Subtractions and additions of sediment are the cause of local bed level changes that in turn alter the original flow field. The discipline of river morphodynamics deals with the interaction between water flow and sediment, which is controlled by the bed shape evolution. Morphodynamic studies use the fundamental techniques of fluid mechanics and applied mathematics to describe these changes and to treat related problems, such as local scour formation, bank erosion, river incision, and river planimetric changes (Parker’s e-book). River morphodynamics became a science with Leonardo da Vinci, who annotated and sketched several morphodynamic phenomena (Manuscript I, 1497), such as bed erosion and deposit formation generated by flow disturbances due to obstacles, channel constrictions, and river bends. Leonardo reported two possible experiments, one on bed excavation by water flow and another on near-bank scour (Marinoni, 1987; Macagno, 1989). Initiation of sediment motion was first described by Albert Brahams (1692–1758), who wrote the two-part book Anfangsgru¨nde der Deich und Wasserbaukunst (Principles of Dike and Hydraulic Engineering) between 1754 and 1757. Brahams suggested that initiation of sediment motion takes place if the near-bed velocity is proportional to the submerged bed material weight to the one-sixth power, using an empirically based proportionality coefficient. Later Shields (1936) proposed a general relationship for initiation of sediment motion based on the analysis of data gathered in numerous experiments. He provided an implicit relation between shear velocity, u*(m s1), and critical shear stress, tc (Pa), at the

137

point of initiation of motion. His relationship is still the one most used for issues dealing with sediment transport. Although sediment transport is the basic process leading to geomorphological changes in rivers, it is the balance between the volume of sediment entrained by the water flow and the volume of deposited sediment that governs the shape of river beds. Pierre Louis George Du Buat (1734–1809), in his Principes d’hydraulique (Du Buat, 1779), realized the importance of bed material for the river cross-sectional shape and conducted experiments to study the cross-section formation in channels excavated in different soil materials ranging from clay to cobbles. However, the first attempt to treat a morphodynamic problem in quantitative terms was made only about one century and a half later by the Austrian Exner (1925), who is consequently considered the founder of morphodynamics. Exner was interested in describing the formation of dunes in river beds, for which he derived one of the existing versions of the conservation laws of bed sediment that are now known as Exner equations. His equation, however, does not describe dune generation, but the evolution of existing dunes:

q zb q qs ð1  pÞ ¼ qt qx

ð3Þ

where p is the soil porosity (–); zb the bed level (positive upward) (m); t the time (s); qs the sediment transport rate per unit of channel width (m2 s1); and x the longitudinal direction (m). By substituting the sediment transport rate, qs, with a monotonic function of flow velocity in Equation (3), the obtained relation reads

  q zb dqs q u ¼ with qs ¼ qs ðuÞ qt du q x

ð4Þ

where u is the flow velocity (m s1). The amount of transported sediment qs increases when the velocity increases, which means that the term

dqs du

ð5Þ

in Equation (4) is always positive. The result is that erosion occurs in areas of accelerating flow, whereas sedimentation occurs in areas of decelerating flow. This could explain why dunes move downstream. Exner had assumed sediment transport capacity to be simply proportional to the flow velocity, whereas in reality sediment transport capacity is related to the flow velocity to the power three or more (Graf, 1971). The combination of Exner’s relation (Equation (3)) to a relation for sediment transport and to the continuity and momentum equations for water flow leads to a fully integrated 1-D morphodynamic model. Several models of this type have been developed after Exner and it is not easy to establish who was the first to do this. Already in 1947, van Bendegom developed a mathematical model describing the geomorphological changes of curved channels in two dimensions (2-D). The model consisted in coupling the 2-D (depth-averaged) momentum and continuity equations for shallow water with the sediment balance equation (Exner’s equation in two dimensions) and a relation describing the sediment transport

138

The Hydrodynamics and Morphodynamics of Rivers

capacity of the flow. He corrected the sediment transport direction to take into account the effects of spiral flow and channel bed slope. van Bendegom carried out the first simulation of 2-D morphological changes of a river bend with fixed banks by hand, since computers were not available then. Bank erosion was finally introduced in 1-D morphodynamic models in the 1980s (Ikeda et al., 1981) and in 2-D models about 10 years later (Mosselman, 1992). Only in recent decades it has been realized that river morphology may be strongly influenced by the presence of aquatic plants and animals, as well as by floodplain vegetation (Tsujimoto, 1999). For a long period, vegetation in open channels was only considered as an additional static flow resistance factor to bed roughness, although already at the end of the nineteenth century some pioneer concepts suggested links between the river geomorphology and plants (Davis, 1899). Over the past few decades the move from empiricism to a more theoretical description of hydrodynamics and morphodynamics has been followed by a move from the expression of theory in equations to computer-based methods. Initially, the latter involved numerical solution of the theoretical equations, but more recently it has been developed with machinelearning techniques for extracting information from measured data which can be seen as a return to empiricism but with vast computing resources compared with past centuries.

2.07.2 State of the Art in Hydrodynamics and Morphodynamics Rivers convey water and sediment through the catchment to the sea. Moving water and sediment are subjected to forces such as gravity, friction, viscosity, turbulence, and momentum. In order to quantify the system we consider physical variables, such as velocity, depth, discharge, sediment concentration, and channel shape. Hydrodynamics and morphodynamics seek to relate these variables to the forces using the concepts of momentum and energy.

•

Process scale (local). This is the spatial scale at which processes, such as sediment entrainment, deposition, and turbulence, occur.

Whatever scale is being considered, the fundamental principles used in fluid dynamics are conservation of mass, momentum (Newton’s second law), and energy. These may need to be simplified according to the scale under consideration, the data available, and the level of detail required in the analysis, but they cannot be violated.

2.07.2.1.1 Mass Conservation of mass is based on the fact that mass can be neither created nor destroyed; therefore, within a general control volume the accumulation of mass is equivalent to the difference between the input and the output. For a definitive derivation the reader is referred to Batchelor (1967) and for a more accessible derivation to Versteeg and Malalasekera (2007). Expressed in partial differential form, conservation of mass is governed by

q q q q ðrÞ þ ðr  uÞ þ ðr  vÞ þ ðr  wÞ ¼ 0 qt qx qy qz

where r is the water density (kg m3); x the longitudinal distance (m); y the transversal distance (m); z the vertical distance (m); t the time (s); u the flow velocity component in longitudinal direction (m s1); v the flow velocity component in transversal direction (m s1); and w the flow velocity component in vertical direction (m s1). Equation (6) states that the change in density r with respect to time within a volume element plus the change in mass flow ðr  uÞ in x-direction plus the change in mass flow ðr  vÞ in y-direction plus the change in mass flow ðr  wÞ in z-direction is equal to zero. In comparison, the equation for the conservation of mass in integral form for an arbitrary volume is

q qt

Z Z Z

r  dV þ

V

2.07.2.1 Fluid Flow The concept of scale, both spatial and temporal, is vital to any study of hydrodynamics or morphodynamics and so in the discussions below we consider the following spatial scales:

•

•

•

Reach scale (entire river reach). A river reach is a large part of the river, which can reasonably be considered as uniform. River reach studies focus on the longitudinal variations of flow field, water depth, and other variables, such as sediment concentration. Often, one value of the variable per river cross section is enough. Cross-section scale (main channel cross section). This is the spatial scale of studies for which the transverse variations of flow field, water depth, roughness, etc., are relevant. In this case it is often sufficient to derive the depth-averaged value of the variable and its variation in transverse direction. Depth scale (water depth). This is the spatial scale of those studies for which the vertical variations of flow field are relevant.

ð6Þ

Z Z

r  u  dS ¼ 0

ð7Þ

S

where the change in density r with respect to time within the control volume plus the change in mass flow r  u over the surface S of the control volume is zero. More compactly, the equation in divergent form is

q ðrÞ þ =  ðr  uÞ ¼ 0 qt

ð8Þ

with the velocity vector u ¼ u  i þ v  j þ w  k in the three directions i, j, k in space.

2.07.2.1.2 Momentum Newton’s second law states that the rate of change of momentum of a body is equal to the force applied. In the case of a fluid, this principle is applied to the general control volume and the net momentum flux (inflow less outflow) is equated to the forces. The forces considered depend on the situation under consideration, but the main ones are gravity, shear stress, and pressure. Again the reader is referred to other

The Hydrodynamics and Morphodynamics of Rivers

texts for detailed derivation (Batchelor, 1967; Versteeg and Malalasekera, 2007).

    Du qp q qu q qu qv r ¼ þ 2m þ l div u þ m þ Dt qy qy qx qx qx qx    q qu qw m þ þ Fx ð9aÞ þ qz qz qx      Dv qp q qu qv q qv þ m þ þ 2m þ l div u r ¼ qy qx qy qx qy Dt qy    q qn qw þ m þ þ Fy ð9bÞ qz qz qy       Dw qp q qu qw q qv qw þ m þ þ m þ r ¼ qz qx qz qx qz qy Dt qy   q qw ð9cÞ þ 2m þ l div u þ Fz qz qz

139

functions as solutions, to a set of algebraic equations that connect values at various discrete points that can be manipulated by a computer. This process is called discretization. Various methods are used for this and the main three are finite difference, finite element, and finite volume. More details can be found elsewhere (Wright, 2005).

2.07.2.2.1 Boundary conditions Whether seeking an analytical or numerical solution, it is necessary to specify boundary conditions for any problem. In open-channel flow, these are specific and tend to be different from those encountered in other fields. In most cases the flow in a reach of river or channel is controlled by a specified discharge at the upstream and downstream boundary, a condition that specifies the depth. The latter includes a fixed depth, a time-varying depth, a critical flow condition, or a depth-discharge relationship.

2.07.2.3 Depth and Process Scales where u, v, and w are the components of velocity in the x, y, and z directions respectively; r the density; p the pressure; m the dynamic viscosity; l the second viscosity; and Fx, Fy, and Fz are the components of body force. Using the divergent form again gives the Navier–Stokes equations as

Du qp r ¼  þ r  ðmruÞ þ Fx qx Dt

ð10Þ

2.07.2.1.3 Energy Conservation of energy comes from the first law of thermodynamics

dE ˙ þ Q˙ ¼W dt

ð11Þ

which states that the change in the total energy E in the vol˙ plus the heat flux Q˙ in the ume element equals the power W volume element. Its application is dependent on the exact situation in which it is applied, and given the large variation in situations it will not be considered in detail here.

2.07.2.2 Numerical Solution It is possible to solve Equations (6)–(11) analytically in a few, simplified cases, and pioneers such as Prandtl were able to obtain significant insight through doing this. However, the full equations are not amenable to closed solutions and only with the advent of digital computing it has become possible to obtain solutions, albeit approximated ones. To derive a form that is suitable for computer solution, the continuous partial derivatives are converted to difference equations for discrete, point values. There are many ways of doing this and specific cases are discussed below in the relevant context. However, numerical techniques for partial differential equations fall into three main categories: finite differences, finite volumes, and finite elements. The initial task, as mentioned above, is to convert the differential equations, which have continuously defined

Viewed at a local scale, the flow is complex and 3-D. It has a predominant downstream flow direction, but the flow can be separated into a boundary layer, where the effects of the boundary and its nature are predominantly felt, and the free stream flow. Within the latter, there are relatively low gradients as the speed of the water increases toward the free surface. The maximum speed is achieved just below the free surface and there is a slight reduction at the surface due to the effects of air resistance and the attenuation of turbulence toward the surface. At channel bends, a particular flow structure is observed. The water higher in the column travels faster than that at a lower position and therefore does not change its direction in as short a distance. This leads to an increase in the water surface elevation at the outer, concave bank, which in turn drives fluid down and along the bed toward the inner, convex bank. In this way, we observe a super-elevation at the outer bend and a secondary circulation. Further counter-rotating circulations may be induced by the main secondary circulation if the bend is sharp (Blanckaert, 2002). The particular configuration of the flow inside river bends should be taken into account for the modeling of sediment transport and river morphodynamics. The complete description of fluid flow, based on the continuum hypothesis which ignores the molecular nature of a fluid, is given by the Navier–Stokes equations described above. For a laminar flow, these equations can be discretized to give a highly accurate representation of the real fluid flow. However, laminar flow rarely occurs in open-channel flows so we must address one of the fundamental phenomena of fluid dynamics: turbulence. As the Reynolds number (Reynolds number is defined by Re ¼ ruL/m, where r is the density, u the velocity, L the representative length scale, and m the viscosity) of a flow increases, random motions are generated that are not suppressed by viscous forces as in laminar flows. The resulting turbulence consists of a hierarchy of eddies of differing sizes. They form an energy cascade which extracts energy from the mean flow into large eddies and in turn smaller eddies extract energy from these which are ultimately dissipated via viscous forces.

140

The Hydrodynamics and Morphodynamics of Rivers

In straight prismatic channels, secondary circulations are present just as in curved ones, but at a much smaller magnitude. Although the main flow is in the downstream direction with no deviation, the effect of the walls on turbulence causes secondary circulations of the order of 1–2% of the main flow (Beaman et al., 2007). Turbulence is perhaps the most important remaining challenge for fluid dynamics generally. In theory, it is possible to predict all the eddy structures from the large ones down to the smallest. This is known as direct numerical simulation (DNS). However, for practical flows this requires computing power that is not available at present and may not be available for many years. A first level of approximation can be made through the use of large eddy simulations (LESs). These use a length scale to differentiate between larger and smaller eddies. The larger eddies are predicted directly through the use of an appropriately fine grid that allows them to be resolved. The smaller eddies are not directly predicted, but are accounted for through what is known as a subgrid scale model (Smagorinsky, 1963). This methodology can be justified physically through the argument that large eddies account for most of the effect on the mean flow and are highly anisotropic whereas the smaller eddies are less important and mostly isotropic. Care is needed in applying these methods as an inappropriate filter or grid size and low accuracy spatio-temporal discretization can produce spurious results. If this is not done, LES is not much more than an inaccurate laminar flow simulation. Although less computationally demanding than DNS, LES still requires fine grids and consequently significant computing resources that still mean it is not a viable, practical solution. In view of the demands of DNS and LES, most turbulence modeling still relies on the concept of Reynolds averaging where the turbulent fluctuations are averaged out and included as additional modeled terms in the Navier–Stokes equations. The most popular option is the k–e model, which is usually the default option in Computational Fluid Dynamics (CFD) software, where k represents the kinetic energy in the turbulent fluctuations and e represents the rate of dissipation of k. Interested readers are referred to CFD texts (Versteeg and Malalasekera, 2007) for further details. Given the complexities and computational demands of 3-D modeling in rivers, it has largely remained a research tool. Notable work has been done by Rastogi and Rodi (1978), Olsen and Stokseth (1995), Hodskinson and Ferguson (1998), and Morvan et al. (2002), and a more comprehensive review is given by Wright (2001).

2.07.2.4 Cross-Section Scale The fully 3-D equations while being a complete representation are computationally expensive to solve and in many situations unnecessarily complex. It is therefore necessary to simplify them and this is often done in the case of open-channel flow. The assumption is made that the flow situation being considered is shallow, that is to say, the lateral length scale is much greater than the vertical one (note: in this regard the Pacific Ocean is shallow in that it is much wider than it is deep!). Once we have assumed shallow water, we can further assume that streamlines are parallel and that there is no

acceleration in the vertical leading to the vertical momentum equation being replaced by an equation for hydrostatic pressure. In turn, once we have assumed that there is no vertical velocity, we can depth-integrate the two horizontal velocities, resulting in three equations: one for conservation of mass and two for momentum in the horizontal. These equations can be derived rigorously by either considering the physical situation or applying the assumptions to the Navier–Stokes equations. These 2-D equations are less time consuming to solve than the Navier–Stokes equations and there is a significant body of research devoted to this. This has culminated in a number of computer codes that are available both commercially and as research codes. These can be classified into those based on the finite difference, finite element, or finite volume methodology. In the present context one significant difference is relevant. The finite element method minimizes the error in the solution to the underlying mathematical equations in a global sense while finite volume minimizes it in a local sense. This means that a finite volume method will always conserve mass at each time step and throughout a simulation. The finite element and finite difference methods will only have true mass conservation once the grid is refined to a level where further refinement makes no further change to the solution. A number of codes based on the finite difference method have been developed and used in practice. Details of each can be found on the developers’ websites. Examples are ISIS2 D (Halcrow), MIKE21 (DHI), TUFLOW (WBM), and Sobek & Delft3d (Deltares). Codes using the finite element method are less common in river applications, but have been popular for flows in estuaries and coastal areas where the geometries can be complex. Examples are TELEMAC-2 D (EDF) (Bates, 1996), SMS produced by Brigham Young University based on codes from the USACE such as RMA2 D (King, 1978), and CCHE2 D produced by NCCHE, University of Mississippi (Wang et al., 1989). Codes using the finite volume method have been developed more recently as their strength in mass conservation and their ability to correctly model transitions have been realized. The latter is based on the use of Godunov-based methods (Sleigh et al., 1998; Alcrudo and Garcia-Navarro, 1993; Bradford and Sanders, 2002) or on the use of total variation diminishing (TVD) schemes (Garcia-Navarro and Saviron, 1992). In recent decades, there has been significant development of unstructured finite volume codes (Anastasiou and Chan, 1997; Sleigh et al., 1998; Olsen, 2000). These can be considered as a combination of finite element and finite volume approaches. They use the same unstructured grids as finite element and solve the mathematical equations in a finite volume manner that ensures conservation. In this way, they ensure physical realism and ease of application. The issue of wetting and drying is a perennially difficult one for 2-D models (Bates and Horritt, 2005). As water levels drop, areas of the domain may become dry and the calculation procedure must remove these from the computation in a way that does not compromise mass conservation or computational stability. Most available codes can deal with this phenomenon, but they all compromise between accuracy and stability. This issue must be carefully examined in results from any 2-D simulation where wetting and drying are

The Hydrodynamics and Morphodynamics of Rivers

significant. There is active research in this area with a number of recent contributions that may well improve matters (Liang, 2008; Lee and Wright, 2009). In assuming a depth-averaged velocity, 2-D models neglect vertical accelerations and make no prediction of vertical velocities. This, in turn, means that they do not predict or model the effects of the secondary circulations described above. The neglect of secondary circulations can lead to inappropriate model predictions for velocity and depth and in turn this can cause inaccuracies in morphological studies where the secondary circulations are a significant contribution to bed/bank erosion. There are a number of amendments to 2-D models to take an account of this phenomenon. The simplest calculates a measure of helical flow from an analysis of the velocity and acceleration vector at a point. This, in turn, is used to calculate a vertical velocity profile and vertical velocities. This approach is adopted in different forms in MIKE21C (DHI 1998), CCHE3D (NCCHE, University of Mississippi; Kodama, 1996), and CH3D (USACE; Engel et al., 1995) among others. A more accurate but computationally expensive method is the layered model (TELEMAC-3D, EDF; Delft3D, Deltares; TRIVAST; Falconer and Lin, 1997). This establishes a number of vertical layers and solves equations for the horizontal velocities in each layer. Subsequently, equations are solved for a vertical velocity based on an analysis of the interactions between each layer and the water depth is calculated appropriately. This is mainly suitable for wide bodies of water with significant vertical variations of velocity, temperature, salinity, or other variables in the vertical such as estuaries, lakes, and coastal zones. Nex and Samuels (1999) applied TELEMAC-3 D to the River Severn. They reported some success and qualitative agreement with measurements. A further development of this technique is to include the treatment of nonhydrostatic pressure variations (Stansby and Zhou, 1998; Casulli and Stelling, 1998). A 2-D model of a river and its floodplains require information about the channel bed topography and the terrain heights of the surrounding floodplain. In the past this required a mixture of time-consuming measurements and interpolation from published, paper-based maps. A significant advance over the past 10–15 years has been the use of remotely sensed data, which offer both increased accuracy and density of data along with reduced collection times. This comes at some expense, but the cost continues to come down. Current techniques such as light detection and ranging (LiDAR) can provide data every 25 cm at accuracies down to 10 cm. More experimental techniques can also be used to measure through the water surface to give detailed and accurate bed topography. Besides providing accurate data for model construction, remote sensing can also provide data on flood extents for use in validation. These procedures are now in regular use in commercial work and continue to be an area of active research. More details can be found in the literature (Horritt et al., 2001; Wright et al., 2008). Remotely sensed data need to be used with careful consideration of accuracy and the level of detail required in specific areas. For example, in modeling the interaction of a main channel with a floodplain it is necessary to have accurate data along the embankments of the main channel, and commonly used LiDAR data can miss these features through the use of a regular rectangular grid.

141

In this case, the LiDAR may need to be supplemented by other techniques such as Global Positioning System (GPS) (Wright et al., 2008)

2.07.2.5 River Reach Scale When considering long river reaches even a 2-D model can become cumbersome. In such cases, the length of the river is of several orders of magnitude greater than the width. It is therefore assumed that lateral variations in velocity and free surface height can be neglected and that the flow direction is entirely along stream. Under these assumptions the equations first formulated by Jean-Claude Barre´ de Saint Venant apply and these have formed the basis for the most widely used commercial river modeling packages. Each of these conceptualizes the river as a series of cross sections. At each the velocity is assumed perpendicular to the cross section. The resistance due to the bed and banks is based on one of the steady-state formulations for normal flow such as Mannings, Che´zy, or Colebrook-White (Chanson, 1999). Early numerical methods for solving this system of equations were pioneered by Abbott and Ionescu (1967) and Preissmann (1961). Both of these methods are essentially parabolic in nature, while the equations are hyperbolic. In view of this more recent methods have drawn on the body of research from compressible gas dynamics which has a similar set of equations. This has produced algorithms that are more robust and able to correctly represent transitions (GarciaNavarro et al., 1999; Crossley et al., 2003), but which are not so straightforward to implement particularly with regard to the incorporation of hydraulic structures such as weirs and sluices. Another recent development that is proving popular in some countries is the linking of 1-D and 2-D models. The former offers efficiency and lower data requirements while the latter can give better results on floodplains. A number of techniques have been proposed for linking these models (Dhonda and Stelling, 2003; Wright et al., 2008), but which one is the most reliable or successful is not yet clear. In fact, there is evidence to suggest that there are considerable differences among the different formulations and even among the different users of the same software package (Kharat, 2009). Although the 1-D approach is based on an analysis of the situation at a cross section, it can be applied to rivers of significant lengths up to hundreds if not thousands of kilometers. Further through the incorporation of junction equations relating flows and depths at confluences and difluences, it can be used to model complex networks of rivers and channels. Over the past three decades, several commercial packages have been developed based on the 1-D shallow water equations (InfoWorks, ISIS, MIKE11, and Sobek, among others). In the US, the USACE Hydrologic Engineering Center has also developed the HEC-RAS software that is freely available. These software packages combine the basic numerical solution with sophisticated tools for data input and graphical output. They are designed to make use of remotely sensed data and to provide 2-D and 3-D output in both steady and animated formats.

142

The Hydrodynamics and Morphodynamics of Rivers

2.07.2.6 Spatial Scales in River Morphodynamics River morphodynamics deals with the shape and, in a wider sense, composition of the river bed. The shape of alluvial rivers is made up by the combination of many geomorphological forms, which can be recognized at specific spatial scales, from small ripples to large bars and meanders. The development of geomorphological forms is related to the balance between entrainment and deposition of sediment over different control volumes and times. In modeling, every factor influencing sediment motion has to be taken into account, but in different ways depending on the spatial and temporal scale of the study (Schumm and Lichty, 1965; Phillips, 1995). In particular, processes that operate at smaller scales are parametrized to take into account their effects at larger spatial and temporal scales. Processes that operate at larger scales may be represented as boundary conditions for the studies focusing on smaller scales. At the largest spatial scale, the one of the entire river basin or single sub-basins, we can recognize the entire river network. Typical river basin-scale issues involve soil erosion, reservoir or lake sedimentation, as well as solid and water discharge formation. Basin-scale studies are characterized by the description of the entire river drainage network or large parts of it, such as the delta or a sub-basin. Geographic information systems, 0-D and 1-D morphodynamic models, as well as 1-D or 2-D runon–runoff models, are the typical tools used. The river basin scale is not further treated here, since its issues generally fall under the other related disciplines of hydrology and physical geography. Lowering the observation point and zooming in on the river system, different river reaches, each one characterized by planform style and sinuosity, are highlighted. A single river reach is characterized by one value of the water discharge, but

changing with time. Depending on the reach characteristics, the typical temporal variations range from hours to days for the discharge; from years to several tens of years for the longitudinal bed slope. A river reach in morphodynamic equilibrium is characterized by a longitudinal bed slope that can be considered constant at a chosen temporal scale (de Vries, 1975). Reach-scale issues mainly deal with the assessment of the environmental impact of human interventions, such as river training, and with the natural river evolution on the long term. For this, morphodynamic studies need to determine bed aggradation and degradation, along the river reach, changes in sinuosity and planform style. The typical tools are 0-D reachaveraged formula (e.g., Che´zy, 1776; Lane, 1955), describing the water flow at reach-scale morphodynamic equilibrium, as well as 1-D cross-sectionally averaged models. Commercial 1-D codes updating the riverbed elevation are: MIKE11 (DHI) and SOBEK-RE (Deltares). By further zooming in on the river, the attention moves to the river corridor, or river belt, the area including main river channel and floodplains. Specific morphological features recognizable at this spatial scale are scroll bars inside river bends (Figure 1), a sign of past bend grow. Corridor-scale studies mainly deal with flood risk, river rehabilitation projects, as well as river planimetric changes. The typical tools are 2-D, depth-averaged, or a combination of 1-D (cross-sectionally averaged) and 2-D (depth-averaged) morphodynamic models. These models often have to include formulations for bank retreat and advance and for the effects of (partly) submerged vegetation on water levels, sediment transport, and deposition. Commercial codes developed for the study of the river morphological changes at this and smaller spatial scales are (among others): MIKE21 (DHI), Delft3 D (Deltares), and SOBEK-1 D-2 D (Deltares). Examples of free 2-D codes are: FaSTMECH (Geomorphology and Sediment Transport

Figure 1 Aerial view of a tributary of the Ob River (Russia). Scroll bars on floodplains and point bars inside river bends are clearly visible. Courtesy of Saskia van Vuren.

The Hydrodynamics and Morphodynamics of Rivers

Laboratory of USGS) and RIC-Nays (Hokkaido University). These two models adopt the user interface IRIC, developed in the Geomorphology and Sediment Transport Laboratory of USGS (USA). Central and multiple bars, either migrating or static, are the characteristic geomorphological features to be studied at the cross-section scale (Figure 2). Typical engineering issues are river navigation and the design of hydraulic works, such as trains of groynes, bridges, and offtakes. Typical tools are 2-D, depth-averaged, models, formulated for curved flow (van Bendegom, 1947), often including bank retreat and advance (Mosselman, 1998). Modeling often regards bar formation, bar migration, and channel widening and narrowing as the natural development or as the effects of human interventions. If the observation point moves from a point above the river to a point inside the river channel, the vertical contour of the river cross section becomes visible (Figure 3). Water-depth variations in transverse direction, due to the presence of local deposits and scours, as well as water-depth variations in longitudinal direction, due to the presence of dunes, are the major morphological features observable at this spatial scale. Typical depth-scale studies deal with scour formation around structures, bank erosion, bank accretion, as well as dune development and migration. Typical tools are either 3-D or 2-D and 1-D vertical morphodynamic models, often focusing on local bed level changes or on vertical variations, of, for instance, salinity, suspended solid concentration, soil stratification, and bank slope. The smallest spatial scale that is relevant for the river morphodynamics is called the process scale. This is the scale of fundamental studies describing processes such as sediment entrainment and deposition, for which phenomenon such as turbulence plays a major role. The typical geomorphological

143

forms to be studied at this small spatial scale are ripples (Figures 4–6). The typical tools are detailed morphodynamics models in one, two, and three dimensions. In morphodynamics, temporal and spatial scales are strongly linked. Phenomena with small spatial scales also have small temporal scales, and phenomena with large spatial scales have large temporal scales (de Vriend, 1991, 1998; Blo¨schl and Sivapalan, 1995). The linkage between spatial and temporal scales is formed by sediment transport. For the development or migration of a small bedform, only a small amount of sediment needs to be displaced, whereas large amounts of sediment are needed for the development of large geomorphological forms, such as bars. Phenomena interact dynamically when they occur more or less on the same scale. Small-scale phenomena, such as ripples, appear as noise in the interactions with phenomena on larger scales, such as bar migration, but they can produce residual effects, such as changes of bed roughness (Figure 5). Their effect on larger scales can be accounted for by parametrization procedures (upscaling). Phenomena operating on much larger spatial and temporal scales can be treated as slowly varying or constant conditions. They define scenarios, described in terms of boundary conditions, when studying their effects on much smaller scales. Thus, basin-scale studies are essential for the generation of the input (boundary conditions) for the morphodynamic studies on smaller spatial scales.

2.07.2.7 Geomorphological Forms in Alluvial River Beds Geomorphological forms in rivers can be caused by the presence of geological forcing, human interventions, and man-made structures, but they also arise as a natural instability of the interface between the flowing water and

Figure 2 Multiple bars in the braided Hii River (Japan). Courtesy of Takashi Hosoda.

144

The Hydrodynamics and Morphodynamics of Rivers Floodplain Floodplain

38

36 (m a.s.l.)

Excavated area

34 Initial bed level 2007 2017

32

2027 30

200

300

400

500 (m)

600

700

Figure 3 River Meuse (the Netherlands): temporal bed level changes during the period 2007–27. On the vertical, the bed elevation in meters above sea level (Villada Arroyave and Crosato, 2010).

Figure 4 Ripples in a straight experimental flume with a sandy bed (the bar shows centimeters). Laboratory of Fluid Mechanics of Delft University of Technology.

sediment. In analogy with the interaction between air moving above water (wind), the instability of the water–sediment interface produces waves of different sizes, which can coexist and interact with each other. Ripples are the smallest ones, originating from the instability of the viscous sublayer near the river bed (Figure 4).

Dunes are the main source of hydraulic resistance of a river and hence a key factor in raising water levels during floods (Figure 7). They are also the first parts of the river bed that need to be dredged to improve navigation. Dune formation and propagation is so intimately linked to sediment transport, that the latter cannot be modeled properly without

The Hydrodynamics and Morphodynamics of Rivers

accounting for dunes (ASCE Task Committee on Flow and Transport over Dunes, 2002). Bars are the largest waves in the river bed; they can be scaled with the channel cross section (Figure 2).

Figure 5 The presence of 3-D ripples acts as noise for the study of alternate bars in this laboratory experiment carried out at the Laboratory for Fluid Mechanics of Delft University of Technology.

Figure 6 2-D ripples in the Het Swin Estuary (the Netherlands).

145

2.07.2.7.1 Ripples and dunes For increasing Froude numbers the river bed is first plane and then covered by ripples and dunes. The flow regime close to the critical Froude number (Fr E 1) is again characterized by plane bed. If the Froude number increases further (supercritical flow), antidunes begin to form with upstream breaking waves over the crest (Simons and Richardson, 1961). Southard and Boguchwal (1990) provided the most extensive bedform phase diagrams showing the possible occurrence of ripples, dunes, antidunes, or plane bed under different sediment size and flow conditions. Bedforms may have either a 2-D or a 3-D pattern. 2-D ripples and dunes have fairly regular spacing, heights, and lengths. Their crest lines tend to be straight or slightly sinuous, and are oriented perpendicular to the mean flow lines (Figure 6). In contrast, 3-D features have irregular spacing, heights, and lengths with highly sinuous or discontinuous crest lines (Ashley, 1990), as in Figures 4 and 5. In general, ripples scale with the sediment diameter while dunes scale with the water depth (Bridge, 2003), but there is no clear distinction between ripples and dunes for limited water depths, as for instance, in flume experiments. Extensive data compilations by Allen (1968) and Flemming (1988) demonstrated that there is a break in the continuum of observed bedforms discriminating ripples from dunes. For instance, ripples are only present for fine sediment with Do1 mm. However, there are no generally valid techniques to divide ripple from dune regimes and some authors choose to make no distinction at all. The first theoretical study of dune instability was carried out by Kennedy (1969). Spectacular progress in knowledge of dune dynamics is linked to the increasing sophistication of numerical modeling (Nelson et al., 1993). Recent models produce detailed simulations of the instantaneous structure of flow over a dune-covered bed. Giri and Shimizu (2006)

146

The Hydrodynamics and Morphodynamics of Rivers

Figure 7 Dunes in the Waal River and Pannerdense Canal (the Netherlands) on 4 November 1998. Flow from right to left. Courtesy of Rijkswaterstaat. Upstream of bifurcation: discharge 9600 m3 s1, water depth 10.7 m, mean grain size 3.3 mm, flow velocity 2.1 m s1, dune height 1.0 m, and dune length 22 m. Analysis by Wilbers, Department of Physical Geography, Utrecht University, Utrecht, The Netherlands.

developed a 2-D model for the prediction of dunes under unsteady flow regime. Nabi et al. (2009) provided the first detailed 3-D model of dune formation.

2.07.2.7.2 Bars Bars are shallow parts of river bed topographies that become visible at low flows and can be either migrating or steady. Bars occur in more or less regular, periodic patterns as a result of interactions between flowing water and sediment. In a river channel one or more parallel rows of bars may be present; the number of rows present is called bar mode m. Alternate bars have mode m equal to 1 (Figure 8); multiple bars have mode larger than 2. Migrating bars as well as steady bars (Crosato and Desta, 2009) develop spontaneously as a result of morphodynamic instability and for this they are often referred to as free bars. Confined sediment deposits caused by local changes of the channel geometry, such as point bars inside river bends (Figure 1), should therefore be distinguished from free bars. The stability analyses performed by, among others, Hansen (1967), Callander (1969), and Engelund (1970) define the conditions that govern the development of bars in alluvial river channels. The width-to-depth ratio of the river channel is the dominating parameter for free bar formation: the larger the width-to-depth ratio, the larger the bar mode. This means that multiple bars form at larger width-to-depth ratios than alternate bars. Moreover, no bars can form for width-to-depth ratios that are smaller than a certain critical value. Parker (1976) and Fredsøe (1978) related the presence or absence of free bars to the channel planform, that is, meandering or braided. By persistently enhancing opposite bank erosion, steady alternate bars (Figure 8, left) are seen as a

m=1

m=2

Figure 8 Left: alternate bars (m ¼ 1). Right: central bars (m ¼ 2).

key ingredient for the evolution of straight water courses into meandering water courses (Olesen, 1984). Multiple bars are a characteristic of braided rivers. The linear theory by Seminara and Tubino (1989) defines marginal stability curves separating the conditions in which a certain number of bars per cross section grows from the conditions in which the same bar mode decays. The river is supposed to select the bar mode with the fastest growth rate, which is a function of the width-to-depth ratio, the Shields parameter, the sediment grain size, and the particle Reynolds number. A single physics-based formula was recently derived by Crosato and Mosselman (2009) from a stability analysis. The formula allows one to compute directly the mode of free bars that develop in an alluvial channel, but it is limited to rivers having width-to-depth ratio smaller than 100. By assuming that meandering rivers are characterized by the

The Hydrodynamics and Morphodynamics of Rivers

presence of alternate bars and braided rivers by multiple bars, the same formula can also be used to determine the type of planform that can be expected to develop after widening or narrowing of a river channel.

2.07.2.8 River Planimetric Changes The study of the river planimetric changes requires the assessment of both bank erosion and bank accretion rates and for braided-anabranched rivers also to the assessment of the stability of channel bifurcations. Meandering rivers have single-thread channels with high sinuosity and almost constant width (Figure 1). They could be regarded as a particular type of braided rivers (Murray and Paola, 1994), those having bar mode equal to 1. River meandering is governed by the interaction between bank accretion, bank erosion, and alluvial bed changes (Figure 9). Bank erosion causes channel widening and enhances opposite bank accretion. Conversely, bank accretion causes river narrowing and enhances opposite bank erosion. The two processes of bank erosion and accretion do not occur contemporarily, and for this reason the river width is subject to continuous fluctuations. However, generally a stable time-averaged width is achieved in the long term. Understanding the process of bank accretion and width formation is therefore a fundamental prerequisite for the modeling of meandering river processes and, more in general, for the modeling of the river morphology. All existing meander migration models (Ikeda et al., 1981; Johannesson and Parker, 1989; Crosato, 1989; Sun et al., 1996; Zolezzi, 1999; Abad and Garcia, 2005; Coulthard and van de Wiel, 2006) assume the rate of bank retreat to be the same as the rate of opposite bank advance. This means that the lateral migration rate of the river channel can be assumed to be equal to the retreat rate of the eroding bank. This is in turn assumed to be proportional to the near-bank flow velocity excess with respect to the normal flow condition, following the approach by Ikeda et al. (1981). Some meander migration models take also into account the effects of the near-bank water depth excess on the bank retreat rate (e.g., Crosato, 1990). The proportionality coefficients in the channel migration formula are supposed to weigh the bank erosion rates.

These coefficients should be a function of the bank characteristics only, but are in fact bulk parameters incorporating the effects of opposite bank advance and some numerical features (Crosato, 2007). Existing theories on river meandering focus on the assessment of bank retreat rates without defining the conditions for the opposite bank to advance with the same speed. However, it is just the balance between the rate of bank advance and the rate of opposite bank retreat that makes the difference between braiding and meandering (Figure 10). A meandering river requires that, in the long term, the bank retreat rate is counterbalanced by the bank advance rate at the other side. If bank retreat exceeds bank advance, the river widens and, by forming central bars or by cutting through the point bar, assumes a multi-thread (braided) pattern. If bank advance exceeds bank retreat, the river narrows and silts up. So far, most research has focused on the processes of bank erosion (e.g., Partheniades, 1962, 1965; Krone, 1962; Thorne, 1988, 1990; Osman and Thorne, 1988; Darby and Thorne, 1996; Rinaldi and Casagli, 1999; Dapporto et al., 2003; Rinaldi et al., 2004) and bed development, whereas the equally important bank accretion has received little attention

(a)

(b)

Figure 10 A sinuous water flow is not sufficient for meandering. (a) straight river planform with bank retreat, but without bank advance. (b) meandering river planform in which bank advance counterbalances bank retreat.

Bank accretion

Bank erosion Bed level changes

Figure 9 Morphological processes shaping the river cross section.

147

148

The Hydrodynamics and Morphodynamics of Rivers

(Parker, 1978a, 1978b; Tsujimoto, 1999; Mosselman et al., 2000). As a result, there are no comprehensive physics-based river width predictors. Bank accretion is governed by the dynamic interaction between riparian vegetation, flow distribution, frequency as well as intensity of low and high flow stages, local sedimentation, soil strengthening and by the interaction between opposite (eroding and accreting) banks. Bank erosion and accretion strongly depend on climate (Crosato, 2008). Climate changes can therefore alter the river cross section and the river pattern. Present knowledge on river morphological processes is insufficient to fully assess these effects. A number of existing 2-D and 3-D morphological models, such as Delft3 D, treat bank accretion as near-bank bed aggradation and bank erosion as near-bank bed degradation. These models are suitable for the prediction of width changes of channels without vegetation and with mildly sloping banks, but fail to predict the morphodynamics of meandering rivers, which are characterized by cohesive banks and riparian vegetation. A few 2-D morphological models simulate bank erosion, but not bank accretion. One example is the model RIPA, which was developed at Delft University of Technology by Mosselman (1992) and further extended by the University of Southampton (Darby et al., 2002). In large valleys or near the sea, the river can split into several channels. In anabranched rivers each anabranch is a distinct, rather permanent, channel with bank lines (Figure 11). The river bed is mainly constituted by loose sediment, such as sand and gravel, whereas silt prevails at the inner parts of bends and in general where the water is calm. Anabranches are commonly formed within deposits of fine material. Vegetation and soil cohesiveness stabilize the river banks and the islands separating the anabranches, so that the planimetric changes are slow if compared to the river bed changes. Studying the morphological changes of this type of rivers requires the assessment of the stability of bifurcations

(Wang et al., 1995; Kleinhans et al., 2008). Experimental and theoretical research started almost 100 years ago (Bulle, 1926; Riad, 1961) and are still going on (Bolla Pittaluga et al., 2003; De Heer and Mosselman, 2004; Ten Brinke, 2005; Bertoldi et al., 2005; Kleinhans et al., 2008). The major difficulty rises in the assessment of sediment distribution between the two branches of the bifurcating channel, which is a function of water discharge distribution, sediment characteristics, channel curvature at the bifurcation point, and presence of bars.

2.07.2.9 Bed Resistance and Vegetation In any study of a river, whether it is experimental, full-scale measurement or numerical in 1-D, 2-D, or 3-D, the irregular geometry of the boundary (bed and banks) cannot be directly represented. Even with full-scale measurements, the boundary cannot be accurately mapped at the scale of the bed material. In all cases, a conceptual representation of the effect of the boundary on the flow is used to account for momentum and energy dissipation. There is much misunderstanding of the nature of these resistance or roughness laws and inconsistencies in their application (Morvan et al., 2008). Clearly, given the importance of boundary resistance determining flow and depth it is necessary to have a clear understanding of the various methods and any limitations on their applicability. If it were possible to solve the Navier–Stokes equations on a grid that was fine enough to resolve the smallest scale of turbulence (the Kolmogorov scale), then there would be no need for a turbulence model or a model of the effect of the boundary. However, this is not yet generally possible and even in 3-D solutions there is a need to simplify the equations. In 3-D a turbulence model is used as well as the resistance model, but in 2-D and 1-D models both these phenomena tend to be included in a resistance term. This in turn can lead to uncertainty in the definition and a lack of rigor in its application. This uncertainty together with lack of rigor

Figure 11 Anabranched planform: the Amazon River near Iquitos, Peru. Courtesy of Erik Mosselman.

The Hydrodynamics and Morphodynamics of Rivers

increases as the dimensionality decreases. Another consequence of the difference between 3-D, 2-D, and 1-D modeling is that the value of a parameter such as roughness height will vary between each dimensionality even if the physical situation under consideration is identical due to the fact that the resistance model incorporates different physical phenomena in each case. This can be seen in Figure 12 from Morvan et al. (2009). The results from Manning’s equation differ from those of the 3-D model as ks is varied indicating that the two are quite different. In fact, the Manning’s equation results are more sensitive to changes in the roughness which is due to the 3-D model representing phenomena such as turbulence and secondary circulations directly rather than in the resistance parametrization. In view of its significance there has been much work in this area over the last century and the reader is referred to Davies and White (1925), Ackers (1958), ASCE (1963), Rouse (1965), Yen (1991, 2002), and Dawson and Fisher (2004). Specific types of roughness are considered by Sayre and Albertson (1963) and ESDU (1979). Reynolds and Schlicting have written useful textbooks on the wider subject (Reynolds, 1974; Schlichting et al., 2004). A good review of the topic in the context of modeling is given in the paper by Morvan et al. (2008). Early work on roughness was performed in pipes by Nikuradase, mentioned above as building on the work of Prandtl. The Darcy–Weisbach equations for pipe flow uses a friction factor that is based on geometry (diameter in the case of pipes), mean velocity, and surface characteristics based on a relative roughness defined by a quantity known as the Nikuradse equivalent sand grain roughness nondimensionalized by the diameter of the pipe. In 1-D openchannel studies the geometric parameter of diameter is replaced by hydraulic radius (area divided by wetted

35

149

perimeter) which leads to discontinuities when the flow moves onto flood plains as the wetted perimeter increases abruptly while the cross-sectional area does not. It is clear that using the theory from pipe flow in open channels raises difficult issues with complex cross sections and using the hydraulic radius to capture geometrical effects is problematic. In practice, many people use hydraulic radius and then adjust the value of the Nikuradse roughness or equivalent to ensure that the frictional head loss per unit channel length matches the bed slope. This demonstrates that the roughness parameter is often related to energy loss in the model as much as any physical measurement of the nature of the surface. In fact, the parameter is a function of local bed geometry, flow regime, cross-section geometry, and turbulence. Given the wide range of effects, it is clear that the parametrization depends on the model used for the overall fluid flow. It is worth considering in a little more detail the nature of the forces acting on the fluid due to presence of the bed. Morvan et al. described these as

• • •

skin drag (e.g., roughness due to surface texture, grain roughness); form drag (e.g., roughness due to surface geometry, bedforms, dunes, separation, etc.); and shape drag (e.g., roughness due to overall channel shape, meanders, bends, etc.).

Skin and form drag can be considered to occur on a plane, but shape drag is due to larger-scale 3-D patterns. Again, it is clear that the way each of these is represented depends on the sort of model used. A resistance parameter such as Manning’s coefficient n or Che´zy used at a reach scale is based on the concept of bed resistance, although in practice it is also calibrated to account for shape drag. In many representations, roughness is characterized by a roughness height. It is often not appreciated that although this quantity has the units of length, it is not a measure of the height of the roughness elements. It is rather a parameter in an analytical model of flow at the wall (i.e., in 3-D):

Q (1-D)

ut 1 þ ¼ lnðEðkþ s Þy Þ u k

Q (3-D)

Mass flow rate

30

Experiment

25

20

15

10 0

0.2

0.4

0.6

0.8

1

ks Figure 12 Variation of the mass flow rate in 1-D model and 3-D model for a trapezoidal channel compared against the measured value (Morvan et al., 2009).

ð12Þ

where Eðkþ s Þ is a function of the nondimensional roughness height, kþ s ¼ ks u =n, in which ks is the roughness height, k the von Karman’s constant usually taken equal to 0.41, and n is the kinematic viscosity. It seems attractive to base our estimates for roughness heights on work such as Nikuradse’s on relatively zsmooth experimental channels. This has led to formulations such as ks ¼ 3:5  D84 or ks ¼ 6:8  D50, where DXX stands for the grain diameter for which xx% of the particles are finer, reported in Clifford et al. (1992). The latter paper makes interesting reading and shows that the grain–roughness relationship is inadequate. This is because there are several momentum loss mechanisms in these flows and they are not represented by such a simple equation. A further complication is that in some 3-D simulations values of the roughness height are derived from these formulas that are in fact greater than the size of the grid perpendicular to the wall. This could suggest that the grid resolves flows at a scale less than the size

150

The Hydrodynamics and Morphodynamics of Rivers

of the roughness which contradicts the fact that the roughness features have been removed to give a smooth planar surface. The above discussion has focused mainly on 3-D models, but the situation when we consider 2-D and 1-D models is even less clear. Continuing the approach of considering surface roughness as the parameter governing resistance, various formulas have been proposed to connect the roughness height with a parameter such as Manning’s n for 1-D models: HR Wallingford tables (Ackers, 1958):

• •

ks ðmmÞ ¼ ðn=0:038Þ6

ð13Þ

•

Massey (Massey 1995):

 ks ðSIÞ ¼ 14:86R=exp10

0:0564R 1=6 n

 ð14Þ

Chow (1959):

 ks ðSIÞ ¼ 12:20R=exp10

It is clear from this discussion that parametrizing resistance in open-channel flows is not straightforward and needs knowledge and experience from numerical and physical modeling. A number of conclusions can be drawn (based on those in Morvan et al. (2009)):

0:0457R 1=6 n

 ð15Þ

Strickler (1923):

ks ðftÞ ¼ ðn=0:0342Þ6

ð16Þ

These differ not only in the numerical values used, but also in the functional form. They also give large ranges for roughness height for small variations in Manning’s n. This indicates the uncertainties in this process, which have led authors to seek better means of characterizing the geometry and surface characteristics in order to approximate resistance. In some cases, particularly with large cross section covering a main channel and floodplains, there are zones with quite different resistances within the cross section. In such cases divided channel method (DCM) can be used where the cross section is divided into panels, and a conveyance is calculated in each one before being combined into a composite value (Knight, 2005). This has been shown to be successful and is incorporated in most commercial software. All these methods assume quasi-straight river reaches, and do not include lateral momentum transfer effects. Thus, they cannot predict accurately either the water level in compound river channels or the proportion of flow between the main channel and floodplains. More recent developments include the effect of flow structure, through the adoption of improved methods (Knight, 2005). These may be grouped under the headings: the DCM, the coherence method (COHM), the Shiono and Knight method (SKM), and the lateral division method (LDM). Several authors have presented examples of these methods applied to fluvial problems (Knight et al., 1989; Knight, 2005). The SKM, for example, uses three parameters rather than just the one used by approaches such as Manning’s or Che´zy. In fully 2-D shallow water models the flow is considered in separate vertical water columns and the variables are depth and two perpendicular velocities or discharges per unit width. In this case, the resistance is applied only to the surface at the base of the water column (the bed) and the roughness height will be different even from a 1-D model and for the same bed material.

•

roughness varies between models, which represent different dimensions and therefore reach-scale roughness is a different concept from local roughness; using roughness to represent features other than sand-grain roughness lessens the validity of the underlying theory and is questionable; models of roughness in 1-D hydraulic models are valid and will continue to be useful when based on sound analysis and calibrated appropriately; and 1-D modelers should focus more on estimating conveyance than establishing one sole value of Manning’s n or Che´zy’s C for a channel.

This shows that the representation of resistance in real rivers is a complex task. It could therefore lead to the conclusions that hydraulic modeling is fraught with difficulty and that it is of little benefit. This is not the case and when used with care they are extremely useful (Knight et al., 2009). If the representation of the resistance due to the nonuniform surface of the bed and banks presents a significant challenge to modelers, the representation of the effects of vegetation is perhaps an even greater one. Further, the need to represent vegetation is becoming greater with the design of more natural channels and the need to model inundation flows across vegetated floodplains. Besides being nonuniform, vegetation experiences changes in its resistance as it deforms as the velocity of the water increases. The effects of vegetation on river processes are many, complex, and difficult to quantify (Fisher and Dawson, 2003; Rinaldi and Darby, 2005; Gurnell et al., 2006). The ability of vegetation to stabilize river banks (Ott, 2000) partly depends upon scale, with both size of vegetation relative to the watercourse and absolute size of vegetation being important (Abernethy and Rutherfurd, 1998). Vegetation stabilization is most effective along small watercourses. On relatively large rivers, fluvial processes tend to dominate (Thorne, 1982; Pizzuto, 1984; Nanson and Hickin, 1986). The effect of vegetation on the conveyance of a channel depends on a number of factors such as density, type, height, and distribution of plants and their development stage (Allmendiger et al., 2005; Dijkstra, 2003). At the local scale, single plants act as roughness elements. Isolated trees and relative small clusters of plants increase turbulence around them leading to local scour, just as bridge piers do. Dense vegetation, instead, reduces the flow velocity between and above plants and sediment transport, enhancing local siltation. In this way, riparian vegetation increases the development of natural levees during floods as well as bank accretion. Rooted plants reduce local soil erosion by binding the soil with the roots (Figure 13) and by covering it. In this way, riparian vegetation decreases bank erosion. Heavy trees, however, can enhance gravitational bank failure by increasing the load on the bank (Ott, 2000). Finally, vegetation causes local accumulation of organic material (falling leaves,

The Hydrodynamics and Morphodynamics of Rivers

Figure 13 Roots protecting the river bank against erosion. Geul River (The Netherlands). Courtesy of Eva Miguel.

branches, and dead plants), which further reinforces the soil cohesion and strength (Baptist, 2005; Baptist and De Jong, 2005; Baptist et al., 2005). At the cross-section scale vegetation affects the river morphodynamics by acting on (Crosato, 2008) (1) river bed degradation/aggradation, (2) bank erosion, and (3) bank accretion by:

•

• • •

Deflecting the water flow. Aquatic and riparian vegetation increase the local hydraulic roughness and for this reason, the flow concentrates where vegetation is absent (Tsujimoto, 1999; Pirim et al., 2000; Rodrigues et al., 2006). This lowers the flow velocity within the plants, where sedimentation increases, and causes bed degradation in the nonvegetated area of the channel, where the flow velocity becomes higher. By deflecting the flow toward the opposite bank, riparian vegetation enhances opposite bank erosion (Dijkstra, 2003). Protecting the vegetated parts of the riverbed and bank against erosion (Figure 13). Accelerating the vertical growth of accreting banks and bars. Raising water levels. By increasing the hydraulic roughness, aquatic vegetation increases the water levels.

At the river-reach scale vegetation affects the water levels as well as the river planform formation (e.g., Murray and Paola, 2003; Jang and Shimizu, 2007; Samir Saleh and Crosato, 2008; Crosato and Samir Saleh, 2010). Murray and Paola

151

studied the effects of soil strengthening by floodplain vegetation on the river planform, whereas Jang and Shimizu and Samir Saleh and Crosato studied the effects of increased hydraulic roughness. All works demonstrated that vegetation decreases the degree of braiding of river systems and might even transform a braiding into a meandering system. Early studies considered the effects of vegetation on flow qualitatively (Powell, 1978; Dawson and Robinson, 1984) and demonstrated that the effects of vegetation varied over the seasons and that the relationship between resistance and vegetation varied greatly with depth. Later, semiquantitive relationships (Stephens et al., 1963; Shih and Rahi, 1982; Pitlo, 1982) were studied and demonstrated that if Manning’s n is used to represent the resistance in a vegetated channel, values of up to 20 times the nonvegetated value can be found, but that such changes were more pronounced in smaller channels. These semiquantitative approaches of increasing the amount of numerical resistance by changing the resistance parameter are still widely used by many practitioners. This is, however, based on the flawed concept resistance due to vegetation, whether emergent or submerged, stems from a boundary layer phenomenon while it is actually a mixing layer phenomenon (Ghisalberti and Nepf, 2002). This implies that the resistance from vegetation depends on depth and can therefore never be fully accounted for by a resistance parameter that is based on a surface representation rather than extending through the water column. These limitations have led to the proposal of more quantitative methods and a number of these were given by Fisher and Dawson (Table 1). The work in Table 1 and that of others (Larsen et al., 1990; Bakry, 1992; Salama and Bakry, 1992; Watson, 1997) indicate that while there may be a relationship between resistance and vegetation, it is complex and there is, as yet, no ideal equation for this relationship. The limitations of this approach have led a number of authors to propose more sophisticated representations based on analyzing the drag coefficient of vegetation. Most work (Wu et al., 1999; Fischer-Antze et al., 2001; Ghisalberti and Nepf, 2002, 2004; Wilson et al., 2003) has focused on laboratory channels which is vital to reduce the uncertainties in full-scale cases and to allow for well-founded fundamental conclusions to be drawn. However, work that has been carried out on real rivers is scarce (Stoesser et al., 2003; Nicholas and McLelland, 2004), which has had little or no measured data for comparison. Stoesser et al. (2003) applied a 3-D model for vegetative resistance on the Restrhein and Nicholas and McLelland (2004) used a 3-D model on the floodplains of a natural river. The drag coefficient is often based on that for a nonflexible cylinder, but this is clearly not the case with vegetation. More recent work has studied the effect of flexibility (Kouwen, 1988; Querner, 1994; Rahmeyer et al., 1996; Fathi-Maghadam and Kouwen, 1997). Further fundamental understanding has been advanced by Japanese researchers and are reviewed by Hasegawa et al. (1999). The reduction-factor approach outlined in Baptist (2005) and Baptist et al. (2007) quantifies the hydraulic effect that vegetation can exert on the flow by considering the distribution of shear stress within the water column rather than

152 Table 1

The Hydrodynamics and Morphodynamics of Rivers Different methods to derive the Manning’s roughness coefficient of vegetated channels (Fisher and Dawson, 2003)

Authors

VRa range (m2 s1)

Discharge (m3 s1)

Areab (m2)

Marshall and Westlake (1990)

0.24–1.3

0.2

Pepper (1970 )

0.58–8.46

2.4

Wessex Scientific Environmental Unit (1987)

0.24–1.3

15

43

Wessex Scientific Environmental Unit (1987)

0.15–1.1

15

43

Wessex Scientific Environmental Unit (1987)

0.15–1.1

15

43

Larsen et al. (1990)

0.025–0.15

0.1

0.7

HR Wallingford (1992)

0.04–0.11

4

3.5

1

Equationc,d K va n ¼ 0:1 þ 0:153 VR K va n ¼ 0:06 þ 0:17 VR K va n ¼ 0:032 þ 0:027 Vd K va n ¼ 0:041 þ 0:022 Vd K va n ¼ 0:029 þ 0:022 Vd K va n ¼ 0:057 þ 0:0036 VR K va n ¼ 0:035 þ 0:0239 VR

a

VR, product of the flow velocity V (m s1) and the hydraulic radius R (m). A, channel cross-sectional area (m2). c Kva, vegetation coverage coefficient. d d, water depth (m). b

considering the forces on individual vegetation stands. In order to include this approach in 2-D and 3-D models, an equivalent value of Che´zy’s roughness coefficient is calculated based on characteristics of the vegetation such as drag and density. Unlike the standard approach, this value changes with vegetation density and depth as the simulation progresses. As observed by Baptist (2005), other 3-D models for the resistance due to vegetation have been developed. The models mentioned earlier by Stoesser et al. (2003) and Nicholas and McLelland (2004) did not add any further source terms to the turbulence model, because they were not certain that this would improve the simulation results. Baptist’s model includes the effects of vegetation in the turbulence closure. This has been shown by Uittenbogaard (2003) to fit laboratory measurements of mean flow, eddy viscosity, Reynolds stress, and turbulence intensity well.

2.07.2.10.1 Incremental changes Incremental changes are as follows:

• • •

2.07.2.10.2 Step changes Step changes are as follows:

• 2.07.2.10 Discussion of Current Research and Future Directions Any discussion of future directions quickly becomes dated and in view of this the authors restrict themselves to outlining the areas where new developments are anticipated or required. As a precursor the overall context for river studies should be mentioned and a significant challenge that is already being addressed is how to position river science and engineering within the overall framework of modern river management which entails full recognition of environmental, societal, and economic issues. Overall the major issue in rivers, as in all studies of the natural environment, is how to account for physical features and phenomena that are not directly incorporated into the models (whether conceptual or numerical). In rivers this means, amongst others, bed resistance, vegetation, turbulence, each of which is a significant challenge in its own right. It is perhaps best to consider future directions as progressing by either increments or step changes.

improvements in the estimation of the parameters for bed resistance and better end-user tools that acknowledge uncertainty and encourage a rigorous approach to calibration; improvements in our understanding of flow through vegetation and the ways in which this can be parameterized; and increased understanding of which models to use in which circumstance which should take account of spatial and temporal scales, uncertainty, and levels of acceptable risk; this includes more knowledge of the role of reduced complexity modeling (Hunter et al., 2007).

• • •

new methods of representing resistance parameterization based on improved encapsulation of knowledge from experimental and full-scale measurement; development of fundamental understanding and models for bank accretion to bring this to the level of current work on bank erosion; development of new paradigms to explicitly acknowledge all sources of uncertainty in modeling; and development of a scientific basis for an understanding of the generation, movement, and impact of floating debris.

References Abad JD and Garcia MH (2005) Hydrodynamics in Kinoshita-generated meandering bends: Importance for river planform evolution. In: Parker G and Garcı´a MH (eds.) River, Coastal and Estuarine Morphodynamics: RCEM 2005, pp. 761--771. London: Taylor and Francis (ISBN 0 415 39270 5). Abbott MB and Ionescu F (1967) On the numerical computation of nearly horizontal flows. Journal of Hydraulic Research 5: 97--117. Abernethy B and Rutherfurd ID (1998) Where along a river’s length will vegetation most effectively stabilise stream banks? Geomorphology 23: 55--75. Ackers P (1958) Hydraulics research paper: Resistance of fluids flowing in channels and ducts. HMSO 1: 1--39.

The Hydrodynamics and Morphodynamics of Rivers Alcrudo F and Garcia-Navarro P (1993) A high-resolution Godunov-type scheme in finite volumes for the 2D shallow-water equations. International Journal for Numerical Methods in Fluids 16(6): 489--505. Allen JRL (1968) The nature and origin of bedform hierarchies. Sedimentology 10: 161--182. Allmendiger NE, Pizzuto JE, Potter N Jr., Johnson TE, and Hession WC (2005) The influence of riparian vegetation on stream width, eastern Pennsylvania. USA GSA Bulletin 117(1/2): 229--243 (doi: 10.1130/B25447.1). Anastasiou K and Chan CT (1997) Solution of the 2D shallow water equations using the finite volume method on unstructured triangular meshes. International Journal for Numerical Methods in Fluids 24: 1225--1245. Anderson JD (2005) Ludwig Prandtl’s boundary layer. Physics Today 58(12): 42--48. ASCE (1963) Friction factors in open channels [Task Force on Friction Factors in Open Channels]. Journal of the Hydraulics Division, Proc. ASCE, Vol 89, HY2, March, 97-143 (Discussion in Journal of the Hydraulics Division, Vol 89, July, Sept. & Nov., 1963 and closure in Vol. 90, HY4 July 1964). ASCE Task Committee on Flow and Transport over Dunes (2002) Flow and transport over dunes. Journal of Hydraulic Engineering 128: 726--728. Ashley GM (1990) Classification of large-scale subaqueous bedforms: A new look at an old problem. Journal of Sedimentary Petrology 60: 160--172. Bakry MF (1992) Effect of submerged weeds on the design: Procedure of earthen Egyptian canals. Irrigation and Drainage Systems 6: 179--188. Baptist MJ (2005) Modelling floodplain biogeomorphology. PhD thesis, Delft University of Technology, Delft, The Netherlands, ISBN 90-407-2582-9 (http:// repository.tudelft.nl/view/ir/uuid%3Ab2739720-e2f6-40e2-b55f-1560f434cbee/). Baptist MJ and De Jong JF (2005) Modelling the influence of vegetation on the morphology of the Allier, France. In: Harby A, et al. (eds.) Proceeings of Final COST 626, Silkeborg, Denmark, pp. 15–22. 19–20 May 2005. Baptist MJ, van den Bosch LV, Dijkstra JT, and Kapinga S (2005) Modelling the effects of vegetation on flow and morphology in rivers. Large Rivers 15(1–), HYPERLINK ‘‘http://journalseek.net/cgi-bin/journalseek/journalsearch.cgi?field=issn&query= 0365-284X’’ Archiv fu¨r Hydrobiologie–Supplement 155(1): 339–357. Baptist MJ, Babovic V, Uthurburu JR, Keijzer M, Uittenbogaard RE, Mynett A, and Verwey A (2007) On inducing equations for vegetation resistance. Journal of Hydraulic Research 45(4): 435--450. Batchelor G (1967) An Introduction to Fluid Dynamics. Cambridge: Cambridge University Press. Bates PD, Anderson MG, Price DA, Hardy RJ, and Smith CN (1996) Analysis and development of hydraulic models for floodplain flows. In: Anderson M, Walling D, and Bates P (eds.) Floodplain Processes. Chichester: Wiley. Bates PD and Horritt MS (2005) Modelling wetting and drying processes in hydraulic models. In: Computational Fluid Dynamics: Applications in Environmental Hydraulics, ch. 6, pp. 121–146. Chichester: Wiley (doi:10.1002/0470015195). Barre´ de Saint Venant AJC (1871) The´orie du mouvement non permanent des eaux avec application aux crues des rivie`res et a` l’introduction des mare´es dans leur lits. Compte rendu des seances de l’Academie des Sciences 73: 147–154 and 237–240 (in French). Beaman F, Morvan HP, and Wright NG (2007) Estimating Parameters for Conveyance in 1D Models of Open Channel Flow from Large Eddy Simulation. IAHR Congress, Venice, Italy. Bertoldi W, Pasetto A, Zanoni L, and Tubino M (2005) Experimental observations on channel bifurcations evolving to an equilibrium state. In: Parker G and Garcia MH (eds.) Proceedings of the 4th Symposium on River, Coastal and Estuarine Morphodynamics (RCEM), pp. 409--420. MN, USA: ASCE. Blanckaert K (2002) Flow and Turbulence in Sharp Open-Channel Bends. PhD Thesis, EPFL, Lausanne, Switzerland. Blo¨schl G and Sivapalan M (1995) Scale issues in hydrological modelling: A review. Hydrological Processes: An International Journal 9: Also in: Kalma JD and Sivapalan M (eds.) Advances in Hydrological Processes, Scale Issues in Hydrological Modelling, John Wiley & Sons, 1995, ISBN 0-471-95847-6. Boguchwal LA and Southard JB (1990) Bed configurations in steady unidirectional water flows. Part 1: Scale model study using fine sands. Journal of Sedimentary Research 60: 649--657. Bolla Pittaluga M, Repetto R, and Tubino M (2003) Channel bifurcation in braided rivers: Equilibrium configurations and stability. Water Resources Research 39(3): 1046. Bradford SF and Sanders BF (2002) Finite-volume model for shallow-water flooding of arbitrary topography. Journal of Hydraulic Engineering 128(3): 289--298. Brahams A (1754). Anfangs-Gru¨nde der Deich und Wasserbaukunst Teil 1 und 2, Unvera¨nderter Nachdruck der Ausgabe Aurich. Tapper, 1767 u. 1773 (in German) Leer: Marschenrat, Schuster, 1989. Bridge JS (2003) Rivers and Floodplains: Forms, Processes and Sedimentary Record, 491pp. Bodmin: Blackwell.

153

Bulle H (1926) Untersuchungen u¨ber die Geschiebeableitung bei der Spaltung von Wasserla¨ufen (Investigations on the Sediment Diversion at the Division of Channels) (in German). Berlin: VDI Verlag. Callander RA (1969) Instability and river channels. Journal of Fluid Mechanics 36(3): 465--480. Casulli V and Stelling GS (1998) Numerical simulation of 3D quasi-hydrostatic freesurface flows. Journal of Hydraulic Engineering 124(7): 678--686. Chanson H (1999) The Hydraulics of Open Channel Flow. Oxford: Elsevier. Chanson H (2009) Jean-Baptiste Be´langer, hydraulic engineer, researcher and academic. In: Ettema R (ed.) The 33rd IAHR Congress: Water Engineering for a Sustainable Environment. Vancouver, BC, Canada. Vancouver: IAHR. Che´zy A (1776) Formule pour trouver la vitesse constant que doit avoir l’eau dans une rigole ou un canal dont la pente est donne´e. Dossier 847 (MS 1915) of the manuscript collection of the Ecole des Ponts et Chausse´es. Reproduced as Appendix 4, pp. 247–251 of Mouret (1921). Chow VT (1959) Open Channel Flow. New York: McGraw-Hill. Clifford NJ, Robert A, and Richards KS (1992) Estimation of flow resistance in gravelbedded rivers: A physical explanation of the multiplier of roughness length. Earth Surface Processes and Landforms 17: 111--126. Coulthard TJ and Van de Wiel MJ (2006) A cellular model of river meandering. Earth Surface Processes and Landforms 31: 123--132 (doi:10.1002/esp.1315). Crosato A (1989) Meander migration prediction. Excerpta 4: 169--198. Crosato A (1990)Simulation of Meandering River Processes. Communications on Hydraulic and Geotechnical Engineering, Delft University of Technology, Report No. 90-3, ISSN 0169-6548. Crosato A (2007) Effects of smoothing and regridding in numerical meander migration models. Water Resources Research 43(1): W01401 (doi:10.1029/2006WR005087). Crosato A (2008) Analysis and Modelling of River Meandering. PhD Thesis, Delft University of Technology. Crosato A and Desta FB (2009) Intrinsic steady alternate bars in alluvial channels. Part 1: Experimental observations and numerical tests. In: Vionnet CA, Garcı´a MH, Latrubesse EM, and Perillo GME (eds.) Proceedings of the River, Coastal and Estuarine Morphodynamics: RCEM 2009, vol. 2, pp. 759--765. London: Taylor and Francis. Crosato A and Mosselman E (2009) Simple physics-based predictor for the number of river bars and the transition between meandering and braiding. Water Resources Research 45: W03424 (doi:10.1029/2008WR007242). Crosato A and Samir Saleh M (2010) Numerical study on the effects of floodplain vegetation on reach-scale river morphodynamics. Earth Surface Processes and Landforms. Wiley-InterScience (in press). Crossley AJ, Wright NG, and Whitlow CD (2003) Local time stepping for modeling open channel flows. Journal of Hydraulic Engineering 129(6): 455--462. Dapporto S, Rinaldi M, Casagli N, and Vannocci P (2003) Mechanisms of riverbank failure along the Arno river, central Italy. Earth Surface Processes and Landforms 28: 1303--1323. Darby SE, Alabyan AM, and Van De Wiel MJ (2002) Numerical simulation of bank erosion and channel migration in meandering rivers. Water Resources Research 38(9): 1163--1174. Darby SE and Thorne CR (1996) Stability analysis for steep, eroding, cohesive riverbanks. Journal of Hydraulic Engineering 122(8): 443--454. Davis WM (1899) The geographical cycle. Geographical Journal 14: 481--504. Davies SJ and White CM (1925) A review of flow in pipes and channels. Reprinted by the Offices of Engineering, London, 1–16. Dawson H and Fisher KR (2004) Roughness Review & Roughness Advisor, UK Defra/ Environment Agency Flood and Coastal Defence R&D Programme. Dawson FH and Robinson WN (1984). Submerged macrophytes and the hydraulic roughness of a lowland chalk stream. Verhandlungen der Internationalen Vereingunen fur Theroetische und Angwandte Limnologie, pp. 1944–1948. De Heer AFM and Mosselman E (2004) Flow structure and bedload distribution at alluvial diversions. In: Greco M, Carravetta A, and Morte RD (eds.) Riverflow 2004: Proceedings of the Second International Conference on Fluvial Hydraulics, pp. 802–807. Naples, Italy, 23–25 June 2004. London: Taylor and Francis. De Vriend HJ (1991) Mathematical modelling and large-scale coastal behaviour. Part 1: Physical processes. Journal of Hydraulic Research 29(6): 727--740. De Vriend HJ (1998) Large-scale coastal morphological predictions: A matter of upscaling? In: Proceedings of the 3rd Conference on Hydroscience and -Engineering. Brandenburg University of Technology at Cottbus Cottbus/Berlin, Germany, 31 August–3 September, 1998 (on CDROM). De Vries M (1975) A morphological time scale for rivers. In: Proceedings of the 16th Congress of the IAHR, vol. 2, paper B3, pp. 17–23. Sa˜o Paulo, Brazil.

154

The Hydrodynamics and Morphodynamics of Rivers

Dhondia JF and Stelling GS (2002) Application of one dimensional–two dimensional integrated hydraulic model for flood simulation and damage assessment. Hydroinformatics 2002. Cardiff: IWA Publishing. Dijkstra JT (2003) The influence of vegetation on scroll bar development. MSc Thesis, Delft University of Technology, Delft, the Netherlands (http://repository.tudelft.nl/ search/ir/?q=dijkstra&faculty=&department=&type=&year=). Du Buat PLG (1779) Principes d’hydraulique. Paris: L’imprimerie de monsieur. Engel J, Hotchkiss R, and Hall B (1995) Three dimensional sediment transport modelling using CH3D computer model Proceedings of the First International Water Resources Engineering Conference. New York: ASCE. Engelund F (1970) Instability of erodible beds. Journal of Fluid Mechanics 42(3): 225--244. Engelund F and Fredsøe J (1982) Sediment ripples and dunes. Annual Review of Fluid Mechanics 14: 13--37. ESDU (1979) Losses caused by friction in straight pipes with systematic roughness elements, Engineering Sciences Data Unit (ESDU), pp. 1–40. London, September. Exner FM (1925) Uber die Wechselwirkung zwischen Wasser und Geschiebe in Flussen. Sitzungber Akad. Wiss Wien, Part IIa, Bd. 134, pp. 165–180 (in German). Falconer RA and Lin B (1997) Three-dimensional modelling of water quality in the humber estuary. Water Research, IAWQ 31(5). Fathi-Maghadam M and Kouwen N (1997) Nonrigid, nonsubmerged, vegetative roughness on floodplains. Journal of Hydraulic Engineering 123(1): 51--57. Fischer-Antze T, Sto¨sser T, Bates PD, and Olsen NRB (2001) 3D numerical modelling of open-channel flow with submerged vegetation. Journal of Hydraulic Research 39(3): 303--310. Fisher K and Dawson H (2003) Reducing uncertainty in river flood conveyance: Roughness review. Project W5A-057, DEFRA/Environment Agency Flood and Coastal Defence R and D Programme. http://www.river-conveyance.net/ces/ documents/RoughnessReviewFinal_July07.pdf (accessed May 2010). Flemming BW (1988) Zur Klassifikation subaquatischer, stro¨mungstransversaler Transportko¨rper. Bochumergeologische und geotechnische Arbeiten 29: 44--47 (in German). Fredsøe J (1974) On the development of dunes in erodible channels. Journal of Fluid Mechanics 64: 1--16. Fredsøe J (1978) Meandering and braiding of rivers. Journal of Fluid Mechanics 84(4): 609--624. Fredsøe J (1982) Shape and dimensions of stationary dunes in rivers. Journal of Hydraulic Engineering 108(HY8): 932--946. Garcia-Navarro P, Fras A, and Villanueva I (1999) Dam-break flow simulation: Some results for one-dimensional models of real cases. Journal of Hydrology 216(3–4): 227--247. Garcia-Navarro P and Saviron JM (1992) McCormack’s method for the numericalsimulation of one-dimensional discontinuous unsteady open channel flow – reply. Journal of Hydraulic Research 30(6): 862--863. Ghisalberti M and Nepf HM (2002) Mixing layers and coherent structures in vegetated aquatic flows. Journal of Geophysical Research-Oceans 107(C2). Ghisalberti M and Nepf HM (2004) The limited growth of vegetated shear layers. Water Resources Research 40(7). Ghisalberti M and Nepf H (2006) The structure of the shear layer in flows over rigid and flexible canopies. Environmental Fluid Mechanics 6(3): 277--301. Giri S (2008) Computational modelling of bed form evolution using detailed hydrodynamics: A brief review on current developments. In: van Os AG and Erdbrink CD (eds.) Proceedings of NCR-Days 2008, NCR-Publications 33-2008, pp. 74–75 (ISSN 1568-234X). Minneapolis, MN: NCR. Giri S and Shimizu Y (2006) Numerical computation of sand dune migration with free surface flow. Water Resources Research 42: W10422 (doi:10.1029/2005WR004588). Gurnell AM, van Oosterhout MP, de Vlieger B, and Goodson JM (2006) Reach-scale impacts of aquatic plant growth on physical habitat. River Research and Applications 22: 667--680. Graf WH (1971) Hydraulics of Sediment Transport. 513pp. New York: McGraw-Hill. Graf WH (1984) Hydraulics of Sediment Transport. Part One: A Short History of Sediment Transport. 521pp. (ISBN 0-918334-56-X). Highlands Ranch, CO: Water Resources Publication. Hansen E (1967) On the formation of meanders as a stability problem. Progress Report 13, 9pp. Lyngby: Coastal Engineering Laboratory, Technical University of Denmark, Basic Research. Hasegawa K, Asai S, Kanetaka S, and Baba H (1999) Flow properties of a deep open experimental channel with a dense vegetation bank. Journal of Hydroscience and Hydraulic Engineering 17(2): 59--70. Hodskinson A and Ferguson R (1998) Numerical modelling of separated flow in river bends: Model testing and experimental investigation of geometric controls on the extent of flow separation at the concave bank. Hydrological Processes 12: 1323--1338.

Horritt MS, Mason D, and Luckman AJ (2001) Flood boundary delineation from synthetic aperture radar imagery using a statistical active contour model. International Journal of Remote Sensing 22(13): 2489--2507. Hunter NM, Bates PD, Horrritt MS, and Wilson MD (2007) Simple spatially-distributed models for predicting flood inundation: A review. Geomorphology 90: 208--225. Ikeda S, Parker G, and Sawai K (1981) Bend theory of river meanders. Part 1: Linear development. Journal of Fluid Mechanics 112: 363--377. Jang C-L and Shimizu Y (2007) Vegetation effects on the morphological behavior of alluvial channels. Journal of Hydraulic Research 45(6): 763--772. Johannesson H and Parker G (1989) Velocity redistribution in meandering rivers. Journal of Hydraulic Engineering 115(8): 1019--1039. Kennedy JF (1969) The formation of sediment ripples, dunes and antidunes. Annual Review of Fluid Mechanics 1: 147--168. Kharat DB (2009) Practical Aspects of Integrated 1D–2D Flood Modelling of Urban Floodplains using LiDAR Topography Data. PhD Thesis, School of the Built Environment, Heriot Watt University, Edinburgh. King IP and Norton WR (1978) Recent application of RMA’s finite element models for two dimensional hydrodynamics and water quality Second International Conference on Finite Elements in Water Resources. London: Pentech Press. Kleinhans MG, Jagers HRA, Mosselman E, and Sloff CJ (2008) Bifurcation dynamics and avulsion duration in meandering rivers by one-dimensional and threedimensional models. Water Resources Research 44: W08454 (doi:10.1029/ 2007WR005912). Knight DW (2005) River flood hydraulics: Theoretical issues and stage-discharge relationships. In: Knight DW and Shamseldin AY (eds.) River Basin Modelling for Flood Risk Mitigation, ch. 17, pp. 301–334. Leyden: Balkema. Knight DW, McGahey C, Lamb R, and Samuels P (2009) Practical Channel Hydraulics: Roughness, Conveyance and Afflux. Taylor and Francis. Knight DW, Shiono K, and Pirt J (1989) Prediction of depth mean velocity and discharge in natural rivers with overbank flow. In: Falconer RA, Goodwin P, and Matthew RGS (eds.) International Conference on Hydraulic and Environmental Modelling of Coastal, Estuarine and River Waters. University of Bradford. Aldershot: Gower Technical Press. Kodama T, Wang S, and Kawaharam M (1996) Model verification on 3D tidal current analysis in Tokyo Bay. IInternational Journal for Numerical Methods in Fluids 22: 43--66. Kouwen N (1988) Field estimation of the biomechanical properties of grass. Journal of Hydraulic Research 26(5): 559--568. Krone RB (1962) Flume studies of the transport of sediment in estuarial shoaling processes. Final Report, Hydraulic Engineering Laboratory and Sanitary Engineering Research Laboratory, Berkeley, CA, prepared for US Army Engineer District, San Francisco, CA, under US Army Contract No. DA-04-203 CIVENG-59-2. Krone RB (1963) A Study of Rheological Properties of Estuarial Sediments. Technical Bulletin No. 7, Committee on Tidal Hydraulics, Corps of Engineers, US Army; prepared by US Army Engineer Waterways Experiment Station, Vicksburg, MS. Lane EW (1955) The importance of fluvial morphology in hydraulic engineering. In: Proceedings of the American Society of Civil Engineers, vol. 81, paper 745, 17pp. San Diego, Reston, VA: ASCE. Larsen T, Frier J, and Vestergaard K (1990) Discharge/stage relations in vegetated Danish streams. In: White WR (ed.) Proceedings of the International Conference on River Flood Hydraulics, paper F1. 17–20 September 1990. Wallingford: Wiley. Lee S-H and Wright NG (2010) Simple and efficient solution of the shallow water equations with source terms. International Journal for Numerical Methods in Fluids 63(3): 313--340. Liang QH and Borthwick AGL (2008) Adaptive quadtree simulation of shallow flows with wet-dry fronts over complex topography. Computers and Fluids 38(2): 221--234. Macagno E (1989) Leonardian Fluid Mechanics in the Manuscript I. IIHR Monograph, No. 111. Iowa City, IA: The University of Iowa. Marinoni A (1987) Il Manoscritto I. Florence: Giunti-Barbera (in Italian). Marshall EJP and Westlake DF (1990) Water velocities around water plants in chalk streams. Folia Geobotanica 25: 279--289. Massey BS (1995) Mechanics of Fluids. Chapman and Hall. Morvan HP, Knight DW, Wright NG, Tang XN, and Crossley AJ (2008) The concept of roughness in fluvial hydraulics and its formulation in 1-D, 2-D & 3-D numerical simulation models. Journal of Hydraulic Research 46(2): 191--208. Morvan H, Pender G, Wright NG, and Ervine DA (2002) Three-dimensional hydrodynamics of meandering compound channels. Journal of Hydraulic Engineering 128(7): 674--682.

The Hydrodynamics and Morphodynamics of Rivers Mosselman E (1992) Mathematical Modelling of Morphological Processes in Rivers with Erodible Cohesive Banks. PhD Thesis, Communications on Hydraulic and Geotechnical Engineering, No. 92-3, Delft University of Technology, ISSN 01696548. Mosselman E (1998) Morphological modelling of rivers with erodible banks. Hydrological Processes 12: 1357--1370. Mosselman E, Shishikura T, and Klaassen GJ (2000) Effect of bank stabilisation on bend scour in anabranches of braided rivers. Physics and Chemistry of the Earth, Part B 25(7–8): 699--704. Murray AB and Paola C (1994) A cellular model of braided rivers. Nature 371: 54--57. Murray AB and Paola C (2003) Modelling the effects of vegetation on channel pattern in bedload rivers. Earth Surface Processes and Landforms 28: 131--143 (doi: 10.1002/esp.428). Nabi M, De Vriend HJ, Mosselman E, Sloff CJ, and Shimizu Y (2009) Simulation of subaqueous dunes using detailed hydrodynamics. In: Vionnet CA, Garcı´a MH, Latrubesse EM, and Perillo GME (eds.) Proceedings of the River, Coastal and Estuarine Morphodynamics: RCEM 2009, pp. 967--974. London: Taylor and Francis. Nanson GC and Hickin EJ (1986) A statistical analysis of bank erosion and channel migration in Western Canada. Bulletin of the Geological Society of America 97(4): 497--504. Nelson JM, McLean SR, and Wolfe SR (1993) Mean flow and turbulence fields over two-dimensional bedforms. Water Resources Research 29: 3935--3953. Nex A and Samuels P (1999) The use of 3 D CFD models in river flood. defence, Report No. SR542, HR Wallingford. Nicholas AP and McLelland SJ (2004) Computational fluid dynamics modelling of three-dimensional processes on natural river floodplains. Journal of Hydraulic Research 42(2): 131--143. Olesen KW (1984) Alternate bars in and meandering of alluvial rivers. In: Elliott CM (ed.) River Meandering, Proceedings of the Conference Rivers ‘83, pp. 873–884. New Orleans, LA, USA, 24–26 October 1983. New York: ASCE. Olsen NRB (2000) Unstructured hexahedral 3D grids for CFD modelling in fluvial geomorphology. Fourth International Conf. Hydroinformatics 2000, Iowa, USA. Olsen NRB and Stokseth S (1995) 3-Dimensional numerical modeling of water-flow in a river with large bed roughness. Journal of Hydraulic Research 33(4): 571--581. Osman AM and Thorne CR (1988) Riverbank stability analysis I: Theory. Journal of Hydraulic Engineering 114(2): 134--150. Ott RA (2000) Factors affecting stream bank and river bank stability, with an emphasis on vegetation influences. Prepared for the Region III Forest Practices Riparian Management Committee, Tanana Chiefs Conference, Inc. Forestry Program, Fairbanks, Alaska. Parker G (1976) On the cause and characteristic scales of meandering and braiding in rivers. Journal of Fluid Mechanics 76(3): 457--479. Parker G (1978a) Self-formed straight rivers with equilibrium banks and mobile bed. Part 1: The sand–silt river. Journal of Fluid Mechanics 89: 109--125. Parker G (1978b) Self-formed straight rivers with equilibrium banks and mobile bed. Part 2: The gravel river. Journal of Fluid Mechanics 89: 127--146. Partheniades E (1962) A Study of Erosion and Deposition of Cohesive Soils in Salt Water. PhD Thesis, University of California, Berkeley, CA, USA. Partheniades E (1965) Erosion and deposition of cohesive soils. Journal of the Hydraulic Division 91(HY1): 105--139. Pepper AT (1970) Investigation of Vegetation and Bend Flow Retardation of a Stretch of the River Ousel. BSc Dissertation, National College of Agricultural Engineering, Silsoe. Phillips JD (1995) Biogeomorphology and landscape evolution: The problem of scale. Geomorphology 13: 337--347. Pitlo RH (1982) Flow resistance of aquatic vegetation. In: Proceedings of the 6th EWRS Symposium on Aquatic Weeds. European Weed Research Society. Pizzuto JE (1984) Bank erodibility of shallow sandbed streams. Earth Surface Processes and Landforms 9: 113--124. Powell KEC (1978) Weed growth – a factor in channel roughness. In: Herschy RW (ed.) Hydrometry, Principles and Practice, pp. 327--352. Chichester: Wiley. Preissmann A (1961) Propagation des intumescences dans les canaux et rivieres. In: Proceedings of the First Congress of the French Association for Computation, pp. 433–442. Grenoble, France. Pirim T, Bennet SJ, and Barkdoll BD (2000) Restoration of degraded stream corridors using vegetation: An experimental study. United States Department of Agriculture, Channel & Watershed Processes Research Unit, National Sedimentation Laboratory, Research Report No 14. Querner EP (1994) Aquatic weed control within an integrated water management framework. In: White WR (ed.) Report 67 of the Agricultural Research Department. Wageningen: Agricultural Research Department.

155

Rahmeyer W, Werth D, and Freeman GE (1996) Flow resistance due to vegetation in compound channels and floodplains. In: Semi-Annual Conference on Multiple Solutions to Floodplain Management. Utah, November 1996. Rastogi A and Rodi W (1978) Predictions of heat and mass transfer in open channels. Journal of Hydraulic Division 104(HY3): 397--420. Reynolds AJ (1974) Turbulent Flows in Engineering. London: Wiley. Riad K (1961) Analytical and Experimental Study of Bed Load Distribution at Alluvial Diversions. Doctoral Thesis, Delft University of Technology, Delft, The Netherlands. Rinaldi M and Casagli N (1999) Stability of streambanks formed in partially saturated soils and effects of negative pore water pressure: The Sieve River (Italy). Geomorphology 26(4): 253--277. Rinaldi M, Casagli N, Dapporto S, and Gargini A (2004) Monitoring and modelling of pore water pressure changes and riverbank stability during flow events. Earth Surface Processes and Landforms 29: 237--254. Rinaldi M and Darby SE (2005) Advances in modelling river bank erosion processes. In: Proceedings of the 6th International Gravel Bed Rivers VI, Lienz, Austria, 5–9 September, 2005. Rodrigues S, Bre´he´ret J-G, Macaire J-J, Moatar F, Nistoran D, and Juge´ P (2006) Flow and sediment dynamics in the vegetated secondary channels of an anabranching river: The Loire River (France). Sedimentary Geology 186: 89--109. Rouse H (1965) Critical analysis of open-channel hydraulics. Journal of Hydraulic Engineering 91: 1--25. Salama MM and Bakry MF (1992) Design of earthen, vegetated open channels. Water Resources Management 6: 149--159. Sayre WW and Albertson ML (1963) Roughness spacing in rigid open channels. Transactions of American Society of Civil Engineers, ASCE 128(1): 343--427. Schlichting H, Gersten K, Krause E, and Oertel HJ (2004) Boundary-Layer Theory. Springer Samir Saleh M and Crosato A (2008) Effects of riparian and floodplain vegetation on river patterns and flow dynamics. In: Gumiero G, Rinaldi M, and Fokkens B (eds.) Proceedings of 4th ECRR International Conference on River Restoration, Italy, Venice S. Servolo Island, 16–21 June 2008, ECRR-CIRF Publication. Printed by Industrie Grafiche Vicentine S.r.l., pp. 807–814. Schumm SA and Lichty RW (1965) Time, space and casuality in geomorphology. American Journal of Science 263: 110--119. Seminara G and Tubino M (1989) Alternate bar and meandering: Free, forced and mixed interactions. In: Ikeda S and Parker G (eds.) River Meandering, Water Resources Monograph, vol. 12, pp. 267–320 (ISBN 0-87590-316-9). Washington, DC: AGU. Shields AF (1936) Application of Similarity Principles and Turbulence Research to Bed-Load Movement. Hydrodynamics Laboratory Publication No. 167, Ott WP and Van Uchelen JC (trans.), US Department of Agriculture, Soil Conservation Service, Cooperative Laboratory, California Institute of Technology, Pasadena, CA. Shih SF and Rahi GS (1982) Seasonal variations of Manning’s roughness coefficient in a subtropical marsh. American Society of Agricultural Engineers 25: 116--119. Simons DB and Richardson EV (1961) Forms of bed roughness in alluvial channels. Journal of Hydraulics Division 87(1): 87--105. Sleigh PA, Gaskell PH, Berzins M, and Wright NG (1998) An unstructured finitevolume algorithm for predicting flow in rivers and estuaries. Computers and Fluids 27(4): 479--508. Smagorinsky J (1963) General circulation experiments with primitive equations. Part 1: Basic experiments. Monthly Weather Review 91: 99--164. Southard JB and Boguchwal LA (1990) Bed configuration in steady unidirectional water flows. Part 2: Synthesis of flume data. Journal of Sedimentary Research 60(5): 658--679. Stansby PK and Zhou JG (1998) Shallow-water flow solver with nonhydrostatic pressure: 2D vertical plane problems. International Journal for Numerical Methods in Fluids 28: 541--563. Stephens JC, Blackburn RD, Seamon DE, and Weldon LW (1963) Flow retardance by channel weed and their control. Journal of the Irrigation and Drainage Division 89(IR2): 31--56. Stoesser T, Wilson CAME, Bates PD, and Dittrich A (2003) Application of a 3D numerical model to a river with vegetated floodplains. Journal of Hydroinformatics 5(2): 99--112. Strickler A (1923) Beitrage auf Frage der eschwindigkeitsformel under der Rauhiskeitszahlen frustrome, kanale under gescholossene Leitungen. Mitteilungen des eidgenossischen Amtes fur Wasserwirtschaft 16. Sun T, Meakin P, and Jøssang T (1996) A simulation model for meandering rivers. Water Resources Research 32(9): 2937--2954. Ten Brinke WBM (2005) The Dutch Rhine: A restrained river. Uitgeverij Veen Magazines, Diemen, The Netherlands, ISBN 9076988 919.

156

The Hydrodynamics and Morphodynamics of Rivers

Thorne CR (1982) Processes and mechanisms of river bank erosion. In: Hey RD, Bathurst JC and Thorne CR (eds.) Gravel-Bed Rivers, pp. 227--259. Chichester: Wiley. Thorne CR (1988) Riverbank stability analysis. II: Applications. Journal of Hydraulic Engineering 114(2): 151--172. Thorne CR (1990) Effects of vegetation on riverbank erosion and stability. In: Thornes JB (ed.) Vegetation and Erosion, pp. 125--144. Chichester: Wiley. Tsujimoto T (1999) Fluvial processes in streams with vegetation. Journal of Hydraulic Research 37(6): 789--803. Uittenbogaard R (2003) Points of view and perspectives of horizontal large-eddy simulation at Delft, CERI, Sapporo, http://www.wldelft.nl/rnd/publ/docs/ Ui_CE_2003.pdf. van Bendegom L (1947) Enige beschouwingen over riviermorfologie en rivierverbetering. De Ingenieur B. Bouw- en Waterbouwkunde 1 59(4): 1–11 (in Dutch). (Some considerations on river morphology and river improvement. English translation, Natural Resources Council Canada, 1963, Technical Translation No. 1054.) Versteeg HK and Malalasekera W (2007) Introduction to Computational Fluid Dynamics: The Finite Volume Method, 503pp. Harlow: Pearson. Villada Arroyave JA and Crosato A (2010) Effects of river floodplain lowering and vegetation cover. In: Proceedings of the Institution of Civil Engineers, Water Management, vol. 163, pp. 1–11 (doi:10.1680/wama2010.163.1.1). Wallingford (1992) The hydraulic roughness of vegetated channels. Report No. SR 305, March 1992. HR Wallingford. Wang SSY, Alonso VV, Brebbia CA, Gray WG, and Pinder GF (1989) Finite elements in water resources. Third International Conference, Finite Elements in Water Resources, Mississippi, USA. Wang ZB, Fokkink RJ, De Vries M, and Langerak A (1995) Stability of river bifurcations in 1D morphodynamic models. Journal of Hydraulic Research 33(6): 739--750. Watson D (1987) Hydraulic effects of aquatic weeds in UK rivers. Regulated Rivers: Research and Management 1: 211--227. Wessex Scientific Environmental Unit (1987) The Effect of Aquatic Macrophytes on the Hydraulic Roughness of a Lowland Chalk River. Wright NG (2001) Conveyance implications for 2D and 3D modelling. Scoping Study for Reducing Uncertainty in River Flood Conveyance. Environment Agency (UK). Wright NG (2005) Introduction to numerical methods for fluid flow. In: Bates P, Ferguson R, and Lane SN (eds.) Computational Fluid Dynamics: Applications in Environmental Hydraulics. Chichester: Wiley.

Wright NG, Villanueva I, Bates PD, et al. (2008) A case study of the use of remotelysensed data for modelling flood inundation on the River Severn, UK. Journal of Hydraulic Engineering 134(5): 533--540. Wu FC, Shen HW, and Chou YJ (1999) Variation of roughness coefficients for unsubmerged and submerged vegetation. Journal of Hydraulic Engineering 125(9): 934--942 (doi:10.1061/(ASCE)0733-9429(1999)). Yen BC (1991) Channel Flow Resistance: Centennial of Manning’s Formula. Colorado, USA: Water Resources Publications. Yen BC (2002) Open channel flow resistance. Journal of Hydraulic Engineering 128(1): 20--39. Zienkiewicz OZ and Cheung YK (1965) Finite elements in the solution of field problems. Engineer 507--510. Zolezzi G (1999) River Meandering Morphodynamics. PhD Thesis, 180pp. Department of Environmental Engineering, University of Genoa.

Relevant Websites http://delftsoftware.wldelft.nl Deltares; Delft Hydraulics Software: SOBEK and Delft3D. http://www.halcrow.com Halcrow; ISIS Software. http://www.hec.usace.army.mil Hydrologic Engineering Center; HEC-RAS Software. http://www.mikebydhi.com MIKE by DHI. http://www.river-conveyance.net Reducing Uncertainty in Estimation of Flood Levels; Conveyance and Afflux Estimation System (CES/AES). http://wwwbrr.cr.usgs.gov US Geological Survey Central Region Research; Geomorphology and Sediment Transport Laboratory of USGS. http://vtchl.uiuc.edu Ven Te Chow Hydrosystems Laboratory; Gary Parker’s e-book. http://www.wallingfordsoftware.com Wallingford Software; InfoWorks Software.

2.08 Lakes and Reservoirs D Uhlmann and L Paul, University of Technology, Dresden, Germany M Hupfer, Leibniz-Institute of Freshwater Ecology and Inland Fisheries, Berlin, Germany R Fischer, Consulting Engineers for Water and Soil Limited, Possendorf, Germany & 2011 Elsevier B.V. All rights reserved.

2.08.1 2.08.1.1 2.08.1.1.1 2.08.1.1.2 2.08.1.1.3 2.08.1.2 2.08.1.3 2.08.1.4 2.08.1.4.1 2.08.1.4.2 2.08.1.5 2.08.1.6 2.08.1.6.1 2.08.1.6.2 2.08.1.6.3 2.08.1.6.4 2.08.1.6.5 2.08.1.6.6 2.08.1.6.7 2.08.1.7 2.08.1.8 2.08.1.9 2.08.2 2.08.2.1 2.08.2.2 2.08.2.3 2.08.3 2.08.3.1 2.08.3.2 2.08.3.3 2.08.3.3.1 2.08.3.3.2 2.08.4 2.08.4.1 2.08.4.2 2.08.4.3 2.08.4.4 References

Morphometry, Hydrodynamics, Chemistry, and Biology of Lakes Origin and Development of Lakes Origin of lakes Lake development with a large contribution of photosynthesis Development of reservoirs Structure and Functioning of Drainage Basins of Natural and Man-Made Lakes Lake and Reservoir Morphometry Influx and Vertical Distribution of Solar Energy Underwater light conditions Heat budget and thermal structure Water Movement Basic Chemistry Systematics of lakes with respect to water quality Ionic balance Inorganic compounds and buffer properties Sequence of microbially mediated redox processes Iron, manganese, and sulfur compounds Nutrients (nitrogen and phosphorus) and trace substances Organic carbon – humic compounds Biotic Structure Photosynthesis: Generation and Consumption of Dissolved Oxygen Oxygen Stratification: Circulation/Quality Types of Lakes Fundamental Properties of Reservoirs Functions of Reservoirs Characteristic Differences between Natural Lakes and Reservoirs Environmental Impacts of Reservoirs Management, Protection, and Rehabilitation of Lakes and Reservoirs Main Water-Quality Problems General Management Strategies Measures for Eutrophication Control External measures Internal measures Current Knowledge Gaps and Future Research Needs Lakes and Reservoirs as Constituents of Their Catchment Areas Responses of Lakes and Reservoirs to Climate Change Biodiversity and Its Role in the Functioning of Lake and Reservoir Ecosystems Integrated Management of Lakes and Reservoirs

2.08.1 Morphometry, Hydrodynamics, Chemistry, and Biology of Lakes 2.08.1.1 Origin and Development of Lakes

157 157 157 158 159 159 160 163 163 166 170 173 173 176 176 177 179 181 184 185 189 192 196 196 199 203 204 204 204 205 205 205 208 209 210 210 210 211

useful in understanding some of the general characteristics of the lake and drawing comparisons with other lakes of similar origins. In general, the forces forming a lake are

2.08.1.1.1 Origin of lakes Lakes are hollows which are filled, at least partially, with water. The nature of the physico-chemical and biological events taking place in a lake is related to its shape and size, as well as to the characteristics of the drainage basin. These characteristics are in turn largely determined by the mode of origin of the lake. Thus, ascertaining the mode of origin of a lake is

1. catastrophic, or sudden in geological terms; 2. regional in nature, often giving rise to several similar lakes forming a lake district; and 3. caused by erosion (of the outlet) and sedimentation of the basin so that lakes become temporary features of the landscape.

157

158

Lakes and Reservoirs

From a geological point of view, not only reservoirs but also lakes are short-lived (with a few exceptions such as Lake Baikal and Lake Tanganyika, the first of which is more than 56 million years old). Lakes are formed in two ways: (1) by the filling with water into a natural depression or (2) by impoundment behind a natural dam. A depression may be formed by a geological deformation (tectonic or volcanic), by elutriation, by ice or water erosion, or by the slow melting of dead ice blocks in the subsoil. Lake pans may also have been deflated by wind action. Natural dams may be formed by glacial moraines, by landslides, or by biogenic CaCO3 deposition. Most of the existing lakes have been formed by glacial activity during the last ice age. The following statements on the genesis of lakes are essentially based upon Wetzel (2001), Hutchinson (1957), and Keller (1962): 1. Lakes formed by glacial ice movement are more in number than lakes formed by other processes. With the retreat of the large Pleistocene glaciers, an immense number of lakes were formed, creating a large variety of glacially formed lake types. Glacial ice-scour lakes occur in extended rock areas where loosened rock materials have been removed by glaciers. Examples are the upland peneplains of Scandinavia, the United Kingdom, and the great Canadian Shield. Due to scouring of preexisting valleys to great depths, the Great Bear Lake and the Great Slave Lake were formed. The most impressive examples of lakes on the North American continent produced by ice erosion of rocks, however, are the Laurentian Great Lakes. 2. In glaciated valleys, chains of smaller lakes could also have evolved by scouring. Morainal damming of pre-glacial valleys created many lakes in the Northern Hemisphere. Dams of moraine material may, in high mountains, occasionally attain a height of more than 120 m (Keller, 1962). Cirque lakes are frequently arranged along a valley in stairways. Many lakes also emerged from cavities in tillite, a metamorphosed old bedrock. The closely related Kettle lakes originated from the melting of ice blocks which had previously been buried for up to several hundred years, in moraine material. 3. In the flat regions of Siberia and in northern America, millions of small, shallow cryogenic lakes, which evolved from the thawing of local permafrost soils, were formed. Biogenic formation of CaCO3 dams occurs when due to biological (photosynthesis) or mechanical (very turbulent flow) removal of (carbonate-balanced) CO2 from Ca (HCO3)2-rich water (cf. Section 2.08.1.6.3), calcareous crusts may be generated with a growth of 1 cm yr1 or more. In this way, barriers which are able to impound a river can be formed. In front of these dams, there are usually waterfalls and behind them there are lakes (Figure 1). One of the organisms which may cause this growth of high-calc-sinter dams is the filamentous cyanobacterium, Schizothrix. 4. The lake basins of tectonic lakes are depressions which have been formed by movements of comparatively deep portions of the Earth’s crust. Most of these lakes are a result

Figure 1 Some of the Plitvice Lakes, Croatia, the dams of which have been formed by biogenic precipitation of CaCO3. The brown color is caused by Fe3þ. Courtesy of Dr. Anita Belanovic.

Figure 2 Diagram of a tectonic lake basin: a depressed fault-block between two upheaved fault-blocks. In the foreground, situation after a considerable period of erosion and deposition. From Wetzel RG (2001) Limnology. Lake and River Ecosystems, 3rd edn. San Diego, CA: Academic Press.

of faulting with single-fault displacements, or exist in downfaulted troughs (Figure 2; Wetzel, 2001). The latter type is called a ‘graben’. Well-known examples are Lake Baikal in Siberia and Lake Tanganyika in equatorial Africa. These lakes have maximum depths of 1620 m and 1435 m, respectively. Both lakes contain a large number of plant and animal species which are endemic, that is, they occur only in these particular water bodies. Both lakes were already in existence in the Mesozoic period. Another

Lakes and Reservoirs

5.

6.

7.

8.

9.

well-known example of a graben lake is Lake Tahoe (California/Nevada). From a moderate uplifting of the seabed connected to tectonic movements, the Caspian Sea and the Aral Sea in Western Asia were formed in the Miocene period. Upwarping of the Earth’s crust also resulted in the formation of other large lakes such as Lake Okeechobee, Florida, and Lake Victoria, Central Africa. Volcanic lakes are formed when depressions that may be formed due to volcanic activity are undrained, and usually are filled with water. The basins and their drainage areas often have a basaltic nature and thus a low concentration of dissolved solids, inclusive of nutrients. Volcanic crater lakes are often circular and are called ‘maar lakes’ if they have small diameters (up to 2000 m) and ‘calderas’ when of a larger size. A well-known example of a maar lake is Crater Lake, Oregon, with a depth of 608 m. Lava streams may flow into a preexisting river valley and form a dam wall. Behind this dam, a lake may be created. Lake origins are also formed by landslides when large quantities of unconsolidated material suddenly move into the floors of valley streams to create dams and lakes (Figure 3; Wetzel, 2001). Such landslides occur frequently in glaciated mountains. The landslides are usually brought about by abnormal events such as excessive rain acting on unstable slopes, or by earthquake activity. Disastrous floods may be caused downstream if such dams break. Solution lakes have been created by the dissolution of carbonate, also of sulfate or other soluble rock. They are mostly connected, similar to many other lake types, with the groundwater. Among several lake types formed by river activity, floodplain lakes are the best known. Oxbow lakes were created from truncated meanders. Deflation lakes result from the erosive effect of wind, mostly in arid areas. They are often ephemeral. The fine structure of sediments/soils in very large dry depressions in North and South Africa, in many cases, reflect climatic changes over the last millennia.

Figure 3 Lake formed by a large landslide into a steep-sided streameroded canyon. From Wetzel RG (2001) Limnology. Lake and River Ecosystems, 3rd edn. San Diego, CA: Academic Press.

159

2.08.1.1.2 Lake development with a large contribution of photosynthesis As soon as in a shallow lake the higher emergent vegetation (see Figure 33; Uhlmann, 1979) starts to predominate, the accumulation of biomass residues, mostly, cellulose and lignin mud, considerably increases. The thickness of the organic sediment layer may be a multiple of the water depth in the senescent stage of a lake (see Figure 41; Kusnezow, 1959). The accelerated sedimentation and resulting shallowness favor both the further spreading of emergent vegetation and the increase in water losses by evapotranspiration. This often results in the final disappearance of the water body. The biomass residues may also originate from floating mats of Sphagnum moss. These peat-forming mosses initially colonize the outer margins of the water body. The drainage patterns of lakes in the flood plains of tropical rivers can be largely altered by massive growths, not only of emergent, but also of floating-leaved vegetation. This can probably also go along with increased sediment accumulation. In the early stages of development of clear-water lakes formed by glaciation, the biotic productivity is limited by the lack of nutrients, the long winter period, and a high removal rate of dissolved organic materials by photolysis. During the long intermediate stage of lake ontogeny, phytoplankton production governs sediment accumulation, with a small contribution of macrophyte biomass. Organic sedimentation here is largely balanced by microbial decomposition. In the later and terminal stages, the proportion of emergent and wetland vegetation and thus the biogenic silting-up rapidly increases, due to the great amount of lignified and cellulosic residues which now accumulate. This also applies to lakes which are shallow initially. There are also other terminal stages in the development of lakes in temperate climates, for example, various types of persistent mire ecosystems and shallow lakes which are durable due to a permanent high inflow and level of groundwater.

2.08.1.1.3 Development of reservoirs Compared with natural lakes, reservoirs are extremely shortlived. They are designed for a lifetime of at least 50 years, but this is, in many cases, only realistic if the deposited silt is collected in pre-reservoirs and mechanically removed at intervals of several decades. Without such countermeasures, a reservoir may be completely filled with sediment within a few years in areas with very heavy erosion. On the other hand, reservoirs with an estimated lifetime of several centuries also exist (Nilsson, 2009). In the first phase of impoundment, the plankton benefit from the release of nutrients, due to the degradation of the submerged terrestrial vegetation. This may even increase the fish yields. The second phase is characterized by an oligotrophication if the inflowing water is poor in nutrients, or by a eutrophication if the water is fertilized by domestic animals or agricultural effluent. The final phase is an advanced deposition of silt whereby conditions for biogenic siltation are substantially improved. In tropical reservoirs, luxuriant growths of floating-leaved plants become possible under conditions of comparatively small water-level fluctuations.

160

Lakes and Reservoirs

2.08.1.2 Structure and Functioning of Drainage Basins of Natural and Man-Made Lakes The drainage basin clearly regulates the characteristics of lakes and reservoirs, which include soil, ionic composition, slope, and, in combination with the climate, vegetation cover (Wetzel, 2001). Soil and vegetation not only influence the runoff, but also the composition and quantity of organic matter that enters the tributaries. The area of the drainage basin as related to the area of the water body is normally o10:1 in the case of lakes, but it is 410:1 for reservoir catchments. The drainage basin of a reservoir is usually large enough to fill the man-made lake within a period of 1 year or less. The influence of the structure and the uses of the drainage basin upon the water quality are therefore quite substantial. Reservoirs are also subjected much more to the stochastics in the relationship between atmospheric precipitation and runoff, than is the case in Pleistocene lakes, which are largely supplied by groundwater. Lakes are often close to the center of their drainage basin, whereas reservoirs are located at the margin. This is predetermined by the morphometry of the territory. Densely forested drainage basins provide a good water quality in reservoirs because of their anti-erosion function. Furthermore, a forest cover promotes the infiltration of rainwater as a pre-treatment step in the context of drinking-water supply. The potable water supply of big cities such as New York, Tokyo, Beijing, Rio de Janeiro, and Los Angeles is completely, or largely, based upon water from forested drainage basins (K. H. Feger, personal communication). From agricultural areas, mainly soil particles, nitrogen and phosphorus from fertilizers, as well as herbicides and pesticides are lost by surface runoff, due to heavy rains. Pastures may not only be sources of N- and P-compounds, but also release cysts of parasitic protists from manure depositions. The nitrate concentration in a reservoir or lake is often an indicator of the state of the environment. In the Saidenbach Reservoir (Germany), the increase in agricultural production, and particularly the application of liquid manure to the catchment, has caused a threefold increase of nitrate concentration in 17 years (rise from 10 mg l1 in 1962 to 30 mg l1 in 1979, W. Horn in Uhlmann and Horn, 2001). Densely populated drainage basins are generally not compatible with the safe operation of drinking-water reservoirs. The introduction of purified domestic effluent (with advanced treatment) into drinking-water reservoirs is extremely problematic not only due to the loads of potentially harmful microorganisms, but also due to unacceptable N and P concentrations in storm-water outlets, subsequent to heavy rains. The allowable P concentrations for purified effluent in flowing waters are normally set at a level which does not affect the ecosystem. However, the very low P concentrations of around 10 mg l1, which are required for drinking-water reservoirs, may not be achieved downstream of wastewatertreatment plants even if these are operated in full accordance with internationally accepted regulations. In temperate climates, drinking-water reservoirs are situated mostly in hilly areas with igneous rock as the mineral subsoil. Consequently, they often have water of low hardness or they can even be weakly acidic. This facilitates the binding

of phosphate to Al- and Fe3þ-complexes (also in the colloidalsize class) and favors an oligotrophic state of the water body. In the past decades, many sites in Europe and North America have become subject to an acidification of soil and water due to atmospheric depositions. Liming of the forests in the drainage basins has been used as a counteractive measure, but it simultaneously increases the trophic state (i.e., phytoplankton production) as is well known from fishponds. In former decades, bogs in the catchments of drinkingwater reservoirs were often drained. If clearing of the drainage ditches is not done on a regular basis, the resulting waterlogging leads not only to an increased leaching of (coniferous) soils, but also to the growth of bogs. This may result in an increased concentration of humic substances/dissolved organic carbon in reservoir waters (Sudbrack et al., 2005). Thus, the costs for water treatment may largely increase. Sometimes reservoir systems are interconnected. Downstream water bodies generally receive better quality water which is improved due to the retention in the upstream water bodies, and they serve as pre-reservoirs for water treatment. Thus, the concentration of imported suspended solids is generally much lower downstream. In many cases, the trophic state and phytoplankton production (Sections 2.08.1.7 and 34.1.8) likewise decrease. The quality may also be improved by introducing water from reservoirs which are situated in a bypass.

2.08.1.3 Lake and Reservoir Morphometry The morphology of lakes, their size and shape, is often related to their origin and age. Lake morphometry is the quantification of characteristic morphological dimensions whose fundamental limnological importance was emphasized by Kalff (2002): Regardless of how lakes are formed, their surface shape, surface area, underwater form, depth and the irregularity of their shorelines have a major impact on turbulence, lake stratification, sedimentation and resuspension, and the extent of littoral-zone wetlands that determine lake functioning.

The determination of morphometric measures requires a bathymetric map of the lake with a scaled outline of the shoreline and submerged contour lines in several depths below the surface. In the past, the depth development of a lake had to be determined by lowering a plumb line from a boat or the frozen surface at many stations. Nowadays, precise bathymetric maps of water bodies are created using digital sonars coupled with a global positioning system (GPS)-receiver and data evaluation using geographic information system GIS-software (Figure 4; Sytsma et al., 2004). The interaction of a lake with the atmosphere (e.g., radiative energy balance, gas exchange, and direct matter import by precipitation) and the impact of driving meteorological forces (particularly the effect of wind on mixing and stratification, surface waves, and water movements) depend primarily on its surface area A0 and form described by the maximum length lmax and width bmax measured perpendicular to lmax. The maximum length is the distance between the two most distant points of the lake surface. It is often measured as a straight line that may cross islands or promontories. Sometimes, it is

Elevation, m

Lakes and Reservoirs

161

1660 1650 1640 1630 1620 1610 1600 1590 1580 1570 1560 1550 1540 1530 1520 0

300

600

900 1200 1500 1800 2100 2400 2700 3000 Surface area, ha

Depth (m) N

0

Elevation, m

Waldo Lake Bathymetric Map

130 Kilometers 0

0.5

1

2

1660 1650 1640 1630 1620 1610 1600 1590 1580 1570 1560 1550 1540 1530 1520 0

(a)

(b)

100 200 300 400 500 600 700 800 900 1000 1100 1200 1300

(c)

Volume, cubic hectometers

Figure 4 Example of a bathymetric study at the Waldo Lake (Oregon, USA): (a) data-collection cruise paths; (b) bathymetric map interpolated from data collected; (c) resulting hypsographic curve Az (above) and volume–depth distribution Vz (below). From Sytsma M, Rueter J, Petersen R, et al. (2004) Waldo Lake Research in 2003. Center for Lakes and Reservoirs, Department of Environmental Sciences and Resources, Department of Civil and Environmental Engineering, Portland State University, Portland, Oregon 97201–0751. http://www.clr.pdx.edu/docs/2003report.pdf (accessed April 2010).

Wind fetch (km) 0

5

10

15

20

25

30

0 A0

5 Zmix (m)

VBL b max l max

b eff

l eff

Temperate zone

10

Arai (1981) Patalas (1984) Kling (1988)

15

Hanna (1990)

20 25

Africa

30 500 m Figure 5 Morphometric characteristics of the Neunzehnhain II reservoir (Germany, 501 42.60 N, 131 09.130 E; VBL: pre-dam). Reservoir and circle have identical areas (A0 ¼ 28.9 ha). The shoreline development DL of the reservoir is 2.0 and that of the circle is 1.0.

measured along the thalweg of the lake. In contrast with bmax, the mean width bmean is given as the quotient of A0 and lmax. However, most important are the effective length leff and width beff (perpendicular to leff) which represent the longest distances from shore to shore, not interrupted by land (Figure 5). The effective lake axis deff is the average of leff and beff. Surface dimensions are used to quantify the wind effect on a lake. The parameter leff is usually called the maximum ‘effective wind fetch’ F (Ha˚kanson, 1981; Kling, 1988; Hanna, pffiffiffiffiffi 1990). Other authors define F as A0 (Arai, 1981) or deff (Patalas, 1984). No matter how F is defined, it was frequently

Figure 6 Empirical relationships between wind fetch and mixing depth zmix established for lakes in Japan, Europe and North America, and in tropical Africa. Data for lakes in Japan from Arai T (1981) Climatic and geomorphological influences on lake temperature. Verhandlungen Internationale Vereinigung fu¨r Theoretische und Angewandte Limnologie 21: 130–134; for lakes in Europe and North America from Patalas K (1984) Mid-summer mixing depths of lakes of different latitudes. Verhandlungen Internationale Vereinigung fu¨r Theoretische und Angewandte Limnologie 22: 97–102 and Hanna M (1990) Evaluation of models predicting mixing depth. Canadian Journal of Fisheries and Aquatic Sciences 47: 940–947; and for lakes in tropical Africa from Kling GW (1988) Comparative transparency, depth of mixing and stability of stratification in lakes of Cameroon, West Africa. Limnology and Oceanography 33: 27–40.

found to correlate significantly with mixing depth zmix (Figure 6; Arai, 1981; Patalas, 1984; Kling, 1988; Hanna, 1990). Furthermore, the maximum height of surface waves and thus, their erosive impact on the shores, sediment resuspension,

162

Lakes and Reservoirs

and transport into deeper regions of the lake are related to F (see Section 2.08.1.5). The length of the shoreline L0 and the dimensionless shoreline development DL with

L0 DL ¼ pffiffiffiffiffiffiffiffi 2 pA0

light conditions, nutrients, oxygen, primary productivity, sediment resuspension, and many others. Thienemann (1927) has already stated that shallow lakes generally tend toward a higher eutrophy than deep lakes. He defined the boundary between eutrophy and oligotrophy at z ¼ 18 m for German lakes. Kalff (2002) named z probably the most useful single morphometric feature available. The shallowness of a lake can be characterized by its relative depth zrel (%), which is the ratio between zmax and the diameter of a circle with area A0:

ð1Þ

characterize the land–water and littoral–pelagial interactions of a lake. DL relates L0 to the circumference of a circle with an area identical to the lake’s surface A0 (Figure 5). Thus, the minimum of DL is 1, and the more the lake’s surface differs from a circular shape, the higher is the DL, indicating stronger linkage of the lake to the drainage basin and more extended shallow littoral zones. Many reservoirs exhibit high DL due to their dendritic surface shape. The depth distribution of a lake is described by the hypsographic curve. The areas Az are dependent on depths z. Az is calculated from determinations of the areas enclosed by k contour lines in different depths, from surface down to the maximum depth zmax, drawn on a bathymetric map. The volume Vi  Viþ1 of the layer between neighboring contour lines at depths zi and ziþ1 can be estimated as follows:

Vi  Viþ1 E 13ðAi þ Aiþ1 þ

pffiffiffiffiffiffiffiffiffiffiffiffiffi Ai Aiþ1 Þðziþ1  zi Þ

zrel ¼ 50zmax

V0 A0

ð4Þ

Characteristic values of zrel lie between 1% (large and shallow lakes) and about 4% (deep lakes). Calderas, maars, fjords, or solution basins may have zrel410%. The record zrel ¼ 374% is held by the Hawaiian volcanic crater lake, Kauhako (A0 ¼ 0.35 ha, zmax ¼ 250 m) (Cole, 1994). The extension of the littoral zone, the potential development of submerged macrophytes, the near-shore sediment transport and quality (water and organic content, particle size; e.g., Ha˚kanson and Boulion, 2002), and the colonization of littoral sediments with benthic organisms, depend on the slope s (%) of the shore. The slope si between two contour lines at depths zi and ziþ1 is calculated as follows (lengths and depths in m, areas in m2; see Figure 7)

ð2Þ

with 0rirk – 1, zk ¼ zmax, Ak ¼ 0, and Vk ¼ 0. The sum of the volumes of the layers below zi is the volume Vi and consequently, the sum of the volumes of all layers is the total volume V0 of the lake. Finally, the volume–depth development Vz for 0rzrzmax can be constructed. It is clear that the accuracy of the curves Az and Vz increases with the increasing number k of contour lines. Maximum depth zmax and average depth z with

z ¼

rffiffiffiffiffi p A0

si ¼ 100

ðLi þ Liþ1 Þðziþ1  zi Þ 2ðAi  Aiþ1 Þ

ð5Þ

The average basin slope s is

s ¼ 50

ð3Þ

k1 zmax X ðLi þ Liþ1 Þ nA0 i¼0

ð6Þ

Lakes of identical A0 and zmax may have different volumes due to different volume–depth distributions. In order to classify lake types, the index DV called ‘volume development’ was defined. DV is the ratio between the lake’s real volume

are very important parameters influencing the vertical distribution and zonation of, for example, temperature, underwater

d

L1

di

d

Am

Am

L2

L

(L1 + L2)/2 d ≈

2A m

=

L1 + L2

2(A1 – A2) L 1 + L2

di tan() =

Z2 – Z1 di

tan() =

Z2 – Z1

 di

Z2 – Z1 d

=

(L1 + L2) (Z2 – Z1) 2(A1 – A2 )

Figure 7 Schematic illustration on how to determine the slope between two contour lines L1 and L2 at depths z1 and z2 (L1, L2, z1, and z2 in m, areas A1 and A2 enclosed by L1 and L2 given in m2, Am ¼ A1  A2, di – distance between L1 and L2 at a certain position, d – average distance).

Lakes and Reservoirs V0 ¼ A0 z (Equation (3)) and the volume 13A0 zmax of an inverted cone with base area equal to the lake’s surface A0 and height coincident with the lake’s zmax:

DV ¼

z 3A0 z ¼3 A0 zmax zmax

0

n ¼ 2DV  3 ðJunge s shape indexÞ

consequences:

•

ð7Þ

A total of 202 out of 243 lakes evaluated by Carpenter (1983) came under the range 1 (V-shaped or cone) rDVr2 (U-shaped or ellipsoid). The U-shaped basins of many old natural lakes are the result of lake aging, that is, the deposition and focusing of large quantities of allochthonous and autochthonous sediments in the deepest parts, over a long time. Relatively young reservoirs, however, often have V-shaped basins. Morphometric models were developed that describe the geometric shape of lakes as quadric surfaces and sinusoids based on DV. Junge (1966) introduced the transformation

•

•

ð8Þ

and derived the following formulas for the relative area Ax and volume Vx using the normalized depth x ¼ z=zmax (z downward positive):

Az ¼ 1  nx2  ð1  nÞx A0

ð9Þ

Vz 6x  3ð1  nÞx 2  2nx 3 ¼1 V0 3þn

ð10Þ

Ax ¼

Vx ¼

•

Junge (1966) found satisfactory agreement between measured and calculated volume–depth distributions for most lakes and ponds considered in the survey. Significant deviations are characteristic of lakes with singular deep pits (DVo1) or large flat-bottom areas (DV42). Based on Junge’s model, lake types can be classified by principal geometric characteristics (Table 1; Junge, 1966). In order to elucidate the most important differences between the basic geometric lake types, circular basins with identical A0 and zmax are assumed. The areas and volumes of the upper (epilimnetic) water layers decrease much faster in the cone than in the ellipsoid (Figure 8; Junge, 1966). This has many very important Table 1

163

•

•

The epilimnion of V-shaped basins is shallower. Thus, the impact of sediment-related processes, such as the extension of the littoral area and its colonization with submerged macrophytes, as well as its role as a habitat for fishes, the influence of benthic organisms, sediment resuspension, and nutrient remobilization at the sediment–water interface on the epilimnetic matter turnover, is potentially highest in V-shaped water bodies. The area of the threshold between epilimnion and hypolimnion is smallest in V-shaped and largest in U-shaped basins. Therefore, the probability that particles and algae settle on the epilimnetic sediment area is highest in Vshaped lakes. The dilution of nutrients released from epilimnetic sediments is greatest in those lakes. Both features were found to influence the primary productivity of lakes (Fee, 1979). The ratio rV between the epilimnion and hypolimnion volumes is highest and increases faster in the cone and thus, the hypolimnetic oxygen balance is more critical in Vshaped basins. Thienemann (1927) had already postulated that lakes with rV 4 1 tend toward a eutrophic state while those with rVo1 are more oligotrophic. Sedimentation from epilimnion into hypolimnion is an important loss factor for phytoplankton, primarily for fast-settling diatoms. The probability of algae settling into the hypolimnion is much higher in U-shaped than in V-shaped lakes. The average epilimnetic light intensity is higher in V-shaped systems, due to the lower volume in the deeper zones compared with U-shaped ones. Light limitation of phytoplankton growth may be more significant in U-shaped water bodies. The transport of sediments into deeper regions of the lake, the so-called sediment focusing, depends on the average basin slope (Blais and Kalff, 1995), which is highest in Ushaped lakes. The total volume V0 of the ellipsoid is twice the volume of the cone and 4/3V0 of the paraboloid (Table 1). Consequently, the cone has a much lower heat capacity. Furthermore, its thermal stability is much lower due to the low depth of the gravity center, if identical vertical temperature distributions are assumed. Both facts may influence the timing of the periods of full turnover and stratification and the beginning of ice covering.

Parameters describing the shape of basic geometric lake types Basic geometric lake type

Parameter

Symbol

Cone

Paraboloid

Ellipsoid

Volume development Junge’s shape index Normalized area Normalized volume Depth of gravity center of the completely mixed lake Total volume (m3)

DV n Ax Vx xgc V0

1 1 (1–x)2 (1–x)3 1/4 A0zmax/3

3/2 0 1–x (1–x)2 1/3 A0zmax/2

2 1 1–x2 1–x(3–x2)/2 3/8 2A0zmax/3

From the morphometry model of Junge CO (1966) Depth distributions for quadric surfaces and other configurations. In: Hrbacek J (ed.) Hydrobiological Studies, Academia Publishing House of the Czechoslovak Academy of Sciences, pp. 257–265. Prague.

164

Lakes and Reservoirs

Cone (V-shaped)

Paraboloid

Ellipsoid (U-shaped)

Ax 0.2

0.4

0.6

0.8

1

0

0

0

0.2

0.2

0.4 0.6 0.8 1

Normalized depth x

Normalized depth x

0

Vx 0.2

0.4

0.6

0.8

1

0.4 0.6 0.8 1

Figure 8 (Top) Bathymetric maps of idealized (identical circular surface and maximum depth) lake types based on the morphometry model. Contour lines are drawn for the normalized depths x ¼ 0, 0.2, 0.4, y, 1. Attention should be paid to the different portions of the gray (epilimnion) and white (hypolimnion) areas (assumed a mixing depth of xmix ¼ 0.4). (Bottom) Normalized hypsographic curves Ax ¼ f(x) (left) and volume–depth distributions cone, paraboloid, and Ellipsoid). The dashed lines mark the assumed mixing depth of Vx ¼ f(x) (right) for idealized lake types ( xmix ¼ 0.4. Bathymetric maps based on the morphometry model of Junge CO (1966) Depth distributions for quadric surfaces and other configurations. In: Hrbacek J (ed.) Hydrobiological Studies, Academia Publishing House of the Czechoslovak Academy of Sciences, pp. 257–265. Prague.

•

V-shaped lakes have a shorter theoretical residence time t ¼ V0 =Qa (a), where Qa (m3 a1) is the mean annual discharge. The dilution of inflowing water-carrying nutrients, suspended matter, and other substances is low in V-shaped lakes, but much higher in U-shaped ones. Hence, the resistance of U-shaped basins against changing external loading is greater. The delay of an aggravation of the trophic state, in the case of increasing nutrient imports, is longer in those lakes. V-shaped water bodies may respond faster to reduced external loading.

It can be concluded that not only the size, but also the shape of lakes considerably influences their physical, chemical, and biological structure and functioning.

2.08.1.4 Influx and Vertical Distribution of Solar Energy Solar radiation is the Earth’s most important natural energy source and has a prominent ecological role. The global solar irradiance IG (W m2) is the solar radiation measured on a horizontal plane at the Earth’s surface and spans wavelengths l between about 200 and 3000 nm. The spectrum is divided into the ranges of ultraviolet (UV) radiation (lo380 nm; may be harmful to organisms), visible light (380 nmrlr750 nm), and infrared radiation (l4750 nm; thermal radiation). IG is

the sum of direct sun radiation and diffuse sky radiation (measured at full cloud cover, radiation reflected from clouds, water and dust particles, and other aerosols suspended in the atmosphere). It depends on latitude (Figure 9; Stras ’kraba, 1980), altitude (thickness of the atmosphere, higher IG at higher altitudes), and penetrability of the atmosphere (higher IG in dry regions compared with wet regions). If measured values of the local global radiation are not available, they can be approximately calculated from daily integrals of the radiation reaching the Earth’s surface on totally cloudless days, observations of sunshine duration and day length (Strasˇkraba, 1980).

2.08.1.4.1 Underwater light conditions The ratio of the irradiance reflected at surfaces to the incident flux is called ‘albedo’ r (in parts of 1). The reflection at water surfaces is mostly mirror like (specular reflection), with respect to the surface normal (angle of incidence a equals angle of reflection). Albedo r of direct sun radiation is a function of a and, therefore, the daily average varies geographically. However, r decreases quickly with decreasing a (ro0.13 if ao701, and ro0.03 if ao451). The reflection of diffuse sky radiation is lower at high a and higher at low a, than that of direct sun radiation. Surface waves reduce r at a 4701. For Central

Lakes and Reservoirs 3000

the changing spectral composition of light, with increasing water depth (Figure 10; Vollenweider, 1961; Uhlmann and Horn, 2001), due to the wavelength-specific transmission

2500 2000

3.4° 10°

1500

20°

Global radiation (J cm–2 d–1)

30° 1000 40°

70°

500

50°

Northern Hemisphere

60°

3000 Southern Hemisphere 2500 0° 2000

10° 20°

1500

30° 1000 40° 500

50°

70°

Tlz ¼ 100

J

F

M A M

J

Ilz ð%Þ Il0

Ilz ¼ Il0 expðkl zÞ J

A

S

O N

ð13Þ

D

Figure 9 Annual variations of daily integrals of the radiation reaching the Earth’s surface on totally cloudless days calculated for selected latitudes (atmospheric transmission factor 0.6). Modified from Strasˇkraba M (1980) The effects of physical variables on freshwater production: Analysis based on models. In: le Cren ED and McConnell RH (eds.) The Functioning of Freshwater Ecosystems, IBP 22, pp. 13–84. Cambridge University Press.

Europe, a daily average of rE0.1 can be assumed for water surfaces. Fresh snow reflects about 80–90%, old snow about 40–70%, and ice c. 25–35% of the irradiation. Radiation entering the water surface changes its direction due to the higher density and lower velocity of propagation in water than in air. This phenomenon is called refraction. The angle of refraction b is lower than a. The opposite applies to radiation, which is scattered from particles in the water back into the air. The radiation is angled away from the surface normal. If the angle of incidence of the backscattered radiation is greater than 491, it is completely reflected at the water– air interface. The photosynthetically active radiation (PAR) spanning 400–700 nm, within the range of visible light, is potentially a growth-limiting factor with regard to plants. Considering the intensity I0þ of PAR above the water surface, the approximation I0þE0.46IG is widely accepted and with an average daily albedo r ¼ 0.1, the intensity I0– of PAR just below the water surface can be expressed as

I0 ¼ 0:414IG

ð12Þ

with the wavelength l and the light intensities Il0– and Ilz just below the surface and at depth z, respectively. While the transmission of pure water is high in the blue–green part of the spectrum (l ¼ 450–500 nm), the range of the most penetrating light component shifts toward green–yellow (l ¼ 530– 580 nm) or even yellow–red (l ¼ 580–650 nm), depending on the concentration of suspended particles and dissolved substances (e.g., humic acids). Thus, the visual images of lakes change correspondingly: clear lakes appear blue–green, and eutrophicated lakes or those influenced by colored dissolved organic substances look greenish or brownish, due to the predominant backscattering of the respective range of the light spectrum. Although some phytoplankton species were found to react specifically to changing light quality (chromatic adaptation), the decrease of absolute light intensity with increasing water depth is far more important for photosynthesis. The light attenuation is described by the Lambert–Beer’s law:

60° 0

165

ð11Þ

Divers realize that colored objects become much paler in greater water depths. This phenomenon is the consequence of

The attenuation coefficient kl (m1) is a measure for the combined effect of absorption (i.e., transformation of radiation energy into heat or biochemical energy) and scattering (i.e., change of propagation direction caused by particles or water-density inhomogeneities) on the intensity of the light of the wavelength l. Although Equation (13) is, strictly speaking, only valid for parallel monochromatic light beams, it can also be applied to relatively narrow spectral bands such as PAR. The average PAR attenuation coefficient kPAR is derived from the spectral attenuation coefficients kl

1 kPAR ¼ 300

7Z00

kl dlE 13ðk450 þ k550 þ k650 Þ

ð14Þ

400

or, nowadays, directly calculated from underwater light measurements using spherical quantum sensors whose spectral response is adapted to the PAR range. The value kPAR is the sum of kW (pure water), kS (dissolved or colloidal matter), kP (phytoplankton), and kD (nonliving particles)

kPAR ¼ kW þ kS þ kP þ kD ¼ kW þ eS CS þ eP CP þ eD CD

ð15Þ

where the ei (l m1 mg1) are the substance-specific attenuation coefficients, and the Ci (mg l1) the substance concentrations, respectively. The mean extinction of the light flux directed downward is a suitable index for evaluating its spectral distribution (Vollenweider, 1961; see Figure 10). The property of substances to preferably absorb and reflect light in specific wavelength ranges is utilized in the remote sensing of waterquality criteria (e.g., the distribution of chlorophyll and water temperature) by air or satellite-borne reflectance measurements.

166

Lakes and Reservoirs 100 Max

Min

Mean

f(kPAR)

f(zSD)

Std

3

D1 80 Transmission (%)

Attenuation coefficient k, (m–1)

4

2

1

D10

60

S1 D100

40 S2 20

0 350 400 450 500 550 600 650 700 750 Wavelength  (nm)

0 400

S5

S3

S10 450

500 550 600 Wavelength  (nm)

650

700

Figure 10 (a) Mean (7standard deviation) and range of variation (min, max) of spectral attenuation coefficients (measured bi-weekly during the icefree seasons from 1975 to 1985) as well as approximations derived from the average attenuation coefficient of PAR (f (kPAR)) and the average Secchi disk transparency (f (zSD)) of the Saidenbach Reservoir (Germany) compared with the spectral standard distribution (Std); (b) average spectral underwater light transmission in the Saidenbach Reservoir (S1: 1m, S2: 2m, y, S10: 10m) and of distilled water (thickness of water layer D1: 1m, D10: 10m, D100: 100m). (a) Std from Vollenweider RA (1961) Photometric studies in inland waters: I. Relations existing in the spectral extinction of light in water. Memorie dell’Istituto Italiano di Idrobiologia 13: 87–113. (b) Curves D1, D10, and D100 from Uhlmann D and Horn W (2001) Hydrobiologie der Binnengewa¨sser. Ein Grundriss fu¨r Ingenieure und Naturwissenschaftler. Stuttgart: Eugen Ulmer; and data of the Saidenbach Reservoir from Paul L (1989) Interrelationships between optical parameters. Acta Hydrophysica, Berlin 33: 41–63.

The transparency or Secchi depth zSD provides a clear impression of the optical properties of standing waters. It is easily measured using a Secchi disk, a white disk, usually 25 cm in diameter, which is lowered on the shady side of a boat down to the depth of its visual disappearance. Relationships such as kPAR zSD ¼ a have been frequently published, since the early work of Poole and Atkins (1929). The value of a must be considered as water-body specific and, is consequently found to be widely scattering between about 1.1 and 4.6. This is quite understandable, because of the fact that the Secchi disk visually disappears, not only because of light attenuation but primarily because the contrast between disk and background becomes imperceptible. This is strongly influenced by both the concentration and size of the suspended particles in the water column above the disk. Thus, the percentage of transmission at zSD must be higher in a turbid lake. Nevertheless, the Secchi depth can be used as a predictor for spectral underwater light distribution. This was shown by Paul (1989), who found highly significant correlations of the type

ki ¼ ai þ

bi zSD

ð16Þ

where i stands for both l and PAR, respectively (Figure 10). The value ai can be considered as a first-order approximation for the lake-specific attenuation coefficients of water without suspended particles. The light requirements of phototrophic organisms are quite different and, moreover, depend on their physiological state. Water plants adapted to low-light intensities are very sensitive to small enhancements of illumination or, on the other hand, their photosynthetic response is inhibited if the light intensity increases sharply. Thus, it is impossible to define a general minimum-compensation light intensity Icomp (W m2) that is necessary to compensate for the oxygen consumption at night (respiration), with the respective photosynthetic oxygen

production in the daytime. Therefore, the underwater light situation is characterized by the depth zeu of the euphotic zone, which is the water layer expanding from surface down to the depth, where the PAR intensity Izeu is 1% of I0:

zeu ¼

4:6 kPAR

ð17Þ

In the euphotic zone (phototrophic layer), the light intensity is considered to be sufficiently high to allow photosynthesis. In the layers below zeu (tropholytic zone), respiration exceeds production and the phytoplankton development is light limited. It has to be borne in mind that zeu is only an approximate guiding principle. It is related to full daylight and sufficient day length. The light intensity Izeu in medium or high latitudes is much lower in winter than in summer and thus, zeu may substantially overestimate the depth of the compensation point in winter, and vice versa in summer. Phytoplankton cells are vertically transported by winddriven and/or convective currents throughout the mixed layer bounded by the mixing depth zmix and are exposed to an average light intensity

Izmix ¼ I0 zmix ¼

zZ mix

expðkPAR zÞdz 0

I0 ð1  expðkPAR zmix ÞÞ kPAR zmix

ð18Þ

or, if kPARzmix 4 3 and Equations (11) and (17) are considered

Izmix E

I0 zeu I0 zeu ¼ E 0:09IG kPAR zmix 4:6zmix zmix

ð19Þ

Hence, if the ratio between mixing depth zmix and euphotic depth zeu exceeds a critical value, phytoplankton development

Lakes and Reservoirs

• • • • •

•

the absorption of short-wave solar radiation, the net exchange of long-wave radiation between lake surface and atmosphere, the conductive exchange of heat at the water surface depending on the temperature difference between water and air, the evaporative heat loss, the heat import and export by inflow and outflow (only important in lakes or reservoirs with short retention times or in lakes significantly fed by the meltwater of glaciers), and the heat exchange at the lake bottom (only important in instances of geothermal activities).

Heat is vertically transported by convective and advective currents and turbulent eddy-diffusion. Currents generated by wind have a particular impact, and heat conduction is of minor importance. Therefore, the thermal structure of a lake greatly depends on the size and shape of the surface and its exposure to wind. The amount of heat stored by a lake, the heat content y (kJ), is calculated as follows:

y ¼ cp r

zZ max

 Az Wz dzE 4186 * V0 W

ð20Þ

0

with area Az (m2) and temperature Wz (1C) at depth z (m), maximum depth zmax, specific heat of water cp ¼ 1 kcal kg1 K1 ¼ 4.1855 kJ kg1 K1, density of water r (kg m3), lake  (1C). The annual volume V0 (m3), and average temperature W heat budget, the quantity of heat gained during warming and released during autumnal cooling, is the difference between the annual maximum and minimum heat content. HalbfaX (1921) described the immense magnitude of a lake’s annual heat turnover, comparing it with the length of a wagon train loaded with an equivalent amount of coal for heating. For the relatively small Klingenberg reservoir (near Dresden, Germany; V0 E 16 * 106 m3 and zmaxE33 m), he calculated the annual heat budget equivalent to the combustion heat of a coal train, 17 km in length. Thus, large lakes may significantly influence local climate. The vertical temperature distribution represents density stratification. Generally, the water density r depends on salinity s (%), pressure P (bar), and temperature W. A set of formulas to precisely calculate r(W,s,P), was provided by Chen and Millero (1986). However, the influence of salinity (normally so1 %) and pressure P with PE0.1z is negligible in freshwater and r is primarily determined by W (Figure 11, see also Figure 12; Chen and Millero, 1986). The unique quality of water, the decrease of density at temperatures o4 1C (anomaly of water), and the fact that the density of ice only is

( )

1000

0.30

999

0.25

998

0.20

997

0.15

996

0.10 |( )–( +1)|

995

0.05

994

Density diff. |(  )–( +1)| (kg m–3)

2.08.1.4.2 Heat budget and thermal structure The heat budget of a lake is determined by seasonal variations of

0.35

1001

Density ( ) (kg m–3)

is severely restricted, due to a very low average light intensity in the mixed water column. This principle is utilized in the artificial destratification of lakes, to control mass development of nuisance algae (see Section 2.08.3.3).

167

0.00 0

5

10 15 20 25 Temperature  (°C)

30

35

Figure 11 Density of water r(W) and density difference Dr ¼ |r(W)  r(W þ 1)| vs. temperature W at sea level and normal air pressure.

about 92% of the density of water at 0 1C, ensures the survival of aquatic organisms even in cold winters. Ice develops and floats at the lake surface and the water temperature in deep layers is not much lower than 4 1C. Therefore, aquatic organisms need to resist a much smaller annual temperature variation than terrestrial organisms. The density difference Dr increases considerably with increasing temperature. For instance, Dr between water temperatures of 19 1C and 4 1C is almost identical with Dr between 28 1C and 22 1C. Thus, relatively small temperature differences in tropical lakes may represent stronger density gradients than in temperate lakes. At the beginning of the warmer seasons, the heat gain in surface water is higher than wind-driven currents that can distribute vertically, and a temperature stratification is established in deep or even shallower, but wind-sheltered lakes, that usually consists of three characteristic layers (Figure 12):

• • •

Epilimnion – the upper warm, less dense, and turbulently mixed layer of almost homogenous temperature and density. Metalimnion – the intermediate stratum with strongly decreasing temperature and increasing density. Hypolimnion – the cold, more dense, and relatively quiescent bottom layer with low temperature and density gradients.

The static stability of temperature stratification can be characterized by density gradients, for example, by the buoyancy or Brunt–Va¨isa¨la¨ frequency N (s1) with

N2 ¼

g dr 2 * 9:81 * ðrzþDz  rz Þ E r dz rzþDz þ rz

ð21Þ

(g (m s2) the local acceleration due to gravity and Dz ¼ 1 m), or by the relative thermal resistance to mixing R (Wetzel and Likens, 1991) with

R¼

rzþDz  rz 10 3 E ðrzþDz  rz Þ rð41 CÞ  rð51 CÞ 8

ð22Þ

168

Lakes and Reservoirs

Temperature  (°C) 5

0

10

Density  (kg m–3)

15

20

998.5

0

Depth z (m)

5

999

999.5

1000

1000.5

0.3

0.4

 ( ,s,P ) Epilimnion

10

Metalimnion

15

Hypolimnion

20



()

25

20

30

Δ()

35 40 0.2 (a)

0.205

0.21

0.215

0.22

Conductivity 20 (mS cm–1)

0 (b)

0.1

0.2

Density diff. Δ (kg

m–3)

Figure 12 Characteristic vertical temperature and density stratification for lakes of the transient zone (about 401–601 N or S) in summer. (a) Vertical profiles of water temperature W and conductivity k20 (mS cm2, related to W ¼ 20 1C) measured in the Saidenbach Reservoir on 10 July 2007. (b) Respective distributions of water density r(W) as a function of W alone, density r(W,s,P) depending on W, salinity sE0.5k20, and pressure PE0.1z, and density difference Dr(W) ¼ r(Wzþ1) – r(Wz). The maximum deviation of r(W) from r(W,s,P) is less than 0.04%, which clearly shows the primary impact of temperature on density in freshwaters. (b) Curve r(W,s,P) from Chen C-TA and Millero FJ (1986) Precise thermodynamic properties of natural waters covering only the limnological range. Limnology and Oceanography 31: 657–662.

Schmidt (1915) defined thermal stability S0 (Nm m2) as the required energy per square meter of the lake surface, to completely mix the stratified water body, without change in its heat content

S0 ¼

¼

g A0 g A0

zZ max

Az ðzfc  zÞðrfc  rz Þdz

0 zZ max 0

Az rz ðz  zfc Þdz ¼

gM ðzst  zfc Þ A0

ð23Þ

with area Az and density rz at depth z, acceleration due to gravity g ¼ 9.81 m s2, surface area A0, depth zfc of the gravity center, and density rfc during the full circulation period. Thus, S0A0 corresponds to the work (to be accomplished by the wind) that is required to lift the total mass M of the lake by the distance zst  zfc, which is the difference between the depths zst and zfc of the gravity center of the stratified, and the completely circulating lake. In the case of a stable stratification (lighter, less dense above heavier water layers with higher density), it is zst4zfc and S040. The stability characterizes the degree of separation of the hypolimnetic water layers from the epilimnetic ones. Deep lakes have a higher S0 than shallow water bodies. The difference zst  zfc usually amounts to only a few millimeters. However, the energy needed for mixing a stably stratified lake is huge, due to the enormous mass to be raised. For instance, the mass of the relatively small Klingenberg reservoir, mentioned above, is about 1600 times the mass of the Eiffel tower in Paris. Thermal stability is an important parameter that has to be considered in planning artificial lake destratification, as a measure to prevent hypolimnetic oxygen depletion and/or the mass development of noxious

phytoplankton by light limitation, resulting from a too high ratio of zmix and zeu. The metalimnion is usually defined as the stratum where the temperature gradient exceeds a certain limit (e.g., 1 K m1). However, considering lakes at different latitudes with quite different temperature ranges, the upper and the lower threshold of the metalimnion should be related to density gradients. In temperate zones, the beginning of a stable summer stratification is often observed when the surface temperature exceeds 10 1C. Accordingly, the depth of the 10 1C-isotherm is a good predictor for the threshold between meta- and hypolimnion. Consequently, the metalimnion could more generally be defined as the layer with density gradients greater than 0.08 kg m4 (see Figure 12). This limit corresponds approximately to the water-density difference between 9 1C and 10 1C and, thus, to N 2 E 0:0008 s 2 (Equation (21)) or to RE 10 (Equation (22)). The level of the maximum density gradient is called ‘thermocline’. Talling (1971) determined the mixing depth zmix of the upper mixed layer as the depth with a temperature 0.5 K below the temperature at a depth of 2 m. In this manner, superficial thermal gradients, which may develop during transitional calm weather periods, are largely excluded. Referring to this principle, but transferred to density gradients, zmix can be defined as the depth with a density 0.08 kg m3 higher than that at the depth of 2 m. Thus, for a given temperature W247.2 1C in z ¼ 2 m, the temperature Wzmix (1C) at the depth zmix amounts to

Wzmix ¼ W2  0:28  3031200 expð2W2 Þ  5:2 expð0:2W2 Þ

ð24Þ

Latitude and altitude determine a lake’s seasonal temperature range and mixing scheme, depending on the regional

Lakes and Reservoirs

•

A sequence of typical vertical temperature profiles of a dimictic water body (Lake Stechlin, Germany) is shown in Figure 15.

) co ve r en

ti

ce

POLYMICTIC

1000 0 90

80

70

ic ictic

om

T

Dim

ic

2000

ict

tic

(p

er

m

an

3000

Warm monomictic

•

4000

Am

•

5000

on

•

6000

m

Oligomictic lakes. They are mostly very deep tropical lakes, with high heat capacities, that are rarely and irregularly mixed (usually under extreme weather situations, e.g., tropical storms). Polymictic lakes. They refer to shallower lakes with low vertical-density gradients that mix frequently, sometimes daily. Monomictic lakes. They are lakes with one mixing period, either in winter at water temperatures Z4 1C (warm monomictic subtropical lakes, or large and deep lakes, in the temperate zone with high heat capacity, that do not freeze), or in summer at water temperatures r4 1C (cold monomictic lakes at high latitudes, where ice cover melts only in summer). Dimictic lakes. They are sufficiently deep lakes of the temperate zone that circulate in spring and autumn and are ice covered in winter. Meromictic lakes. They refer to partially mixed lakes with a deep-water layer enriched by dissolved salts, or are sufficiently wind sheltered, small but deep lakes.

ld

•

Co

1. Amictic lakes. They are permanently frozen lakes at high latitudes and/or altitudes that never overturn; and 2. Holomictic lakes. They refer to lakes that mix at least once per year, further specified as:

An inverse temperature distribution (colder above warmer water) is observed during the winter stagnation when the lake is ice covered (e.g., 15 February 2006 in Figure 15). Vertical mixing is strongly reduced due to the cutoff of wind action by the ice. Some convective mixing just below the ice is possible on sunny days, if the ice is clear and irradiance heats the uppermost water layers. After the disappearance of the ice cover, complete mixing of the entire water body is likely. As long as the water temperature is lower than 4 1C, warming at the surface provokes an increase in the density and mixing is induced, even in dead calm. After the water has reached the temperature of maximum density (profile from 7 April 2006 in Figure 15), further heating produces less-dense water at the surface and mixing requires sufficient wind energy. Thus, mixing becomes more and more episodic and depends on the actual weather

Altitude (m)

air temperature range (Figure 13; Stras ’kraba 1980). For latitudes up to about 401 N or S, the bottom temperature WB of deep lakes corresponds approximately to the minimum annual water temperature and decreases from very high values in the tropics, to 4 1C. At higher latitudes, WB remains constant at the temperature of the density maximum. The resulting mixing type depends on the absolute temperature range and the seasonal surface temperature variation. Increasing distance from the equator and increasing altitude have the same effect (Figure 14; Hutchinson and Lo‘ ffler, 1956). The following thermal lake types are distinguished depending on the principal mixing behavior:

169

T

OLIGOMICTIC

T

60 50 40 30 Latitude (°N or °S)

20

10

0

Figure 14 Scheme of the distribution of thermal lake types depending on latitude and altitude. Modified from Hutchinson GE and Lo¨ffler H (1956) The thermal classification of lakes. Proceedings of the National Academy of Sciences of the United States of America 42: 84–86. T, transitional regions.

Water temperature (°C)

30

20

5° N 35° N 50° N 72° N

10

0 J

F

M

A

M

J

J

A

S

O

N

D

Figure 13 Trends of seasonal variations of surface (open symbols) and corresponding bottom temperatures (filled symbols) of medium-sized lakes at low elevations calculated for selected northern latitudes from empirical equations provided by Strasˇkraba M (1980) The effects of physical variables on freshwater production: Analysis based on models. In: le Cren ED and McConnell RH (eds.) The Functioning of Freshwater Ecosystems, IBP 22, pp. 13–84. Cambridge University Press.

170

Lakes and Reservoirs Temperature (°C) 0

5

10

15

20

25

0 5

Depth z (m)

10 15 20 25

15.02.06 07.04.06 11.05.06

30

01.08.06 15.11.06

35

19.12.06

40 Figure 15 Sequence of temperature profiles characterizing the seasonal change between mixing and stratification typical for a dimictic lake (upper 40 m of Lake Stechlin, 131 020 E, 531 090 N, Germany; maximum depth of the lake is 69 m). Courtesy of Dr. P. Kasprzak, IGB Berlin.

situation. Inconsistent and relatively cold weather (typical April-weather) may prolong the period of spring full circulation and foster the warming of the whole water column (increase in temperature of the deep-water layers to more than 5 1C). Conversely, warm and calm weather immediately after the temperature homogeneity at 4 1C, may quickly form density gradients at the surface, which even strong winds cannot equalize any further. Thus, the spring full circulation period is short and the deep-water layers remain relatively cold (as was observed on 11 May 2006, Figure 15). Once a stable thermocline is established and the summer stagnation has started, further increasing air and, consequently, surface water temperatures, strengthen the temperature and density differences (1 August 2006 in Figure 15) and, thus, the thermal, hydrodynamic, chemical, and biological decoupling between the illuminated, warm, flushed, wind-mixed epilimnion and the usually dark, cold, quiescent hypolimnion takes place. The thermal structure fundamentally influences the temporal development and spatial distribution of biological and chemical food-web components. The water column is subdivided into two reaction spaces with completely different physical, chemical, and biological properties. Therefore, Ruttner (1962) characterized thermics as the pivotal point of lake limnology. After midsummer, irradiation and air-temperature decline and successive cooling and mixing increase the depth of the epilimnion, and the metalimnion slowly propagates downward. The metalimnetic density gradients decrease, and heat and matter exchange by eddy-diffusion, between epi- and hypolimnion, increases. For instance, the maximum density gradients of the temperature profile from 15 August 2006 in Lake Stechlin (Figure 15) were much smaller than

0.08 kg m4 and thus, by definition, the stratification could no longer be considered as stable. Eventually, the stratification disappeared (19 December 2006) and the lake went into the phase of autumn full circulation. Further cooling favors convective overturn, until the water temperature reaches 4 1C. From then on, the ice cover on an entire lake may be established in a single, calm, and frosty night and winter stagnation will be initiated. If this happens early, the deep-water temperature remains relatively high (B4 1C) all through the winter. Paradoxically, the temperature of the water column may decrease much more in a mild winter with a late ice-up. Climate change is expected to significantly influence seasonal temperature development, the duration of the mixing and stagnation periods, the solubility of gases, and the exchange of heat and matter between water and sediment (Blenckner et al., 2002). Milder winters result in later freeze-up and earlier ice break-up and, in extreme cases, ice cover and winter stagnation do not even develop. Thus, formerly dimictic lakes may become monomictic. Recent model simulations predict opposite effects of climate change in some regions of the temperate latitudes, for example, in Northern Atlantic regions (Hansen et al., 2004). Decreasing temperatures are forecast, due to changes in the thermohaline circulation of the ocean. Therefore, climate change will influence the duration of summer stagnation, the epilimnetic temperatures and density gradients, and, thus, the hypolimnetic oxygen budget of stagnant water bodies. It will also likely affect the phytoplankton species composition, succession, and abundance.

2.08.1.5 Water Movement Unlike rivers, lakes are identified as stagnant or standing water bodies. However, natural waters are never completely quiescent. Horizontal and vertical water movements of quite different spatial and temporal scales transport dissolved and particulate materials and heat. They influence the gas exchange with the atmosphere and affect the basin morphology, due to erosion and deposition of sediments. Therefore, knowledge about the hydrodynamic structure is important for the understanding of the matter turnover of lakes. Water flows in lakes and reservoirs are largely turbulent, that is, chaotic, swirling, multidirectional, and disordered. Unidirectional and smooth laminar flows can only be observed at very low flow velocities, for example, in thin boundary layers between water and sediment in deep, stratified lakes or in the metalimnion of wind-sheltered, small basins during calm weather. Wind, solar radiation, and in- and outflows are the most important forces generating water movement. In large lakes, air-pressure differences along the surface, the Coriolis force, resulting from the Earth’s rotation, and the gravitational attraction of the sun and moon, may also cause or influence water movement. The spatial and temporal variations in wind force are of the greatest importance for the formation of nonperiodic currents. Wind acts at the water surface and, thus, the size and shape of the lake and its orientation to the prevailing wind direction are decisive factors. The wind exposure of a lake is described as

Lakes and Reservoirs

a wind fetch, defined as the unobstructed distance that wind can travel over water in a constant direction. The kinetic energy of the wind is proportional to u3, where u (m s1) is the wind velocity, normally measured 10 m above the surface. The velocity of wind-driven currents is about 0.02u and is independent of the height of surface waves. In the open water of large and deep lakes, the Coriolis force causes a deflection of the wind drift to the prevailing wind direction of about 451 to the right in the Northern hemisphere, and to the left in the Southern hemisphere, respectively. This deflection increases with water depth and therefore the currents in the deepest water layers may flow opposite to the wind direction. This phenomenon is called the ‘Ekman spiral’. In smaller lakes, the water feels the shore and the bottom, and boundary effects influence the flow-field. Currents parallel to the shores prevail. A downwind drift of water masses, unavoidably causes the leeward drift of a corresponding amount of water and, consequently, large-scale horizontal and, for example, in the mixed epilimnion, vertical circular motions are formed (Figure 16; Hutter K (1983)). Such gyres may produce inhomogeneous (patchy) distributions of chemical or biological constituents (e.g., patchiness of phytoplankton or waterquality parameters). The circulation patterns are strongly influenced by lake-basin irregularities (e.g., islands and bays). Attentive observers may occasionally notice streaks of foam or debris (windrows) at uniform distances from one another, at the surface of lakes on windy days, that are deflected 51–151 to the right of the wind direction (in the Northern Hemisphere). This appearance is an indication of the Langmuir circulation, a wind-driven helical circulation system, rotating clockwise and counterclockwise alternatively, that is initiated at wind speeds of more than about 3 m s1 (Figure 17). Air bubbles, produced by breaking waves and floating materials, flow from the upwelling range (divergence zone) to the range of the downwelling motion (convergence zone) and concentrate at the surface. The distances between the windrows increase with increasing wind speed. The diameter of the vortices is about half of the distance between the streaks, but never larger than the depth of the epilimnion. Langmuir cells may significantly affect the development of the phytoplankton, which are passively transported vertically, through the underwater light field, within short time intervals (Vincent, 1980). Patchiness of zooplankton may also be caused by Langmuir circulation (Malone and McQueen, 1983). Propagating surface-gravity waves imply a horizontal transport of water. However, wind waves only cause surface water particles to move in circular orbits, with almost no drift of water. Wind waves are characterized by their height H from trough to crest, length L from crest to crest, and period of oscillation (Figure 18). Wave formations and their dimensions, depend on wind speed, wind fetch F, wind duration, and water depth. In the open water, the maximum height Hmax of thepwaves can be estimated from fetch F as ffiffiffi Hmax ¼ 0:105 F. Hmax is identical to the diameter of the orbital motion at the surface. The diameters Dz of the orbits shrink with increasing depth z and no vertical displacement of water parcels, attributed to surface waves is found below zE0.5 L (Figure 18). Waves approaching the shore regions or in shallow lakes with zmaxo0.5 L feel the bottom, and the shape of the orbital motions close to the bottom becomes

171

Figure 16 Qualitative distribution of the mean steady state transport in Lake Zu¨rich for spatially uniform constant winds blowing from N, S, W, E, and SE. Modified from Hutter K (1983) Stro¨mungsdynamische Untersuchungen im Zu¨rich- und im Luganersee – Ein Vergleich von Feldmessungen mit Resultaten theoretischer Modelle. Schweizerische Zeitschrift fur Hydrologie 45: 101–144.

more and more elliptical. Lightweight particles are resuspended and washed downward into the deeper regions of the basin. With further decreasing water depth, the waves at the surface become higher and steeper, the wavelengths shorter, and their erosive impact increases. Finally, if zo0.05 L, the waves break and strong erosion of the shore may be observed. Thus, wind waves strongly affect the development of the shorelines and the littoral zones of lakes. Strong wind, persistently blowing from a constant direction, pushes the upper warm water masses of a stratified lake to the downwind side and generates a tilt of the whole surface and thus, an unstable position. Due to the restoring force of gravity, the water flows back and, due to inertia, a swinging, oscillating motion of the surface is caused, which is called a surface or ‘external seiche’. Periods of external seiches are

172

Lakes and Reservoirs

Wind

Top view

Windrow

Wind

Windrow

Surface Cross section Downwelling (fast)

Upwelling Downwelling (fast) (slow)

Figure 17 Schematic representation of Langmuir circulation cells. Air bubbles and debris are flowing from the divergence (upwelling) zone to the convergence (downwelling) zone and create streaks (windrows) of almost constant distance at the surface that are nearly parallel to the wind direction.

Period of oscillation Time 0

Depth

H/2

L/2 Figure 18 Circular motion of water particles in five layers from surface down to the depth L/2 at five moments during one period of oscillation of a wind wave. L, wave length in cm; H, wave height in cm.

rather short – seconds in small lakes and minutes or a few hours in large basins. The amplitudes vary between a few centimeters in small water bodies and about 2 m in large lakes. However, internal seiches, that is, the periodic up- and downwelling of water layers of different density and depth, forming standing waves of much larger amplitudes and longer periods of oscillation, are more important in terms of matter transport and impact on phytoplankton development. This phenomenon is not visible to observers at the surface, but becomes evident from considerable periodic temperature variations in the depths of the metalimnion and below. The example shown in Figure 19 indicates superimposed, onenodal, internal seiches of different periods of oscillation (about 24 h and 8 h) in a reservoir. The temperature variations at the West and East stations almost mirror each other, while they are comparably low at the central station, which is apparently close to the position of the wave node of the oscillation. Large horizontal, but low vertical water movements are observed in the nodal areas. The opposite applies to the crest regions, where up- and downwelling prevails. However, these vertical movements generate highly turbulent currents along the sloped bottom and create so-called internal surges, similar to breaking surface waves (Mortimer and Horn, 1982). The

temperature stratification at the sediment surface periodically varies from stable (warm water over colder sediment; situation 1 in Figure 19) to unstable (colder water over warmer sediment that fosters the release of interstitial water; situation 2 in Figure 19). The velocities of the vertical movements are usually in the range of several millimeters per second, while those of the horizontal current components may be up to a hundred times higher. The vertical displacements of the layers from their stable positions depend on the size of the lake basin, density gradient, depth of and vertical distance to the thermocline, and can be higher than 10 m. Sudden changes of wind direction or periodic (e.g., diurnal) fluctuations of wind speed may cause phase shifts of the oscillations in different depths, resulting in the interference of waves with several lakespecific periods of oscillation (Figure 20). Internal seiches may be observed almost permanently in stratified lakes during the summer stagnation (Figure 21). Even in calm weather, the oscillations continue for a long time (days or even weeks) with, however, decreasing amplitudes after their excitement has ceased. Wave structures rotating around large lake basins, may occur under the influence of the Coriolis force and wind-driven, horizontal large-scale circulations and, finally, highly complex current patterns result.

Lakes and Reservoirs West

Centre

Temperature (°C)

East

1

6.5

173

2

6.0

5.5

5.0

(a)

4.5 060513

060514

West

060515

060517 060516 Date (yymmdd)

Centre

1

060518

060519

East

2

(b)

Figure 19 (a) Periodic temperature variations measured in the Saidenbach reservoir (Germany) at three stations (West about 50 m in front of the dam, Center B1 km, and East about 2 km apart from the West station) in time intervals of 30 min in May 2006. (b) Diagram of the position of a water interface at the two moments marked in the graph above. Arrows qualitatively indicate prevailing water movements.

Internal seiches have a great impact on the turbulent vertical exchange of heat and the transport of materials. They resuspend sediment particles, accelerate their dislocation into the deepest regions of the lake, and enhance the release of dissolved substances from the sediment. The periodical transport of phytoplankton cells, throughout the vertical light field in the crest regions of the internal waves, substantially increases the photosynthesis rate in water layers at the base of the euphotic zone (Paul, 1987). Discharge-related currents, especially floods, may generate basin-wide water movements, particularly in lakes with short retention times. As the import of nutrients, allochthonous particulate matter and other substances by the tributaries, is most important to the materials budget, knowledge about the seasonal variability of the depth of inflow and the propagation of the inflowing water is crucial for the understanding of the trophic situation and the availability of nutrients in the euphotic zone. In reservoirs, the balance between inflow and outflow from different depths, decisively determines the development of the fill-level and the volume of the hypolimnion during summer stratification. The withdrawal of water from the deep-water layers of reservoirs causes currents in the hypolimnion, which is more or less quiescent in natural lakes. The entrainment depth to which the inflowing water plunges characteristically varies seasonally, depending on the

temperature (density) distribution in the lake and the temperature (density) of the tributaries (Figure 22; Carmack et al., 1979). Surface inflow is observed when the density of the river water is lower than that of the lake; underflow occurs in the reverse instance. Interflows are typical in situations in which the river density is between that of the lake’s surface and the bottom of the lake. Hydraulic short-circuiting, that is, the longitudinal distribution of river water from the mouth of the tributary to the dam in a relatively thin metalimnetic layer, within a very short time (a few hours), has frequently been observed in reservoirs. Such events are critical in the case of drinking-water reservoirs, because harmful substances (e.g., turbidity and microbial pollution) in high concentrations may contaminate the raw water (Clasen and Bernhardt, 1983). Intrusion far below the depth of the respective lake temperature can be observed, if the density of the inflowing water is considerably enhanced due to very high flood-induced turbidity, caused by suspended mineral particles. Such turbidity currents may import oxygen into the hypolimnion of seldom fully circulating (e.g., deep pre-alpine) lakes (Lambert et al., 1984; DeCesare et al., 2006). Turbidity currents are a special form of density currents. Density currents are, in general, water movement, caused by density differences. They can also result from water-temperature differences, as a consequence of differential heating or

174

Lakes and Reservoirs Time (h) 0

48

24

72

96

120

0

Depth (m)

15

20

3

2

1

3

2

Figure 20 Vertical and temporal temperature variability caused by superimposed internal waves of different periods of oscillation and phase shifts in the upper 20 m water layer of the West station c. 50 m in front of the dam (zmax ¼ 45 m) of the Saidenbach reservoir observed from 18 May 2005 0 a.m. to 22 May 2005 12 p.m. (top left). For the marked points in time, wave modes and principal water movements are schematically shown.

30 25 Temperature (°C)

W_1 W_3

20

W_5 W_7.5

15

W_10 W_15 W_20

10

W_30

5 0 060501

060531

060630

060731

060830

060930

061030

Date (yymmdd) Figure 21 Results of short-term temperature records (measuring interval 30 min) at different depths (m, indicated by the numbers given in the legend) at station West (B50 m in front of the dam) of the Saidenbach reservoir (Germany) during the summer stratification in 2006 (dates are given in the yymmdd-format). The permanent temperature fluctuations at depths below the epilimnion show the ubiquitary nature of internal seiches. Those at the surface may also result from short-term changes of irradiation and air temperature.

cooling in lake segments. For instance, shallow bays on the margins of lakes, may heat up during the day and cool down at night, more rapidly than the open water. The resulting density differences generate convective exchange of water and of dissolved materials between littoral and pelagic zones (Wells and Sherman, 2001). Strongly increased conductivity (salinity), for example, due to thaw salt from roads in winter, may also generate density currents and vertical temperature inversions. Convective currents are generally upwelling movements of less dense, lighter water (e.g., plumes of heated waste water), or downwelling movements of denser, heavier

water (resulting from surface cooling in summer or warming in spring when the deeper water layers have temperatures lower than 4 1C). As mentioned above, water movement in lakes is mostly turbulent. Turbulence results from friction between water layers moving with different velocities (Baumert et al., 2005). Shear forces produce vortices and, if they collapse, they dissipate the energy of motion and cause mixing of water. The spatial dimensions of these vortices, that is, the intensity of eddy-diffusion, decrease with increasing density gradient and/ or reduction of velocity differences between adjacent water

Lakes and Reservoirs

Summer

Winter

TS > TR

TR < TL

TR > TB (a)

TB > 4 °C

175

TL < 4 °C (e)

Early autumn

Early spring

T S > TR TR > TB

TL < 4 °C

TB > 4 °C (b)

TR ~ 4 °C (f)

Middle autumn

Middle spring

TL > 4 °C

TL < 4 °C

TR ~ 4 °C (c)

TR > 4 °C (g)

Late autumn

Late spring

TL > 4 °C TR < 4 °C

TR < TL TL > 4 ° C

(d)

(h)

Figure 22 Schematic representations of seasonal riverine circulation patterns of the Kamloops Lake, British Columbia. Stippled areas denote river water; dashed areas denote lake and river water mixtures involved in cabbeling process (mixing of water of identical density but slightly different temperature and salinity). From Carmack EC, Gray CBJ, Pharo CH, and Daley RJ (1979) Importance of lake–river interaction on seasonal patterns in the general circulation of Kamloops Lake, British Columbia. Limnology and Oceanography 24: 634–644. TR, river temperature; TL, lake water temperature; TS, lake surface temperature; and TB lake bottom temperature.

strata. Turbulence transports heat, dissolved substances, and gases vertically between the epilimnion and the hypolimnion. This can be observed by small-scale vertical temperature inversions in temporally and spatially highly resolved, thermal microstructure measurements (Wu¨est et al., 2000).

2.08.1.6 Basic Chemistry 2.08.1.6.1 Systematics of lakes with respect to water quality There are approximately 8 million natural lakes with surface areas of 41 ha, on the Earth (Ryanzhin et al., 2001). The majority of them are freshwater lakes, which are of vital importance for humankind, animals, and plants. A global model based on the Pareto distribution, shows that the global extent of natural lakes Z0.1 ha is about 304 million lakes (Downing et al., 2006).

Some of the different types of lakes classified on the basis of their water quality are as follows: 1. Soft- and hard-water lakes. As many lakes are connected with the groundwater, their water chemistry is influenced by the geological substrate of the watershed. Hard-water lakes dominate when the catchment is rich in calcium. Soft-water lakes are characterized by the low content of the hardness components, calcium and magnesium. Lakes with a calcium deficit are normally fed by rainwater without soil contact, or exist in Ca-deficient, sandy, outwash plains. In calcareous, oligotrophic (i.e., clear water) lakes that are mainly fed by groundwater, such as Lake Stechlin, Northern Germany, submerged plants use HCO 3 /CO2 as a C source for photosynthesis þ CO2  H2 O3 HCO 3 þH

ð25Þ

176

Lakes and Reservoirs 2 þ HCO 3 3 CO3 þ H

ð26Þ

Suspensions or deposits of hardly soluble CaCO3 (biogenic decalcification) may be formed by CO2 uptake and Hþ consumption (pH increase)

Ca 2þ þ CO2 3 3 CaCO3 ðsÞ

ð27Þ

Another example of hard-water lakes are the acidic hard-water lakes. Here, CaCO3 formation does not occur and sulfate is the dominating anion. These lakes are, in many cases, impacted or man made (mining lakes). Acid mine drainage (AMD), a result of the mining and milling of sulfur-bearing coal and ores, plays a dominant role in surface-water chemistry and pollution, in many areas of the world. Oxidation of disulfide minerals (e.g., pyrite or marcasite (FeS2)) occurs mostly from the reactions of tailing and mining wastes with oxygen and water in underground workings, tailings, open pits, and waste rock dumps. This produces acidic water, rich in metals, commonly referred to as AMD. The most noticeable environmental change is the pollution of flowing water with severe impacts on aquatic life. The following reactions result from acidified runoff on regulated mine sites. Representative species of bacteria engaged in these processes are also mentioned 1.1. Iron disulfide (pyrite and marcasite) is oxidized to sulfate by oxygen with a very high energy yield for bacteria (Thiobacillus ferrooxydans and Thiobacillus thiooxydans)

2FeS2 ðsÞ þ 7O2 þ 2H2 O þ ) 2Fe2þ þ 4SO2 4 þ 4H

ð28Þ

1.2. Ferrous iron is oxidized to ferric iron

14Fe 2þ þ 3:5O2 þ 14Hþ ) 14Fe3þ þ 7H2 O ðslow under acidic conditionsÞ ð29Þ 1.3. Sulfur/sulfide is oxidized with ferric ions to sulfate

FeS2 ðsÞ þ 14Fe3þ þ 8H2 O þ ) 15Fe2þ þ 2SO2 4 þ 16H

ð30Þ

1.4. Ferric iron is hydrolyzed to ferric hydroxide

Fe 3þ þ 3H2 O ) FeðOHÞ3 ðsÞ þ 3Hþ

ð31Þ

In the presence of nitrate, the FeS2 oxidation proceeds by autotrophic denitrification (Thiobacillus denitrificans) þ 5FeS2 ðsÞ þ 14NO 3 þ 4H

) 5Fe2þ þ 10SO2 4 þ 7N2 þ 2H2 O

ð32Þ

whereas Fe(II) is further oxidized to Fe(III) by species such as Gallionella ferruginea

10Fe 2þ þ 2N 3 þ 14H2 O ) 10FeOOHðsÞ þ N2 þ 18Hþ

ð33Þ

Reactions (28), (30), (31), and (33) lead to an enormous production of acid. Sulfur concentrations in many surface waters have increased greatly as a result of acid mine run off and SO2 emissions. 2. Saline lakes. In semiarid and arid climates, lakes may show a high concentration of dissolved solids, due to the surplus of evapotranspiration above the runoff into the lakes. In relation to their main ingredients, salt lakes may be subdivided into soda, chloride, and sulfate lakes. The majority of saline lakes, in terms of area, are chloride lakes. Historically, they are remnants of isolated seawater bodies in continental locations. 3. Soda lakes. Lakes with a very high alkalinity level, mainly due to soda (Na2CO3), occur in southeast Europe and are common in the East African rift valley. The water in soda lakes becomes alkaline with a pH of approximately 10, due to the alkalis, carbonate, and bicarbonate. The water tastes bitter and feels oily. These salts accumulate in lakes without discharge, if the subsoil consists of carbonate or volcanic rock, and whose water budget is characterized by high evaporation rates. Therefore, soda lakes are usually found in semi-deserts and steppe areas. Mono Lake (California) contains about 280 million tons of dissolved solids and, depending on its seasonally fluctuating water level, is 2–3 times more salty than the ocean. It is also rich in borate and potassium. Periodic eruptions of volcanic ash have also considerably contributed to Lake Mono’s chemical mix. Soda lakes are often rich in biomass, provided they are not too deep. Due to the high pH values and salt concentration, alkaliphilic/alkalitolerant and simultaneously, halophilic organisms, are characteristic. The limited biodiversity essentially comprises specialized bacteria (among others: cyanobacteria such as Spirulina and Archaea) and algae. They may appear in great abundance and reduce the Secchi disk transparency to a few centimeters. Soda lakes thus rank among the most productive ecosystems. Special protophytes (flagellates) are characterized by accessory-colored pigments (carotenoids, phycobiline, and rhodopsin). They are responsible for the conspicuous coloration of numerous soda lakes. Many sodium carbonate lakes are utilized for the production of natural soda. 4. Bog lakes. These lakes are normally poor in electrolytes. Bog lakes are found in all geographic latitudes of the humid climate zones, from the wetlands in the hills to the plains, to the marshes adjacent to large rivers. They are among the aquatic systems with high species diversity. In bogs and bog lakes, the production of organic C compounds is greater than microbial mineralization. The slow and incomplete decomposition of vegetation residues, under continuous water surplus from rainfall or soil water, is accompanied by a high oxygen deficit, resulting in peat deposition and siltation (see Section 2.08.1.6). Dystrophic lakes are poor in nutrients and calcium as well as phytoplankton, and are mostly strongly acidic and rich in dissolved humic materials. They are clear, but mostly brownish. Their watershed is often small; therefore, it is not remarkable that some species typical of bog lakes are also found in acidic mine lakes.

Lakes and Reservoirs

5. Crater lakes or volcanic lakes. Crater lakes covering active (fumarolic) volcanic vents are sometimes termed volcanic lakes – a cap of meteoric water over the vent of an active volcano. The chemistry of the water may be dominated by high-temperature volcanic gas components or by a lower temperature fluid that has interacted extensively with volcanic rock. Precipitation of minerals such as gypsum (CaSO4* H2O) and silica (SiO2) can determine the concentration of Ca and Si (Kusakabe, 1994). The water of these lakes may be extremely acidic (e.g., pH B 0.3). Lakes located in dormant or extinct volcanoes tend to contain freshwater, and the water clarity in such lakes may be exceptionally high due to the lack of inflowing streams and sediments. Crater lakes form as incoming precipitation fills the depression. The lake deepens until equilibrium is reached between water inflow, losses due to evaporation, subsurface drainage, and possibly also surface outflow, if the lake fills the crater up to the lowest point of its rim. Surface outflow can erode the deposits damming the lake, lowering its level. If the dam erodes rapidly, a breakout flood can be produced.

2.08.1.6.2 Ionic balance

balance of water

NaAlSi3 O8 ðsÞ þ5:5H2 O ) Naþ þ OH ðalbiteÞ

þ 2H4 SiO4 þ 12Al2 Si2 O5 ðOHÞ4 ðsÞ

Main inorganic compounds and buffering properties present in lakes and reservoirs are as follows: 1. Alkalines. The alkali ions Naþ and Kþ are mostly discharged in the K- and Na-feldspar weathering processes and represent an important part of the ion

NaAlSi3 O8 ðsÞ þ 4:5H2 O þ CO2  H2 O 1 ) Naþ þ HCO 3 þ 2H4 SiO4 þ 2Al2 Si2 O5 ðOHÞ4 ð35Þ 3KAlSi3 O8 þ2CO2  H2 O þ 12H2 O ) 2Kþ ðK-feldsparÞ þ 2HCO 3 þ 6H4 SiO4 þ KAl3 Si3 O10 ðOHÞ2 ðmica-illiteÞ

 CaCO3 þH2 O ) Ca2þ þ HCO 3 þ OH

ð37Þ

CaCO3 þ CO2 H2 O ) Ca2þ þ 2HCO 3

ð38Þ

ðcalciteÞ

CaSO4 ðanhydriteÞ

) Ca 2þ þ SO2 4

CaMgðCO3 Þ2 þ2H2 O ) Ca2þ þ Mg2þ ðdolomiteÞ  þ 2HCO 3 þ 2OH

4% 15%

17%

73%

(a)

Figure 23 Major anions (a) and cations (b) in freshwater systems.

ð40Þ

3. Carbonate species (CO2/HCO3–/CO32). The reactive inorganic forms of environmental carbon are carbon dioxide (CO2*H2O), carbonic acid (H2CO3), bicarbonate (HCO 3) and carbonate (CO2 3 ). Carbon dioxide plays a fundamental role in determining the pH in lakes. An important element in acid–base chemistry is the bicarbonate ion, HCO 3 , which may act as either an acid or a base. Aqueous

HCO3– SO42– Cl – Other

–

ð39Þ

The predominant source of magnesium is dolomite:

1.25

+

+5 100

Lower limit, braided sand bed channels

101

(a)

102

103

Median annual flood

+ i:

Some mosses but low abundance of all groups ii: High abundance patch submerged, some mosses

10−1

iii: High abundance branched emergents iv: High abundance linear submerged, some linear emergents v: High abundance linear emergents

++

Slope

10−2

10−3

+ + + ++ + + + + + +++ + ++ + ++ ++++ + + + + ++ + ++ ++ + + + + ++ + + +++ + + + ++ + ++ + + +

vi: High abundance linear and patch submerged and linear emergents

++ + + +

+ +

+ + + + ++ + +

+ + 10−4 100 (b)

101

102

103

Median annual flood

Figure 13 Mean bed-sediment size (a) and the assemblage of macrophyte morphotypes (b) found on a sample of 467 British river reaches in relation to valley gradient (slope in m m1), the median annual flood (in m3 s1), and thresholds of channel style identified by Church M (2002) Geomorphic thresholds in riverine landscapes. Freshwater Biology 47: 541–557. Adapted from Gurnell AM, O’Hare JM, O’Hare MT, Dunbar MJ, and Scarlett PD (2010) Associations between assemblages of aquatic plant morphotypes and channel geomorphological properties within British rivers. Geomorphology 116: 135–144.

Thus, channel changes to flow regulation below dams involves the complex interaction of sediment-transport processes and riparian-vegetation growth (e.g., Petts and Gurnell, 2005, Section 2.10.3.2). One major concern is the impact of flow regulation on channel sedimentation at salmon-spawning grounds, which could impact upon the intra-gravel environment for egg development over winter and for fry emergence in late winter (Milhous, 1982, 1998; Reiser et al., 1989). In particular, deoxygenated conditions in spawning gravels can cause poor egg survival (Malcolm et al., 2005). Consequently,

there have been considerable efforts to quantify the volume, magnitude, duration, and timing of sediment-maintenance flows that flush fines without eroding the important underlying gravels (Wu and Chou, 2004).

2.10.4.2.2 Habitat-suitability criteria Medium and low flows, together experienced for about 90% of the time along most rivers, sustain a diversity of hydraulic habitats. The complex channel morphology of natural rivers

258

Hydrology and Ecology of River Systems

creates a heterogeneous flow environment, described most simply by patterns of velocity and depth. These hydraulic habitats map onto the mesohabitats defined above, but represent the varying velocity-depth conditions within different parts of the river channel to which mobile species can respond as discharge varies. Different species of animals have been observed to have preferred habitats or tolerate different habitats in terms of velocities and depths, to which the caliber of the substratum has been added as a third key habitat characteristic in defining habitat suitability (Bovee, 1978; Gore, 1978). Thus, these associations between the three habitat properties and the preferences of biota have been developed into habitat-suitability criteria (Figure 11(b)), which define the probability of use of habitats by a specific life stage or a particular species. The underlying premise behind the concept of habitatsuitability criteria is that populations, and thus biodiversity in rivers, are limited by habitat events (Stalnaker et al., 1996). Habitat-suitability criteria describe how individuals of a species select the most favorable conditions in a stream but will also use less favorable conditions, with the preference for use decreasing where conditions are less favorable. Simple indices are based on the frequency of occurrence of actual habitat conditions used by a target organism in a particular reach. The ratio of the proportion of habitat utilized to available habitat area within the reach defines the habitat preference. More complex, composite indices may be defined (e.g., Vadas and Orth, 2001; Ahmadi-Nedushan et al., 2006) but these involve several assumptions (Bovee, 1986) not least that all physical variables are equally important and independent. This habitat-suitability-criteria concept has been challenged over the past 30 years because of the lack of concordance between changes in suitable habitat and fish populations, its simplified approach to hydraulic habitat characterization (e.g., Gore and Nestler, 1988), and lack of biological realism (e.g., Orth, 1987), but it remains central to biological-response models that seek to explain and predict spatial and temporal distributions of instream biotic populations.

2.10.4.2.3 Models of biological responses to changing flows A very widely used model called Physical HABitat SIMulation (PHABSIM, Tharme, 2003) integrates the changing hydraulic conditions associated with variations in discharge with the habitat preferences of one or more selected species (Figure 11(b)). The method relies on three principles (Stalnaker, 1994): the chosen species exhibits preferences within a range of habitat conditions that it can tolerate; these ranges can be defined for each species; and the area of stream providing these conditions can be quantified as a function of discharge and channel structure. The primary approach uses a simple 1-D hydraulic model, but this fails to predict spatial patterns of velocity in natural rivers, although it is useful for determining average velocity variations with changing discharge. This weakness has been overcome by the increasing use of 2-D hydraulic models that can describe the spatial and temporal heterogeneity of hydraulic conditions and provide a link to mesohabitat patterns (Bovee, 1996; Hardy, 1998; Stewart et al., 2005; Crowder and Diplas, 2006).

Considerable efforts have been spent on attempts to assess the ecological credibility of PHABSIM by demonstrating the biological significance of carrying capacity as a limiting factor of population size (Lamouroux et al., 1999; Kondolf et al., 2000). However, validation of the approach in biological terms has proved difficult not least in establishing discrete relationships between biological populations and the weighted usable area (WUA) from empirically derived habitat-suitability curves. From a practical perspective, there is no doubt that the accumulated experience of using PHABSIM means that its strengths and weaknesses are well understood. Parasiewicz (2003) advanced a PHABSIM derivative, MesoHABSIM. By mapping mesohabitats at different flows along extensive sections of a river and establishing the suitability of each mesohabitat for the dominant members of the fish community, it is possible to derive rating curves to describe changes in relative areas of suitable habitat in response to flow. MesoHABSIM focuses on mesoscale approaches to build on strengths of PHABSIM protocols while providing options for addressing large spatial scales appropriate for water-resource planning (Jacobson, 2008). A rational framework for modeling fish-community response to changing habitat conditions developed by Bain and Meixler (2008) is appropriate for integrating with physical-habitat modeling (Parasiewicz, 2008). The fish-collection survey is the most effort-intensive component of MesoHABSIM, but literature-based evidence and expert opinion can be used, and a regional approach allows transfer of habitat-use models among rivers of similar type (Parasiewicz, 2007). However, the challenge to relate habitat use to changing flows remains elusive (Petts, 2009). The temporal dynamics of habitat quantity may be a major factor determining fish-population responses in riverine environments (Stalnaker et al., 1996), but there is limited evidence that this is manifest by different patterns of habitat use and a large number of empirical case studies have been unable to develop general relationships (Poff and Zimmerman, 2009). The biomass of a species or a particular life stage within a community can vary because of biological processes such as reproduction, energetics, and mortality that may be influenced by one or more unspecified environmental factors, which undoubtedly blur any simple relationships between species abundance and habitat criteria. Considering trout, for example, recruitment has been shown to be strongly influenced by winter flows (Cattane´o et al., 2002; Lobon-Cervia, 2003; Mitro et al., 2003), but Sabaton et al. (1997) and Gouraud et al. (2001) demonstrated the impact of summer low flows that limit adult-trout biomass and spring flows that limit young-ofthe-year numbers between emergence and their first summer, supporting the findings of Capra et al. (2003) that postemergence high flows have a major impact on the density of 0 þ fishes. For unionid mussels, Morales et al. (2006) predicted community development as a function of individual growth and reproduction, biotic interaction involving host fish and intra- and inter-species food competition, and habitat criteria (substrate stability), and demonstrated that for lowdensity species, even a small level of habitat modification could have a substantial impact on population survival. For common floodplain fishes, Halls and Welcomme (2004) advanced an age-structured population-dynamics model

Hydrology and Ecology of River Systems

incorporating density-dependent growth, mortality, and recruitment to show the importance of high flood duration, large area of inundation, and a slow rate of flood recession. Thus, biological interactions and flow variability, especially the length of time since the last major flood or drought, may confound attempts to demonstrate simple relationships between habitat availability and fish stocks (Sabaton et al., 2008) and provide a challenge to the future development of predictive tools.

2.10.5 Managing River Flows to Protect Riverine Ecosystems In this chapter, we have developed three interdependent themes. In Section 2.10.2, we identified those hydrological characteristics of river networks that have been found to be of high ecological importance, primarily focusing on flow regimes, flow extremes, and hydrological connectivity. In Section 2.10.3, we demonstrated that these hydrological characteristics form the main control on the geomorphological style of river corridors and thus their shifting habitat mosaic. Complex interactions occur among river flows, fluvial sediments, and vegetation within naturally adjusting corridors, providing resilient, bio-complex river environments. In Section 2.10.4, we showed how the flow regime and the style and dynamics of the river corridor and river system define the ecohydraulic and mesohabitat complexity of the aquatic ecosystem and its three-dimensional connectivity. Throughout, we have touched on the effects of human interventions. In this section, we conclude our discussion of hydrology and ecology by considering how river flows can be managed to protect riverine ecosystems as well as support human needs, an area that is attracting enormous attention from researchers, managers, and policymakers as the world’s rivers come under increasing human pressure (e.g., Annear et al., 2004; Naiman et al., 2002). Many rivers today have flow regimes that differ markedly from the climate-driven regime because of impoundments; the magnitude, frequency, and timing of floods have altered with land-use change; and moderate- to low-flow percentiles have been changed in various ways along the length of a river as a consequence of both abstractions and discharges from wastewater-treatment works (Figure 14). River-flow regulation to control flooding and provide water for human use has had deep-seated impacts on river-corridor ecosystems (e.g., Ward and Stanford, 1979; Petts 1984; Tockner and Stanford, 2002). Direct surface-water abstractions and structural flood-alleviation measures, the construction of all types of surface reservoirs, and development of groundwater resources, including the conjunctive management of surface and groundwater, change the river flow regime and thus induce changes in river-channel characteristics (size, form, and style) and the river corridor and channel-habitat mosaic. The importance of managing flows to sustain riverine ecosystems and especially populations of native species has been demonstrated by the impacts of flow regulation below dams upon river-channel characteristics (Petts and Gurnell, 2005, Section 2.10.3.2) and biota (Petts, 1984, 2007). The deep impact of flow regulation is supported by the observation that regulated rivers regain normative attributes with sufficient distance

259

below a dam and that depressed populations of native species can recover if the natural magnitudes and variations in flows are reestablished. Thus, Armitage’s (2006) long-term study of the impacts of Cow Green Dam on the River Tees in Northern England revealed that a narrower range of environmental conditions and increased flow stability led to a dynamically fragile community (indicated by observed changes in community diversity and abundance) which is very susceptible to perturbations because it has developed in their absence. Periphyton and reservoir plankton play an important role in structuring the faunal composition by creating an environment where biotic interactions are more likely. This does not require restoration to some pristine state, but the recovery of some large portion of the lost capacity to sustain native biodiversity and bioproduction is possible by management of processes that maintain normative habitat conditions (Stanford et al., 1996). Naiman et al. (2002) summarized the fundamental ecological principles for understanding hydrology–ecology relationships along rivers, focusing on the climatically driven variability of flows at least from season to season and from year to year. The two linked general principles are 1. that the natural-flow regime shapes the evolution of aquatic biota and ecological processes; and 2. that every river has a characteristic flow regime and an associated biotic community. Four further principles were elaborated by Bunn and Arthington (2002): 1. Flow is a major determinant of physical habitat in rivers, which in turn is a major determinant of biotic composition. 2. Maintenance of the natural patterns of connectivity between habitats (a) along a river and (b) between a river and its riparian zone and floodplain is essential to the viability of populations of many riverine species. 3. Aquatic species have evolved life-history strategies primarily in response to the natural-flow regime and the habitats that are available at different times of the year and in both wet and dry years. 4. The invasion and success of exotic and introduced species along river corridors is facilitated by regulation of the flow regime, especially with the loss of natural wet–dry cycles. These principles underpin three elements of regulated river management: the determination of (1) benchmark flows, (2) ecologically acceptable hydrographs, and (3) ecologically acceptable flow-duration curves (Figure 15). These three elements inform short-term and local operational rules; seasonal and short series of annual flow management; and long-term water-resource planning, respectively. The science and application of environmental flows has attracted considerable attention and Tharme (2003) identified over 200 approaches that have been described for advising on environmental flows in 44 countries. These range from reconnaissance-level assessments relying on ecologically informed hydrological methodologies to approaches using complex hydrodynamic habitat modeling. In some areas, such as Australia and southern Africa, a lack of ecological data and

260

Hydrology and Ecology of River Systems

R. Tame

0

Distance from source (km)

River Manifold

Distance from source (km)

Artifical flow component R. Dove

10

20

Net discharge to river > 5 MId

0

Net abstractions from river > 5 MId 10

WRW Water reclamation works PWS Public water supply

20

Nature flow component Net discharge to river > 5 MId

30 30 40

Minworth A.S. WRW Coleshill WRW

River Blythe 40

50

River Churnet

River Churnet

Minworth P.F. WRW 50

60

River Anker River Tean 60

70 Hilton Brook

300

80

70

Eggington PWS

200

100

0

100

300

200

100

Flow (MId)

0

100

200

300

400

500

600

Flow (MId)

Figure 14 Residual flow diagrams representing the river condition under dry weather flow (DWF) (the average of the annual series of the minimum weekly/7 consecutive days) for a relatively natural (River Dove) and heavily influenced river (River Tame) within the River Trent catchment, UK. The diagrams show downstream flow increments from tributaries (natural flow component) and water-reclamation works and abstractions (artificial flow component). The total discharge at any point along the river is the combination of the natural and artificial flow components. Adapted from Pirt J (1983) The Estimation of Riverflows in Ungauged Catchments. Unpublished Doctoral Thesis, 2 volumes, Loughborough University of Technology.

process models, and political pressure to deliver environmental flow recommendations in short timeframes, often less than 1 year, has led to the use of scientific panels to set environmental flows (Cottingham et al., 2002). For more than 40 years, tools have been advanced defining benchmark flows to allocate flow to meet in-river needs (Petts and Maddock, 1994) of which PHABSIM has been most widely used. In the 1960s and 1970s, early attempts to set instream flows for rivers focused on the annual minimum flow expressed as a hydrological statistic, commonly as either a flow-duration statistic (such as the 95th percentile flow) or as a fixed percentage of the average daily flow (ADF), with several studies proposing 20% ADF to protect aquatic habitat in streams (e.g., Tennant, 1976). However, recognition of the threat to fisheries of confining flow management to annual minimum flows led to more complex and hydrologically rational approaches (Stalnaker 1979, 1994; Stalnaker et al., 1996). By the early 1990s, the science and management of regulated rivers had expanded from the determination of instream flows to environmental flows and many schemes applied more complex flow-habitat models to address wider

issues than the instream needs (the hydraulic habitats) of a single species. Three general approaches to the allocation of flows to support river-ecosystem needs are being advanced and have achieved some success (Arthington et al., 2003): hydrological methods, hydraulic models, and holistic approaches. These approaches address the sustainability of communities and ecosystems, the access of aquatic biota to seasonal floodplain and riparian habitats as well as the need for high flows to sustain the geomorphological dynamics of the river corridor and floodplain habitats (RRA, 2003). They enabled advancement of an ecologically acceptable flow regime concept (Figure 15; Petts, 1996; Petts et al., 1999). This recognized that different life stages and different species benefit from different flows at different times of the year, and in different years. Rivers must be protected in wet years as well as drought years because high flows provide optimum conditions for some species and are also responsible for sustaining the quality and diversity of in-channel and riparian habitats. Societal demands for river-ecosystem protection have accelerated the development of innovative, locally applicable

Hydrology and Ecology of River Systems

261

Flow time series Year 1

Year 2

Year 3

Year n

Ecological assessment

Event analysis • Magnitude • Duration Alternatives e.g., physical habitat

• Frequency

% time exceeded

Synthesis

Benchmark flows

Flow duration Naturalized

Gauged Flow

Acceptable hydrographs

Flow

Wet

Normal Dry

Jan

Time

Dec

% time exceeded

Cost− benefit Acceptable minimum regime

Flow

EAFR hydrographs

EAFR flow duration

Operational rules

Water resources allocation

Figure 15 A general procedure for deriving an ecologically acceptable flow regime represented as one or more (e.g., wet and dry years) hydrographs for defining operational rules and as a flow-duration curve for assessing abstractable volumes. The procedure allows the evaluation of alternatives including physical-habitat improvements as part of the decision-making process. Adapted from Petts GE and Amoros C (eds.) (1996) Fluvial Hydrosystems, 308pp. London: Chapman and Hall.

methods and tools especially within regions having limited databases. However, there are also many examples where sophisticated, science-based models are being applied to specific problems. For example, Grand et al. (2006) used a cell-based model of backwater geometry, a pond-based temperature model, and a model of invertebrate production to investigate the effects of within-day flow fluctuations caused by hydropower operations on nursery habitats for larval and juvenile Colorado pike minnow (Ptychocheilus lucius) along the Green River below Flaming Gorge dam, Utah, USA. As noted by Parasiewicz (2001), if community structure reflects habitat structure, then securing habitat for the most common species might preserve the most profound characteristics of the ecosystem and provide survival conditions for the majority of the aquatic community. Progress in developing models that link physical-habitat dynamics and population biology of large organisms such as fish may have been constrained by the

difficulty in merging the space- and timescales appropriate to both physical and biological sciences (Petts et al., 2006). In the hydrological approaches, flow is considered as a simple proxy for a number of related parameters that may have a key influence on the range of aquatic, wetland, and riparian habitats along the river corridor. Thus, Extence et al. (1999) developed a scoring system as an indicator of hydrological stress based upon surveys of macroinvertebrate fauna that has been shown to be sensitive to particular hydrological indices (Wood et al., 2001a; Monk et al., 2008). A range of hydrologic parameters for each year of flow record can be used to characterize inter-annual variation before (reference period) and after flow regulation/abstraction (Richter et al., 1996, 1997). Of the hydrological approaches, White et al. (2005) used wavelet analysis to assess dam operations in reconstructing desired flow characteristics. This method provides an easy-to-interpret approach for investigating hydrological

262

Hydrology and Ecology of River Systems

change when the management history is uncertain and timescales of important cycles are unknown. White et al. (2005) used wavelet analysis to detect the hydrological implications of management practices over a range of timescales and suggested that the method could provide a powerful data-mining technique for assessing hydrological changes. The statistical characterization of ecologically relevant hydrograph parameters could be used to define the variability of the dimensions of the flow regime within which artificial influences should be contained (Richter et al., 1997). To date, such approaches have been used in water resources and environmental management in the USA (e.g., Richter et al., 1998, 2006; Mathews and Richter, 2007) and elsewhere (e.g., Shiau and Wu, 2004, 2006). Thus, Galat and Lipkin (2000) recommended changes in reservoir management to return regulated flows to within the pattern of natural variability, thereby simulating a natural riverine ecosystem. They argued that naturalization of the flow regime would benefit not only the ecological system but also the economic value of the Missouri River, once the products of agriculture, electric-power generation, and transportation are integrated with the socioecological benefits of a naturalized flow regime. In a follow-up study, Jacobson and Galat (2008) focused on developing a flow regime to support the endangered pallid sturgeon (Scaphirhynchus albus). Specific hydrograph requirements for pallid sturgeon reproduction were unknown; so much of the design process was based on hydrological parameters extracted from the reference natural-flow regime. Three issues often hinder the apparently simple and reasonable application of such hydrological approaches. First, standards need to be set to apply an appropriate record length. At least 12 years data are required for statistical integrity and longer records are needed to incorporate variable weather patterns over decadal timescales and to provide for actual scales of variability in the magnitude and timing of flows and the natural frequencies of these flows. The flow regime is a complex concept. Flow regimes typical of each hydro-climatic region across the globe represent average conditions created by combining a small number of flow-regime types, particular to each hydro-climatic region. Variations of the flow regime from year to year within the British temperate maritime hydro-climatic zone based upon analysis of 80 station-years of flow data (1977–97) from four major rivers across the United Kingdom showed that the typical flow regime with a high-flow season from December to February and a low-flow season from June to August occurred in only 51% of the years (Harris et al., 2000). Of the other flow years, three variants of the typical flow regime were differentiated:

• • •

Subtype A. Twenty-eight percent had a dominant peak in November often with a secondary peak in April. Subtype B. Sixteen percent had a dominant peak in March with a secondary peak in December. Subtype C. Six percent were characterized by winter drought with no dominant peak and typically a very dry January.

Second is the issue of naturalizing the gauged flow regime. In many areas, the pristine catchment has no relevance to the modern day. The hydrology of catchments characterized by long-term human interference – such as urban conurbations

and intensive agriculture – bears little resemblance to the hydrologic character of unmodified catchments in a given hydro-climatic region. The concept for such catchments may be to produce functionally diverse, self-regulating ecological systems (Petts et al., 2000). In reality, this requires determination of the flow regime that would be sustained under current or future catchment conditions in the absence of existing dams, reservoirs, diversions, and abstractions. Third, the linkages between flow regime and ecological health are complex in both time and space. The natural dynamic character relates not only to flow variability but also to water quality, especially temperature variations; sediment dynamics and channel dynamics (that are also influenced by patterns of woody vegetation growth); changes in food/energy supply; and interactions between biological populations. Across the UK, most regulated rivers are supported in summer by compensation flows that maintain minimum flows or they may even experience enhanced flows during dry summers where the river supports abstractions from the lower river. Under one scenario, the main ecological impact of flow regulation below reservoirs would not be during a summer drought but during the late summer, autumn, and early winter following a subtype C flow year when the need to fill reservoir storage could eliminate high flows along the main river. Under such circumstances, there would be inadequate river flows to stimulate up-river fish migrations and spawning grounds could be impacted by siltation caused by fine-sediment loaded tributary spates. Water temperature is a particularly important parameter and a river’s thermal regime is a key component of environmental flows. Harris et al. (2000) and Olden and Naiman (2009) have encouraged ecologists and water managers to broaden their perspective on environmental flows to include both flow and thermal regimes in assessing e-flow needs. Assessments should include the comprehensive characterization of seasonality and variability in stream temperatures in the face of artificial influences on flow and potential impacts of climate change. From a scientific perspective, we need to more clearly elucidate the relative roles of altered flow and temperature in shaping ecological patterns and processes in riverine ecosystems. In the absence of universal relationships between flows and biotic responses, King et al. (2003) advanced a value-based system, Downstream Response to Imposed Flow Transformation (DRIFT). This provided a data-management tool for many types and sources of information, predictive models, theoretical principles, and expert knowledge of a panel of scientists. The approach was developed to link the productivity of large floodplain rivers to their flow characteristics in countries or river basins where data scarcity constrains prediction of ecological responses to flow regulation. It was also produced in a developing region with severe water shortage and uncertainties about river-linked ecological processes and where riparian subsistence populations are important in the decision-making process. DRIFT supports the scientific-panel approach to recommend environmental flows within an adaptive management framework. It is based around four modules: (1) the biophysical module describes the present nature and functioning of the ecosystem; (2) the sociological module identifies subsistence users at risk from flow

Hydrology and Ecology of River Systems

abstraction or regulation; (3) a module that combines the outputs from the first two to develop biophysical and subsistence scenarios; and (4) a module to address mitigation and compensation costs. Arthington et al. (2003) used DRIFT to establish the environmental flow requirements of rivers in Lesotho and contended that the methodology can provide a best-practice framework for conducting scientific-panel studies. Linking environmental variables to dam-release rules has been shown to achieve significant water savings (Harman and Stewardson, 2005). However, there is no single best method or approach (IUCN, 2003). Given the variety of water resource contexts, the range of environmental settings and variety of species concerned, there has been an increasing tendency to use evidence-based expert judgment. In reality, all environmental flow assessments provide the evidence available to larger or smaller, expert panels, – the decision-making process that evaluates tradeoffs between water users. From a flowmanagement perspective, these tradeoffs include those between magnitude, frequency, duration, timing, and rate of change (Poff and Ward, 1989), and the evidence is often hierarchically structured to include three orders (Petts, 1980) or levels (Young et al., 2004), broadly linking primary processes, habitat structure, and biota. Moreover, Jacobson and Galat’s (2008) experience with flow-regime design using a hydrological approach demonstrated lack of confidence by stakeholders in the value of the natural-flow regime as a measure of ecosystem benefit. The lack of confidence resulted from the lack of fundamental scientific documentation, as might have been provided by a more complex hydrological– hydraulic–biological model. Stakeholders desired proof of ecological benefits commensurate with certainty of economic losses. This conflict between demands for more biologically accountable models and political actions to set environmental flows has also been highlighted by Arthington et al. (2006). Despite considerable progress in understanding how flow variability sustains river ecosystems, there is a growing temptation to ignore natural-system complexity in favor of simplistic, static, environmental flow rules to resolve pressing river-management issues. Arthington et al. (2006) argue that such approaches are misguided and will ultimately contribute to further degradation of river ecosystems. In the absence of detailed empirical information of environmental flow requirements for rivers, they proposed a generic approach that incorporates essential aspects of natural-flow variability shared across particular classes of rivers that can be validated with empirical biological data and other information in a calibration process. They further argue that this approach can bridge the gap between simple hydrological rules of thumb and more comprehensive environmental flow assessments and experimental flow-restoration projects. Thus, in the USA, Poff et al. (2009) achieved a consensus view from a panel of international scientists on a framework for assessing environmental flow needs that combines a regional hydrological approach and ecological response relationships for each river type based initially on the literature, existing data, and expert knowledge. Stakeholders and decision makers then explicitly evaluate acceptable risk as a balance between perceived value of the ecological goals, the economic costs involved, and the scientific uncertainties. The main risk is a perceived lack of

263

incentive for what could be considered to be costly monitoring and longer-term research to develop evidence of biota– flow relationships for supporting adaptive management. In conclusion, it is impossible to return regulated rivers to an unimpacted state or even to define what such an unimpacted state might be, given the long history of human intervention across the Earth’s surface superimposed on a background of global environmental change. However, it is possible to combine scientific understanding and expert judgment to establish river flows that can support river ecosystems, where the river channel and riparian zone are also managed. At least in a set of reaches distributed across the river network, river flows, sediment, and vegetation need to interact relatively freely to provide refugia for species and resilient sites from which other areas of the river network can be recolonized. Sustaining river ecosystems needs to be based on the maintenance of appropriate river flows coupled with restoration of the potential for reaches within the river network to adjust to those flows.

References Abbe TB and Montgomery DR (2003) Patterns and processes of wood debris accumulation in the Queets river basin, Washington. Geomorphology 51: 81--107. Agostinho AA, Gomes LC, Verissimo S, and Okada EK (2004) Flood regime, dam regulation and fish in the Upper Parana River: Effects on assemblage attributes, reproduction and recruitment. Reviews in Fish Biology and Fisheries 14(1): 11--19. Ahmadi-Nedushan B, St-Hilaire A, Be´rube´ M, Robichaud E, Thie´monge N, and Bobe´e B (2006) A review of statistical methods for the evaluation of aquatic habitat suitability for instream flow assessment. River Research and Applications 22(5): 503--523. Allan JD (2004) Landscapes and riverscapes: The influence of land use on stream ecosystems. Annual Review of Ecology, Evolution and Systematics 35: 257--284. Allen MR and Ingram WJ (2002) Constraints on future changes in climate and the hydrologic cycle. Nature 419(6903): 224--232. Amoros C and Roux AL (1988) Interaction between water bodies within the floodplains of large rivers: Function and development of connectivity. In: Schreiber KL (ed.) Connectivity in Landscape Ecology, Proceedings of the 2nd International Seminar of the International Association for Landscape Ecology, Mu¨nster, 1987, Munstersche Geographische Arbeiten, pp. 125–130. Paderborn: Schoningh. Anderson KE, Paul AJ, McCauley E, Jackson LJ, Post JR, and Nisbet RM (2006) Instream flow needs in streams and rivers: The importance of understanding ecological dynamics. Frontiers in Ecology and the Environment 4(6): 309--318. Andersson E, Nilsson C, and Johansson ME (2000) Plant dispersal in boreal rivers and its relation to the diversity of riparian flora. Journal of Biogeography 27: 1095--1106. Amitage PD (2006) Long-term faunal changes in a regulated and an unregulated stream – cow green thrity years on. River Research and Applications 22(9): 947–966. Annear T, Chisholm I, Beecher H, et al. (2004) Instream Flow for Riverine Resource Stewardship, revised edition. Cheyenne, WY: Instream Flow Council. Arnell NW (2003) Effects of IPCC SRES emissions scenarios on river runoff: A global perspective. Hydrology and Earth System Sciences 7(5): 619--641. Arora VK and Boer GJ (2001) Effects of simulated climate change on the hydrology of major river basins. Journal of Geophysical Research – Atmospheres 106(D4): 3335--3348. Arscott DB, Tockner K, van der Nat D, and Ward JV (2002) Aquatic habitat dynamics along a braided alpine river ecosystem (Tagliamento River, Northeast Italy). Ecosystems 5(8): 802--814. Arscott DB, Tockner K, and Ward JV (2001) Thermal heterogeneity along a braided floodplain river (Tagliamento River, northeastern Italy). Canadian Journal of Fisheries and Aquatic Sciences 58(12): 2359--2373. Arscott DB, Tockner K, and Ward JV (2003) Spatio-temporal patterns of benthic invertebrates along the continuum of a braided Alpine river. Archiv fu¨r Hydrobiologie 158(4): 431--460.

264

Hydrology and Ecology of River Systems

Arsenault D, Boucher E, and Bouchon E (2007) Asynchronous forest-stream coupling in a fire-prone boreal landscape: Insights from woody debris. Journal of Ecology 95: 789--801. Arthington AH, Bunn SE, Poff NL, and Naiman RJ (2006) The challenge of providing environmental flow rules to sustain river ecosystems. Ecological Applications 16(4): 1311--1318. Arthington AH, Rall JL, Kennard MJ, and Pusey BJ (2003) Environmental flow requirements of fish in Lesotho rivers using the DRIFT methodology. River Research and Applications 19(5–6): 641--666. Arthurton R, Barker S, Rast W, and Huber M (2007) Water. In: Ashton P (ed.) Global Environmental Outlook 4 (GEO-4), p. 42pp. Malta: UNEP. Asselman NEM and Middelkoop H (1995) Floodplain sedimentation: Quantities, patterns and processes. Earth Surface Processes and Landforms 20: 481--499. Bain MB and Meixler MS (2008) A target fish community to guide river restoration. River Research and Applications 24(4): 453--458. Barsoum N and Hughes FMR (1998) Regeneration response of black poplar to changing river levels. In: Wheater H and Kirby C (eds.) Hydrology in a Changing Environment, pp. 397--412. Chichester: Wiley Baxter CV and Hauer FR (2000) Geomorphology, hyporheic exchange, and selection of spawning habitat by bull trout (Salvelinus confluentus). Canadian Journal of Fish and Aquatic Sciences 57: 1470--1481. Bayley PB (1991) The flood-pulse advantage and the restoration of river-floodplain systems. Regulated Rivers: Research and Management 6: 75--86. Begin ZB (1981) The relationship between flow-shear stress and stream pattern. Journal of Hydrology 52(3–4): 307--319. Benda L, Poff NL, Miller D, et al. (2004) The network dynamics hypothesis: How channel networks structure riverine habitats. BioScience 54(5): 413--427. Benedetti MM (2003) Controls on overbank deposition in the Upper Mississippi River. Geomorphology 56: 271--290. Bertoldi W, Gurnell A, Surian N, et al. (2009) Understanding reference processes: Linkages between river flows, sediment dynamics and vegetated landfroms along the Tagliamento River, Italy. River Research and Applications 25: 501--516. Birken AS and Cooper DJ (2006) Processes of Tamarix invasion and floodplain development along the lower Green River, Utah. Ecological Applications 16(3): 1103--1120. Bledsoe BP and Watson CC (2001) Logistic analysis of channel pattern thresholds: Meandering, braiding, and incising. Geomorphology 38(3–4): 281--300. Boedeltje G, Bakker JP, Bekker RM, Van Groenendael JM, and Soesbergen M (2003) Plant dispersal in a lowland stream in relation to occurrence and three specific lifehistory traits of the species in the species pool. Journal of Ecology 91(5): 855--866. Boedeltje G, Bakker JP, Ten Brinke A, Van Groenendael JM, and Soesbergen M (2004) Dispersal phenology of hydrochorous plants in relation to discharge, seed release time and buoyancy of seeds: The flood pulse concept supported. Journal of Ecology 92(5): 786--796. Bovee KD (1978) The incremental method of assessing habitat potential for coolwater species, with management implications. American Fisheries Society Special Publication 11: 340--346. Bovee KD (1986) Development and evaluation of habitat suitability criteria for use in the instream flow incremental methodology. Instream Flow Information Paper, 21, US Fish and Wildlife Service Biological Report No. 86. Bovee KD (1996) Perspectives on two-dimensional river habitat models: The PHABSIM experience Proceedings of the Second International Symposium on Habitat Hydraulics, pp. 149--162. Que´bec, QC: INRS-Eau. Bray DI (1982) Regime equations for gravel-bed rivers. In: Hey RD, Bathurst JC, and Thorne CR (eds.) Gravel-bed Rivers: Fluvial Processes, Engineering and Management, pp. 517--552. Chichester: Wiley. Brewer PA and Lewin J (1998) Planform cyclicity in an unstable reach: Complex fluvial response to environmental change. Earth Surface Processes and Landforms 23: 989--1008. Brierley GJ and Fryirs KA (2005) Geomorphology and River Management: Application of the River Styles Framework. 398pp. Malden, MA: Blackwell. Brooks AP and Brierley GJ (2002) Mediated equilibrium: The influence of riparian vegetation and wood on the long-term evolution and behaviour of a near-pristine river. Earth Surface Processes and Landforms 27(4): 343--367. Brooks AP, Brierley GJ, and Millar RG (2003) The long-term control of vegetation and woody debris on channel and flood-plain evolution: Insights from a paired catchment study in southeastern Australia. Geomorphology 51(1–3): 7--29. Brunke M and Gonser T (1997) The ecological significance of exchange processes between rivers and groundwater. Freshwater Biology 37: 1--33. Bruns DA, Minshall GW, Gushing CE, Cummins KW, Brock JT, and Vannote RL (1984) Tributaries as modifiers of the river-continuum concept: Analysis by polar ordination and regression models. Archiv fu¨r Hydrobiologie 99: 208--220.

Bunn SE and Arthington AH (2002) Basic principles and ecological consequences of altered flow regimes for aquatic biodiversity. Environmental Management 30(4): 492--507. Caissie D (2006) The thermal regime of rivers: A review. Freshwater Biology 51(8): 1389--1406. Capra H, Sabaton C, Gouraud V, Souchon Y, and Lim P (2003) A population dynamics model and habitat simulation as a tool to predict brown trout demography in natural and bypassed stream reaches. River Research and Applications 19(5–6): 551--568. Carson MA (1984a) Observations on the meandering–braided river transition, the Canterbury Plains, New Zealand: Part one. New Zealand Geographer 40(1): 12--19. Carson MA (1984b) Observations on the meandering–braided river transition, the Canterbury Plains, New Zealand: Part two. New Zealand Geographer 40(2): 89--99. Cattane´o F, Lamouroux N, Breil P, and Capra H (2002) The influence of hydrological and biotic processes on brown trout (Salmo trutta) population dynamics. Canadian Journal of Fisheries and Aquatic Sciences 59(1): 12--22. Church M (1992) Channel morphology and typology. In: Callow C and Petts G (eds.) The Rivers Handbook: Hydrological and Ecological Principles, pp. 126--143. Oxford: Blackwell. Church M (2002) Geomorphic thresholds in riverine landscapes. Freshwater Biology 47: 541--557. Corenblit D, Steiger J, Gurnell AM, Tabacchi E, and Roques L (2010) Control of sediment dynamics by vegetation as a key ecological function within fluvial corridors. Earth Surface Processes and Landforms 34(13): 1790–1810. Corenblit D, Tabacchi E, Steiger J, and Gurnell AM (2007) Reciprocal interactions and adjustments between fluvial landforms and vegetation dynamics in river corridors: A review of complementary approaches. Earth-Science Reviews 84(1–2): 56--86. Cottingham P, Thoms MC, and Quinn GP (2002) Scientific panels and their use in environmental flow assessment in Australia. Australian Journal of Water Resources 5: 103--111. Coulthard TJ (2005) Effects of vegetation on braided stream pattern and dynamics. Water Resources Research 41: W04003. Craig LS, Palmer MA, Richardson DC, et al. (2008) Stream restoration strategies for reducing river nitrogen loads. Frontiers in Ecology and the Environment 6(10): 529--538. Crowder DW and Diplas P (2006) Applying spatial hydraulic principles to quantify stream habitat. River Research and Applications 22(1): 79--89. Dakova S, Uzunov Y, and Mandadjiev D (2000) Low flow – the river’s ecosystem limiting factor. Ecological Engineering 16: 167--174. Danielopol DL (1989) Groundwater fauna associated with riverine aquifers. Journal of the North American Benthological Society 8(1): 18--35. Danvind M and Nilsson C (1997) Seed floating ability and distribution of alpine plants along a northern Swedish river. Journal of Vegetation Science 8(2): 271--276. De´camps H and Tabacchi E (1994) Species richness in vegetation along river margins. In: Giller PS, Hildrew AG, and Raffaelli DG (eds.) Aquatic Ecology: Scale, Pattern and Process, pp. 1--20. Oxford: Blackwell. Desloges JR and Church MA (1989) Wandering gravel-bed rivers. Canadian Geographer 33(4): 360--364. Dettinger MD and Diaz HF (2000) Global characteristics of stream flow seasonality and variability. Journal of Hydrometeorology 1: 289--310. Doyle MW, Stanley EH, Strayer DL, Jacobson RB, and Schmidt JC (2005) Effective discharge analysis of ecological processes in streams. Water Resources Research 41: W11411. Edwards PJ, Kollmann J, Gurnell AM, Petts GE, Tockner K, and Ward JV (1999) A conceptual model of vegetation dynamics on gravel bars of a large Alpine river. Wetlands Ecology and Management 7: 141--153. Elwood JW, Newbold JD, O’Neill RV, and Van Winkle W (1983) Resource spiraling: An operational paradigm for analyzing lotic ecosystems. In: Fontaine TD and Bartell SM (eds.) Dynamics of Lotic Ecosystems, pp. 3--27. Ann Arbor, MI: Ann Arbor Science. Emery JC, Gurnell AM, Clifford NJ, Petts GE, Morrissey IP, and Soar PJ (2003) Classifying the hydraulic performance of riffle-pool bedforms for habitat assessment and river rehabilitation design. River Research and Applications 19(5– 6): 533--549. Extence CA, Balbi DM, and Chadd RP (1999) River flow indexing using British benthic macroinvertebrates: A framework for setting hydroecological objectives. Regulated Rivers 15: 543–574. Ferguson RI (1987) Hydraulic and sedimentary controls of channel pattern. In: Richards K (ed.) River Channels, pp. 129--158. Oxford: Blackwell. Fisher SG (1990) Recovery processes in lotic ecosystems: Limits of successional theory. Environmental Management 14: 725--736. Francis RA (2007) Size and position matter: Riparian plant establishment from fluvially deposited trees. Earth Surface Processes and Landforms 32(8): 1239--1243.

Hydrology and Ecology of River Systems Francis RA, Gurnell AM, Petts GE, and Edwards PJ (2005) Survival and growth responses of Populus nigra, Salix elaeagnos and Alnus incana cuttings to varying levels of hydric stress. Forest Ecology and Management 210: 291--301. Francis RA, Gurnell AM, Petts GE, and Edwards PJ (2006) Riparian tree establishment on gravel bars: Interactions between plant growth strategy and the physical environment. In: Sambrook Smith GH, Best JL, Bristow CS, and Petts GE (eds.) Braided Rivers: Process, Deposits, Ecology and Management, International Association of Sedimentologists Special Publication 36, pp. 361–380. VIC: Blackwell. Francis RA, Tibaldeschi P, and McDougall L (2008) Fluvially-deposited large wood and riparian plant diversity. Wetlands Ecology and Management 16(5): 371--382. Frissell CA, Liss WJ, Warren CE, and Hurley MD (1986) A hierarchical framework for stream habitat classification: Viewing streams in a watershed context. Environmental Management 10: 199--214. Galat DL and Lipkin R (2000) Restoring ecological integrity of great rivers: Historical hydrographs aid in defining reference conditions for the Missouri river. Hydrobiologia 422: 29--48. Gibert J, Dole-Olivier M-J, Marmonier P, and Vervier P (1990) Surface water– groundwater ecotones. In: Naiman RJ and De´camps H (eds.) The Ecology and Management of Aquatic–Terrestrial Ecotones, pp. 199--226. Paris: UNESCO. Glenn EP and Nagler PL (2005) Comparative ecophysiology of Tamarix ramosissima and native trees in western US riparian zones. Journal of Arid Environments 61(3): 419--446. Gomez B, Eden DN, Peacock DH, and Pinkney EJ (1998) Floodplain construction by recent, rapid vertical accretion: Waipaoa river, New Zealand. Earth Surface Processes and Landforms 23: 405--413. Gomi T, Sidle RC, and Richardson JS (2002) Understanding processes and downstream linkages of headwater systems. BioScience 52(10): 905--916. Goodson JM, Gurnell AM, Angold PG, and Morrissey IP (2002) Riparian seed banks along the lower River Dove, UK: Their structure and ecological implications. Geomorphology 47(1): 45--60. Goodson JM, Gurnell AM, Angold PG, and Morrissey IP (2003) Evidence for hydrochory and the deposition of viable seeds within winter flow-deposited sediments: The River Dove, Derbyshire, UK. River Research and Applications 19(4): 317--334. Gore JA (1978) A technique for predicting in-stream flow requirements of benthic macroinvertebrates. Freshwater Biology 8(2): 141--151. Gore JA and Nestler JM (1988) Instream flow studies in perspective. Regulated Rivers: Research and Management 2: 93--101. Gouraud V, Baglinie`re JL, Baran P, Sabaton C, Lim P, and Ombredane D (2001) Factors regulating brown trout populations in two French rivers: Application of a dynamic population model. Regulated Rivers: Research and Management 17(4–5): 557--569. Gradzinski R, Baryla J, Doktor M, et al. (2003) Vegetation-controlled modern anastomosing system of the upper Narew River (NE Poland) and its sediments. Sedimentary Geology 157(3–4): 253--276. Graf WL (1978) Fluvial adjustments to the spread of tamarisk in the Colorado Plateau region. Geological Society of America Bulletin 89: 1491--1501. Gran K and Paola C (2001) Riparian vegetation controls on braided stream dynamics. Water Resources Research 37: 3275--3283. Grand TC, Railsback SF, Hayse JW, and LaGory KE (2006) A physical habitat model for predicting the effects of flow fluctuations in nursery habitats of the endangered Colorado pikeminnow (Ptychocheilus lucius). River Research and Applications 22(10): 1125--1142. Gregory KJ and Walling DE (1973) Drainage Basin Form and Process. 456pp. London: Edaward Arnold. Gregory SV, Swanson FJ, McKee WA, and Cummins KW (1991) An ecosystem perspective of riparian zones. BioScience 41(8): 540--551. Gurnell AM (1995) Vegetation along river corridors: Hydrogeomorphological interactions. In: Gurnell AM and Petts GE (eds.) Changing River Channels, pp. 237--260. Chichester: Wiley. Gurnell AM (2003) Wood storage and mobility. In: Gregory SV, Staley K, and Gurnell AM (eds.) The Ecology and Management of Wood in World Rivers, American Fisheries Society American Fisheries Society Symposium 37, pp. 75–91. Bethesda, MD. Gurnell AM (2003) Wood storage and mobility. In: Gregory SV, Staley K, Gurnell AM (eds.) The Ecology and Management of Wood in World Rivers, American Fisheries Society American Fisheries Society Symposium 37, Bethesda, Maryland, 75–91. Gurnell AM (2010) Wood in fluvial systems. In: Wohl, E (ed.) Treatise in Geomorphology, Volume 9, Fluvial Geomorphology, chap. 9.11. Elsevier (in press). Gurnell AM, Blackall TD, and Petts GE (2008) Characteristics of freshly deposited sand and finer sediments along an island-braided, gravel-bed river: The roles of water, wind and trees. Geomorphology 99: 254--269.

265

Gurnell AM, O’Hare JM, O’Hare MT, Dunbar MJ, and Scarlett PD (2010) Associations between assemblages of aquatic plant morphotypes and channel geomorphological properties within British rivers. Geomorphology 116: 135–144. Gurnell AM, Lee M, and Souch C (2007) Urban rivers: Hydrology, geomorphology, ecology and opportunities for change. Geography Compass 1(5): 1118--1137. Gurnell AM and Petts G (2006) Trees as riparian engineers: The Tagliamento River, Italy. Earth Surface Processes and Landforms 31(12): 1558--1574. Gurnell AM, Petts GE, Hannah DM, et al. (2000) Wood storage within the active zone of a large European gravel-bed river. Geomorphology 34(1–2): 55--72. Gurnell AM, Petts GE, Hannah DM, et al. (2001) Riparian vegetation and island formation along the gravel-bed Fiume Tagliamento, Italy. Earth Surface Processes and Landforms 26(1): 31--62. Gurnell AM, Piegay H, Swanson FJ, and Gregory SV (2002) Large wood and fluvial processes. Freshwater Biology 47(4): 601--619. Gurnell AM, Thompson K, Goodson J, and Moggridge H (2008) Propagule deposition along river margins: Linking hydrology and ecology. Journal of Ecology 96: 553--565. Gurnell AM, Tockner K, Edwards PJ, and Petts GE (2005) Effects of deposited wood on biocomplexity of river corridors. Frontiers in Ecology and Environment 3(7): 377--382. Gurnell AM, Surian N, and Zanoni L (2009) Multi-thread river channels: A perspective on changing European alpine river systems. Aquatic Sciences 71(3): 253–265. Hack JT (1960) Interpretation of erosional topography in humid temperate regions. American Journal of Science 258A: 80--97. Haines AT, Finlayson BL, and McMahon TA (1988) The global classification of flow regimes. Applied Geography 8: 255--272. Halls AS and Welcomme RL (2004) Dynamics of river fish populations in response to hydrological conditions: A simulation study. River Research and Applications 20(8): 985--1000. Hardy TB (1998) The future of habitat modeling and instream flow assessment techniques. Regulated Rivers: Research and Management 14(5): 405--420. Harman C and Stewardson M (2005) Optimizing dam release rules to meet environmental flow targets. River Research and Applications 21(2–3): 113--129. Harner MJ and Stanford JA (2003) Differences in cottonwood growth between a losing and a gaining reach of an alluvial floodplain. Ecology 84(6): 1453--1458. Harper DM, Smith CD, and Barham PJ (1992) Habitats as the building blocks for river conservation assessment. In: Boon PJ, Calow P, and Petts GE (eds.) River Conservation and Management, pp. 311--319. Chichester: Wiley. Harper DM, Smith CD, and Barham PJ (1995) The ecological basis for management of the natural river environment. In: Harper DM and Erguson AGD (eds.) The Ecological Basis for River Management, pp. 219--238. Chichester: Wiley. Harris NM, Gurnell AM, Hannah DM, and Petts GE (2000) Classification of river regimes: A context for hydroecology. Hydrological Processes 14(16–17): 2831--2848. Harvey GL, Clifford NJ, and Gurnell AM (2008) Towards an ecologically meaningful classification of the flow biotope for river inventory, rehabilitation, design and appraisal purposes. Journal of Environmental Management 88(4): 638--650. Hawkins CP, et al. (1993) A hierarchical approach to classifying stream habitat features. Fisheries 18: 3--12. Hickin EJ and Nanson G (1984) Lateral migration rates of river bends. Journal of Hydraulic Engineering 110: 1557--1567. Hooke JM (2003) River meander behaviour and instability: A framework for analysis. Transactions of the Institute of British Geographers NS 28: 238--253. Horton RE (1945) Erosional development of streams and their drainage basins: Hydrophysical approach to quantitative morphology. Geological Society of America Bulletin 56(3): 275--370. Huang HQ and Nanson GC (2006) Why some alluvial rivers develop an anabranching pattern. Water Resources Research 43: W07441. Huet M (1959) Profiles and biology of Western European streams as related to fish management. Transactions of the American Fisheries Society 88(3): 19--26. Huntington TG (2006) Evidence for intensification of the global water cycle: Review and synthesis. Journal of Hydrology 319(1–4): 83--95. Hynes HBN (1970) The Ecology of Running Waters. Toronto: University of Toronto. Hynes HBN (1975) The stream and its valley. Vereinigung fur Theoretische und Angewandte Limnologie Verhandlungen 19: 1--15. Hynes HBN (1983) Groundwater and stream ecology. Hydrobiologia 100: 93--99. Illies J and Botosaneanu L (1963) Proble`mes et me´thodes de la classification et de la zonation e´cologique des eaux courantes, conside´re´es surtout du point de vue faunistique. Mitteilungen der IVL 12: 57pp. IUCN (2003) Flow: The Essentials of Environmental Flows. Gland: IUCN. Jacobson RA (2008) Applications of MesoHABSIM using fish community targets. River Research and Applications 24(4): 434--438.

266

Hydrology and Ecology of River Systems

Jacobson RB and Galat DL (2008) Design of a naturalized flow regime – an example from the lower Missouri River, USA. Ecohydrology 1(2): 81--104. Johansson ME, Nilsson C, and Nilsson E (1996) Do rivers function as corridors for plant dispersal? Journal of Vegetation Science 7(4): 593--598. Johnson WC (1997) Equilibrium response of riparian vegetation to flow regulation in the Platte river, Nebraska. Regulated Rivers: Research and Management 13(5): 403--415. Johnson WC (2000) Tree recruitment and survival in rivers: Influence of hydrological processes. Hydrological Processes 14(16–17): 3051--3074. Junk WJ, Bayley PB, and Sparks RE (1989) The flood-pulse concept in river-floodplain systems. In: Dodge DP (ed.) Proceedings of the International Large River Symposium, Canadian Special Publication of Fisheries and Aquatic Sciences, vol. 106, pp. 110–127. Ottawa, ON: Canadian Government Publishing Centre. Junk WJ and Wantzen KM (2004) The flood pulse concept: New aspects, approaches, and applications – an update. In: Welcomme RL and Petr T (eds.) Proceedings of the Second International Symposium on the Management of Large Rivers for Fisheries, Food and Agriculture Organization & Mekong River Commission, vol. 2, pp. 117–149. FAO Regional Office for Asia and the Pacific, Bangkok. RAP Publication. Karaus U, Alder L, and Tockner K (2005) Concave islands: Habitat heterogeneity of parafluvial ponds in a gravel-bed river. Wetlands 25(1): 26--37. Karrenberg S, Edwards PJ, and Kollmann J (2002) The life history of Salicaceae living in the active zone of floodplains. Freshwater Biology 47: 733--748. Kemp JL, Harper DM, and Crosa GA (2000) The habitat-scale ecohydraulics of rivers. Ecological Engineering 16: 17--29. Kennen JG, Kauffman LJ, Ayers MA, Wolcock DM, and Colarullo SJ (2008) Use of an integrated flow model to estimate ecologically relevant hydrologic characteristics at stream biomonitoring sites. Ecological Modelling 211: 57--76. King J, Brown C, and Sabet H (2003) A scenario-based holistic approach to environmental flow assessments for rivers. River Research and Applications 19(5–6): 619--639. Kingsford RT and Auld KM (2005) Waterbird breeding and environmental flow management in the Macquarie Marshes, arid Australia. River Research and Applications 21(2–3): 187--200. Knighton D (1998) Fluvial Forms and Processes: A New Perspective. London: Arnold. Kollmann J, Vieli M, Edwards PJ, Tockner K, and Ward JV (1999) Interactions between vegetation development and island formation in the Alpine river Tagliamento. Applied Vegetation Science 2: 25--36. Kondolf GM, Larsen EW, and Williams JG (2000) Measuring and modelling the hydraulic environment for assessing instream flows. North American Journal of Fisheries Management 20: 1016--1028. Krasovskaia I (1996) Sensitivity of the stability of river flow regimes to small fluctuations in temperature. Hydrological Sciences Journal des Sciences Hydrologiques 41(2): 251--264. Krasovskaia I and Saelthun NR (1997) Sensitivity of the stability of Scandinavian river flow regimes to a predicted temperature rise. Hydrological Sciences Journal des Sciences Hydrologiques 42(5): 693--711. Kundzewicz ZW, Mata LJ, Arnell NW, et al. (2007) Freshwater resources and their management. In: Parry ML, Canziani OF, Palutikof JP, van der Linden PJ, and Hanson CE (eds.) Climate Change 2007: Impacts, Adaptation and Vulnerability, Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, pp. 173–210. Cambridge: Cambridge University Press. Lake PS (2003) Ecological effects of perturbation by drought in flowing waters. Freshwater Biology 48(7): 1161--1172. Lake PS (2007) Flow-generated disturbances and ecological responses: Floods and droughts. In: Wood PJ, Hannah DM, and Sadler JP (eds.) Hydroecology and Ecohydrology: Past, Present and Future, pp, pp. 75--92. Chichester: Wiley. Lamouroux N, Capra H, Pouilly M, and Souchon Y (1999) Fish habitat preferences in large streams of southern France. Freshwater Biology 42(4). 673–673 Lane EW (1957) A study of the shape of channels formed by natural streams flowing in erodible material. US Army Corps of Engineers, Missouri River Division, Sediment Series 9: 106. Latterell JJ, Bechtold JS, O’Keefe TC, Van Pelt R, and Naiman RJ (2006) Dynamic patch mosaics and channel movement in an unconfined river valley of the Olympic Mountains. Freshwater Biology 51: 523--544. Leopold LB and Maddock T (1953) The hydraulic geometry of stream channels and some physiographic implications. US Geological Survey Professional Paper 252: 57pp. Leopold LB and Wolman MG (1957) River channel patterns – braided, meandering and straight. US Geological Survey Professional Paper 282B: 39--85. Leopold LB, Wolman MG, and Miller JP (1963) Fluvial Processes in Geomorphology. 522pp. San Francisco: W.H. Freeman.

Lite SJ and Stromberg JC (2005) Surface water and ground-water thresholds for maintaining Populus-Salix forests, San Pedro River, Arizona. Biological Conservation 125: 153--167. Lobon-Cervia J (2003) Spatio-temporal dynamics of brown trout production in a Cantabrian stream: Effects of density and habitat quality. Transactions of the American Fisheries Society 132: 621--637. Lytle DA and Poff NL (2004) Adaptation to natural flow regimes. Trends in Ecology and Evolution 19(2): 94--100. Mack GH and Leeder MR (1998) Channel shifting of the Rio Grande, southern Rio Grande Rift; implications for alluvial stratigraphic models. Sedimentary Geology 117: 207--219. Malard F, Ferreira D, Dole0 dec S, and Ward JV (2003) Influence of groundwater upwelling on the distribution of the hyporheos in a headwater river flood plain. Archiv fu¨r Hydrobiologie 157: 89--116. Malard F, Uehlinge U, Zah R, and Tockner K (2006) Flood-pulse and riverscape dynamics in a braided glacial river. Ecology 87(3): 704--716. Malcolm IA, Soulsby C, Youngson AF, and Hannah DM (2005) Catchment-scale controls on groundwater–surface water interactions in the hydorheic zone: Implications for salmon embryo survival. River Research and Applications 21: 977--990. Maser C and Sedell JR (1994) From the Forest to the Sea: The Ecology of Wood in Streams, Rivers, Estuaries and Oceans. Florida: St. Lucie, Delray Beach. Mathews R and Richter BD (2007) Application of the Indicators of Hydrologic Alteration software in environmental flow setting. Journal of the American Water Resources Association 43: 1400--1413. McMahon TA and Finlayson BL (2003) Droughts and anti-droughts: The low flow hydrology of Australian rivers. Freshwater Biology 48(7): 1147--1160. Merritt DM and Cooper DJ (2000) Riparian vegetation and channel change in response to river regulation: A comparative study of regulated and unregulated streams in Green River basin, USA. Regulated Rivers: Research and Management 16: 543--564. Milhous RT (1982) Effect of sediment transport and flow regulation on the ecology of gravel-bed rivers. In: Hey RD, Bathurst JC, and Thorne CR (eds.) Gravel-Bed Rivers: Fluvial Processes, Engineering and Management, pp. 819--842. Chichester: Wiley. Milhous RT (1998) Modelling of instream flow needs: The link between sediment and aquatic habitat. Regulated Rivers: Research and Management 14(1): 79--94. Milly PCD, Wetherald RT, Dunne KA, and Delworth TL (2002) Increasing risk of great floods in a changing climate. Nature 415(6871): 514--517. Milner AM (1994) System recovery. In: Calow P and Petts GE (eds.) Rivers Handbook, pp. 76--98. Oxford: Blackwell. Mitro MG, Zale AV, and Rich BA (2003) The relation between age-0 rainbow trout (Oncorhynchus mykiss) abundance and winter discharge in a regulated river. Canadian Journal of Fisheries and Aquatic Sciences 60: 135--139. Moggridge HL and Gurnell AM (2009) Controls on the sexual and asexual regeneration of Salicaceae along a highly dynamic, braided river system. Aquatic Sciences 71(3): 305–317. Mollard JD (1973) Air photo interpretation of fluvial features. In: Proceedings of the 7th Canadian Hydrology Symposium, pp. 341–380. Monk WA, Wood PJ, Hannah DM, and Wilson DA (2007) Selection of river flow indices for the assessment of hydroecological change. River Research and Applications 23(1): 113--122. Monk WA, Wood PJ, Hannah DM, and Wilson DA (2008) Macroinvertebrate community response to inter-annual and regional flow regime dynamics. River Research and Applications 24: 988–1001. Monk WA, Wood PJ, Hannah DM, Wilson DA, Extence CA, and Chadd RP (2006) Flow variability and macroinvertebrate community response within riverine systems. River Research and Applications 22(5): 595--615. Montgomery DR (1999) Process domains and the river continuum. Journal of the American Water Resources Association 35: 397--410. Montgomery DR and Abbe TB (2006) Influence of logjam-formed hard points on the formation of valley-bottom landforms in an old-growth forest valley, Queets River, Washington, USA. Quaternary Research 65(1): 147--155. Morales Y, Weber LJ, Mynett AE, and Newton TJ (2006) Mussel dynamics model: A hydroinformatics tool for analyzing the effect of different stressors on the dynamics of freshwater mussel communities. Ecological Modelling 197: 448--460. Murray DR (1986) Seed dispersal by water. In: Murray DR (ed.) Seed Dispersal, pp. 49--85. San Diego, CA: Academic Press. Naiman RJ, Bunn SE, Nilsson CE, Petts GE, Pinay G, and Thompson LC (2002) Legitimizing fluvial ecosystems as users of water: An overview. Environmental Management 30(4): 455--467. Naiman RJ and De´camps H (eds.) (1990) The Ecology and Management of Aquatic– Terrestrial Ecotones. Paris: UNESCO.

Hydrology and Ecology of River Systems Naiman RJ, Lonzarich DG, Beechie TJ, and Ralph SC (1992) General principles of classification and the assessment of conservation potential in rivers. In: Boon PJ, Calow P, and Petts GE (eds.) River Conservation and Management, pp. 93--123. Chichester: Wiley. Nanson GC and Knighton AD (1996) Anabranching rivers: Their causes, character and classification. Earth Surface Processes and Landforms 21(3): 217--239. Nanson GC and Croke JC (1992) A genetic classification of floodplains. Geomorphology 4(6): 459--486. Newbold JD, Elwood JW, O’Neill RV, and Van Winkle W (1981) Nutrient spiralling in streams: The concept and its field measurement. Canadian Journal of Fisheries and Aquatic Sciences 38: 860--863. Newbold JD, Mulholland PJ, Elwood JW, and O’Neill RV (1982) Organic spiralling in stream ecosystems. Oikos 38(3): 266--272. Newson MD, Harper DM, Padmore CL, Kemp JL, and Vogel B (1998) A cost-effective approach for linking habitats, flow types and species requirements. Aquatic Conservation: Marine and Freshwater Ecosystems 8(4): 431--446. Newson MD and Newson CL (2000) Geomorphology, ecology and river channel habitat: Mesoscale to basin-scale challenges. Progress in Physical Geography 24: 195--217. Nilsson C, Andersson E, Merritt DM, and Johansson ME (2002) Differences in riparian flora between riverbanks and river lakeshores explained by dispersal traits. Ecology 83(10): 2878--2887. Nilsson C, Gardfjell M, and Grelsson G (1991) Importance of hydrochory in structuring plant communities along rivers. Canadian Journal of Botany Revue Canadienne de Botanique 69(12): 2631--2633. Nilsson C and Svedmark M (2002) Basic principles and ecological consequences of changing water regimes: Riparian plant communities. Environmental Management 30(4): 468--480. Olden JD and Naiman RJ (2009) Incorporating thermal regimes into environmental flows assessments: Modifying dam operations to restore freshwater ecosystem integrity. Freshwater Biology 55(1): 86--107. Olden JD and Poff NL (2003) Redundancy and the choice of hydrologic indices for characterizing streamflow regimes. River Research and Applications 19(2): 101--121. Orghidan T (1959) Ein neuer Lebensraum des unterirdischen Wassers: Der hyporheische Biotop. Archiv fu¨r Hydrobiologie 55: 392--414. Orsborn JF and Allman CH (eds.) (1976) Proceedings of the Symposium and Specialty Conference on Instream Flow Needs, I and II. Bethesda, MD: American Fisheries Society. Orth DJ (1987) Ecological considerations in the development and application of instream flow-habitat models. Regulated Rivers: Research and Management 1(2): 171--181. Osmundson DB, Ryel RJ, Lamarra VL, and Pitlick J (2002) Flow-sediment-biota relations: Implications for river regulation effects on native fish abundance. Ecological Applications 12(6): 1719--1739. Osterkamp WR (1978) Gradient, discharge and particle-size relations of alluvial channels in Kansas, with observations of braiding. American Journal of Science 278: 1253--1268. Padmore CL (1997) Biotopes and their hydraulics: A method for determining the physical component of freshwater habitat quality. In: Boon PJ and Howell DL (eds.) Freshwater Quality: Defining the Indefinable, pp. 251--257. Edinburgh: HMSO. Paetzold A, Schubert CJ, and Tockner K (2005) Aquatic terrestrial linkages along a braided-river: Riparian arthropods feeding on aquatic insects. Ecosystems 8: 748--759. Page KJ, Nanson GC, and Frazier PS (2003) Floodplain formation by oblique accretion on the Murrumbidgee River, Australia. Journal of Sedimentary Research 73: 5--14. Parasiewicz P (2001) MesoHABSIM: A concept for application of instream flow models in river restoration planning. Fisheries 26(9): 6--13. Parasiewicz P (2003) Upscaling: Integrating habitat model into river management. Canadian Water Resources Journal 28: 283--299. Parasiewicz P (2007) The MesoHABSIM model revisited. River Research and Applications 23(8): 893--903. Parasiewicz P (2008) Application of MesoHABSIM and Target Fish Community approaches to restoration of the Quinebaug River, Connecticut and Massachusetts, USA. River Research and Applications 24(4): 459--471. Parde´ M (1955) Fleuves et rivie`res. 224pp. Paris: Armand Colin. Pardo I and Armitage PD (1997) Species assemblages as descriptors of mesohabitats. Hydrobiologia 344: 111--128. Patel JA (2007) Evaluation of low flow estimation techniques for ungauged catchments. Water and Environment Journal 21(1): 41--46. Pettit NE and Froend RH (2001) Availability of seed for recruitment of riparian vegetation: A comparison of a tropical and a temperate river ecosystem in Australia. Australian Journal of Botany 49(4): 515--528.

267

Petts GE (1980) Long-term consequences of upstream impoundment. Environmental Conservation 7: 325--332. Petts GE (1984) Impounded Rivers. Chichester: Wiley. Petts GE (1996) Water allocation to protect river ecosystems. Regulated Rivers: Research and Management 12: 353--365. Petts GE (2009) Instream-flow science for sustainable river management. Journal of the American Water Resources Association 45: 1071–1086. Petts GE and Amoros C (eds.) (1996) Fluvial Hydrosystems, 308pp. London: Chapman and Hall. Petts GE and Bickerton MA (1994) Influence of water abstraction on macroinvertebrate community gradients within a glacial stream system: La Borge d’Arolla, Valais, Switzerland. Freshwater Biology 32: 375--386. Petts GE, Bickerton MA, Crawford C, Lerner DN, and Evans D (1999) Flow management to sustain groundwater-dominated stream ecosystems. Hydrological Processes 13(3): 497--513. Petts GE and Gurnell AM (2005) Dams and geomorphology: Research progress and future directions. Geomorphology 71: 27--47. Petts GE and Maddock I (1994) Flow allocation for in-river needs Rivers Handbook, pp. 289--307. Oxford: Blackwell. Petts GE, Nestler J, and Kennedy R (2006) Advancing science for water resources management. Hydrobiologia 565(1): 277--288. Petts GE, Sparks R, and Campbell I (2000) River restoration in developed economies. In: Boon PJ, Davies BR, and Petts GE (eds.) Global Perspectives on River Conservation, pp. 493--508. Chichester: Wiley. Poff NL, Allan D, Bain MB, et al. (1997) The natural flow regime: A paradigm for river conservation and restoration. BioScience 47(11): 769--784. Poff NL, Arthington AH, Bunn SE, et al. (2009) The ecological limits of hydrologic alteration (ELOHA): A new framework for developing regional environmental flow standards. Freshwater Biology 55(1): 147--170. Poff NL, Bledsoe BP, and Cuhaciyan CO (2006) Hydrologic variation with land use across the contiguous United States: Geomorphic and ecological consequences for stream ecosystems. Geomorphology 79: 264--285. Poff NL and Ward JV (1989) Implication of streamflow variability and predictability for lotic community structure: A regional analysis of streamflow patterns. Canadian Journal of Fisheries and Aquatic Science 46: 1805--1818. Poff NL and Zimmerman JKH (2009) Ecological responses to altered flow regimes: A literature review to inform environmental flows science and management. Freshwater Biology 55: 194–205. Poole GC (2002) Fluvial landscape ecology: Addressing uniqueness within the river discontinuum. Freshwater Biology 47: 641--660. Poole GC, O’Daniel SJ, Jones KL, et al. (2008) Hydrologic spiralling: The role of multiple interactive flow paths in stream ecology. River Research and Applications 24(7): 1018--1031. Poole GC, Stanford JA, Running SW, and Frissell CA (2006) Multiscale geomorphic drivers of groundwater flow paths: Subsurface hydrologic dynamics and hyporheic habitat diversity. Journal of the North American Benthological Society 25: 288--303. Power ME, Sun A, Parker G, Dietrich WE, and Wootton JT (1995) Hydraulic food-chain models. BioScience 45(3): 159--167. Pringle CM (2001) Hydrologic connectivity and the management of biological reserves: A global perspective. Ecological Applications 11(4): 981--998. Pringle CM (2003) What is hydrologic connectivity and why is it ecologically important? Hydrological Processes 17(13): 2685--2689. Pringle CM, Naiman RJ, Bretschko G, et al. (1988) Patch dynamics in lotic systems: The stream as a mosaic. Journal of the North American Benthological Society 7(4): 503--524. Raven PJ, Fox P, Everard M, Holmes NTH, and Dawson FH (1997) River habitat survey: A new system for classifying rivers according to their habitat quality. In: Boon PJ and Howell DL (eds.) Freshwater Quality: Defining the Indefinable?, pp. 215--234. Edinburgh: HMSO. Reiser DW, Ramey MP, and Wesche TA (1989) Flushing flows. In: Gore JA and Petts GE (eds.) Alternatives in Regulated River Management, pp. 91--138. Boca Raton, FL: CRC. Resh VH, Brown AV, Covich AP, et al. (1988) The role of disturbance in stream ecology. Journal of the North American Benthological Society 7: 443--455. Rice SP, Greenwood MT, and Joyce CB (2001) Tributaries, sediment sources and the longitudinal organization of macroinvertebrate fauna along river systems. Canadian Journal of Fisheries and Aquatic Sciences 58: 824--840. Richter BD, Baumgartner JV, Braun DP, and Powell J (1998) A spatial assessment of hydrologic alteration within a river network. Regulated Rivers: Research and Management 14(4): 329--340. Richter BD, Baumgartner JV, Powell J, and Braun DP (1996) A method for assessing hydrologic alteration within ecosystems. Conservation Biology 10(4): 1163--1174.

268

Hydrology and Ecology of River Systems

Richter BD, Baumgartner JV, Wigington R, and Braun DP (1997) How much water does a river need? Freshwater Biology 37(1): 231--249. Richter BD, Warner AT, Meyer JL, and Lutz K (2006) A collaborative and adaptive process for developing environmental flow recommendations. River Research and Applications 22(3): 297--318. RRA (2003) Environmental flows for river systems. River Research and Applications, Special Issue 19(5–6): 375--681. Sabaton C, Siegler L, Gouraud V, Bagliniere JL, and Manne S (1997) Presentation and first applications of a dynamic population model for brown trout (Salmo trutta L.): Aid to river management. Fisheries Management and Ecology 4: 425--438. Sabaton C, Souchon Y, Capra H, Gouraud V, Lascaux J-M, and Tissot L (2008) Longterm brown trout populations responses to flow manipulation. River Research and Applications 24(5): 476--505. Scarnecchia DL and Roper BB (2000) Large scale differential summer habitat use of three anadromous salmonids in a large river basin in Oregon, USA. Fisheries Management and Ecology 7: 197--209. Schramm HL and Eggleton NA (2006) Applicability of the flood-pulse concept to temperate floodplain river ecosystems: Thermal and temporal components. River Research and Applications 22: 543--553. Schumm SA (1977) The Fluvial System. 338pp. New York: Wiley. Schumm SA (1985) Patterns of alluvial rivers. Annual Reviews of Earth and Planetary Science 13: 5--27. Sedell JR, Richey JE, and Swanson FJ (1989) The river continuum concept: A basis for the expected ecosystem behavior of very large rivers? In: Dodge DP (ed.) Proceedings of the International Large River Symposium, Canadian Special Publication of Fisheries and Aquatic Sciences, vol. 106, pp. 49–55. Ottawa, ON: Canadian Government Publishing Centre. Sherrard JJ and Erskine WD (1991) Complex response of a sand-bed stream to upstream impoundment. Regulated Rivers: Research and Management 6: 53--70. Shiau J-T and Wu FC (2004) Assessment of hydrologic alterations caused by Chi-Chi diversion weir in Chou-Shui Creek, Taiwan: Opportunities for restoring natural flow conditions. River Research and Applications 20(4): 401--412. Shiau J-T and Wu FC (2006) Compromise programming methodology for determining instream flow under multiobjective water allocation criteria. Journal of the American Water Resources Association 42: 1179--1191. Smith JD (2004) The role of riparian shrubs in preventing floodplain unravelling along the Clark Fork of the Columbia River in Deer Lodge Valley, Montana. In: Bennett SJ and Simon A (eds.) Riparian Vegetation and Fluvial Geomorphology, Water Science and Application, pp. 71–85. Washington, DC: American Geophysical Union. Snelder TH and Biggs BJF (2002) Multiscale river environment classification for water resources management. Journal of the American Water Resources Association 38(5): 1225--1239. Snelder TH, Biggs BJF, and Woods RA (2005) Improved eco-hydrological classification of rivers. River Research and Applications 21(6): 609--628. Stalnaker CB (1979) The use of habitat structure preferenda for establishing flow regimes necessary for maintenance of fish habitat. In: Ward JV and Stanford JA (eds.) The Ecology of Regulated Streams, pp. 321--338. New York: Plenum. Stalnaker CB (1994) Evolution of instream flow habitat modelling. In: Calow P and Petts GE (eds.) Rivers Handbook, pp. 276--288. Oxford: Blackwell. Stalnaker CB, Bovee KD, and Waddle TJ (1996) Importance of the temporal aspects of habitat dynamics to fish population studies. Regulated Rivers: Research and Management 12(2–3): 145--153. Stanford JA, Lorang MS, and Hauer FR (2005) The shifting habitat mosaic of river ecosystems. Proceedings – International Association of Theoretical and Applied Limnology 29: 123--136. Stanford JA and Ward JV (1988) The hyporheic habitat of river ecosystems. Nature 335: 64--66. Stanford JA and Ward JV (1993) An ecosystem perspective of alluvial rivers: Connectivity and the hyporheic corridor. Journal of the North American Benthological Society 12: 48--60. Stanford JA, Ward JV, Liss WJ, et al. (1996) A general protocl for restoration of regulated rivers. Regulated Rivers: Research and Management 12: 391--413. Statzner B, Gore JA, and Resh VH (1988) Hydraulic stream ecology – observed patterns and potential applications. Journal of the North American Benthological Society 7(4): 307--360. Statzner B and Higler B (1986) Stream hydraulics as a major determinant of benthic invertebrate zonation patterns. Freshwater Biology 16: 127--139. Stewart G, Anderson R, and Wohl E (2005) Two-dimensional modelling of habitat suitability as a function of discharge on two Colorado rivers. River Research and Applications 21(10): 1061--1074. Stromberg JC, Beauchamp VB, Dixon MD, Lite SJ, and Paradzick C (2007) Importance of low-flow and high-flow characteristics to restoration of riparian vegetation along rivers in and south-western United States. Freshwater Biology 52(4): 651--679.

Tal M, Gran K, Murray AB, Paola C, and Hicks DM (2004) Riparian vegetation as a primary control on channel characteristics in multi-thread rivers In: Bennett SJ and Simon A (eds.) Riparian Vegetation and Fluvial Geomorphology, Water Science and Application, pp. 43–58., Washington, DC: American Geophysical Union. Tennant DL (1976) Instream flow requirements for fish, wildlife, recreation and related environmental resources. Fisheries Management and Ecology 1: 6--10. Tharme RE (2003) A global perspective on environmental flow assessment: Emerging trends in the development and application of environmental flow methodologies for rivers. River Research and Applications 19(5–6): 397--441. Thorp JH, Thoms MC, and Delong MD (2006) The riverine ecosystem synthesis: Biocomplexity in river networks across space and time. River Research and Applications 22: 123--147. Tickner D, Armitage PD, Bickerton MA, and Hall KA (2000) Assessing stream quality using information on mesohabitat distribution and character. Aquatic Conservation – Marine and Freshwater Ecosystems 10(3): 179--196. Tockner K, Malard F, and Ward JV (2000) An extension of the flood pulse concept. Hydrological Processes 14(16–17): 2861--2883. Tockner K and Stanford JA (2002) Riverine flood plains: Present state and future trends. Environmental Conservation 29(3): 308--330. Tooth S and Nanson GC (1999) Anabranching rivers on the Northern Plains of arid central Australia. Geomorphology 29: 211--233. Tooth S and Nanson GC (2000) The role of vegetation in the formation of anabranching channels in an ephemeral river, Northern Plains, arid central Australia. Hydrological Processes 14: 3099--3117. Townsend CR (1989) The patch dynamics concept of stream community ecology. Journal of the North American Benthological Society 8: 36--50. Truscott A-M, Soulsby C, Palmer SCF, Newell L, and Hulme PE (2006) The dispersal characteristics of the invasive plant Mimulus guttatus and the ecological significance of increased occurrence of high-flow events. Journal of Ecology 94: 1080--1091. Vadas RL and Orth DJ (2001) Formulation of habitat suitability models for stream fish guilds: Do the standard methods work. Transactions of the American Fisheries Society 130: 217--235. van den Berg JH (1995) Prediction of alluvial channel pattern of perennial rivers. Geomorphology 12(4): 259--279. Vannote RL, Minshall GW, Cummins KW, Sedell JR, and Cushing CE (1980) The river continuum concept. Canadian Journal of Fisheries and Aquatic Sciences 37: 130--137. Vogt K, Rasran L, and Jensen K (2004) Water-borne seed transport and seed deposition during flooding in a small river-valley in Northern Germany. Flora 199(5): 377--388. Vo¨ro¨smarty CJ and Sahagian D (2000) Anthropogenic disturbance of the terrestrial water cycle. BioScience 50(9): 753--765. Vo¨ro¨smarty CJ, Sharma KP, Fekete BM, et al. (1997) The storage and aging of continental runoff in large reservoir systems of the world. Ambio 26(4): 210--219. Wadeson LA (1994) A geomorphological approach to the identification and classification of instream flow environments. South African Journal of Aquatic Sciences 20: 1--24. Ward JV (1985) Thermal characteristics of running waters. Hydrobiologia 125: 31--46. Ward JV (1989) The four-dimensional nature of lotic ecosystems. Journal of the North American Benthological Society 8: 2--8. Ward JV, Malard F, Tockner K, and Uehlinger U (1999) Influence of ground water on surface water conditions in a glacial flood plain of the Swiss Alps. Hydrological Processes 13: 277--293. Ward JV and Stanford JA (Eds.) (1979) The Ecology of Regulated Streams. New York: Plenum Press. Ward JV and Stanford JA (1982) Thermal responses in the evolutionary ecology of aquatic insects. Annual Review of Entomology 27: 97--117. Ward JV and Stanford JA (1983b) The intermediate disturbance hypothesis: An explanation for biotic diversity patterns in lotic ecosystems. In: Fontaine TD and Bartell SM (eds.) Dynamics of Lotic Ecosystems, pp. 347--356. Ann Arbor, MI: Ann Arbor Science. Ward JV and Stanford JA (1983a) The serial discontinuity concept of lotic systems. In: Fontaine TD and Bartell SM (eds.) Dynamics of Lotic Ecosystems, pp. 29--42. Ann Arbor, MI: Ann Arbor Science. Ward JV and Stanford JA (1995) Ecological connectivity in alluvial river ecosystems and its disruption by flow regulation. Regulated Rivers: Research and Management 11: 105--120. Ward JV, Tockner K, Arscott DB, and Claret C (2002) Riverine landscape diversity. Freshwater Biology 47(4): 517--539.

Hydrology and Ecology of River Systems Webster JR, Wade JB, and Patten BC (1975) Nutrient recycling and the stability of ecosystems. In: Howell FG, Gentry JB, and Smith MH (eds.) Mineral Cycling in Southeastern Ecosystems, pp. 1--27. Springfield, VA: National Technical Information Service. Welcomme RL (1979) Fisheries Ecology of Floodplain Rivers. London: Longman. Welcomme RL, Winemiller KO, and Cowx IG (2006) Fish environmental guilds as a tool for assessment of ecological condition of rivers. River Research and Applications 22(3): 377--396. White MA, Schmidt JC, and Topping DJ (2005) Application of wavelet analysis for monitoring the hydrologic effects of dam operation: Glen Canyon Dam and the Colorado River at Lees Ferry, Arizona. River Research and Applications 21(5): 551--565. Williams GP and Wolman MG (1984) Downstream effects of dams on alluvial rivers. US Geological Survey Professional Paper 1286: 83.

269

Wood PJ, Hannah DM, Agnew MD, and Petts GE (2001) Scales of hydroecological variability within a groundwater-dominated stream. Regulated Rivers: Research and Management 17(4–5): 347--367. Wu F-C and Chou Y-J (2004) Tradeoffs associated with sediment-maintenance flushing flows: A simulation approach to exploring non-inferior options. River Research and Applications 20(5): 591--604. Young WJ, Chessman BC, Erskine WD, et al. (2004) Improving expert panel assessments through the use of a Composite River Condition Index – the case of the rivers affected by the Snowy Mountains hydro-electric scheme, Australia. River Research and Applications 20(6): 733--750. Zanoni L, Gurnell AM, Drake N, and Surian N (2008) Island dynamics in a braided river from an analysis of historical maps and air photographs. River Research and Applications 24(8): 1141--1159.

2.11 Hydrology and Biogeochemistry Linkages NE Peters, US Geological Survey, Atlanta, GA, USA JK Bo¨hlke, US Geological Survey, Reston, VA, USA PD Brooks, University of Arizona, Tucson, AZ, USA TP Burt, Durham University, Durham, UK MN Gooseff, Pennsylvania State University, University Park, PA, USA DP Hamilton, University of Waikato, Hamilton, New Zealand PJ Mulholland, Oak Ridge National Laboratory, Oak Ridge, TN, USA NT Roulet, McGill University, Montreal, QC, Canada JV Turner, CSIRO Land and Water, Wembley, WA, Australia & 2011 Elsevier B.V. All rights reserved.

2.11.1 2.11.2 2.11.3 2.11.3.1 2.11.3.1.1 2.11.3.2 2.11.3.3 2.11.3.4 2.11.3.4.1 2.11.3.4.2 2.11.3.5 2.11.4 2.11.5 2.11.6 2.11.7 2.11.8 2.11.9 2.11.10 References

Introduction Hydrological Pathways on Drainage Basin Slopes Mountain Environments Precipitation Snow Change in Storage Evaporation and Transpiration Stream Flow Nitrate isotopes in stream water Transit time and residence time Groundwater Recharge Within-River Processes Wetland Processes Lakes Groundwater Acidic Atmospheric Deposition – Acid Rain Summary and Future Considerations Additional Reading

2.11.1 Introduction Biota depend on water, energy, and nutrient transfers within and between ecosystems, which result in complex interactions between hydrology and biogeochemistry. The hydrological cycle and the variation in rates and magnitudes of water transfers along pathways in turn affect biogeochemical interactions. Nutrient uptake by biota varies markedly depending on availability of water, the pathways by which it moves through ecosystems, and the ecosystem type (aquatic, terrestrial), climate, and many geomorphological factors, such as slope and soil type. Variations in flow rates along pathways, reaction rates, composition (mineralogy, chemistry, and biology) and characteristics of interacting materials, chemical composition of the water, and temperature are major factors affecting biogeochemical processes. Most chemical cycles either affect or are affected by biological activity. Research since the early 1990s has revealed that even mobile and conservative elements, such as chlorine, are ¨ berg et al., 1996; White and actively cycled by biota (O ¨ berg, 2003; Lovett et al., 2005; O ¨ berg Broadley, 2001; O and Sande´n, 2005). However, the cycles of nutrients, carbon, and other biogeochemical components are intricately linked

271 272 274 275 275 276 276 276 278 278 280 282 285 287 291 293 294 295 295

to hydrology and biogeochemistry, and are the foci for most of the discussion in this chapter. Scientists have been challenged by the task of determining and quantifying the background processes that affect hydrological and biogeochemical linkages because human activities that accompany population growth and the associated requirements of obtaining and consuming natural resources have accelerated landscape changes (Peters and Meybeck, 2000; Meybeck, 2001; Peters et al., 2005). Human impacts on ecosystems are evident everywhere on Earth. For example, deforestation, channelization, dams and river regulation, land drainage, agriculture, energy generation, and urbanization and management of these activities have had major impacts on hydrology and biogeochemistry (Poff et al., 1997; Friedman et al., 1998; Peters and Meybeck, 2000; Meybeck 2001; Paul and Meyer, 2001; Meyer et al., 2005; Peters et al., 2005; Poff et al., 2006; Palmer et al., 2008; Peters, 2008). Human activities and resource management have evolved and it has been widely recognized that while point-source pollution has become more manageable, diffuse or nonpoint sources of pollutants are dominating contamination of ecosystems, and are much more difficult to identify, quantify, and control (Novotny, 2003; Campbell and Novotny, 2004; Loague and

271

272

Hydrology and Biogeochemistry Linkages

Corwin, 2005). Freshwaters are experiencing declines in biodiversity far greater than those in the most affected terrestrial ecosystems and human threats to global freshwater biodiversity and ecosystem services include overexploitation, water pollution, flow modification, destruction or degradation of habitat, and invasion by exotic species; the combined and interacting influences have resulted in population declines and range reduction of freshwater biodiversity worldwide (Dudgeon et al., 2005). Ecosystem processes, including water, nitrogen, carbon, and phosphorus cycling, changed more rapidly in the second half of the twentieth century than at any time in recorded human history. Human modifications of ecosystems have changed not only the structure of the systems (such as what habitats or species are present in a particular location), but their processes and functioning as well (Millenium Ecosystem Assessment, 2005). The capacity of ecosystems to provide services derives directly from the operation of natural biogeochemical cycles, which in some cases have been modified substantially. Ecosystem services are the human benefits provided by ecosystems. Furthermore, invasive and exotic species continue to affect ecosystems (Mooney and Hobbs, 2000; Mooney et al., 2005), and the increasing presence of recalcitrant endocrine disruptors and pharmaceuticals in stream water has the potential to alter ecosystems and change biogeochemical cycles (McMaster, 2001; Boyd et al., 2003; Stackelberg et al., 2004). Although our understanding of how invasive species affect ecosystem processes is not well understood (Gordon, 1998; Levine et al., 2003), researchers have reported a wide range of effects on hydrology and biogeochemistry, such as by invasive earthworms in northern temperate forests (Bohlen et al., 2004), invasive vegetation in riparian zones (De´camps et al., 2004) caused by damming and river regulation (Nilsson and Berggren, 2000), and invasive species in grasslands (Scott et al., 2001; Hook et al., 2004). However, it is beyond the scope of this chapter to provide detailed information about hydrological and biogeochemical linkages for these and the myriad of other human activities that affect the landscape. The objective of this chapter is to provide an overview of the linkages between hydrology and biogeochemistry in terrestrial and aquatic systems by tracking water flow from headwaters to rivers in larger drainage basins, including groundwater, wetlands, and lakes. The selection of foci was arbitrary and determined largely by the expertise of the co-authors, but provides continuity from a hydrological-cycle perspective and with a bias toward a northern temperate hydroclimate, a geographic region with some of the most detailed process research. To focus the discussion, the chapter is sectioned topically and these topics include hydrological pathways on drainage basin slopes, mountain environments, within-river (or in-stream) processes, wetlands, lakes, and groundwater (and groundwater–surface water interactions). In particular, this chapter provides a view of the linkages among the hydrosphere, biosphere, lithosphere, and chemosphere of processes that affect nutrient cycles, particularly nitrogen and carbon. In addition to the general discussion of nutrient cycling, an example is given of the effects of human activities on these linkages through the widespread impacts of acidic atmospheric deposition. Topics discussed in Chapters (see also Chapter 1.10 Predicting Future Demands for

Water and Chapter 1.09 Implementation of Ambiguous Water-Quality Policies) overlap with some of the material in this chapter, hence will not be discussed here in detail.

2.11.2 Hydrological Pathways on Drainage Basin Slopes

Although the river and the hill-side waste sheet do not resemble each other at first sight, they are only the extreme members of a continuous series, and when this generalization is appreciated, one may fairly extend the ‘river’ all over its basin and up to its very divides. Ordinarily treated, the river is like the veins of a leaf; broadly viewed, it is like the entire leaf (Davis, 1899).

The drainage basin (also known as watershed in the USA and catchment elsewhere) has long been recognized as the fundamental unit of analysis for the sciences of hydrology and geomorphology. It is usually a clearly defined and unambiguous topographic unit, which acts as an open system for inputs of precipitation and outputs of river discharge and evaporation (Chorley, 1971). The topographic, hydraulic, and hydrological unity of the drainage basin provided the basis of the morphometric stream ordering system of Horton (1945), as elaborated by Strahler (1964). Schumm (1977) had generalized sediment transport within the drainage basin into three zones: source area, transfer zone, and sediment sink or area of deposition. This is also a convenient subdivision for analyzing solute transport through the drainage basin, including in-stream cycling, and thus closely accords with the concepts of river continuum (Vannote et al., 1980) and nutrient spiraling (Webster and Patten, 1979). The source zone includes low-order headwater basins, which comprise most of the basin area, and where stream biogeochemical dynamics are primarily controlled by flushing of solutes and organic matter into the stream. Further downstream, in higher-order reaches, the channel becomes increasingly isolated from the surrounding land. Although concentrated flow may occasionally occur on hillslopes in rills and gullies and in the subsurface through cracks and pipes, there is generally a clear division between diffuse flow on slopes and concentrated flow in the river channel. The nature of these diffuse flows has important implications for the residence time of water in the catchment and the transit time of water moving to the stream channel. In headwater basins, the occurrence of runoff leads to sediment and solute removal from hillslopes. Therefore, detailing these processes, source areas, and pathways is relevant. Stream flow may be divided into base flow, that is, stream discharge that is not attributable to direct runoff from precipitation or melting snow and generally is maintained by groundwater discharge during rain- and snowmelt-free periods, and runoff (rainstorm and snowmelt) events. The physical characteristics of the soil and bedrock determine the pathways by which hillslope runoff will reach the channel. The paths taken by water (Figure 1) determine many of the characteristics of the landscape, human uses of the land, and strategies required for wise land-use management (Dunne and Leopold, 1978). There are essentially two models of storm

Hydrology and Biogeochemistry Linkages

273

Infiltration-excess overland flow

rm

flo w

Direct runoff from saturated soil

flo w

su Su b

rn tu Re flow

Bedrock

rfa

ce

sto

Soil

r te wa d n Grou

Figure 1 Schematic of the hydrologic pathways and connections between uplands and streams in headwater catchments.

Storm-runoff mechanisms

Partial-area model

Variable-source area model

Infiltration-excess or Saturation-excess overland flow Hortonian overland flow

Rapid throughflow of new water via macropores or pipes

Subsurface stormflow

Displacement of old water through the soil matrix

Figure 2 Relations among hydrological pathways/mechanisms delivering stormflow from hillslopes to streams.

runoff generation (Figure 2): the partial area and variablesource area models. Horton (1933, 1945) argued that storm runoff is mainly produced by infiltration excess-overland flow. When rainfall intensity exceeds the infiltration capacity of the soil, the excess begins to fill up surface depressions; once these are full, water overflows downslope and surface runoff begins. Horton claimed that infiltration-excess overland flow would occur uniformly across the catchment area, but Betson (1964) showed that this is not necessarily widespread even within a small basin. Betson proposed the partial area model of overland flow generation in which surface runoff is produced only from small areas of the basin during any given storm event. Despite its localized extent, infiltration-excess overland flow can generate large flood peaks and is often associated with soil erosion (Herwitz, 1986). Overland flow moves rapidly and

because its residence time is short, it is usually characterized as new water. Hewlett (1961) proposed the variable-source area model to describe runoff production in areas of permeable soil where subsurface runoff can account for much if not all the storm runoff leaving a basin. The production of significant quantities of subsurface storm flow (otherwise called ‘throughflow’ or ‘interflow’) requires the development of saturated conditions within the soil profile, in effect an ephemeral perched water table. Subsurface storm flow may be generated by movement of water through the soil matrix, by flow in macropores (large diameter conduits in the soil, created by agents such as plant roots, soil cracks, or soil fauna), or by a mixture of both (Beven and Germann, 1982). Given its longer residence time within the soil matrix, throughflow usually has a high solute content and therefore appears relatively old compared to precipitation, whereas macropore flow can be sufficiently rapid to retain the characteristics of new water as evidenced from laboratory (Wildenschild et al., 1994; McIntosh et al., 1999) and field studies (Richard and Steenhuis, 1988; Everts et al., 1989; Jardine et al., 1990; Luxmoore et al., 1990; Peters and Ratcliffe, 1998). If the soil profile becomes completely saturated, saturation-excess overland flow (Dunne and Black, 1970a) will be produced, consisting of a mixture of return flow (exfiltrating old soil water) and direct runoff (new rainfall unable to infiltrate the saturated surface). As the extent of the zone of saturation varies seasonally and during storms, Hewlett (1961) coined the phrase variable-source area to describe his model of storm flow generation. Subsurface storm flow dominates the storm hydrograph where deep-permeable soils overlie less-permeable soil or bedrock, and where steep hillslopes abut the stream. Soils may also contain an impeding layer with can cause a perched water table. Soil saturation is

274

Hydrology and Biogeochemistry Linkages

more likely to occur in soils of moderate hydraulic conductivity, in areas of reduced soil-moisture storage, and on lower slope angles. The role of topography is particularly important in determining where soil saturation occurs and is favored in hillslope hollows, in places where the soil profile is shallow, and at the foot of slopes especially where the slope becomes less steep. Considerable erosion can occur in these areas where groundwater discharges to the surface at the base of a hillslope, referred to as groundwater sapping (Higgins, 1984). Flow in stream channels reflects different source waters. In large seasonally snow-covered basins, such as the Fraser River in western Canada or rivers draining the Rocky Mountains, the Himalayas, or the Alps, snowmelt provides the main annualdischarge response. Depending on the nature of the aquifer, peak baseflow discharge may significantly lag behind precipitation inputs. Groundwater provides the prolonged recession flow associated with the falling limb of the annual hydrograph. Buttle (1994) reviewed processes responsible for conveying pre-event (old) water rapidly to the channel during storm events: groundwater ridging, translatory flow (pistonflow displacement of pre-storm groundwater), macropore flow, saturation-excess overland flow, kinematic waves, and release of water from surface storage. Not all processes occur in all catchments, of course, but whichever predominate, the focus seems inevitably on near-stream zones (Cirmo and McDonnell, 1997; Burt and Pinay, 2005), including upwelling and discharge of deeper groundwater (Winter, 1999; Cle´ment et al., 2003; Lischeid et al., 2007). Bishop et al. (2004) argued that the chemistry of water moving downslope is modified at any given point by soil chemistry, and riparian soils have a particularly important influence on stream-water chemistry because they are the last soils to contact the water before it discharges (Hooper et al., 1998; Buttle, 2005). Riparian zones are not chemically inert and, therefore, whatever the nature and sources of water flowing into the near-stream zone, it should rapidly acquire a chemical signature determined by the nature of the riparian zone (Cirmo and McDonnell, 1997) or mix with upwelling deeper groundwater (Cle´ment et al., 2003). Dilute (new) event water may reach the stream channel quickly as infiltration-excess overland flow, as direct runoff from saturated areas, or where macropores discharge at the channel bank. The extent to which the chemical composition of flow lines is reset within the near-stream zone depends on the residence time of water there, the size of the riparian aquifer storage, and the amount of new water moving through and mixing with the riparian-zone water. In addition, riparianzone vegetation can alter the hydrologic cycle (Figure 3), including water partitioning through plant physiology (uptake). In addition, the vegetation growth function and structure are affected by water quality (Cirmo and McDonnell, 1997; Tabacchi et al., 2000; De´camps et al., 2004). Because of variable contributions of water from various pathways, differences exist in the delivery of carbon and nutrients to streams during base flow compared to storm flow (Buffam et al., 2001). Much contemporary research is concerned with the ability of riparian zones to buffer rivers from upslope pollution inputs, nitrate in particular. Biological processes tend to affect nitrate more than most other solutes. Nitrate may be produced in the near-stream zone by

8 7 1

2

3 4

5 6 Figure 3 The main physical impacts of riparian vegetation on water cycling: 1, interaction with over-bank flow with stems, branches, and leaves (turbulence); 2, flow diversion by log jams; 3, change in the infiltration rate of flood waters and rainfall by litter; 4, increase of turbulence as a consequence of root exposure; 5, increase of substrate macroporosity by roots; 6, increase of the capillary fringe by fine roots; 7, stem flow (the concentration of rainfall by leaves, branches, and stems); and 8, condensation of atmospheric water and interception of dew by leaves. From Tabacchi E, Lambs L, Guilloy H, PlantyTabacchi AM, Muller E, and Decamps H (2000) Impacts of riparian vegetation on hydrological processes. Hydrological Processes 14(16–17): 2959–2976.

mineralization, and rising water tables can then flush this nitrate into the stream (Triska et al., 1989). However, nitrate may also be removed, temporarily by uptake and immobilization, or permanently by denitrification. For forests, stand age affects N-uptake rates and thus N transformation and leaching rates in soils (Stevens et al., 1994; Emmett et al., 1993). Note, however, that channel flow remains a mixture of different source waters, and if significant amounts of water bypass saturated riparian soils, either by flowing across the surface or through permeable strata below the floodplain alluvium, the riparian zone will be ineffective in removing nitrogen (Burt and Pinay, 2005). Contrasts in stream water nitrate response to similar inputs in seemingly comparable watersheds over various timescales also provide insight into the importance of coupling biogeochemical reactions and hydrological pathways (Reynolds et al., 1992; Christopher et al., 2008). Given the several potential mechanisms for stream flow generation, all of these mechanisms point to the fact that streams and groundwater are intricately linked (Winter et al., 1999).

2.11.3 Mountain Environments The wide ranges of elevation and aspect that characterize mountain environments result in tremendous variability in

Hydrology and Biogeochemistry Linkages

how hydrology and biogeochemical cycles are linked in space and time. Topographical complexity affects the amount of precipitation, solar radiation, temperature, and the lateral redistribution of water, resulting in highly heterogeneous biogeochemical processes. Energy and water balance at the land surface are intimately related to ecosystem productivity directly through transpiration, growth, acclimation, and assembly (McDowell et al., 2008), and indirectly through changes in surface physical characteristics, such as albedo and roughness (Bonan and Levis, 2006). Together, these factors affect both in situ biogeochemical reactions and the transport of biogeochemical solutes through the landscape and into surface water. The spatial variability in elevation, aspect, vegetation, soils, and precipitation, including snow cover, associated with mountain systems results in a high degree of both spatial and temporal variability in the coupling of hydrology and biogeochemistry. The coupling of hydrology and biogeochemistry in mountain environments can be observed by addressing a fundamental question, ‘‘What happens to precipitation?’’ (Penman, 1961), which can be evaluated by assessing the components of the water balance equation:

P ¼ DS þ E þ T þ Q þ R

ð1Þ

where P is the input of hydrometeors mainly precipitation, and also includes fog, rim ice, and cloud water, DS is the change in near-surface water storage, E is the evaporation/ sublimation, T is the transpiration, Q is the runoff, and R is the groundwater recharge. The units used for each waterbalance component are typically given in depth per unit time, such as millimeter per day or year, for a drainage/catchment/ watershed area. These terms are implicitly linked to biogeochemical cycling and suggest mechanisms for explicitly linking hydrology and biogeochemistry in mountain catchments. The remainder of the discussion on mountain environments is organized around these components of the water balance.

2.11.3.1 Precipitation Precipitation (P) typically increases with elevation, while temperature decreases (Barros and Lettenmaier, 1994; GarciaMartino et al., 1996). In mountain catchments, soils are wetter for longer at higher elevations for similar landscape positions (Band et al., 2001), for example,, riparian zones, stream channels, and hillslopes, and soil types as at lower elevations; but soils generally are thinner at higher elevations. Both increased soil-water availability and decreased temperature at high elevations reduce water stress on vegetation while simultaneously slowing soil heterotrophic activity, resulting in a smaller percentage of fixed carbon being respired from soils (Schlesinger, 1997). Thicker soils and higher temperatures at low elevations favor higher carbon storage and respiration. Fog and cloud water deposition may be a large percentage of annual water input to some forests in coastal ecosystems (Dawson, 1998; Klemm et al., 2005; Scholl et al., 2007) and to high-elevation forests (Lovett, 1984; Lovett and Kinsman, 1990; Clark et al., 1998; Heath and Huebert, 1999; Herckes et al., 2002; Scholl et al., 2007). Fog and cloud water typically have higher solute concentrations than precipitation (Lovett,

275

1984; Asbury et al., 1994; Reynolds et al., 1996; Weathers and Likens, 1997; Clark et al., 1998; Oyarzu´n et al., 2004) and therefore have higher solute-deposition per unit volume of water. Furthermore, fog and cloud water deposition at forest edges is notably higher than within the forest (Weathers et al., 1995; Ewing et al., 2009). The presence of fog or cloud water affects the plant physiology and biogeochemistry (Joslin and Wolfe, 1992; Bruijnzeel and Veneklaas, 1998; Dawson, 1998; Burgess and Dawson, 2004). Atmospheric washout (scavenging of aerosols and gases by hydrometeors including rain, snow, sleet, freezing rain, and hail (Pruppacher et al., 1983)) affects solute concentrations temporally during rainstorms with concentrations typically decreasing in time. Rainfall also washes off solutes concentrated by evaporation and dry atmospheric deposition that accumulate on vegetation producing much higher solute concentrations in throughfall and stemflow at the onset of rainstorms and decreasing thereafter (Peters and Ratcliffe, 1998).

2.11.3.1.1 Snow Snow is a special case of precipitation that has important implications for biogeochemistry in mountain environments. Seasonal snow cover has been shown to be an important hydrological control on biogeochemistry in many mountain catchments. Both dry and wet deposition are stored in winter snowpacks and released to soil and surface water in the spring (Jeffries, 1989; Peters and Driscoll, 1989; Bales et al., 1993; Williams et al., 1995). Snow cover also insulates soils from low air temperatures during winter (Peters, 1984; Brooks et al., 1996, 2005), providing an environment where soil microorganisms actively cycle carbon and nutrients during winter. Microbial biomass has been shown to reach annual maximum values under snow cover (Brooks et al., 1997, 1998), and these maximum values are also associated with changes in species composition (Lipson et al., 1999). Overwinter soil respiration may return 20–50% of the carbon fixed during the previous growing season to the atmosphere as CO2 (Brooks et al., 2005). Variability in the amount of winter CO2 loss is associated with the timing and amount of snow cover, soil frost, and labile-carbon availability (Brooks and Williams, 1999; Brooks et al., 1999a; Groffman et al., 1999; Grogan and Chapin, 1999; Groffman et al., 2001b). Similarly, overwinter nitrogen mineralization and immobilization in microbial biomass has been shown to be an important source of plant N at the initiation of the growing season. As with CO2 efflux, the magnitude of overwinter N cycling is related to the timing of snow cover and soil frost (Brooks et al., 1995; Groffman et al., 2001a). Consequently, natural variations in seasonal snow cover and climate change can have major impacts on soil processes (Edwards et al., 2007). In many forested mountain catchments, snow–vegetation– energy interactions define snow amount and timing of water availability to terrestrial ecosystems during the growing season (Liston, 2004; Molotch and Bales, 2005; Liston and Elder, 2006; Molotch and Bales, 2006; Veatch et al., 2009). Net ecosystem carbon uptake is dominated by fixation during the snowmelt season when water is not limiting (Monson et al., 2002), and soil respiration and N cycling are also strongly

276

Hydrology and Biogeochemistry Linkages

controlled by the timing of snow cover (Brooks et al., 1997; Grogan and Chapin, 1999; Groffman et al., 2001 b; Grogan and Jonasson, 2003). Finally, snowmelt transports large, but variable amounts of nutrients, organic matter, and inorganic carbon out of the terrestrial ecosystem (Rascher et al., 1987; Hornberger et al., 1994; Campbell et al., 1995; Boyer et al., 1997; Brooks and Williams, 1999; Brooks et al., 1999b; Heuer et al., 1999; Campbell and Law, 2000). Solute concentrations vary markedly in meltwater and preferentially elute during snowmelt (Johannessen et al., 1975; Johannessen and Henriksen, 1978; Tranter et al., 1986; Berg, 1992). As observed for rainfall and throughfall, early meltwater is high in dissolved solutes that have accumulated in the snowpack and are concentrated in brines around snow crystals during snow metamorphism (Tranter and Jones, 2001). The combination of active abiotic and biotic processes under the snowpack and elution of solutes in meltwater may result in high stream concentrations during snowmelt (Peters and Leavesley, 1995; Mitchell, 2001; Driscoll et al., 2005). The snowpack is biologically active (Hoham and Duval, 2001), particularly when liquid water is present and temperatures and temperature ranges are optimum (Jones, 1999; Hoham and Duval, 2001). Light is also an important constraint to the distribution and reproduction of some biological components such as snow algae (Hoham et al., 1998, 2000), and these algae in turn affect the timing and magnitude of snowmelt (Hoham and Duval, 2001). In addition to snowpack biological activity, physical characteristics, such as atmospheric conditions (wind, vapor pressure, and temperature), can affect gaseous transfers of nutrients (CO2 and NOx) to and from snowpacks (Pomeroy et al., 1999; Tranter and Jones, 2001).

2.11.3.2 Change in Storage The first-order controls on near-surface storage of water (DS) are soil and slope. Soil characteristics and landform are major controls on the partitioning of precipitation into infiltration (Philip, 1967), storage (Beven, 1982), recharge (Gee and Hillel, 1988; Phillips et al., 2004), runoff (see Section 2.11.1), and stream flow (Dunne and Black, 1970b). By affecting the amount and timing of water availability to vegetation and soil microbes, these characteristics interact with climate to affect both potential productivity (carbon fixation) and soil microbial processes. Soils are typically shallower, less weathered, and coarser on ridge lines and higher elevations, and deeper and finer in depressions. Fine-textured soils retain water, reduce diffusion, and result in an environment where anaerobic or facultatively anaerobic processes dominate biogeochemical cycling (Pusch et al., 1998; Hill and Cardaci, 2004). Similarly, topographic depressions and areas of hydrological convergence have higher soil moisture, which may result in either episodic or continuous anaerobic conditions. As aerobic respiration decreases so does carbon mineralization, which can result in increases in denitrification and methanogenesis (Jones et al., 1995; Baker et al., 1999). Several factors affect denitrification across time and space as shown schematically in Figure 4 and discussed with respect to each of the environments presented in this chapter.

2.11.3.3 Evaporation and Transpiration Although the absolute magnitude of evaporation and transpiration (E and T) is controlled by climate, the partitioning between E and T is controlled by vegetation water use, and thus is directly linked to carbon fixation and input into the ecosystem. Vegetation serves both as the carbon pump, bringing organic matter into ecosystems, and as the water pump, removing water from ecosystems, and thus controls the amount of chemical energy in organic matter and the amount of water in the environment. Organic matter and water in turn are the primary controls on soil biogeochemical processes, affecting carbon balance, nutrient cycling, and mineral weathering (Amundson et al., 2007). Both the type and amount of vegetation within an ecosystem vary predictably with elevation and latitude (Holland and Steyn, 1975), as the effect of increasing elevation can generally be equated to that of increasing latitude. Similarly, vegetation varies with aspect in relation to both temperature and water availability (Grace, 1989). For example, the tree line often is limited by temperature and typically extends to a higher elevation on south-facing slopes than north-facing slopes in the Northern Hemisphere (Treml and Banas, 2008). In seasonally water-stressed environments, north-facing and topographically shaded slopes may have more dense vegetation cover and different species assemblages than southfacing or nonshaded slopes (Zhang et al., 2009). These differences arise from the interaction between energy and water. By changing the structure and productivity of the land surface, which control the input of carbon to soil and plant N demand from soil, the potential magnitude of soil biogeochemical processes is affected. Moreover, the root zone and rhizosphere in the soil are particularly active biologically and biogeochemically and the presence of roots affects hydrology. The rhizosphere, which is the dynamic interface among plant roots, soil microbes and fauna, and the soil, is attributed to the evolution of soil, that is, the alteration of primary and secondary soil minerals (Cardon and Whitbeck, 2007). The redistribution of soil moisture by plants is not typically considered in biogeochemical-cycling research or in typical hydrological-process assessment (Burgess et al., 1998; Caldwell et al., 1998; Jackson et al., 2000; Meinzer et al., 2001). Plant physiologists with the aid of stable isotopes have made major advances in understanding how plants use water (Ehleringer and Dawson, 1992; Dawson, 1993; Emerman and Dawson, 1996; Dawson and Ehleringer, 1998; Moreira et al., 2000). A general pattern is that roots transport water from deep moist horizons to shallow drier surface-soil horizons, particularly at night, by a process called hydraulic lift (Caldwell et al., 1998). It is not surprising that plant redistribution of water would predominate in dry land, as observed in arid areas, for example, for phreatophytes (Hultine et al., 2003) and sage brush (Richards and Caldwell, 1987; Caldwell and Richards, 1989). However, plant redistribution of water has also been documented for sugar maples in northern temperate forests (Dawson, 1996; Emerman and Dawson, 1996), for three species representing each of three canopy niches in Amazonia (Oliveira et al., 2005), and for blue oaks in the foothills of the Sierra Nevada (Millikin Ishikawa and Bledsoe, 2000). Roots not only move water from depth to surface soils, but can

Space mm

cm

m

km

1000 km

s

Aquatic sediments and

min

soil microsites

Permanently

h

stratified aquatic systems

Low-O2 river reaches

d Time

Periodically anoxic waters and associated sediments mo

yr Periodically anoxic soils cent Groundwater

Marine oxygen-minimum zones OMZs

(a)

Denitrification system types Group A. Diffusion dominated

Group B. Advection dominated

Group C. Periodic anoxia

Distinct nitrification and denitrification layers dictated by O2 concentrations

Stable suboxic/anoxic water or sediment mass into which nitrate is advected

Periodic anoxia caused by soil moisture changes or water stratification

mm-cm scale (Meters for aquatic systems)

Time O2 Nitrification Denitrification

Low-O2 region

Advective input NO3

Dry soils or well-mixed water High O2 nitrification

Low O2

Water-saturated soils or stratified water Low O2 denitrification

High O2

• Water saturated soils and sediments (lakes, wetlands, rivers, estuaries, and continental shelf) • Soil microsites • Anoxic bottom water (lakes, enclosed seas, and fjords)

• OMZs, groundwater, and river reaches

• Terrestrial soils • Periodically stratified lakes, estuaries, and continental shelf

(b)

Figure 4 (a) Classification of systems according to the magnitude of temporal and spatial separation between nitrification and denitrification. Diffusion-dominated systems are indicated in gray, advection-dominated systems are indicated with heavy outlines, and systems with periodic anoxia are indicated by dashed lines. (b) Schematic groupings of systems according to mechanism of nitrate delivery to denitrification zone. Vertical profiles of oxygen concentrations are indicated. Adapted from Seitzinger S, Harrison JA, Bo¨hlke JK, et al. (2006) Denitrification across landscapes and waterscapes: A synthesis. Ecological Applications 16(5): 2064–2090.

278

Hydrology and Biogeochemistry Linkages

also move water from surface to depth as documented for perennial grasses in northwestern South Africa (Schulze et al., 1998); inverse hydraulic lift may be an important mechanism to facilitate root growth in dry desert soil layers below the upper soil zone to which precipitation penetrates. In addition, meteorological conditions through the effect of shade and clouds on soil water potential can affect hydraulic lift (Williams et al., 1993). Furthermore, water hydraulically lifted by large trees may be used by smaller trees and shrubs as shown by isotope tracers (Dawson, 1996); the authors concluded that small trees and shrubs use soil water and large trees use groundwater. Clearly, these plant–water relations should have major implications for biogeochemistry. However, there is a lack of hard evidence linking biogeochemical cycles to water redistribution by plants, although some of these studies cited above discuss the implications of water redistribution on nutrient cycling.

2.11.3.4 Stream Flow Stream water integrates the myriad of catchment hydrological and biogeochemical processes. The catchment is a natural spatial domain to study the coupling of hydrological and biogeochemical cycles and simultaneous measurements of catchment discharge (Q) and hydrochemistry capture the integrated catchment hydrological and biogeochemical behavior. Hydrological and biogeochemical processes are tightly coupled at the Earth’s surface and hydrological fluxes within catchments provide an opportunity to balance material fluxes at defined spatial and temporal scales. Catchments link the atmosphere, plants, soil surface, subsurface, groundwater, and streams through the convergence and interaction of material and energy flows. Quantifying the variability in the contributions of different water sources to stream flow allows inferences to be drawn about hydrological pathways and biogeochemical processes in these source areas (Peters and Driscoll, 1987b; Hooper et al., 1990; McDonnell et al., 1991). For N export, the magnitude of individual responses to a runoff event appears to be related to soil microbial processes, while seasonal to decadal trends are related to vegetation demand (Bormann and Likens, 1994; Likens and Bormann, 1995). A common observation across mountain catchments is that large fractions of both C and N are exported in stream flow in response to rainstorms and snowmelt runoff (Hood et al., 2006), presumably because of changes in routing through the catchment (McGlynn et al., 2003; Bishop et al., 2004; McGuire and McDonnell, 2006). Shallower soils, higher hydraulic conductivity, and proximity to the stream channel increase the likelihood of increased export of biogeochemically active solutes during runoff events.

2.11.3.4.1 Nitrate isotopes in stream water The ability to differentiate sources of nitrate in streams has advanced with the application of multiple isotope techniques for d15N, d18O, and d17O in nitrate (Durka et al., 1994; Kendall, 1998; Burns and Kendall, 2002; Mayer et al., 2002; Fukada et al., 2003; Michalski et al., 2004; Kendall et al., 2007). Nitrate isotopic variations can be related to sources

of nitrate, and they also are modified by subsequent reactions such as denitrification and assimilation. Isotope studies emphasizing either source variations or cycling processes may be complicated by these overlapping effects, but they can be useful in some situations. For example, temporal isotopic studies have provided important constraints on the transmission of atmospheric nitrate through watersheds. Oxygenisotope data indicate that atmospheric nitrate, commonly, is only a minor component of stream nitrate during a range of flow conditions, including snowmelt events, highlighting the importance of nitrification sources in runoff (Burns and Kendall, 2002; Campbell et al., 2002; Buda and DeWalle, 2009). Exceptions include peak flows during storm events, especially in watersheds containing impervious ground cover such as urban areas, where atmospheric nitrate can be a substantial component of stream nitrate (Kendall et al., 2007). Spatially distributed isotopic data in stream networks can provide supporting evidence for varying nitrate sources in different subwatersheds (Mayer et al., 2002; Lindsey et al., 2003). For example, during moderate base-flow conditions in the predominantly agricultural Mahantango WE-38 watershed in Pennsylvania, USA (Lindsey et al., 2003), small streams with low nitrate concentrations and low d15N values from forested upland watersheds joined other streams with higher nitrate concentrations and higher d15N values draining cropland, whereas a few streams with exceptionally high nitrate concentrations and high d15N values drained areas with animal feedlots or pastures. Nitrate apparently was transmitted through the watershed without major isotopic modification after infiltrating through soils. In contrast to the above, it has been suggested that isotopic indicators could be incorporated into conceptual, analytic, or numerical models of flow, transport, and denitrification to evaluate nitrogen losses within watersheds (Sebilo et al., 2003). This approach requires information or assumptions about the upscaling properties of isotope-fractionation effects from micro- to diffuse scales (scale at which mixing of water and solutes is relatively complete). In principle, the regionalscale status of diffuse denitrification could be evaluated by monitoring d 15 NNO3 and d 18 ONO3 in stream flow to determine the catchment-scale status and dynamics of denitrification. A hypothetical example of such an approach is illustrated in Figure 5, which shows predictions of the d 15 NNO3 isotopic behavior of nitrate in stream flow as distributed within an idealized model framework. The stable isotope predictions are based on the riparian nitrate model (RNM), which operates as a filter (plug-in) module within a node-link catchment-scale model (E2); E2 is capable of simulating the hydrological behavior of catchments (Rassam et al., 2006, 2008). The RNM– E2 modeling framework has been augmented with Rayleigh fractionation algorithms at each node within the model domain to track the isotope shifts because of denitrification. The modeling example considers a simple homogeneous riparianzone soil with decreasing available carbon and related microbial activity with depth and denitrification only occurring in the riparian-zone soils. As denitrification proceeds in the groundwater prior to discharge, the residual nitrate becomes relatively enriched in 15N as nitrate concentration decreases (Marriotti et al., 1988; Kendall, 1998). Figure 5 illustrates the RNM–E2 simulated dynamics of nitrate concentrations and

Hydrology and Biogeochemistry Linkages 20

26 Quickflow Baseflow N load

24 22 20

18 16

16

12

14

10

12

8

10 8

6

6

4

4

N load kg s−1

14

18 Flow (m3 s−1)

279

2

2 0

0

−2 2 Ju l-8 2 26 -J ul -8 15 2 -A ug -8 2

-8

65

δ15N-NO3 for 50% denitrification

60

δ15N-NO3 for 90% denitrification

55

δ15N-NO3 for 95% denitrification

50

δ15N-NO3 for 99% denitrification

45 δ15N-NO3

6-

-J un

16

27

-M

ay -

82

82

2

ay M

7-

2

pr -8

-8

-A

ar -M

17

(a)

28

8-

M

ar -8

2

−2

40 35 30 25 20

Zero enrichment

15 10 5

-8

Ju l-8 2 26 -J ul -8 15 2 -A ug -8 2

2

6-

2

-J un

16

82 27

-M

ay -8

2

ay -

7-

M

pr -8

2

-A

17

ar -8 -M

28

(b)

8-

M

ar -8

2

0

Figure 5 Linkages between stream nitrogen (N) loading, variation of d15N of nitrate (d15N–NO3) corresponding to varying amounts of dentrification, and hydrology of the Maroochy catchment southeast Queensland, Australia during the wet season 1982; (a) temporal variations in hydrologic response through quick and baseflow components, and N loading; and (b) concurrent predictions of the streamwater d15N–NO3. From Rassam DW, Knight JH, Turner J, and Pagendam D (2006) Groundwater surface water interactions: Modelling denitrification and d15NNO3–d18ONO3 fractionation during bank storage. In: Institution of Engineers Australia (ed.) Proceedings of the 30th Hydrology and Water Resources Symposium, pp. 157–161. Launceston, TAS, Australia, 4–7 December 2006. Sandy Bay, TAS: Conference Design.

d15N as the stream flow source switches between base flow and rapid delivery of new water during storm flow. The d15N of the stream nitrate increases as the extent of denitrification in the catchment increases. Similar results could be generated for d18ONO3 depending on the denitrification fractionation factors used (Bo¨ttcher et al., 1990; Fukada et al., 2003; Granger et al., 2008). In these models, isotopic variations in nitrate sources are assumed to be negligible.

2.11.3.4.2 Transit time and residence time The distributed hydrological response to precipitation encompasses the spatial and temporal variations of water fluxes in landscapes, and, therefore, is directly related to the variability in biogeochemical cycling described earlier. The hydrological response in a mountainous catchment is controlled largely by the near-surface landscape properties (landform and soil characteristics) that function as hydrological filters (Meybeck and

280

Hydrology and Biogeochemistry Linkages

Vo¨ro¨smarty, 2005); yet, variations are often nonlinear and difficult to represent (Beven and Germann, 1982; Troch et al., 2003). Mean hydrological transit time, or the mean age of water discharging to a stream channel, is an important hydrological descriptor, but its relation to storage, flow pathways, mixing, and sources of water is complex and model dependent (Maloszewski et al., 1983; McGuire et al., 2005; McGuire and McDonnell, 2006; Soulsby et al., 2006). Feedbacks between vegetation structure, soils, biogeochemistry, and landform development result in a range of transit-time distributions throughout a catchment, that is, rates of water movement from the catchment to the channel at various locations along the channel (McGuire and McDonnell, 2006; McDonnell et al., 2007). For example, based on a simple agedistribution model (Tetzlaff et al., 2007), slope was found to be the dominant control of mean transit time in steep Scottish catchments, but soil permeability was a more important control in flat lowland catchments (Tetzlaff et al., 2009). In another study, aspect appeared to be a dominant control on mean transit time, with north-facing slopes having longer transit times than south-facing slopes (Broxton et al., 2009). From a biogeochemical perspective, the residence time, that is, length of time that water remains in a catchment

Soil column: cross section

before it becomes stream flow, has a pronounced effect on potential solute concentrations and export. As water moves from upland areas in the catchment, downslope interactions with other water sources may occur in areas of convergent flow (Anderson and Burt, 1978), promoting physical mixing of waters, exchange of solutes, and increased rates of oxidation– reduction (redox) reactions or hot spots (Fisher et al., 1998; McClain et al., 2003). These hot spots are where high rates of nutrient modifications occur, for example, in riparian zones (Burt and Pinay, 2005), the hyporheic zone, and wetlands (Figure 6) with associated chemical transformations (Figure 7). Episodic transport during hydrological events, such as snowmelt or heavy rain, can reduce the importance of hot spots by reducing residence time, resulting in increased nutrient flux (Boyer et al., 1997; Stanley et al., 1997). Even after water has entered a stream channel, its residence time may be greater than that predicted by stream velocity because of hyporheic exchange, that is, the movement of stream water into the subsurface and back to the stream at a location downstream. The typical rough texture of headwater mountain-basin geomorphology (e.g., pool-riffle sequences) often drives exchange of water through the bed and riparian sediments (Harvey and Bencala, 1993; Kasahara and

Catena: cross section

1m 10 m

(b)

(a)

Upland−stream toposequence: cross section Upland Riparian

100 m

(c)

Sub-basin: plan view Large wetlands 100 km

(e)

Larger river basin: plan view Wetland River/upland contact 10 km

(d)

Figure 6 Hot spots of denitrification occur at multiple spatial scales. (a) Hot spots in a meter of soil may occur along root channels where moisture and organic-matter content are high. (b) Topographic depressions that accumulate organic matter and retain moisture may be hot spots within a catena. (c) Along a toposequence from upland to river, the soil–stream interface may represent a hot spot where high-nitrate groundwater intercepts organic-rich soils. (d) At the scale of sub-basins, the occurrence of hot spots may be dictated by the spatial configuration of upland–wetland or upland–river contact zones. (e) The percentage of land occupied by wetlands determines denitrification hot spots at the scale of large river basins. Adapted from McClain ME, Boyer EW, Dent CL, et al. (2003) Biogeochemical hot spots and hot moments at the interface of terrestrial and aquatic ecosystems. Ecosystems 6(4): 301–312.

Hydrology and Biogeochemistry Linkages

Desert

Terrace

Stream

Wetland

Cropland

281

Forested hillslope

O1–O5 Water table R2 R1 Oxic Anoxic

R5

R6 R5 R4 R3 R2

R2

R3 R4 R6

Reduction reactions

R3 R4 R5 R6 R1

Oxygen reduction Denitrification Mn(IV) reduction

Iron reduction Sulfate reduction

R5

CH4 fermentation

R6

R1

R1 R2

R3

R4

O1 Org matter Oxidation O2 Sulfide Oxidation O3

Iron Oxidation

O4 Oxidation reactions –20

–10

Nitrification

O5 Mn(II) Oxidation 0 pe

10

20

Figure 7 A schematic diagram of some common oxidation and reduction reactions in riparian zones and their relative redox potentials at pH 7. Reaction sequences in waters recharging from, or discharging to, a stream depend on the origin (initial redox status) of the water, the composition of the substrate, and the interaction with biota (bacteria, roots, etc.). Adapted from Dahm CN, Grimm NB, Marmonier P, Valett HM, and Vervier P (1998) Nutrient dynamics at the interface between surface waters and groundwaters. Freshwater Biology 40(3): 427–451; and Appelo CAJ and Postma D (2007) Geochemistry, Groundwater, and Pollution, 2nd edn., 649pp. Rotterdam: AA Balkema.

Wondzell, 2003). The changing morphology along a river network exerts a strong control on gradients that drive exchange, and therefore, the amount of water that flows through the hyporheic zone (Kasahara and Wondzell, 2003). This exchange not only increases stream water residence time in the basin, but also moves nutrients and other solutes into the

subsurface: (1) fueling biogeochemical processes in the shallow subsurface around streams, (2) providing a subsurface habitat that is a mix of surface and groundwater conditions, and (3) generating patches of varying conditions on the streambed in downwelling (where stream water enters the bed) and upwelling (where hyporheic water returns to the

282

Hydrology and Biogeochemistry Linkages

Catchment-scale gradients

Headwaters

Montane floodplain Piedmont floodplain Coastal plain

(a)

Reach-scale gradients Downwelling zone Upwelling zone

(b)

Biofilms

Sediment-scale gradients

Hyporheic invertebrates

(c)

Figure 8 Lateral diagrammatic view of the hyporheic zone (HZ) at three spatial scales. At the catchment scale (a), the hyporheic corridor concept predicts gradients in relative size of the HZ, hydrologic retention, and sediment size (126). At the reach scale (b), upwelling and downwelling zones alternate, generating gradients in nutrients, dissolved gases, and subsurface fauna. At the sediment scale (c), microbial and chemical processes occur on particle surfaces, creating microscale gradients. Arrows indicate water flow paths. From Boulton AJ, Findlay S, Marmonier P, Stanley EH, and Valett HM (1998) The functional significance of the hyporheic zone in streams and rivers. Annual Review Ecology and Systematics 29: 59–81.

channel) locations (Figure 8). Thus, hyporheic exchange is an important hydrological process that directly contributes to many biological functions such as macroinvertebrate population dynamics (Boulton et al., 1998; Malard et al., 2002) and salmonid spawning (Baxter and Hauer, 2000), and biogeochemically important reactions (Figure 5) such as denitrification and retention of dissolved organic carbon (DOC; Baker et al., 1999), water-temperature buffering (Arrigoni et al., 2008), and buffering of heavy metal transport (Fuller and Harvey, 2000).

2.11.3.5 Groundwater Recharge Groundwater recharge (R) is directly linked to mineral weathering rates through the delivery of water, DOC, and CO2 to the subsurface; the export of weathering products to surface water; and CO2 release to the atmosphere (Richey et al., 2002). Consequently, groundwaters with longer transit times typically

contain lower DOC concentrations because of microbial degradation, and higher concentrations of conservative alkalinity-associated DIC (Szramek and Walter, 2004), base cations, and silica (Rademacher et al., 2001). Climate affects water–rock interactions. For example, increasing temperature and high rainfall tend to increase rates of chemical weathering (White and Blum, 1995; White and Brantley, 2003). In addition, weathering is controlled by lithology (Meybeck, 1987; Bluth and Kump, 1994; Meybeck and Vo¨ro¨smarty, 2005). In the long term, linkages or feedbacks between biota and earth materials have modified the near-surface environment of Earth or ‘critical zone’ (Brantley et al., 2007), and, in turn, the chemistry of streams and rivers shows the evidence of biological processes (Amundson et al., 2007). Groundwater, including recharge and discharge, is an important topic and is presented in a separate section, and is also discussed in each of the following sections where it interfaces with the main topic of the section.

Hydrology and Biogeochemistry Linkages

2.11.4 Within-River Processes Biogeochemical processes within streams and rivers can result in high rates of nutrient uptake, chemical transformation, and release of dissolved materials to water. These biogeochemical processes are largely the result of organisms, such as bacteria and fungi, algae, and higher plants, attached to hard surfaces or to organic and inorganic sediments on the streambed. In larger rivers, biogeochemical processes associated with suspended algae and microbes attached to suspended particles can be important. These processes can significantly alter the flux and chemical form of several biologically active solutes, particularly carbon, nitrogen, and phosphorus. Most of the organisms responsible for biogeochemical processes in streams and rivers are stationary, associated with the streambed, yet solutes taken up or released to water are under strong advective forces of downstream water flow. Thus, nutrient cycling has a distinctive spatial or longitudinal component along the axis of stream flow. Biological communities generally change from dominance of shredders, associated with leaf litter and woody debris accumulations in turbulent, higher-velocity streams in headwaters, to collector–gatherers in larger more quiescent streams with lower gradients downstream. Nutrient spirals tend to lengthen from upstream where streambed-surface:water-volume ratios tend to be high to downstream where surface:volume ratios are lower (Figure 9). The concept of nutrient spiraling was proposed as a framework to study nutrient cycling in streams and explicitly considers the simultaneous processes of biological uptake, transformation, or remineralization and hydrological transport downstream (Webster and Patten, 1979; Newbold et al., 1983). As an example of the nature of these reactions and interactions, a schematic of inorganic nitrogen transformations and interactions with streambed biota and hyporheic zone is shown in Figure 10 (Peterson et al., 2001). Several metrics quantifying nutrient spiraling have been defined, including uptake length – the average distance traveled by a nutrient atom in water from where it enters the stream to where it is taken up by biota. Methods for field measurement of uptake length have been developed, including the experimental addition of isotopic nutrient tracers, such as 15N for studies of nitrate and ammonium uptake and denitrification (Newbold et al., 1981; Stream Solute Workshop, 1990; Mulholland et al., 2000; Peterson et al., 2001; Bo¨hlke et al., 2004; Mulholland et al., 2004; Bo¨hlke et al., 2009). Distinctive temporal patterns characterize within-river biogeochemical processes. These patterns are related to seasonality in biological processes and in hydrology. Biological seasonality is largely controlled by the regulation of inputs of light and organic matter by terrestrial ecosystems bordering streams and rivers. In streams draining deciduous forests, nutrient-uptake rates often have two peaks each year: (1) early spring prior to leaf-out when uptake and growth rates of attached algae are high because of high light levels reaching the stream and (2) autumn after leaf-fall when uptake rates by bacteria and fungi associated with leaf decomposition are high (Roberts and Mulholland, 2007). In small, heavily shaded streams, summer is often a period of relatively low rates of nutrient uptake, despite high water temperatures, because algal growth is severely limited by low light levels and activity

283

of heterotrophic microbes is limited by the lack of easily decomposable organic matter in the streambed. However, streams that do not have dense forest canopies (e.g., those in more arid climates and agricultural landscapes) often have late spring or summer peaks in nutrient uptake and cycling because light levels are high and stream flows are more stable (Arango et al., 2008). Late spring and summer peaks in nutrient uptake may also be common in larger rivers because of high light levels and longer water-residence times under lower flows that permit development of larger communities of algae in the water column; however, little is known about rates and temporal dynamics of nutrient uptake in large rivers, that is, with drainage areas greater than 250 000 km2. There is considerable spatial variation in within-river biogeochemical processes, largely controlled by the hydrological regime, channel morphology, catchment land use, and characteristics of stream-bank (riparian) vegetation (Tabacchi et al., 2000). For example, desert streams in monsoonal climates can develop very high rates of nutrient uptake and cycling as algal communities develop during long periods of low, stable stream flow after the monsoon season ends (Grimm et al., 2005). In addition, within-river seasonal variations in macrophyte growth and related geomorphology can affect sediment respiration, nitrification, and denitrification rates (Duff et al., 2002). Furthermore, Duff et al. (2002) showed that as the rooted aquatic macrophyte communities matured, pore water became chemically reduced and nutrient levels increased by one to two orders of magnitude above background in the root zone. These levels were significantly higher than those found in either groundwater or surface water, indicating that streambeds can serve as a nutrient reservoir. Streams with very flashy hydrographs tend to have lower rates of biogeochemical processes because attached organisms are frequently scoured from the streambed during high flows. Cross-site studies have been particularly valuable for identifying broad-scale controls on nutrient uptake and cycling. Stream discharge is often the strongest predictor of nutrient-uptake length in streams, with longer uptake lengths (lower rates of uptake relative to downstream transport) under higher discharge (Peterson et al., 2001; Marti et al., 2004). In relatively undisturbed catchments, land use and riparian vegetation are also important determinants of nutrient uptake, but indirectly via their effects on light regime and stream primary productivity and nutrient inputs (Tabacchi et al., 2000; Hall et al., 2009). Stream algae and microbes are able to increase rates of uptake with increasing nutrient concentrations, although they become somewhat less efficient at removing nutrients from water as concentrations increase (Dodds et al., 2002). As discussed previously, subsurface (hyporheic) zones within streambed sediment accumulations are important sites for biogeochemical processes (Dahm et al., 1998). Water exchange between surface and subsurface zones and between the main channel and backwaters is an important mechanism for increasing rates of nutrient cycling (Triska et al., 1989; Jones and Holmes, 1996), and particularly for rates of denitrification (Duff and Triska, 1990; McMahon and Bo¨hlke, 1996; Mulholland et al., 2009). In-stream processes can be important for the retention of nutrients within river networks and landscapes, thus reducing the potential for eutrophication and harmful algal blooms in

284

Hydrology and Biogeochemistry Linkages

Shredders Grazers Predators

Microbes

1 Collectors

Trout

2

Periphyton

Smallmouth bass

Coarse particulate matter

Collectors

Microbes

Shredders

Fine particulate matter

3 Vascular hydrophytes

Predators Grazers

Stream size (order)

4

5

Perch

Periphyton

6 Fine particulate matter 7

Coarse particulate matter

Fine particulate matter

8 Phytoplankton

Predators

9

10 11

Microbes Collectors

Catfish Zooplankton

12 Relative channel width Figure 9 Nutrient spiraling along the river continuum. In general, nutrient spirals are produced by the simultaneous processes of nutrient cycling (uptake from water by biota and subsequent release back to water) and downstream transport. Adapted from Vannote RL, Minshall GW, Cummins KW, Sedell JR, and Cushing CE (1980) The river continuum concept. Canadian Journal of Fisheries and Aquatic Sciences 37(1): 130–137.

Hydrology and Biogeochemistry Linkages

Streamflow input (NH4+ and NO3−)

Net NH4+ regeneration

Stream bottom (hyporheic zone)

Direct nitrification

Streamflow output (NH4+ and NO3−)

Assimilatory NO3− uptake

Biota Reuptake

Groundwater discharge (NH4+ and NO3−)

NO3−

NH4+

Assimilatory NH4+ uptake

Net NO3− regeneration

285

NH4+

Mineralization

Reuptake

Indirect nitrification

NO3− Denitrification

N2

Figure 10 Simplified schematic diagram of selected processes affecting inorganic N cycling in streams. Biota includes bacteria distributed within the sediments. Although not shown, N species can transfer between stream and hyporheic zone without interacting with biota. Modified from Peterson BJ, Wollheim WM, Mulholland PJ, et al. (2001) Control of nitrogen export from watersheds by headwater streams. Science 292(5514): 86–90.

estuaries and coastal waters. Streams receiving relatively low nutrient inputs can retain more than half of their inputs over a 1-km stream length (Peterson et al., 2001). In addition, increasing nitrogen retention by in-stream processes in landscapes recovering from past disturbances has been reported (Bernhardt et al., 2005). Nutrient cycling is a serial process in streams and rivers, and large cumulative stream length results in long residence times and potentially high rates of nutrient retention in river networks (Wollheim et al., 2006; Alexander et al., 2007; Mulholland et al., 2008). Although small, shallow streams in the network can have particularly high rates of nutrient uptake because of high surface:volume ratios (Alexander et al., 2000), nutrient uptake can also be high in larger streams and rivers (Ensign and Doyle, 2006; Alexander et al., 2007; Tank et al., 2008). However, high nutrient inputs can overwhelm the capacity of biological-uptake processes within streams, resulting in low retention efficiency and large losses to downstream ecosystems (Mulholland et al., 2008; Bo¨hlke et al., 2009). Humans have had large impacts on within-river biogeochemistry. Agriculture and urbanization have been among the most widespread effects on streams. Agriculture often results in stream channelization and other modifications that reduce geomorphologic and hydrodynamic complexity and organic-matter storage and thus the rates of biogeochemical processes; in effect, farmland streams become more of a conduit and less of a barrier to runoff (Burt and Pinay, 2005) and this loss of landscape structure may account, for example, for persistently high concentrations of nitrate in river water (Burt et al., 2008). However, during extended periods of low stable flow in summer, nutrient uptake rates can be high because of high level of light and nutrient concentrations (Bernot et al., 2006). Urbanization can also result in substantial changes to in-stream biogeochemical processes because of many of the same morphological, hydrodynamic, and organic-matter impacts as agriculture, although greater extent of impervious surfaces can result in much flashier

urban-stream hydrographs (Boyd et al., 1993; Finkenbine et al., 2000; Rose and Peters, 2001). Rapid runoff or flashiness generally reduces nutrient-uptake rates in urban streams (Paul and Meyer, 2001; Groffman et al., 2004; Meyer et al., 2005; Walsh et al., 2005) and riparian zones (Groffman et al., 2002). In contrast, modifications to stream networks in heavily urbanized areas, such as detention basins and artificial lakes, can enhance nutrient uptake and retention (Grimm et al., 2005).

2.11.5 Wetland Processes Wetlands are defined differently by country and agency of use, but all definitions recognize the persistence of near-saturated conditions at or above the mineral sediments, hydric soils, and the presence of plants adapted for generally saturated conditions (Mitsch and Gosselink, 2007). Wetlands occur in all climatic and geographic regions of the world but they are more prominent in regions where precipitation exceeds potential evaporation and topographically flat areas where drainage rates are slow. However, given a supply of water, for example, rivers, streams, and groundwater, they can occur even in the most arid regions. The dominant physical factor of wetlands is the presence and persistence of near-saturated conditions. Whether the wetland is in the tropics or the high Arctic, waterlogged conditions are a function of the hydrology of the catchments in which the wetland is located, but the presence of a wetland also alters the hydrology of a catchment. Some wetlands provide temporary storage attenuating flood peaks and sustaining flows during drier periods (Mitsch and Gosselink, 2007); however, wetlands that also accumulate partially decomposed plants, for example, peat, have a poor ability to attenuate runoff because of their hydraulic properties (Bay, 1969; Holden and Burt, 2003; Holden et al., 2006). The role of wetlands in the hydrology and biogeochemistry of catchments is determined, in part, by the position of the

286

Hydrology and Biogeochemistry Linkages

wetland in the landscape. Given topographic depressions and an excess of precipitation over evapotranspiration, wetlands can exist in the headwaters of catchments, in valley floors, or on plateaus adjacent to the drainage divides. While these wetlands might seem isolated from the catchment, their outflow provides water downstream, particularly during periods of storm runoff (Holden and Burt, 2003). The hydrological interaction between streams and wetlands can be very complex, particularly for wetlands adjacent to streams in higherorder drainage basins with low stream gradients. In particular, groundwater can become important in creating saturated conditions that are needed to sustain wetlands (Winter and Woo, 1990). The isolated wetland’s hydrology is mainly a function of precipitation and antecedent water storage, making runoff strongly event related (Verry et al., 1988; Evans et al., 1999). In contrast, wetlands that are well connected to groundwater can have a less variable storage dynamic on the short term, for example, wetlands located at break points on hillslopes where they receive groundwater discharge (Winter, 1999). The magnitude and rate of water input to a wetland relative to the wetland’s storage capacity control residence time, and residence time plus the mixture of biological and chemical reactants determine the biogeochemical dynamics. For example, valley-bottom riparian wetlands can receive groundwater discharge that far exceeds precipitation input and this leads to relatively constant levels of water storage and rapid turnover of stored water (Roulet, 1990). Residence time also can be greatly increased by near-shore processes that effectively cycle wetland water from surface water to groundwater and back again. This is because of evapotranspiration (ET) drawing the water table below wetland stage, reversing hydraulic gradients, and allowing wetland water to flow into groundwater troughs that ring the wetlands, a common characteristic of prairie wetlands (Rosenberry and Winter, 1997). In contrast, water exchange in some wetlands, such as raised peat bogs, is confined to a thin hydrologically active layer near the surface (Ingram, 1978) and leads to a twocompartment system – one compartment with a short residence time (hours to days) and the other compartment a deeper groundwater with long residence times (hundreds of years) (Fraser et al., 2001). Water-exchange rates are an important factor affecting wetland biogeochemistry because they affect the magnitude and supply rate of chemical inputs  (e.g., DO, SO2 4 , and NO3 ) relative to the supply of reactants (e.g., Fe2þ, Mn2þ, NH4þ, organic matter, and microbes) in the wetland. For example, headwater and isolated wetlands receive their chemical inputs from precipitation alone; hence, they tend to be nutrient-poor systems, and if they accumulate peat, the surface vegetation can become isolated from the source of minerals in the underlying substrate and result in acidic conditions (Damman, 1986; Wilcox et al., 1986). In contrast, wetlands that receive water that has contacted other land covers and soils, such as marshes, valley-bottom swamps, or prairie pot holes, can receive a large influx of nutrients and cations, which can result in mineral and nutrient-rich productive ecosystems (Bedford et al., 1999). In wetlands, the temporal and spatial variations in chemical composition and biogeochemical transformations can be evaluated using chemical thermodynamics with a particular

emphasis on redox conditions (Hedin et al., 1998). Because of saturated conditions, oxygen is limited in most wetlands. The diffusion of oxygen in water is 10 000 times slower than that in air. Consequently, if there are processes that consume oxygen, such as decomposition of organic matter, the wetland substrate becomes progressively more reduced. In wetlands with short water-residence times, that is, the water is renewed regularly, and wetlands where the surface is flooded, oxic conditions prevail, at least near the surface; biogeochemistry of these wetlands is dominated by oxygen, nitrate, and iron reduction. However, when residence times are longer and oxygen consumption exceeds supply, wetlands become progressively more reduced, and sulfate reduction and eventually methanogenesis become common (Reddy and DeLaune, 2008). In addition to the importance of hydrologic fluxes of oxygenated waters for oxygen input to wetlands, wetland plants oxygenate their roots, and differences in the ability of wetland plants to aerate their submerged tissues under different flooding regimes play a major role in controlling plant distribution (Sorrell et al., 2000; Pezeshki, 2001). Temporal and spatial variations in saturation are important in controlling temporal and spatial dynamics of wetland biogeochemistry. Spatial and temporal dynamics of the inputs and the distribution of plants are also important controls on wetland biogeochemistry. The wetland setting is based on hydrology, the wetland salinity is based on climate, and biogeochemical response is based on a combination of the two. Recently, biogeochemists have begun to refer to locations on the landscape that show steep redox gradients, which have high biogeochemical transformation rates, as hot spots and the times when redox conditions change quickly at a location as hot moments (McClain et al., 2003). This conceptualization works well across many scales in wetland settings. For example, in unsaturated wetland sediments, the interface between saturated pores and adjacent air-filled pores could be a hot spot for chemical transformations. At a larger scale, transformations from oxidizing conditions upgradient of a riparian wetland to reducing conditions within a wetland are also likely hot spots. A hot moment may occur when the wetland water table rises rapidly during a hydrological event resulting in a large decrease in oxygen availability. Wetland ecosystems can also be viewed as a wetland continuum (Euliss et al., 2004), similar to the stream-continuum concept (Vannote et al., 1980). The concept places wetlands in two dimensions (Figure 11); one in relation to groundwater interaction, that is, recharge and discharge, and the other with respect to climate condition, that is, atmospheric water, from dry or drought conditions to wet or flood/deluge conditions. The wetland continuum provides a framework for organizing and interpreting biological data by incorporating the dynamic changes these systems undergo as a result of normal climatic variation. There are many examples of specific linkages between biogeochemistry and hydrology in wetlands. In almost all settings, carbon availability is the main driver of wetland biogeochemistry. This is reflected in the large accumulation of organic matter commonly observed in wetland soils, and in the importance of wetlands as a major source of DOC downstream (Hinton et al., 1998; Freeman et al., 2001; Worrall et al., 2004; Roulet et al., 2007; Nilsson et al., 2008) and

Hydrology and Biogeochemistry Linkages

287

Drought

Hydrologic relation to atmospheric water

Deluge

The wetland continuum

Recharge

Hydrologic relation to groundwater

Discharge

Terrestrial perennials

Wetland annuals

Robust wetland perennials

Terrestrial annuals

Early-season wetland perennials

Submersed wetland perennials

Figure 11 The wetland continuum, a wetland classification based on hydrology with respect to groundwater interactions from recharge to discharge and climate conditions from drought to deluge. Potential plant communities in wetlands at four discrete points along this axis are depicted. Adapted from Euliss NH, Jr., LaBaugh JW, Fredrickson LH, et al. (2004) The wetland continuum: A conceptual framework for interpreting biological studies. Wetlands 24: 448–458.

emissions of CO2, methane (CH4), and N gases (N2 and N2O). Wetland CH4 and N2O fluxes are extremely high compared to those from other landscapes (Bowden, 1986; Robertson, 2001; Svensson et al., 2001; Rosenberry et al., 2006). Wetland CH4 fluxes have been linked to water-table depths and temperature (MacDonald et al., 1998), which is the basis for a one-dimensional methane flux model (Figure 12) for natural wetlands (Walter and Heimann, 2000; Walter et al., 2001). Wetlands can accumulate carbon in the form of peat deposits and export water with high DOC concentrations (Moore, 2003); therefore, area of wetlands is often strongly correlated with rates of DOC export (Dillon and Molot, 1997; Xenopoulos et al., 2003). However, this correlation is not universal even in landscapes with significant wetland cover (Frost et al., 2006). Even in catchments where wetlands are only a small fraction of the catchment area, they can still have a profound effect on catchment biogeochemistry producing relatively high DOC concentrations in drainage waters, because they are often the last point of contact before the water enters a stream or river (Bishop et al., 2004). In this case, the role of the wetland can be quite dynamic depending on temporal variations in the hydrological connection of the

wetland with the adjacent hillslope (McGlynn et al., 2003; Burt and Pinay, 2005). Wetlands can be more effective in removing nutrients from circulating waters than lakes or rivers (Saunders and Kalff, 2001), and retaining constituents from atmospheric deposition, but removal is not a universal conclusion. Riparian wetlands, in particular, can be quite effective in removing and retaining nutrients and nitrogen in particular (Jansson et al., 1994), but the dynamics and hydrological setting play a critical role in the effectiveness of nutrient removal (Cirmo and McDonnell, 1997). Nutrients retained under one condition, such as periods when the water table is receding, can be rapidly mobilized upon rewetting, and the retention can be much less throughout the year based on an event or on a seasonal basis (Devito et al., 1989). Wetlands, through dynamic coupling with uplands and the atmosphere, can sometimes act as biogeochemical hot spot sources instead of sinks. For example, wetlands are sources of methyl-mercury (St. Louis et al., 1996; Babiarz et al., 1998), and the hydrological coupling of wetlands to upland sources of water (Branfireun et al., 1996; St. Louis et al., 1996; Galloway and Branfireun, 2004) and sources of sulfur, either through hydrological input or by atmospheric deposition

Hydrology and Biogeochemistry Linkages

CH4 emission

Climate

Soil surface

Rooting depth

CH4 production

NPP

Diffusion

CH4 concentration

Soil temperature

Oxic soil

Ebullition

CH4 oxidation

Water table

Atmosphere

Vegetation

Plant-mediated transport

288

Anoxic soil

Relative pore space

Soil depth Figure 12 Schematic of the one-dimensional methane model. The processes leading to methane emission to the atmosphere occur in the soil between soil depth and soil surface. Methane production takes place in the anoxic soil below the water table; the methane production rate depends on soil temperature and net primary productivity (NPP). Methane oxidation occurs in the oxic soil above the water table and depends on temperature. The model calculates methane concentrations in each (1 cm thick) soil layer. Transport occurs by diffusion through water/air-filled soil pores, ebullition to the water table, and plant-mediated transport from layers above the rooting depth. From Walter BP, Heimann M, and Matthews E (2001) Modeling modern methane emissions from natural wetlands, 1. Model description and results. Geophysical Research Letters 106(D24): 34189–34206.

(Branfireun et al., 1999), is a critical factor. In addition, the reducing conditions in wetlands are active areas for redox reactions involving iron, aluminum, and manganese. In summary, the key link between hydrology and biogeochemistry in wetlands originates from the dynamics of water storage and related residence time on the supply of oxygen and chemical reactants.

2.11.6 Lakes Figure 13 shows the major sources, transformations, and sinks of nutrients in lakes. There are three major initial points of entry of nutrients into lakes: surface-water inflows, groundwater inflows, and atmospheric inputs, including wet and dry deposition. Surface-water inflows may include both point sources and nonpoint sources, while atmospheric inputs and most groundwater contributions are nonpoint sources. Occasionally, other nonpoint sources of nutrients, such as guano from abundant bird populations (O’Sullivan, 1995), direct wastewater or pollutant inputs, or salmon-spawning migrations, may make significant contributions to the total lakeinput nutrient load (Naiman et al., 2002). In the latter case, the loss of sockeye salmon cadavers from some oligotrophic lakes in British Columbia, Canada, as a result of human effects on spawning runs, was determined to have significantly affected the total load of phosphorus (Stockner and MacIsaac,

1996). An active lake-fertilization program was used in several of these lakes to stimulate primary productivity and initiate a trophic cascade that would ultimately provide improved growth conditions for juvenile salmon (Stockner and MacIsaac, 1996). While reduced levels of primary production in lakes – sometimes termed (re) oligotrophication – may occasionally occur in highland regions or in response to major nutrient load-reduction strategies, the recent history of human influences on lakes has been characterized by nutrient concentrations that are substantially higher than natural or background concentrations. These high nutrient concentrations have led to eutrophication and some of its undesirable consequences such as harmful algal blooms, oxygen depletion of bottom waters, and, occasionally, fish kills. The species and transformations of nutrients around the point of entry of an inflow to a lake may be highly important in determining the biogeochemical effects of the inflow. Nutrients in groundwater entering a lake may pass through alternating zones of oxidizing and reducing conditions – the latter are often prevalent in the organic-rich sediments deposited in sheltered areas of the lake bed (Schuster et al., 2003). This heterogeneous environment can stimulate a diverse microbial flora associated with the rich array of microenvironments and redox conditions. Reducing conditions can (1) stimulate denitrification of nitrate when this nutrient is present, (2) lead to a buildup of ammonium as nitrification is inhibited, (3) result in dissolution of phosphate

Hydrology and Biogeochemistry Linkages

289

Groundwater inflows

Stream inflow Stream inflow

Point source

Dry atmospheric deposition Wet atmospheric deposition N fixation Denitrification

Periphyton

Trophic cascade

Macrophytes Sedimentation

Recycling

Sedimentation

Figure 13 Sources, sinks, and transformations of nutrients in a lake (not including outflows).

in association with reduction of oxidized forms of iron and manganese, and, (4) under strongly reducing conditions, may lead to sulfate reduction and methanogenesis (Reeburgh, 1983). Thus, the rich microbial consortia associated with the land–water transition can strongly influence the availability and species of nutrients, including carbon, nitrogen, and phosphorus. Processes in the littoral zone of lakes also can profoundly alter the composition of surface inflows. This zone may support a rich benthos and can include emergent and submerged macrophytes and periphyton that may have a large demand for nutrients. Dense beds of submerged macrophytes can also create quiescent conditions suitable for sedimentation of particulate forms of nutrients and associated inorganic sediments (Madsen et al., 2001). The effects of large surface inflows may be strongly dependent on their water-column insertion depths, that is, the depth at which a surface inflow enters the water column of the lake because of density differences, which in turn are regulated by the relative temperature of the inflow and water column, or occasionally also by the salinity or sediment concentration of the inflow (see Chapter 1.08 Managing Agricultural Water). In the case of a thermally stratified lake, an inflow that is warmer than lake water into which it intrudes will promote a surface overflow in which inorganic nutrients in the inflow will be available in the pelagic zone to the resident microscopic suspended plants (i.e., phytoplankton). An inflow that is cooler and therefore denser than water throughout the entire water column of the lake is likely to create an underflow, in which nutrients may not be immediately available to the phytoplankton resident in

the photic zone. When these inflows are large relative to the lake volume, they can have important secondary effects such as oxygenation of bottom waters (Hamilton et al., 1995). Many lake inflows have a temperature intermediate between those of the surface and bottom of the water column, and therefore create an interflow that can propagate horizontally at discrete depths in stratified lakes. Thus, nutrients, suspended sediments, contaminants, and microbes (notably pathogens) that are often present at much higher concentrations in storm flow, may be rapidly dispersed across a stratified lake in an interflow (Chung et al., 2009). In very simple terms, a box-type model can be used to describe steady-state concentrations of nutrients in a lake in which losses are because of sedimentation and outflows. This type of model has been widely used to describe lake-water concentrations of phosphorus (TPlake), but is not commonly used for nitrogen because atmospheric transformations (i.e., N fixation and denitrification) and related gaseous transfers between the lake and the atmosphere are more difficult to quantify. Vollenweider (1969) was first to apply the model in the form

TPlake ¼

L zðr þ sÞ

ð2Þ

where L is the areal loading rate of TP, z is the mean lake depth, r is the hydraulic flushing rate given by the inflow discharge divided by the lake volume (i.e., r ¼ 1/tw) where tw is the residence time, and s is a first-order decay rate for TP to account for sedimentation losses. Calculations are typically based on annualized values. Various extensions and

290

Hydrology and Biogeochemistry Linkages

simplifications of this model have been made, for example, s can be approximated by 10 ðmÞ=z, and Vollenweider and Kerekes (1982) produced the following modified model:

TPin TPlake ¼ 1:55 pffiffiffiffiffi0:82 1  tw

ð3Þ

where TPin is the inflow total phosphorus concentration. Equation (3) is based on data for 87 lakes, which show hydraulic flushing rate (or residence time) to be the primary factor driving differences in total phosphorus concentrations between the inflows and the lake. Many, mostly empirical, relationships in turn have been used to derive annual mean and peak concentrations of phytoplankton chlorophyll a and primary production as well as Secchi disk transparency from TPlake (Vollenweider and Kerekes, 1982). The focus on predictive models for total phosphorus reflects a long-held paradigm that phosphorus generally limits productivity in freshwater ecosystems (Likens, 1972; Schindler et al., 2008) and that increases in this nutrient can contribute to lake eutrophication (Lean, 1973). An argument has been made that a shortfall in nitrogen supply compared with the requirements for balanced growth – the Redfield ratio (Sterner and Elser, 2002) – can be compensated for by heterocystous cyanobacteria (blue-green algae) that will fix atmospheric nitrogen (i.e., N2 dissolved in water) when this nutrient becomes limiting. Earlier work on this subject (Lean, 1973; Smith, 1983) has remained topical but has recently been reignited by Schindler et al. (2008) who declared that controlling nitrogen inputs alone could exacerbate eutrophication by increasing the dominance of N-fixing cyanobacteria and the probability of harmful algal blooms. In contrast, recent research suggests that human activities have changed biogeochemical dynamics (Bergstro¨m and Jansson, 2006; Elser et al., 2009a, 2009b) and that phytoplankton biomass yield in most of the lakes in the Northern Hemisphere was limited by N in their natural state. Furthermore, there are advocates for control on both nitrogen and phosphorus loads (Lewis and Wurtsbaugh, 2008) on the basis that N fixation often fails to compensate sufficiently for N limitation in lake phytoplankton, that experimental systems manipulated with additional nutrients have often been found to be similarly controlled by N and P, and that high background loads of P that saturate demand necessarily dictate that another nutrient will be limiting (Lewis et al., 2008). Debate about the relative merits of N versus P control will surely continue into the foreseeable future, but considerations should also be given to the connectivity of inland waters to estuarine and coastal waters for which N limitation is the norm. In some cases, silica has been reported as a limiting nutrient for diatom production (Tilman et al., 1982). In contrast, only rarely is inorganic carbon supply limiting to primary production. These cases may arise for lakes that are poorly buffered and where high rates of photosynthesis remove carbon dioxide and raise pH to a level where bicarbonate (pHE7–9) or even carbonate (pHEZ10) predominate; the latter form is not available to plants, while only some plants can take up bicarbonate (Wetzel, 2001). The models given by Equations (2) and (3) do not reveal the mechanisms by which nutrients are regenerated and transformed within lakes. As a result of surface wave action,

internal waves, or unidirectional currents, the lake bed may be subject to water motions that can disturb pore water and resuspend particulate material, thus increasing concentrations of sediments and nutrients in the overlying water column (Hamilton and Mitchell, 1997). Dissolved nutrients may also be recycled from bottom sediments to the water column as a result of concentration gradients between these two media. These gradients may be enhanced by reducing conditions in the bottom sediments and sometimes in bottom waters, which result in dissolution of iron- or manganese-bound forms of phosphorus and a buildup of ammonium as nitrification shuts down. The pioneering work of Mortimer (1941, 1942) pointed to the key role of dissolved oxygen (DO) and redox potential in waters overlying the sediments. Thus, conditions that stimulate sediment nutrient releases through the benthos also control the benthic macroinvertebrate communities, which burrow deeply into layered sediments and accelerate nutrient cycling through bioturbation and fecal production (Covich et al., 1999). DO concentrations in bottom waters of deep lakes are closely linked to the availability of labile organic matter, the duration of density stratification, and the pool of DO in the bottom waters. Lakes may thermally stratify for periods of minutes to years, creating vertical density gradients that persist only temporarily in shallow lakes, seasonally (monomictic or dimictic lakes) in deeper systems, and not at all in permanently ice-covered lakes or at high altitude (amictic lakes) (Lampert and Sommer, 2007). These mixing patterns, of which there are several variants, dictate the renewal periods of oxygen to bottom waters and therefore play a key role in determining nutrient-release rates from bottom sediments based on their oxidation status. A high rate of supply of organic matter to the bottom waters (hypolimnion) of a stratified lake can completely remove DO from this layer and occurs in deep, eutrophic lakes. In contrast, in oligotrophic lakes that mix seasonally, DO generally remains present in bottom waters between periods of mixing when oxygen is renewed to levels approximating saturation, and nutrientregeneration rates from bottom sediments are markedly lower than in deep eutrophic lakes. Eutrophic lakes in which the hypolimnion becomes anoxic may support high rates of denitrification and can also emit considerable quantities of both methane and nitrous oxide (Seitzinger et al., 2006). Another situation relevant to eutrophication is when iron and manganese sequester phosphorus under oxic conditions, removing more P to the sediments than in lakes without high iron and manganese concentrations (Dean et al., 2003). Relatively high groundwater discharge to the lake from lithologies containing high amounts of iron and manganese can produce high concentrations of iron and manganese in the lake, that is, a combined lithology and hydrological control on P removal. Redox-sensitive species transformations of iron and manganese mean, however, that under anoxic conditions, P may be released with dissolution of iron and manganese from inorganic sediments. A key driver of the fluxes of organic matter to the deeper waters and sediments of lakes is the phytoplankton productivity of surface waters, as organic material synthesized in the photic zone eventually falls into bottom waters and sediments if it is not oxidized during settling or removed via outflows. In stratified lakes, most production of phytoplankton biomass

Hydrology and Biogeochemistry Linkages

Climatic dominance

occurs in the surface mixed layer (epilimnion) although it can occur at greater depths in oligotrophic lakes with high water clarity. Most nutrients in the epilimnion occur either as organic forms within the biomass, including not only phytoplankton, but also other microorganisms (e.g., bacteria and fungi) and higher trophic levels (e.g., zooplankton or fish), as well as in dead organic matter (detritus) and dissolved nutrient species. Occasionally, in very turbid waters with high concentrations of inorganic suspended solids, concentrations of phosphorus in particulate inorganic form may be the dominant component of the total phosphorus concentration (Grobblelaar and House, 1995). Heterotrophic microorganisms (mostly bacteria and fungi) recycle organic nutrients into forms that can be taken up again by autotrophs. Generally, only a very small fraction of the total nutrients is in a bioavailable inorganic form. The concentration and nature of nonliving particulate organic matter critically influence rates of primary production by controlling rates of regeneration of inorganic nutrients. In large and/or eutrophic lakes, much of this organic matter is generated within the lake itself (autochthonous production), while in smaller lakes, organic matter within the lake may be heavily subsidized by external inputs from the catchment (allochthonous production). Thus, the species and concentrations of nutrients in a lake vary from the interplay of many complex processes, including loading rates, mixing and stratification, redox-associated transformations, and uptake and partitioning of nutrients through the biota.

Dissolution

Biogenic dominance

Concentration

Lithologic dominance

Groundwater can have various meanings depending on the context and timescale of interest. This section is mainly concerned with the part of the natural hydrological cycle beginning at the top of the saturated zone where groundwater recharge occurs and ending in a discharge area where groundwater becomes surface water (Figure 14). Timescales of groundwater movement from recharge to discharge range from minutes (e.g., near-stream response to a rainstorm) to millions of years (e.g., fossil groundwater beneath an arid landscape). The biogeochemistry of groundwater is driven by abiotic and microbially mediated reactions that in part result from the physical transport of aqueous reactants into contact with subsurface materials with which they are not in equilibrium. In this way, groundwater movement and biogeochemistry affect the development and distribution of microbial communities in the subsurface. In turn, groundwater is an important route for delivery of water and solutes, including nutrients and toxins, from the land surface to streams with a range of residence times and chemical compositions that are different from those of surface runoff. The relative proportions of runoff, and shallow and deep groundwater discharge, can change at various time scales, causing marked changes in stream chemistry as a function of flow. Groundwater controls on stream chemistry may be at least as important as in-stream biogeochemical controls in many situations.

Evapotranspiration

Water reuse Fertilizers

Closed system relative to soil gases

2.11.7 Groundwater

Atmospheric gases

Airborne constituents

Soluble minerals and soil gases Insoluble residues Open system relative to soil gases

291

• Dissolution • Precipitation

Runoff infiltration Soil reactions

• Ion exchange • Redox reactions • Sorption Water table

Springs Lakes and rivers

Short residence time

• Dissolution • Precipitation • Ion exchange • Sorption • Redox reactions Long residence time • Gas generation and consumption

Brackish seeps

Ocean

e

ing

n zo

Seawater

ix

Volcanic or magmatic CO2

M

High temperature and pressure Figure 14 (A) schematic of the hydrochemical cycle. From Back W, Baedecker MJ, and Wood WW (1993) Scales in chemical hydrogeology: A historical perspective. In: Alley WM (ed.) Regional Ground-Water Quality, pp. 111–129. New York: van Nostrand Reinhold.

292

Hydrology and Biogeochemistry Linkages

Groundwater recharge beneath an unsaturated zone typically contains oxidants (electron acceptors, e.g., nitrate, sulfate, and dissolved atmospheric oxygen (O2)), whereas recharge beneath a surface-water body is more likely to contain higher concentrations of reduced species (electron donors, e.g., ammonium, methane, and dissolved organic matter (DOM)). In some cases, electron acceptors and electron donors are carried into the saturated zone together during recharge (e.g., O2 and DOM), in which case they may react with each other until one or the other is consumed. In many cases, however, solid phases in the aquifer are the predominant sources of electron donors (e.g., organic matter, ferrous iron minerals, and sulfide minerals) or electron acceptors (e.g., ferric iron minerals and sulfate minerals), and the rates and progress of biogeochemical reactions affecting solutes in recharge are largely controlled by the geology of the subsurface (abundance and reactivity of solid phases) and water–rock contact (McMahon and Chapelle, 2008). Many of these reactions are catalyzed by bacteria taking advantage of chemical potential energy caused by the juxtaposition of chemical species that are not in equilibrium (Appelo and Postma, 2007; Stumm and Morgan, 1996). Reaeration of groundwater is limited by downward advection and loss of communication with overlying air in recharge areas. Where groundwater flow is largely unidirectional (advection7longitudinal dispersion), redox zones generally tend to follow relatively simple progressions with age, for example, if starting with oxic recharge, 4þ 2þ 2 reduction of O2, NO 3 Mn , Fe , SO4 , and CO2 (Figure 7), but the progression through this sequence may be more or less complete depending on groundwater flow time, element availability, and aquifer reactivity (Edmunds et al., 1984). Reduction of O2 and NO 3 typically progress more rapidly in organic-rich mudstones and unweathered glacial deposits containing reactive rock fragments, whereas these reactions commonly are slow in highly evolved siliciclastic sediments and some carbonate rocks. Thus, discharging groundwater can be oxic or highly reduced depending on the hydrogeologic setting. Starting with anoxic recharge, other types of reactions may be important and redox processes affecting aqueous 2þ species (e.g., oxidation of CH4, H2S, NHþ 4 , and Fe ) may be reversed. Reversed redox sequences along groundwater flowpaths are well documented in anthropogenic contaminant plumes, for example, near landfills and organic spill sites (Baedecker et al., 1993; Christensen et al., 2001), and they also occur in aquifers underlying recharge areas in wetlands and lakes containing organic-rich bed sediments (e.g., Katz et al., 1995). Other reactions involving dissolution and precipitation of inorganic solid phases cause concentration gradients in non-redox-sensitive constituents such as SiO2, Na, Mg, Ca, Sr, and Mg. For unconsolidated water-table aquifers with typical recharge rates in humid to semiarid environments, groundwater is commonly stratified, being youngest near the water table and progressively older downward. Groundwater chemistry therefore also may be stratified in recharge areas as a result of changing conditions in recharge composition (e.g., changing composition of atmospheric deposition or addition of anthropogenic contaminants) and biogeochemical reactions in the aquifer (Back et al., 1993; Bo¨hlke, 2002). In discharge areas, and where groundwater flowpaths are confined between

impermeable units, these patterns can change. In discharge areas, preexisting gradients of groundwater age and chemistry may become horizontal or overturned as flow vectors turn upward (Bo¨hlke et al., 2002). These spatial patterns may be confused with locally generated biogeochemical gradients in the absence of detailed information. Aquifer heterogeneity can result in complex reaction zones including bidirectional transport of reactants and products across aquifer–aquitard contacts (McMahon, 2001). In karst and fractured rock aquifers, groundwater flowpaths and biogeochemical mass transfers may be especially complex because of coexisting high-permeability conduits and massive low-permeability units (Bakalowicz, 2005). Complex patterns of solute transport and redox progression are typical near shallow water tables (Scholl et al., 2006) and near sediment-surface water interfaces, such as hyporheic zones, lake beds, and wetlands, where flow reversals and(or) diffusion are important, for example, beneath forested wetlands (Alewell et al., 2006). Biogeochemical processes affecting groundwater chemistry operate over a large range of timescales. For example, measured rates of oxygen reduction and denitrification range over at least 8–10 orders of magnitude. Because of this, and because these reactions commonly are limited by the distributions of reactive solid phases, groundwater chemical gradients may be either sharp boundaries (flux-controlled) or gradual transitions (kinetically controlled). Rates of biogeochemical reactions in aquifers have been measured by various techniques, in part reflecting the range of timescales involved. Laboratory experiments with groundwater and aquifer material can be used to study reaction potential on short timescales. In situ tracer injection experiments including isotopically labeled reactants can be used for intermediate timescales. Examples include single-well or push–pull tests and natural-gradient tracer breakthrough experiments (Istok et al., 1997; Smith et al., 2004; Kellogg et al., 2005; Bo¨hlke et al., 2006). Laboratory experiments and single-well injections commonly indicate higher reaction rates than in situ natural gradient measurements, presumably in part because of physical disruption or other forms of biogeochemical stimulation (Smith et al., 1996, 2006). Groundwater dating of reaction-zone chemical gradients may be the only practical empirical method of measurement at longer timescales. Examples include the use of modern atmospheric environmental tracers, for example, tritium and chlorofluorocarbons for reaction zones on 0–60year timescales (Bo¨hlke et al., 2002; Green et al., 2008) and 14 C for reaction zones on 103–104-year timescales (Vogel et al., 1981; Plummer et al., 1990). Because of aquifer heterogeneity, groundwater ages and reaction rates are evaluated most reliably from field data using solute transport models that account for dispersion and sample mixing (Scholl, 2000; Weissmann et al., 2002; Green et al., 2010). In discharge areas, groundwater can interact with sediments and plants in riparian wetlands, streambeds, and estuaries, where organic matter and other reactants may be more abundant than elsewhere in the saturated zone. Reactions in these areas are strongly dependent on groundwater flowpaths. Near-stream geomorphology and vertical components of groundwater flow largely determine whether groundwater interacts with shallow riparian soils and plants or bypasses

Hydrology and Biogeochemistry Linkages

those potential reaction sites before discharging to streams. Slow diffuse flow through reactive material may cause important changes in groundwater chemistry just prior to discharge, whereas rapid flow through permeable layers and macropores may avoid such reactions (Burt et al., 1999; Angier et al., 2005). The relative importance of these flowpaths and reactions for overall mass balance in discharge areas is difficult to assess in watershed-process studies. Direct discharge of groundwater to estuaries is a potential source of land-derived water and nutrients to coastal waters, but is difficult to quantify and may exhibit complex patterns of physical and chemical interaction with salty pore water (Manheim et al., 2004; Andersen et al., 2005). Discharge also may be affected or enhanced by bioirrigation, the augmentation of flow across the sediment–water interface by filter feeders living in estuaries (Martin et al., 2006; Meysman et al., 2006). Groundwater discharge is an important component of stream flow and solute loads, especially in low-order streams (Alexander et al., 2007). Total stream flow typically is dominated by groundwater discharge except briefly during intense runoff events. Even during flood peaks, the fraction of stream flow delivered from the land surface to the stream without moving into the subsurface typically is small (Buttle, 1994; Buttle and Peters, 1997; Bishop et al., 2004; Burt and Pinay, 2005). As groundwater is stratified in age and chemistry, the age and chemistry of groundwater contributing to stream flow can be quite variable and complex. Old groundwater (decades to millennia) may discharge upward from below the streambed, while younger groundwater (months to years) discharges from shallower flowpaths. Because of variations in subsurface hydraulic properties, age distributions in discharge may be difficult to define and mean age alone may be a poor indicator of the assemblage of watershed transit times. Changing conditions throughout a drainage basin can cause changes in the proportions of different groundwater types and the proportions of groundwater and runoff, contributing to stream flow over timescales ranging from hours to days (storms and snowmelt events) to months (seasonal effects of precipitation and evapotranspiration) to years (interannual and longer climate variation or land-use changes). For example, seasonal variation of nitrate and sulfate concentrations with stream flow may depend on production, consumption, and storage at different timescales in the unsaturated and saturated zones (Lynch and Corbett, 1989; Shanley and Peters, 1993; Huntington et al., 1994; Peters, 1994; Bo¨hlke et al., 2007). At interannual and decadal timescales, responses of streams to changes in loadings of nonpoint-source contaminants can be complex, subdued, or delayed because of changing inputs, groundwater residence times, and water–rock interactions (Bo¨hlke and Denver, 1995; Burt et al., 2008). Surficial aquifers in unconsolidated sediments, such as coastal plains, alluvial valleys, and glacial outwash deposits, commonly have mean groundwater residence times of the order of decades. As changes in land use, agricultural-fertilizer use, and atmospheric deposition commonly occur on decadal timescales, many aquifers contain transient records of anthropogenic nonpoint-source contaminants. This means that the mass flux of a constituent in annual recharge may be different from the mass flux in annual discharge even where the constituent

293

behaves conservatively and the water balance is in steady state. This is a common feature of agricultural drainage basins and an important limitation on watershed nitrogen balances and export predictions (Bo¨hlke, 2002). Commonly, it is difficult to resolve the effects of temporal changes in recharge chemistry from progressive biogeochemical reactions in aquifers without detailed study.

2.11.8 Acidic Atmospheric Deposition – Acid Rain Many linkages between hydrology and biogeochemistry were revealed from research conducted to understand the effects of acid rain on terrestrial and aquatic ecosystems beginning in the 1970s. Some of the most import linkages were demonstrated through studies of deleterious effects on biota, particularly forests and fish. Decreases in pH and increases in dissolved inorganic aluminum concentrations have diminished species diversity and abundance of plankton, invertebrates, and fish in acid-impacted surface waters (Schindler, 1988). Acid rain effects on ecosystems include forest decline (Pitelka, 1994; DeHayes et al., 1999), bird population declines and changes (Graveland, 1998), and aquaticbiota declines including algae, macroinvertebrates, and fish (Schindler, 1988). Extremely high deposition of N species (wet and dry deposition) has had a range of effects on forests from fertilization to changes in N mineralization and increased N leaching through soils to surface water (Vitousek et al., 1982; Aber, 1992; Aber et al., 1995, 1998; Emmett et al., 1998a, 1998b; Emmett, 1999; Mitchell, 2001). Vegetation filters atmospheric contaminants and dry deposition can be a major input to ecosystems (Reynolds et al., 1994; Peters et al., 1998). Acidification causes base cations and metals, particularly inorganic aluminum, to be mobilized, which in turn, has deleterious effects on aquatic biota, such as fish (Driscoll et al., 1980). For example, aluminum precipitates on fish gills ultimately affecting blood pH and decreasing the capacity of hemoglobin to transfer oxygen (Fromm, 1980). The loss of nutrient base cations, such as calcium, from soils affects forest growth and health (DeHayes et al., 1999), and subsequent decreases in receiving waters affect aquatic biota (Holt and Yan, 2003; Keller et al., 2003; Jeziorski and Yan, 2006; Jeziorski et al., 2008; Cairns and Yan, 2009). A knock-on effect of the S emissions and deposition associated with acid rain is that increased inputs of sulfate decrease methane production in wetlands (Schimel, 2004). Hydrology is a major driver that delivers acid rain through terrestrial vegetation, soils, and groundwater to streams and lakes. Acid-neutralizing capacity (or alkalinity) is generated by mineral weathering, but base-poor lithologies for which weathering rates are relatively low, including quartzites, sandstones, and granitoid metamorphic and igneous rocks, are particular susceptible to the addition of acids (Schnoor and Stumm, 1986; Reynolds et al., 2001). Glaciated terrain on these lithologies is susceptible to acidic deposition, particularly where glacial deposits are thin and relatively impermeable. For example, US lakes and streams with comparable-sized drainage basins in the west-central Adirondack Mountains, NY, receiving similar acidic deposition,

294

Hydrology and Biogeochemistry Linkages

responded differently in neutralizing the acidity simply because of differences in the thickness and distribution of surficial material (Peters and Murdoch, 1985; Peters and Driscoll, 1987a). The residence time of water combined with the weathering rate of surficial materials determines the amount of alkalinity in the water. Longer residence time associated with long hydrological pathways results in higher alkalinities compared to those of short hydrological pathways. For moderate to high acidic-deposition rates, streams become chronically acidic when residence times are short and similarly for lakes where the drainage basin:lake area is high. However, even drainage basins with large deposits of thick till or stratified drift and generally long residence time may experience episodic acidification during snowmelt, large rainstorms, or sea-salt episodes (Wright et al., 1988; Heath et al., 1992; Hindar et al., 1995; Larssen and Holme, 2005). Through cleaner fuel technologies and emission-control systems, acid rain generally has decreased throughout North America and Europe since the, 1970s, but acidity remains higher than the inferred pre-industrial conditions (Stoddard et al., 1999; Jeffries et al., 2003; Wright et al., 2005). In addition, liming of lakes and watersheds can restore pH values, but other changes resulting from acid deposition, such as those in soil chemistry and related biota, are not reversible on short timescales (Schindler, 1999). For example, weathering is a primary source for soil base cations and the process of restoring the original soil cation-exchange complex may take several hundred years before full recovery occurs (Cosby et al., 1985; Driscoll et al., 2001). The persistence of surfacewater acidification even with reductions in acid deposition has been attributed to losses of exchangeable base cations in the soil (Lawrence et al., 1999; Lawrence, 2002), which is reflected in many soil organisms and other biota, such as birds (Graveland, 1998; Hamburg et al., 2003). A multitracer assessment of red spruce, a species showing recent growth reductions and decreases in plant-available calcium in the northeastern US, suggests a progressive shallowing of effective depth of base-cation uptake by fine roots (Bullen and Bailey, 2005). Forest harvesting also decreases exchangeable basecation pools (Federer et al., 1989; Watmough et al., 2003). The reductions in atmospheric emissions have targeted S and although substantial – for example, greater than 40% in the northeastern United States and eastern Canada and greater than 60% in Norway – have not been sufficient for surface waters to recover chemically and biologically (Stoddard et al., 1999; Aherne et al., 2003; Jeffries et al., 2003; Skjelkva˚le et al., 2003; Watmough and Dillon, 2003; Larssen, 2005; Wright et al., 2005) and the biological recovery may be hampered by other environmental factors such as drought and increasing water temperatures (Arnott and Yan, 2002; Ashforth and Yan, 2008). Remediation has resulted in restoration of some aquatic biota, such as fish, but restoration of surfacewater chemistry to pre-industrial conditions may not be possible and the trajectory of biogeochemical and species evolution has likewise changed (Schindler, 1999). The restoration may also be exacerbated by other environmental factors such as climate-change affects on the frequency and severity of sea-salt episodes and drought, turnover of organic carbon, and mineralization of nitrogen (Skjelkva˚le et al., 2003; Wright et al., 2006).

2.11.9 Summary and Future Considerations The hydrosphere, biosphere, lithosphere, and chemosphere are intricately linked through a wide range of spatial and temporal scales. For a comprehensive understanding of biogeochemical cycling, an understanding of the hydrologicalprocesses is required. In addition to the wealth of information linking hydrology and biogeochemistry across different aspects of the hydrological cycle, there is a wealth of information on in-stream hydrological variability and biogeochemical processing in streams and rivers. Section 2.11.2 provided an overview of hydrological processes in headwaters with respect to stream flow generation. Mechanisms delivering water from hillslopes to stream channels were presented and discussed with respect to the relative contributions of old water, that is, water stored in the basin soils and groundwater, and new water, that is, associated with precipitation and snowmelt. The relative importance of biogeochemical processes along hydrological pathways was highlighted with a particular focus on the importance of nearstream (riparian) saturated zones in resetting the chemical signature of water flowing into the riparian zone. The riparian zone in many basins effectively buffers upslope nutrient inputs, but may also alter nutrient concentrations and fluxes through N cycling processes, such as mineralization, denitrification, and uptake by riparian vegetation. Section 2.11.3 discussed processes affecting the components of the water budget, snow formation, and ablation processes, and those in the soil below snow-cover overwinter and during snowmelt. Microbes remain active in soils under the snowpack where water is not limited. The coupling of these nutrient transformations and snow-meltwater fluxes can result in delivery of large quantities of nutrients, organic matter, and carbon export from terrestrial ecosystems. Furthermore, solutes in snowpacks preferentially elute during melting, which in turn, can stimulate biological activity within the snowpack, for example, snow algae. The presence and rate of water movement combined with the organic-matter composition and temperature of soils largely determine the nature of the biogeochemical reactions, for example, aerobic versus anaerobic. Vegetation and solar radiation control soil water content through evaporation and transpiration and the vegetation is in part controlled by soil type and thickness, aspect, and elevation. Tree roots can redistribute water in soils affecting nutrient uptake. Plant–soil relations are intricately linked to biogeochemical cycling through the rhizosphere. Downstream mixing affects water and solute transit times, which are intricately linked to hydrological pathways through soils and groundwater and in streams with riparian and hyporheic zones. These pathway contributions, in turn, are controlled by the magnitude and intensity of rainfall and snowmelt. Section 2.11.4 presented the concept of nutrient spiraling including the concept of nutrient-uptake length and the importance of temperature and stream flow variability on biogeochemistry. The effects of stream–groundwater interactions through hyporheic and riparian zones were also discussed. Hyporheic zone processes tend to have a larger effect per unit area on the water column in shallow upper reaches, but continuing losses through large river networks can have

Hydrology and Biogeochemistry Linkages

large cumulative effects.. Field studies involving isotopes (including isotopically labeled compounds) have elucidated the within-river transformations of nitrogen species including how these are affected by seasonality, stream flow, light penetration, and terrestrial organic matter and nutrient inputs from near-stream ecosystems. Spatial variations in within-river processes are also controlled by hydrology, channel morphology, catchment land use, and riparian vegetation. Section 2.11.5 contrasted important processes in hydrologically isolated wetlands with those temporally connected to streams and rivers. The exchange of water, sediments, and nutrients in wetlands with adjacent catchment areas, groundwater, and streams has a major effect on biogeochemical processes. Residence time is a key driver of biogeochemical dynamics ranging from rapid turnover rates in valley-bottom riparian wetlands with high groundwater discharge to extremely slow turnover rates in a thin active layer at the surface of raised peat bogs. The near-saturated conditions of wetlands with typically high organic contents control the redox potential, which drives the biogeochemical processes. Oxygen typically limits degradation rates in wetlands and carbon is the main driver of wetland biogeochemistry. Furthermore, the temporal and spatial variability of residence time and related turnover rates therefore dictate the biogeochemical processes. Section 2.11.6 discussed atmospheric, stream, and groundwater nutrient inputs, stratification and within-lake processes, interactions with sediments, and limiting nutrients. The nutrients associated with groundwater discharge to lakes are affected by the composition of sediments, which may alternate from oxidized to reduced conditions. Differences in sediment composition control redox conditions and, subsequently, aerobic or anaerobic reactions that affect nutrient transformations and species. Plants in littoral zones, such as emergent and submerged macrophytes and periphyton, can also alter lake nutrient composition by trapping particulates and through nutrient uptake (growth) and release (decay). Phosphorus generally limits productivity in freshwater ecosystems, but with excess phosphorus, nitrogen may be limiting; however, nitrogen can be supplemented by blooms of N-fixing blue-green algae. Although recent research suggests that surface waters were N limited prior to industrialization, the science is still contentious about the relative importance (or limitations) of N and P in controlling biological productivity of freshwaters. Lake stratification controls mixing of top and bottom waters, thus affecting biogeochemical processes. The nutrient status and productivity of surface waters determines light penetration and subsequent supply of organic matter and nutrients to bottom waters. Section 2.11.7 presented information about typical reactions controlled by hydrological pathways, lithology (mineralogy) and biota, the importance of residence time in biogeochemical evolution, and linkages between groundwater and surface water. Biogeochemistry of groundwater is largely related to microbially mediated redox reactions that result from physical transport of aqueous reactants into contact with subsurface materials with which they are not in equilibrium, where microbial communities develop to catalyze reactions in exchange for energy. Redox conditions in groundwater vary depending on landscape position, with oxidizing conditions prevailing in headwaters and beneath the unsaturated zone

295

and more reducing conditions occurring in lowlands and under streams and lakes. Redox conditions may also be affected by lithology. Consequently, discharging groundwater can be oxic or highly reduced depending on the hydrogeologic setting. Biogeochemical processes affecting groundwater chemistry operate over a large range of timescales (e.g., 8–10 orders of magnitude for oxygen reduction and denitrification). Stream flow is typically dominated by groundwater discharge, even during floods. But because groundwater can vary markedly in age and chemistry, the discharging mixture of groundwater contributed from a wide range of hydrological pathways can cause stream water composition and delivery of nutrients to aquatic ecosystems to likewise vary markedly in time and space. Examples are given of the effects of human activities on hydrology and biogeochemistry linkages in each of the sections and in a separate section on acidic atmospheric deposition. Although much research has been conducted in assessing the linkages between hydrology and biogeochemistry, many challenges remain, particularly in linking observations across a wide range of temporal and spatial scales. Vegetation, soils, hydrology, and biogeochemistry develop and respond together; yet, our efforts to study these linkages are often narrowly focused, resulting in high levels of site-specific knowledge, but slower progress in extrapolating to larger spatial scales and in developing meaningful generalizations. The need for morecomprehensive interdisciplinary studies is warranted to link terrestrial vegetation and soils in headwaters through riparian zones/floodplains to streams. These interdisciplinary studies would incorporate in-stream processes including interactions with the hyporheic zone, across scales and hydroclimatic zones. Understanding hydrological and biogeochemical processes also requires knowledge of the biological components and their functioning within these studies. Advances in technology continue to provide smaller and more robust sensors, smaller data-acquisition packages with innovative data-transmission capabilities, and better analytical instrumentation for accurate and precise measurement of low elemental and solute concentrations on small samples. In addition, new tools are evolving in the areas of nanotechnology, remote sensing, and biosensor technology, which are providing new and innovative ways to evaluate processes linking hydrology and biogeochemistry. In addition, computer-technology advances and new visualization software with much higher computation and processing speeds provide a platform for innovative designs in data analysis and modeling. Interdisciplinary research incorporating some of these new technologies for data collection and processing coupled with the computer processing and visualization may provide new ways of data mining and testing of hydrological, biological, and biogeochemical process interactions.

2.11.10 Additional Reading The literature is comprehensive with information about hydrology and biogeochemistry linkages. For some additional details about general water-quality characteristics, see Meybeck et al. (2005) and Peters et al. (2005); for nitrogen cycling,

296

Hydrology and Biogeochemistry Linkages

see Burt et al. (1993), Heathwaite et al. (1996), Cirmo and McDonnell (1997), Edwards and Wetzel (2005), Goulding et al. (1998), Lohse et al. (2009), Mitchell (2001), Vollenweider and Kerekes (1982), and Wetzel (2001); and for stream– groundwater interactions, see Burt and Pinay (2005), Dahm et al. (1998), Jones and Holmes (1996), Jones and Mulholland (2000), Rosenberry and Labaugh (2008), Winter (1999), Winter and Woo (1990), and Winter et al. (1999). Finally, Lohse et al. (2009) provide an overview of linkages between hydrology and biogeochemistry, and Belnap et al. (2005) discuss hydrology and microbial linkages in arid and semiarid watersheds.

References Aber JD (1992) Nitrogen cycling and nitrogen saturation in temperate forest ecosystems. Trends in Ecology and Evolution 7(7): 220--224. Aber JD, Magill A, Mcnulty SG, et al. (1995) Forest biogeochemistry and primary production altered by nitrogen saturation. Water, Air and Soil Pollution 85(3): 1665--1670. Aber JD, McDowell W, Nadelhoffer K, et al. (1998) Nitrogen saturation in temperate forest ecosystems. BioScience 48(11): 921--934. Aherne J, Dillon PJ, and Cosby BJ (2003) Acidification and recovery of aquatic ecosystems in south central Ontario, Canada: Regional application of the MAGIC model. Hydrology and Earth System Science 7(4): 561--573. Alewell C, Paul S, Lischeid G, Ku¨sel K, and Gehre M (2006) Characterizing the redox status in three different forested wetlands, with geochemical data. Environmental Science and Technology 40(24): 7609--7615. Alexander RB, Boyer EW, Smith RA, Schwarz GE, and Moore RB (2007) The role of headwater streams in downstream water quality. Journal of the American Water Resources Association 43: 41--59. Alexander RB, Smith RA, and Schwartz GE (2000) Effect of stream channel size on the delivery of nitrogen to the Gulf of Mexico. Nature 403: 758--761 (doi:10.1038/ 35001562). Amundson R, Richter DD, Humphreys GS, Jobba´gy EG, and Je´roˆme G (2007) Coupling between biota and earth materials in the critical zone. Elements 3(5): 327--332 (doi:10.2113/gselements.3.5.327). Andersen MS, Nyvang V, Jakobsen R, and Postma D (2005) Geochemical processes and solute transport at the seawater/freshwater interface of a sandy aquifer. Geochimica et Cosmochimica Acta 69(16): 3979--3994 (doi:10.1016/ j.gca.2005.03.017). Anderson MG and Burt TP (1978) The role of topography in controlling throughflow generation. Earth Surface Processes 3(4): 331--344. Angier JT, McCarty GW, and Prestegaard KL (2005) Hydrology of a first-order riparian zone and stream, mid-Atlantic coastal plain, Maryland. Journal of Hydrology 309: 149--166 (doi:10.1016/j.jhydrol.2004.11.017). Appelo CAJ and Postma D (2007) Geochemistry, Groundwater, and Pollution, 2nd edn. Rotterdam: AA Balkema. 649pp. Arango CP, Tank JL, Johnson LT, and Hamilton SK (2008) Assimilatory uptake rather than nitrification and denitrification determines nitrogen removal patterns in streams of varying land use. Limnology and Oceanography 53(6): 2558-2572. Arnott SE and Yan ND (2002) The influence of drought and re-acidification on zooplankton emergence from resting stages. Ecological Applications 12(1): 138--153. Arrigoni AS, Poole GC, Mertes LAK, O’Daniel SJ, Woessner WW, and Thomas SA (2008) Buffered, lagged, or cooled? Disentangling hyporheic influences on temperature cycles in stream channels. Water Resources Research 44: W09418 (doi:09410.01029/02007WR006480). Asbury CE, Mcdowell WH, Trinidadpizarro R, and Berrios S (1994) Solute deposition from cloud water to the canopy of a Puerto Rican montane forest. Atmospheric Environment 28(10): 1773--1780. Ashforth D and Yan ND (2008) The interactive effects of calcium concentration and temperature on the survival and reproduction of Daphnia pulex at high and low food concentrations. Limnology and Oceanography 53(2): 420--432. Babiarz CL, Hurley JP, Benoit JM, Shafer MM, Andren AW, and Webb DA (1998) Seasonal influences on partitioning and transport of total and methylmercury in rivers from contrasting watersheds. Biogeochemistry 41: 237--257.

Back W, Baedecker MJ, and Wood WW (1993) Scales in chemical hydrogeology: A historical perspective. In: Alley WM (ed.) Regional Ground-Water Quality, pp. 111--129. New York: van Nostrand Reinhold. Baedecker MJ, Cozzarelli IM, Eganhouse RP, Siegel DI, and Bennett PC (1993) Crude oil in a shallow sand and gravel aquifer – III. Biogeochemical reactions and mass balance modeling in anoxic groundwater. Applied Geochemistry 8(6): 569--586. Bakalowicz M (2005) Karst groundwater: A challenge for new resources. Hydrogeology Journal 13(1): 148--160. Baker MA, Dahm CN, and Valett HM (1999) Acetate retention and metabolism in the hyporheic zone of a mountain stream. Limnology and Oceanography 44(6): 1530--1539. Bales RC, Davis RE, and Williams MW (1993) Tracer release in melting snow: Diurnal and seasonal patterns. Hydrological Processes 7(4): 389--401. Band LE, Tague CL, Groffman P, and Belt K (2001) Forest ecosystem processes at the watershed scale: Hydrological and ecological controls of nitrogen export. Hydrological Processes 15(10): 2013--2028. Barros AP and Lettenmaier DP (1994) Dynamic modelling of orographically induced precipitation. Reviews of Geophysics 32(3): 265--284. Baxter CV and Hauer FR (2000) Geomorphology, hyporheic exchange, and selection of spawning habitat by bull trout (Salvelinus confluentus). Canadian Journal of Fisheries and Aquatic Sciences 57(7): 1470--1481. Bay RR (1969) Runoff from small peatland watersheds. Journal of Hydrology 9: 90--102. Bedford BL, Walbridge MR, and Aldous A (1999) Patterns in nutrient availability and plant diversity of temperate North American wetlands. Ecology 80(7): 2151--2169. Belnap J, Welter JR, Grimm NB, Barger N, and Ludwig JA (2005) Linkages between microbial and hydrologic process in arid and semiarid watersheds. Ecology 86(2): 298--307. Berg NH (1992) Ion elution and release sequence from deep snowpacks in the central Sierra Nevada, California. Water, Air and Soil Pollution 61: 139--168. Bergstro¨m A-K and Jansson M (2006) Atmospheric nitrogen deposition has caused nitrogen enrichment and eutrophication of lakes in the northern hemisphere. Global Change Biology 12(4): 635--643 (doi:10.1111/j.1365-486.2006.01129.x). Bernhardt ES, Likens GE, Hall RO, et al. (2005) Can’t see the forest for the stream? In-stream processing and terrestrial nitrogen exports. BioScience 55(3): 219--230. Bernot MJ, Tank JL, Royer TV, and David MB (2006) Nutrient uptake in streams draining agricultural catchments of the midwestern United States. Freshwater Biology 51(3): 499--509 (doi:10.1111/j.1365-2427.2006.01508.x). Betson RP (1964) What is watershed runoff? Journal of Geophysical Research 69(8): 1542--1552. Beven K (1982) On subsurface stormflow: Predictions with simple theory for saturated and unsaturated flows. Water Resources Research 18(6): 1627--1633. Beven K and Germann P (1982) Macropores and water flow in soils. Water Resources Research 18(5): 1311--1325. Bishop K, Seiber J, Kohler S, and Laudon H (2004) Resolving the double paradox of rapidly mobilised old water with highly variable responses in runoff chemistry. Hydrological Processes 18: 185--189. Bluth GJS and Kump LR (1994) Lithologic and climatologic controls of river chemistry. Geochimica et Cosmochimica Acta 58(10): 2341--2359. Bohlen PJ, Scheu S, Hale CM, et al. (2004) Non-native invasive earthworms as agents of change in northern temperate forests. Frontiers in Ecology and the Environment 2(8): 427--435. Bonan GB and Levis S (2006) Evaluating aspects of the community land and atmosphere models (CLM3 and CAM3) using a dynamic global vegetation model. Journal of Climate 19(11): 2290--2301. Bormann FH and Likens GE (1994) Pattern and Process in a Forested Ecosystem. New York: Springer. 253pp. Boulton AJ, Findlay S, Marmonier P, Stanley EH, and Valett HM (1998) The functional significance of the hyporheic zone in streams and rivers. Annual Review Ecology and Systematics 29: 59--81. Bowden WB (1986) Gaseous nitrogen emissions from undisturbed terrestrial ecosystems: An assessment of their impacts on local and global nitrogen budgets. Biogeochemistry 2: 249--279. Boyd GR, Reemtsma H, Grimm DA, and Mitra S (2003) Pharmaceuticals and personal care products (PPCPs) in surface and treated waters of Louisiana, USA and Ontario, Canada. Science of the Total Environment 311: 135--149. Boyd MJ, Bufill MC, and Knee RM (1993) Pervious and impervious runoff in urban catchments. Hydrological Sciences Journal/Journal des Sciences Hydrologiques 38(6): 463--479. Boyer EW, Hornberger GM, Bencala KE, and McKnight DM (1997) Response characteristics of DOC flushing in an alpine catchment. Hydrological Processes 11: 1635--1647.

Hydrology and Biogeochemistry Linkages

Bo¨hlke JK (2002) Groundwater recharge and agricultural contamination. Hydrogeology Journal 10: 153--179 (doi:10.1007/s10040-001-0183-3). Bo¨hlke JK, Antweiler RC, Harvey JW, et al. (2009) Multi-scale measurements and modeling of denitrification in streams with varying flow and nitrate concentration in the upper Mississippi River basin, USA. Biogeochemistry 93: 117--141 (doi:10.1007/s10533-008-9282-8). Bo¨hlke JK and Denver JM (1995) Combined use of groundwater dating, chemical, and isotopic analyses to resolve the history and fate of nitrate contamination in two agricultural watersheds, Atlantic coastal plain, Maryland. Water Resources Research 31(9): 2319--2339. Bo¨hlke JK, Harvey JW, and Voytek MA (2004) Reach-scale isotope tracer experiment to quantify denitrification and related processes in a nitrate-rich stream, midcontinent United States. Limnology and Oceanography 49(3): 821--838. Bo¨hlke JK, O’Connell ME, and Prestegaard KL (2007) Ground-water stratification and delivery of nitrate to an incised stream in varying flow conditions. Journal of Environmental Quality 36(3): 664--680. Bo¨hlke JK, Smith RL, and Miller DN (2006) Ammonium transport and reaction in contaminated ground water: Application of isotopic tracers and isotope fractionation studies. Water Resources Research 42: W05411 (doi:10,1029/ 2005WR004349). Bo¨hlke JK, Wanty R, Tuttle M, Delin G, and Landon M (2002) Denitrification in the recharge area and discharge area of a transient agricultural nitrate plume in a glacial outwash sand aquifer, Minnesota. Water Resources Research 38(7): 10.1--10.26 (doi:10.1029/2001WR000663). Bo¨ttcher J, Strebel O, Voerkelius S, and Schmidt H-L (1990) Using isotope fractionation of nitrate-nitrogen and nitrate-oxygen for evaluation of microbial denitrification in a sandy aquifer. Journal of Hydrology 114: 413--424. Branfireun B, Heyes A, and Roulet NT (1996) The hydrology and methyl mercury dynamics of a headwater wetland. Water Resources Research 32: 1785--1794. Branfireun BA, Routlet NT, Kelly CA, and Rudd JWM (1999) In situ sulphate stimulation of mercury methylation in a boreal peatland: Toward a link between acid rain and methylmercury contamination in remote environments. Global Biogeochemical Cycles 13(3): 743--750. Brantley SL, Goldhaber MB, and Ragnarsdottir KV (2007) Crossing disciplines and scales to understand the critical zone. Elements 3(5): 207--314. Brooks P, McKnight D, and Bencala K (1999a) The relationship between soil heterotrophic activity, soil dissolved organic carbon (DOC) leachate, and catchment-scale DOC export in headwater catchments. Water Resources Research 35(6): 1895--1902. Brooks PD, Campbell DH, Tonnessen KA, and Heuer K (1999b) Natural variability in N export from headwater catchments: Snow cover controls on ecosystem N retention. Hydrological Processes 13(14): 2191--2201. Brooks PD, McKnight D, and Elder K (2005) Carbon limitation of soil respiration under winter snowpacks: Potential feedbacks between growing season and winter carbon fluxes. Global Change Biology 11(2): 231--238. Brooks PD, Schmidt SK, and Williams MW (1997) Winter production of CO2 and N2O from alpine Tundra: Environmental controls and relationship to inter-system C and N fluxes. Oecologia 110(3): 403--413. Brooks PD and Williams MW (1999) Snowpack controls on nitrogen cycling and export in seasonally snow-covered catchments. Hydrological Processes 13(14– 15): 2177--2190. Brooks PD, Williams MW, and Schmidt SK (1995) Snowpack controls on soil nitrogen dynamics in the Colorado alpine. In: Tonnessen KA, Williams MW, and Tranter M (eds.) Biogeochemistry of Seasonally Snow-Covered Catchments, vol. 228, pp. 283--292. Wallingford: IAHS. Brooks PD, Williams MW, and Schmidt SK (1996) Microbial activity under alpine snowpacks, Niwot Ridge, Colorado. Biogeochemistry 32(2): 92--113. Brooks PD, Williams MW, and Schmidt SK (1998) Inorganic nitrogen and microbial biomass dynamics before and during spring snowmelt. Biogeochemistry 43(1): 1--15. Broxton PD, Troch PA, and Lyon SW (2009) On the role of aspect to quantify water transit times in small mountainous catchments. Water Resources Research 45: W08427 (doi:10.1029/2008WR007438). Bruijnzeel LA and Veneklaas EJ (1998) Climatic conditions and tropical montane forest productivity: The fog has not lifted yet. Ecology 79(1): 3--9. Buda AR and DeWalle DR (2009) Dynamics of stream nitrate sources and flow pathways during stormflows on urban, forest and agricultural watersheds in central Pennsylvania, USA. Hydrological Processes 23(22): 3292--3305. Buffam I, Galloway JN, Blum LK, and McGlathery KJ (2001) A stormflow/baseflow comparison of dissolved organic matter concentrations and bioavailability in an Appalachian stream. Biogeochemistry 53(3): 269--306. Bullen TD and Bailey SW (2005) Identifying calcium sources at an acid depositionimpacted spruce forest: A strontium isotope, alkaline earth element multi-tracer approach. Biogeochemistry 74(1): 63--99.

297

Burgess SSO, Adams MA, Turner NC, and Ong CK (1998) The redistribution of soil water by tree root systems. Oecologia 115: 306--311. Burgess SSO and Dawson TE (2004) The contribution of fog to the water relations of Sequoia sempervirens (D. Don): Foliar uptake and prevention of dehydration. Plant, Cell and Environment 27(8): 1023--1034. Burns DA and Kendall C (2002) Analysis of d15N and d18O to differentiate NO 3 sources in runoff at two watersheds in the Catskill Mountains of New York. Water Resources Research 38(5): 1051 (doi:10.1029/2001WR000292). Burt TP, Heathwaite AL, and Trudgill ST (1993) Nitrate: Processes, Patterns and Management. Chichester: Wiley. Burt TP, Howden NJK, Worrall F, and Whelan MJ (2008) Importance of long-term monitoring for detecting environmental change: Lessons from a lowland river in south east England. Biogeosciences 5(6): 1529--1535. Burt TP, Matchett LS, Goulding KWT, Webster CP, and Haycock NE (1999) Denitrification in riparian buffer zones: The role of floodplain sediments. Hydrological Processes 13(10): 1431--1463. Burt TP and Pinay G (2005) Linking hydrology and biogeochemistry in complex landscapes. Progress in Physical Geography 29(3): 297--316. Buttle JM (1994) Isotope hydrograph separations and rapid delivery of pre-event water from drainage basins. Progress in Physical Geography 18(1): 16--41. Buttle JM (2005). Isotope hydrograph separation of runoff sources. In: Anderson MG, McDonnell JJ (eds.) Encyclopedia of Hydrological Sciences. vol. 3, Part 8, ch. 116, pp. 1763–1774. Chichester: Wiley. Buttle JM and Peters DL (1997) Inferring hydrological processes in a temperate basin using isotopic and geochemical hydrograph separation: A re-evaluation. Hydrological Processes 11(6): 557--573. Cairns A and Yan ND (2009) A review of the influence of low ambient calcium concentrations on freshwater daphniids, gammarids, and crayfish. Environmental Reviews 17: 67--79. Caldwell MM and Richards JH (1989) Hydraulic lift: Water efflux from upper roots improves effectiveness of water uptake by deep roots. Oecologia 79: 1--5. Caldwell MM, Dawson TE, and Richards JH (1998) Hydraulic lift: Consequences of water eflux from the roots of plants. Oecologia 113: 151--161. Campbell DH, Clow DW, Ingersoll GP, Mast MA, Spahr NE, and Turk JT (1995) Nitrogen deposition and release in alpine watersheds, Loch Vale, Colorado, USA. In: Tonnessen KA, Williams MW, and Tranter M (eds.) Biogeochemistry of Seasonally Snow-Covered Catchments, vol. 228, pp. 243--253. Wallingford: IAHS. Campbell DH, Kendall C, Change CCY, SIlva SR, and Tonnessen KA (2002) Pathways for nitrate release from an alpine watershed: Determination using q15N and q18O. Water Resources Research 38(5). (doi:10.1029/2001WR000294). Campbell JL and Law BE (2000) Dissolved organic nitrogen budgets for upland, forested ecosystems in New England. Biogeochemistry 49(2): 123--142. Campbell NS and Novotny V (2004) Water Quality – Prevention, Identification, and Management of Diffuse Pollution. New York: IWA Publishing. 322pp. Cardon ZG and Whitbeck JL (2007) The Rhizosphere: An Ecological Perspective. New York, NY: Elsevier. 212pp. Chorley RJ (1971) The drainage basin as the fundamental geomorphic unit. In: Chorley RJ (ed.) Introduction to Physical Hydrology, pp. 37--59. London: Methuen. Christensen TH, Kjeldsen P, Bjerg PL, et al. (2001) Biogecochemistry of landfill leachate plumes. Applied Geochemistry 16(7–8): 659--718. Christopher SF, Mitchell MJ, McHale MR, Boyer EW, Burns DA, and Kendall C (2008) Factors controlling nitrogen release from two forested catchments with contrasting hydrochemical responses. Hydrological Processes 22(1): 46--62. Chung SW, Hipsey MR, and Imberger J (2009) Modelling the propagation of turbidity density inflows into a stratified lake: Daecheong Reservoir, Korea. Environmental Modelling and Software 24(12): 1467--1482 (doi:10.1016/j.envsoft.2009.05.016). Cirmo CP and McDonnell JJ (1997) Linking the hydrologic and biogeochemical controls of nitrogen transport in near-stream zones of temperate-forested catchments: A review. Journal of Hydrology 199(12): 88--120. Clark KL, Nadkarni NM, Schaefer D, and Gholz HL (1998) Cloud water and precipitation chemistry in a tropical montane forest, Monteverde, Costa Rica. Atmospheric Environment 32(9): 1595--1603. Cle´ment J-C, Aquilina L, Bour O, Plaine K, Burt TP, and Pinay G (2003) Hydrological flowpaths and nitrate removal rates within a riparian floodplain along a fourth-order stream in Brittany (France). Hydrological Processes 17(6): 1177--1195. Cosby BJ, Wright RF, Hornberger GM, and Galloway JN (1985) Modeling the effects of acid deposition: Estimation of long-term water quality responses in a small forested catchment. Water Resources Research 21(11): 1591--1601. Covich AP, Palmer MA, and Crowl TA (1999) The role of benthic invertebrate species in freshwater ecosystems: Zoobenthic species influence energy flows and nutrient cycling. BioScience 49(2): 119--127.

298

Hydrology and Biogeochemistry Linkages

Dahm CN, Grimm NB, Marmonier P, Valett HM, and Vervier P (1998) Nutrient dynamics at the interface between surface waters and groundwaters. Freshwater Biology 40(3): 427--451. Damman A (1986) Hydrology, development, and biogeochemistry of ombrogenous peat bogs with special reference to nutrient relocation in a western Newfoundland bog. Canadian Journal of Botany 64(2). 384–294 Davis WM (1899) The geographical cycle. Geographical Journal 14: 481--504. Dawson TE (1993) Hydraulic lift and water use by plants: Implications for water balance, performance and plant–plant interactions. Oecologia 95(4): 565--574. Dawson TE (1996) Determining water use by trees and forests from isotopic, energy balance and transpiration analyses: The roles of tree size and hydraulic lift. Tree Physiology 16: 263--272. Dawson TE (1998) Fog in the California redwood forest: Ecosystem inputs and use by plants. Oecologia 117(4): 476--485. Dawson TE and Ehleringer JR (1998) Plants, isotopes and water use: A cachment-scale perspective. In: Kendall C and McDonnell JJ (eds.) Isotope Tracers in Catchment Hydrology, pp. 165--202. New York: Elservier. Dean WE, Neff BP, Rosenberry DO, Winter TC, and Parkhurst RS (2003) The significance of ground water to the accumulation of iron and manganese in the sediments of two hydrologically distinct lakes in North-Central Minnesota: A geological perspective. Ground Water 41: 951--963. DeHayes DH, Schaberg PG, Hawley GJ, and Strimbeck GR (1999) Acid rain impacts on calcium nutrition and forest health. BioScience 49(10): 789--800. De´camps H, Pinay G, Naiman RJ, et al. (2004) Riparian zones; where biogeochemistry meets biodiversity in management practice. Polish Journal of Ecology 52(1): 3--18. Devito KJ, Dillon PJ, and Lazerte BJ (1989) Phosphorus and nitrogen retention in five Precambrian shield wetlands. Biogeochemistry 8(3): 185--204. Dillon PJ and Molot LA (1997) Dissolved organic and inorganic carbon mass balances in central Ontario lakes. Biogeochemistry 36(1): 29--42. Dodds WK, Lobez AJ, Bowden WB, et al. (2002) Nitrogen uptake as a function of concentration in streams. Journal of the North American Benthological Society 21: 206--220. Driscoll CT, Baker JP, Bisogni JJJ, and Schofield CL (1980) Effect of aluminium speciation on fish in dilute acidified waters. Nature 284: 161--164. Driscoll CT, Fallon-Lambert K, Chen L (2005) Acidic deposition: Sources and effects. In: Anderson MG, McDonnell JJ (eds), Encyclopedia of Hydrological Sciences, vol. 3, Part 8, ch. 95, pp. 1441–1457. Chichester: Wiley. Driscoll CT, Lawrence GB, Bulger AJ, et al. (2001) Acidic deposition in the northeastern United States: Sources and inputs, ecosystem effects, and management strategies. BioScience 51(3): 180--198. Dudgeon D, Artheington AH, Gessner MO, et al. (2005) Freshwater biodiversity: Importance, threats, status and conservation challenges. Biological Reviews 81: 163--182. Duff JH, Hendricks SP, Jackman AP, and Triska FJ (2002) The effect of Elodea canadensis beds on porewater chemistry, microbial respiration, and nutrient retention in the Shingobee River, Minnesota, North America. Verhandlungen Internationale Vereinigung fur Theoretische und Angewandte Limnologie 28: 1--9. Duff JH and Triska FJ (1990) Denitrification in sediments from the hyporheic zone adjacent to a small forested stream. Canadian Journal of Fisheries and Aquatic Sciences 47(6): 1140--1147. Dunne T and Black RD (1970a) An experimental investigation of runoff production in permeable soils. Water Resources Research 6(2): 478--490. Dunne T and Black RD (1970b) Partial area contributions to storm runoff in a small New-England watershed. Water Resources Research 6: 1296--1305. Dunne T and Leopold LB (1978) Water in Environmental Planning. New York, NY: WH Freeman. 818pp. Durka W, Schulze E-D, Gebauer G, and Voerkeliust S (1994) Effects of forest decline on uptake and leaching of deposited nitrate determined from 15N and 18O measurements. Nature 372(6508): 765--767 (doi:10.1038/372765a0). Edmunds WM, Miles DL, and Cook JM (1984) A comparative study of sequential redox processes in three British aquifers from the United Kingdom. In: Eriksson E (ed.) Hydrochemical Balances of Freshwater Systems, vol. 150, pp. 55--70. Wallingford: IAHS. Edwards AC, Scalenghe R, and Freppaz M (2007) Changes in the seasonal snow cover of alpine regions and its effect on soil processes: A review. Quarternary International 162–163: 172--181. Edwards AC and Wetzel RL (2005) Nutrient cycling. In: Anderson MG and McDonnell JJ (eds.), Encyclopedia of Hydrological Sciences, vol. 3, Part 8, ch. 96, pp. 1459– 1477. Chichester: Wiley. Ehleringer JR and Dawson TE (1992) Water uptake by plants – perspectives from stable isotope composition. Plant, Cell and Environment 15(9): 1073--1082.

Elser JJ, Andersen T, Baron JS, et al. (2009a) Shifts in lake N:P stoichiometry and nutrient limitation driven by atmospheric nitrogen deposition. Science 326(5954): 835--837. Elser JJ, Kyle M, Steger L, Nydick KR, and Baron JS (2009b) Nutrient availability and phytoplankton nutrient limitation across a gradient of atmospheric nitrogen deposition. Ecology 90(11): 3062--3073. Emerman SH and Dawson TE (1996) Hydraulic lift and its influence on the water content of the rhizosphere: An example from sugar maple. Acer saccharum. Oecologia, 108: 273--278. Emmett BA (1999) The impact of nitrogen on forest soils and feedbacks on tree growth. Water, Air and Soil Pollution 116(1–2): 65--74. Emmett BA, Boxman D, Bredemeier M, et al. (1998a) Predicting the effects of atmospheric nitrogen deposition in conifer stands: Evidence from the NITREX ecosystem-scale experiments. Ecosystems 1(4): 352--360. Emmett BA, Reynolds B, Silgram M, Sparks TH, and Woods C (1998b) The consequences of chronic nitrogen additions on N cycling and soilwater chemistry in a Sitka spruce stand, North Wales. Forest Ecology and Management 101(1–3): 165--175. Emmett BA, Reynolds B, Stevens PA, et al. (1993) Nitrate leaching from afforested Welsh catchments: Interaction between stand age and nitrogen deposition. AMBIO 22(6): 386--394. Ensign SH and Doyle MW (2006) Nutrient spiraling in streams and river networks. Journal of Geophysical Research 111: G04009 (doi:10.1029/2005JG000114). Euliss NH Jr., LaBaugh JW, Fredrickson LH, et al. (2004) The wetland continuum: A conceptual framework for interpreting biological studies. Wetlands 24: 448--458. Evans MG, Burt TP, Holden J, and Adamson JK (1999) Runoff generation and water table fluctuations in blanket peat: Evidence from UK data spanning the dry summer of 1995. Journal of Hydrology 221: 141--160. Everts CJ, Kanwar RS, Alexander ECJ, and Alexander SC (1989) Comparison of tracer mobilities under laboratory and field conditions. Journal of Environmental Quality 18(4): 491--498. Ewing HA, Weathers KC, Templer PH, et al. (2009) Fog water and ecosystem function: Heterogeneity in a California Redwood Forest. Ecosystems 12(3): 417--433. Federer CA, Hornbeck JW, Tritton LM, Martin CW, Pierce RS, and Smith CT (1989) Long-term depletion of calcium and other nutrients in eastern US forests. Environmental Management 13(5): 593--601. Finkenbine JK, Atwater JW, and Mavinic DS (2000) Stream health after urbanization. Journal of the American Water Resources Association 36(5): 1149--1160. Fisher SG, Grimm NB, Marti E, Holmes RM, and Jones JB (1998) Material spiraling in stream corridors: A telescoping ecosystem model. Ecosystems 1(1): 19--34. Fraser CJD, Roulet NT, and Lafleur M (2001) Groundwater flow patterns in a large peatland. Journal of Hydrology 246: 142--154. Freeman C, Evans CD, Monteith DT, Reynolds B, and Fenner N (2001) Export of organic carbon from peat soils. Nature 412(6849): 785 (doi:10.1038/35090628). Friedman JM, Osterkamp WR, Scott ML, and Auble GT (1998) Downstream effects of dams on channel geometry and bottomland vegetation: Regional patterns in the Great Plains. Wetlands 18(4): 619--633. Fromm PO (1980) A review of some physiological and toxicological responses of freshwater fish to acid stress. Environmental Biology of Fishes 5(1): 79--93. Frost PC, Larson JH, Johnston CA, et al. (2006) Landscape predictors of stream dissolved organic matter concentration and physicochemistry in a Lake Superior river watershed. Aquatic Sciences 68(1): 40--51 (doi:10.1007/s00027-005-0802-5). Fukada T, Hiscock KM, Dennis PF, and Grischek T (2003) A dual isotope approach to identify denitrification in groundwater at a river-bank infiltration site. Water Research 37: 3070--3078. Fuller CC and Harvey JW (2000) Reactive uptake of trace metals in the hyporheic zone of a mining-contaminated stream, Pinal Creek, Arizona. Environmental Science and Technology 34(7): 1150--1155. Galloway ME and Branfireun BA (2004) Mercury dynamics of a temperate forest wetland. Science of the Total Environment 325: 239--254. Garcia-Martino AR, Warner GS, Scatena FN, and Civco DL (1996) Rainfall, runoff and elevation relationships in the Luquillo Mountains of Puerto Rico. Caribbean Journal of Science 32(4): 413--424. Gee GW and Hillel D (1988) Groundwater recharge in arid regions – review and critique of estimation methods. Hydrological Processes 2(3): 255--266. Gordon DR (1998) Effects of invasive, non-indigenous plant species on ecosystem processes: Lessons from Florida. Ecological Applications 8(4): 975--989. Goulding KWT, Bailey NJ, Bradbury NJ, et al. (1998) Nitrogen deposition and its contribution to nitrogen cycling and associated processes. New Phytologist 139: 49--58. Grace J (1989) Tree lines. Philosophical Transactions of the Royal Society, London B 324: 233--245.

Hydrology and Biogeochemistry Linkages

Granger J, Sigman DM, Lehmann MF, and Torell PD (2008) Nitrogen and oxygen isotope fractionation during dissimilatory nitrate reduction by denitrifying bacteria. Limnology and Oceanography 53(6): 2533--2545. Graveland J (1998) Effects of acid rain on bird populations. Environmental Reviews 6(1): 41--54. Green CT, Bo¨hlke JK, Bekins B and Phillips, S. (2010) Mixing effects on apparent reaction rates and isotope fractionation during denitrification in a heterogeneous aquifer. Water Resources Research 46 (doi:0.1029/2009WR008903). Green CT, Puckett LJ, Bo¨hlke JK, et al. (2008) Limited occurrence of denitrification in four shallow aquifers in agricultural areas of the United States. Journal of Environmental Quality 37: 994--1007. Grimm NB, Sheibley RW, Crenshaw CL, Dahm CN, Roach WJ, and Zeglin LH (2005) N retention and transformation in urban streams. Journal of the North American Benthological Society 24: 626--642. Grobblelaar JU and House WA (1995) Phosphorus as a limiting resource in inland waters; interactions with nitrogen. In: Tiessen H (ed.) Phosphorus in the Global Environment: Transfers, Cycles and Management, pp. 255--276. Chichester: Wiley. Groffman PM, Boulware NJ, Zipperer WC, Puyat RV, Band LE, and Colosimo MF (2002) Soil nitrogen cycle processes in urban riparian zones. Environmental Science and Technology 36(21): 4547--4552. Groffman PM, Driscoll CT, Fahey TJ, Hardy JP, Fitzhugh RD, and Tierney GL (2001a) Colder soils in a warmer world: A snow manipulation study in a northern hardwood forest ecosystem. Biogeochemistry 56(2): 135--150. Groffman PM, Driscoll CT, Fahey TJ, Hardy JP, Fitzhugh RD, and Tierney GL (2001b) Effects of mild winter freezing on soil nitrogen and carbon dynamics in a northern hardwood forest. Biogeochemistry 56(2): 191--213. Groffman PM, Hardy JP, Nolan SS, Fitzhugh RD, Driscoll CT, and Fahay TJ (1999) Snow depth, soil frost and nutrient loss in a northern hardwood forest. Hydrological Processes 13(14): 2275--2286. Groffman PM, Law NL, Belt KT, Band LE, and Fisher GT (2004) Nitrogen fluxes and retention in urban watershed ecosystems. Ecosystems 7(4): 393--403. Grogan P and Chapin FS (1999) Arctic soil respiration: Effects of climate and vegetation depend on season. Ecosystems 2(5): 451--459. Grogan P and Jonasson S (2003) Controls on annual nitrogen cycling in the understory of a subarctic birch forest. Ecology 84(1): 202--218. Hall ROJ, Tank JL, Sobota DJ, et al. (2009) Nitrate removal in stream ecosystems measured by 15N addition experiments: Total uptake. Limnology and Oceanography 54: 653--665. Hamburg SP, Yanai RD, Arthur MA, Blum JD, and Siccama TG (2003) Biotic control of calcium cycling in northern hardwood forests: Acid rain and aging forests. Ecosystems 6(4): 399--406 (doi:10.1007/s10021-002-0174-9). Hamilton DP and Mitchell SF (1997) Wave-induced shear stresses, plant nutrients and chlorophyll in seven shallow lakes. Freshwater Biology 38: 159--168. Hamilton DP, Schladow SG, and Fisher IH (1995) Controlling the indirect effects of flow diversions on water quality in an Australian reservoir. Environment International 21(5): 583--590. Harvey JW and Bencala KE (1993) The effect of streambed topography on surface– subsurface water exchange in mountain catchments. Water Resources Research 29(1): 89--98. Heath JA and Huebert BJ (1999) Cloudwater deposition as a source of fixed nitrogen in a Hawaiian montane forest. Biogeochemistry 44(2): 119--134. Heath RH, Kahl SA, Noerton SA, and Fernandez IJ (1992) Episodic stream acidification caused by atmophseric deposition of sea salts at Acadia National Park, Maine, United States. Water Resources Research 28(4): 1081--1088. Heathwaite AL, Johnes PJ, and Peters NE (1996) Trends in nutrients. Hydrological Processes 10(2): 263--293. Hedin LO, von Fischer JC, Ostrom NE, Kennedy BP, Brown MG, and Robertson GP (1998) Thermodynamic constraints on nitrogen transformations and other biogeochemical processes at soil–stream interfaces. Ecology 79(2): 684--703. Herckes P, Wendling R, Sauret N, Mirabel P, and Wortham H (2002) Cloudwater studies at a high elevation site in the Vosges Mountains (France). Environmental Pollution 117(1): 169--177. Herwitz SR (1986) Infiltration-excess caused by stemflow in a cyclone-prone tropical rainforest. Earth Surface Processes and Landforms 11(4): 401--412. Heuer K, Brooks PD, and Tonnessen KA (1999) Nitrogen dynamics in two high elevation catchments during spring snowmelt 1996, Rocky Mountains, Colorado. Hydrological Processes 13(14): 2203--2214. Hewlett JD (1961) Watershed management. Report for 1961 Southeastern Forest Experiment Station. Asheville, NC: US Department of Agriculture, Forest Service. Higgins CG (1984) Piping and sapping: Development of landforms by groundwater outflow. In: La Fleur RG (ed.) Groundwater as a Geomorphic Agent, pp. 18--58. London: Allen and Unwin.

299

Hill AR and Cardaci M (2004) Denitrification and organic carbon availability in riparian wetland soils and subsurface sediments. Soil Science Society of America Journal 68(1): 320--325. Hindar A, Henriksen A, Kaste Ø, and Tørseth K (1995) Extreme acidification in small catchments in southwestern Norway associated with a sea salt episode. Water, Air and Soil Pollution 85(2): 547--552. Hinton MJ, Schiff SL, and English MC (1998) Sources and flowpaths of dissolved organic carbon during storms in two forested watersheds of the Precambrian Shield. Biogeochemistry 41(2): 175--197. Hoham RW and Duval B (2001) Microbial ecology of snow and freshwater ice with emphasis on snow algae. In: Jones HG, Pomeroy JW, Walker DA, and Hoham RW (eds.) Snow Ecology: An Interdisciplinary Examination of Snow-Covered Ecosystems, pp. 166--226. Cambridge: Cambridge University Press. Hoham RW, Marcarelli AM, Rogers HS, et al. (2000) The importance of light and photoperiod in sexual reproduction and geographical distribution in the green snow alga, Chloromonas sp.-D (Chlorophyceae, Volvocales). Hydrological Processes 14(18): 3309--3321. Hoham RW, Schlag EM, Kang JY, et al. (1998) The effects of irradiance levels and spectral composition on mating strategies in the snow alga, Chloromonas sp.-D, from the Tughill Plateau, New York State. Hydrological Processes 12(10–11): 1627--1639. Holden J and Burt TP (2003) Runoff production in blanket peat covered catchments. Water Resources Research 39(7). SWC 6.1–6.9. Holden J, Burt TP, Evans MG, and Horton M (2006) Impact of land drainage on peatland hydrology. Journal of Environmental Quality 35: 1764--1778. Holland PG and Steyn DG (1975) Vegetational responses to latitudinal variations in slope angle and aspect. Journal of Biogeography 2(3): 179--183. Holt C and Yan ND (2003) Recovery of crustacean zooplankton communities from acidification in Killarney Park, Ontario, 1971–2000: pH 6 as a recovery goal. AMBIO 32(3): 203--207. Hood EW, Gooseff MN, and Johnson SL (2006) Changes in the character of stream water dissolved organic carbon during flushing in three small watersheds, Oregon. Journal of Geophysical Research – Biosciences 111: G01007 (doi:10.1029/ 2005JG000082). Hook PB, Olson BE, and Wraith JM (2004) Effects of the invasive forb Centaurea maculosa on grassland carbon and nitrogen pools in Montana, USA. Ecosystems 7(6): 686--694. Hooper RP, Aulenbach BT, Burns DA, et al. (1998) Riparian control of stream-water chemistry: Implications for hydrochemical basin models. In: Kovar K, Tappeiner U, Peters NE, and Craig RG (eds.) Hydrology, Water Resources and Ecology in Headwaters, vol. 248, pp. 451--458. Wallingford: IAHS. Hooper RP, Christophersen N, and Peters NE (1990) Modelling streamwater chemistry as a mixture of soilwater end-members – an application to the Panola Mountain catchment, Georgia, U. S. A. Journal of Hydrology 116(1): 321--343. Hornberger GM, Bencala KE, and Mcknight DM (1994) Hydrological controls on dissolved organic carbon during snowmelt in the Snake River near Montezuma, Colorado. Biogeochemistry 25(3): 147--165. Horton RE (1933) The role of infiltration in the hydrologic cycle. Transactions of the American Geophysical Union 14: 446--460. Horton RE (1945) Erosional development of streams and their drainage basins: Hydrophysical approach to quantitative morphology. Bulletin of the Geological Society of America 56: 275--370. Hultine KR, Williams DG, Burgess SSO, and Keefer TO (2003) Contrasting patterns of hydraulic redistribution in three desert phreatophytes. Oecologia 135: 167--175. Huntington TG, Hooper RP, and Aulenbach BT (1994) Hydrologic processes controlling sulfate mobility in a small forested watershed. Water Resources Research 30(2): 283--295. Ingram HAP (1978) Soil layers in mires: Function and terminology. Journal of Soil Science 29: 224--227. Istok JD, Humphrey MD, Schroth MH, Hyman MR, and O’Reilly KT (1997) Single-well, ‘‘Push–Pull’’ test for in situ determination of microbial activities. Ground Water 35(4): 619--631. Jackson RB, Sperry JS, and Dawson TE (2000) Root water uptake and transport: Using physiological processes in global predictions. Trends in Plant Science 5(11): 482--488. Jansson M, Andersson R, Berggren H, and Leonardson L (1994) Wetlands and lakes as nitrogen traps. AMBIO 23(6): 320--326. Jardine PM, Wilson GV, and Luxmoore RJ (1990) Unsaturated solute transport through a forest soil during rain storm events. Geoderma 46: 103--118. Jeffries DS (1989) Snowpack storage of pollutants, release during melting, and impact on receiving waters. In: Norton SA, Lindberg SE, and Page AL (eds.) Acidic Precipitation, Volume 4: Soils, Aquatic Processes and Lake Acidification, pp. 107--154. New York: Springer.

300

Hydrology and Biogeochemistry Linkages

Jeffries DS, Clair TA, Couture S, et al. (2003) Assessing the recovery of lakes in southeastern Canada from the effects of acidic deposition. AMBIO 32(3): 176--182. Jeziorski A and Yan ND (2006) Species identity and aqueous calcium concentrations as determinants of calcium concentrations of freshwater crustacean zooplankton. Canadian Journal of Fisheries and Aquatic Sciences 63(5): 1007--1013. Jeziorski A, Yan ND, Paterson AM, et al. (2008) The widespread threat of calcium decline in fresh waters. Science 322(5902): 1374--1377. Johannessen M, Dale T, Gjessing ET, Henriksen A, and Wright RF (1975) Acid precipitation in Norway: The regional distribution of contaminants in snow and the chemical concentration processes during snowmelt. In: Nye JF, Miiller F, and Oeschger H (eds.) Isotopes and Impurities in Snow and Ice, vol. 118, pp. 116--120. Wallingford: IAHS. Johannessen M and Henriksen A (1978) Chemistry of snow meltwater: Changes in concentration during melting. Water Resources Research 14(4): 615--619. Jones HG (1999) The ecology of snow-covered systems: A brief overview of nutrient cycling and life in the cold. Hydrological Processes 13: 2135--2147. Jones JB, Holmes RM, Fisher SG, Grimm NB, and Greene DM (1995) Methanogenesis in Arizona, USA dryland streams. Biogeochemistry 31(3): 155--173. Jones JB and Mulholland PJ (2000) Streams and Ground Waters. San Diego, CA: Academic Press. 425pp. Jones JBJ and Holmes RM (1996) Surface–subsurface interactions in stream ecosystems. Trends in Ecology and Evolution 11(6): 239--242. Joslin JD and Wolfe MH (1992) Red spruce soil solution chemistry and root distribution across a cloud water deposition gradient. Canadian Journal of Forest Research 22(6): 893--904. Kasahara T and Wondzell SM (2003) Geomorphic controls on hyporheic exchange flow in mountain streams. Water Resources Research 39: (doi, 10.1029/2002WR001386). Katz BG, Plummer LN, Busenberg E, Revesz KM, Jones BF, and Lee TM (1995) Chemical evolution of groundwater near a sinkhole lake, northern Florida, 2. Chemical patterns, mass transfer modeling, and rates of mass transfer reactions. Water Resources Research 31(6): 1565--1584. Keller W, Yan ND, Somers KM, and Heneberry JH (2003) Crustacean zooplankton communities in lakes recovering from acidification. Canadian Journal of Fisheries and Aquatic Sciences 60(11): 1307--1313 (doi: 10.1139/f03-111). Kellogg DQ, Gold AJ, Groffman PM, Addy K, Stolt MH, and Blazejewski G (2005) In situ ground water denitrification in stratified, permeable soil underlying riparian wetlands. Journal of Environmental Quality 34(2): 524–533. Kendall C (1998) Tracing nitrogen sources and cycling in catchments. In: Kendall C and McDonnell JJ (eds.) Isotope Tracers in Catchment Hydrology, pp. 519--576. New York: Elservier. Kendall C, Elliott EM, and Wankel SD (2007) Tracing anthropogenic inputs of nitrogen to ecosystems. In: Michener RH and Lajtha K (eds.) Stable Isotopes in Ecology and Environmental Science, 2nd edn., pp. 375--449. Malden, MA: Blackwell. Klemm O, Wrzesinsky T, and Scheer C (2005) Fog water flux at a canopy top: Direct measurement versus one-dimensional model. Atmospheric Environment 39(29): 5375--5386. Lampert W and Sommer U (2007) Limnoecology. The Ecology of Lakes and Streams. Oxford: Oxford University Press. 336pp. Larssen T (2005) Model prognoses for future acidification recovery of surface waters in Norway using long-term monitoring data. Environmental Science and Technology 39: 7970--7979. Larssen T and Holme J (2005) Afforestation, seasalt episodes and acidification – a paired catchment study in western Norway. Environmental Pollution 139(3): 440--450. Lawrence GB (2002) Persistent episodic acidification of streams linked to acid rain effects on soil. Atmospheric Environment 36(10): 1589--1598. Lawrence GB, David MB, Lovett GM, et al. (1999) Soil calcium status and the response of stream chemistry to changing acidic deposition rates. Ecological Applications 9(3): 1059--1072. Lean DRS (1973) Phosphorus dynamics in lake water. Science 79: 678--680. Levine JM, Vila` M, D’Antonio CM, Dukes JS, Grigulis K, and Lavorel S (2003) Mechanisms underlying the impacts of exotic plant invasions. Proceedings of the Royal Society of London B 270(1517): 775--781. Lewis WM, Saunders JS, and McCutchan JH (2008) Application of a nutrient saturation concept to the control of algal growth in lakes. Lake and Reservoir Management 24: 41--46. Lewis WM and Wurtsbaugh WA (2008) Control of lacustrine phytoplankton by nutrients: Erosion of the phosphorus paradigm. International Review of Hydrobiology 93(4–5): 446--465. Likens GE (1972) Nutrients and eutrophication – the limiting-nutrient controversy. Limnology and Oceanography (Special Symposium) 1: 1--328. Likens GE and Bormann FH (1995) Biogeochemistry of a Forested Ecosystem, 2nd edn. New York: Springer. 170pp.

Lindsey BD, Phillips SW, Donnelly CA, et al. (2003) Residence times and nitrate transport in ground water discharging to streams in the Chesapeake Bay Watershed. US Geological Survey Water Resources Invesitagions Report 03-4035, 201pp. Denver, CO: US Geological Survey Branch of Information Services. Lipson GE, Schmidt SK, and Monson RK (1999) Links between microbial population dynamics and nitrogen availability in an alpine ecosystem. Ecology 80(5): 1623--1631. Lischeid G, Kolb A, Alewell C, and Paul S (2007) Impact of redox and transport processes in a riparian wetland on stream water quality in the Fichtelgebirge region, southern Germany. Hydrological Processes 21(1): 123--132. Liston GE (2004) Representing subgrid snow cover heterogeneities in regional and global models. Journal of Climate 17(6): 1381--1397. Liston GE and Elder K (2006) A distributed snow-evolution modeling system (SnowModel). Journal of Hydrometeorology 7(6): 1259--1276. Loague K and Corwin DL (2005) Point and nonpoint source pollution. In: Anderson, MG and McDonnell JJ (eds.) Encyclopedia of Hydrological Sciences, vol. 3, Part 8, ch. 95 (doi:10.1002/0470848944.hsa097). Chichester: Wiley. Lohse KA, Brooks PD, McIntosh JC, Meixner T, and Huxman TE (2009) Interactions between biogeochemistry and hydrologic systems. Annual Review of Environment and Resources 34: 65--96. Lovett GM (1984) Rates and mechanisms of cloud water deposition to a subalpine balsam fir forest. Atmospheric Environment 18(2): 361--371. Lovett GM and Kinsman JD (1990) Atmospheric pollutant deposition to high elevation ecosystems. Atmospheric Environment 24A: 2767--2786. Lovett GM, Likens GE, Buso DC, Driscoll CT, and Bailey SW (2005) The biogeochemistry of chlorine at Hubbard Brook, New Hampshire, USA. Biogeochemistry 72: 191--232. Luxmoore RJ, Jardine PM, Wilson GV, Jones JR, and Zelazny LW (1990) Physical and chemical controls of preferred path flow through a forested hillslope. Geoderma 46: 139--154. Lynch JA and Corbett ES (1989) Hydrologic control of sulfate mobility in a forested watershed. Water Resources Research 25(7): 1695--1703. MacDonald JA, Fowler D, Hargreaves KJ, Skiba U, Leith ID, and Murray MB (1998) Methane emission rates from a northern wetland; response to temperature, water table and transport. Atmospheric Environment 32(1): 3219--3227. Madsen JD, Chambers PA, James WF, Koch EW, and Westlake DF (2001) The interaction between water movement, sediment dynamics and submerged macrophytes. Hydrobiologia 444: 71--84. Malard F, Tockner K, Dole-Olivier M-J, and Ward JV (2002) A landscape perspective of surface–subsurface hydrological exchanges in river corridors. Freshwater Biology 47(4): 621--640. Maloszewski P, Rauert W, Stichler W, and Herrmann A (1983) Application of flow models in an alpine catchment area using tritium and deuterium data. Journal of Hydrology 66: 319--330. Manheim FT, Krantz DE, and Bratton JF (2004) Studying ground water under Delmarva coastal bays using electrical resistivity. Ground Water 42(7): 1052--1068. Mariotti A, Landreau A, and Simon B (1988) 15N isotope biogeochemistry and natural denitrification process in groundwater: Application to the chalk aquifer of northern France. Geochimica et Cosmochimica Acta 52(7): 1869–1878. Marti E, Aumatell J, Gode L, Poch M, and Sabater F (2004) Nutrient retention efficiency in streams receiving inputs from wastewater treatment plants. Journal of Environmental Quality 33(1): 285--293. Martin JB, Cable JE, Jaeger J, Hartl K, and Smith CG (2006) Thermal and chemical evidence for rapid water exchange across the sediment–water interface by bioirrigation in the Indian River Lagoon, Florida. Limnology and Oceanography 51(3): 1332--1341. Mayer B, Boyer EW, Goodale C, et al. (2002) Sources of nitrate in rivers draining sixteen watersheds in the northeastern U.S.: Isotopic constraints. Biogeochemistry 57/58: 171--192. McClain ME, Boyer EW, Dent CL, et al. (2003) Biogeochemical hot spots and hot moments at the interface of terrestrial and aquatic ecosystems. Ecosystems 6(4): 301--312. McDonnell JJ, Sivapalan M, Vache´ K, et al. (2007) Moving beyond heterogeneity and process complexity: A new vision for watershed hydrology. Water Resources Research 43: W07301 (doi: 10.1029/2006WR005467). McDonnell JJ, Stewart MK, and Owen IF (1991) Effect of catchment-scale subsurface mixing on stream isotopic response. Water Resources Research 27(12): 3065--3073. McDowell NG, White S, and Pockman WT (2008) Transpiration and stomatal conductance across a steep climate gradient in the southern Rocky Mountains. Ecohydrology 1(3): 193--204. McGlynn BL, McDonnell J, Stewart M, and Seibert J (2003) On the relationships between catchment scale and streamwater mean residence time. Hydrological Processes 17(1): 175--181.

Hydrology and Biogeochemistry Linkages

McGuire KJ and McDonnell JJ (2006) A review and evaluation of catchment transit time modeling. Journal of Hydrology 330(3–4): 543--563. McGuire KJ, McDonnell JJ, Weiler M, et al. (2005) The role of topography on catchment-scale water residence time. Water Resource Research 41: W05002 (doi: 10.1029/2004WR003657). McIntosh J, McDonnell JJ, and Peters NE (1999) Tracer and hydrometric study of preferential flow in large undisturbed soil cores from the Georgia Piedmont, USA. Hydrological Processes 13(2): 139--155. McMahon PB and Bo¨hlke JK (1996) Denitrification and mixing in a stream–aquifer system: Effects on nitrate loading to surface water. Journal of Hydrology 186: 105--128. McMahon PB and Chapelle FH (2008) Redox processes and water quality of selected principal aquifer systems. Ground Water 46: 259--271. McMahon PG (2001) Aquifer/aquitard interfaces: Mixing zones that enhance biogeochemical reactions. Hydrogeology Journal 9(1): 34--43. McMaster ME (2001) A review of the evidence for endocrine disruption in Canadian aquatic ecosystems. Water Quality Research Journal of Canada 36(2): 215--231. Meinzer FC, Clearwater MJ, and Goldstein G (2001) Water transport in trees: Current perspectives, new insights and some controversies. Environmental and Experimental Botany 45: 239--262. Meybeck M (1987) Global chemical weathering of surficial rocks estimated from river dissolved loads. American Journal of Science 287: 401--428. Meybeck M (2001) River basins under anthropocene conditions. In: von Bodungen B and Turner RK (eds.) Science and Integrated Coastal Management, pp. 275--294. Berlin: Dahlem University Press. Meybeck M, Peters NE, and Chapman DV (2005) Water quality. In: Anderson MG and McDonnell JJ (eds.) Encyclopedia of Hydrological Sciences, vol. 3, part 8, ch. 91, pp. 1373–1385. Chichester: Wiley. Meybeck M and Vo¨ro¨smarty CJ (2005) Fluvial filtering of land-to-ocean fluxes: From natural Holocene variations to Anthropocene. CR Geoscience 337: 107--123 (doi: 10.1016/j.crte.2004.09.016). Meyer JL, Paul MJ, and Taulbee WK (2005) Stream ecosystem function in urbanizing landscapes. Journal of the North American Benthological Society 24(3): 602--612. Meysman FJR, Galaktionov OS, Gribsholt B, and Middelburg JJ (2006) Bioirrigation in permeable sediments: Advective pore-water transport induced by burrow ventilation. Limnology and Oceanography 51(1): 142--156. Michalski G, Meixner T, Fenn M, et al. (2004) Tracing atmospheric nitrate depsition in a complex semiarid ecosystem using D17O. Environmental Science and Technology 38(7): 2175--2181. Millenium Ecosystem Assessment E (2005) Ecosystems and Human Well-Being: Synthesis. Washington, DC: Island Press. 137pp. Millikin Ishikawa C and Bledsoe CS (2000) Seasonal and diurnal patterns of soil water potential in the rhizosphere of blue oaks: Evidence for hydraulic lift. Oecologia 125: 459--465. Mitchell MJ (2001) Linkages of nitrate losses in watersheds to hydrological processes. Hydrological Processes 15(17): 3305--3307. Mitsch WJ and Gosselink J (2007) Wetlands. Hoboken, NJ: Wiley. Molotch NP and Bales RC (2005) Scaling snow observations from the point to the grid element: Implications for observation network design. Water Resources Research 41: W11421 (doi: 10.1029/2005WR004229). Molotch NP and Bales RC (2006) SNOTEL representativeness in the Rio Grande headwaters on the basis of physiographics and remotely sensed snow cover persistence. Hydrological Processes 20(4): 723--739. Monson RK, Turnipseed AA, Sparks JP, et al. (2002) Carbon sequestration in a highelevation, subalpine forest. Global Change Biology 8(5): 459--478. Mooney HA and Hobbs RJ (2000) Invasive Species in a Changing World. Washington, DC: Island Press. 457pp. Mooney HA, Mack RN, McNeely JK, et al. (eds.) (2005) Invasive Alien Species: A New Synthesis. Washington, DC: Island Press. Moore TR (2003) Dissolved organic carbon in a boreal landscape. Global Biogeochemical Cycles 17(4): 1109 (doi: 10.1029.2003GB002050). Moreira MZ, Sternberg LdSL, and Nepstad DC (2000) Vertical patterns of soil water uptake by plants in a primary forest and an abandoned pasture in the eastern Amazon: An isotopic approach. Plant and Soil 222: 95--107. Mortimer CH (1941) The exchange of dissolved substances between mud and water in lakes. I and II. Journal of Ecology 29(2): 280--329. Mortimer CH (1942) The exchange of dissolved substances between mud and water in lakes. III and IV. Journal of Ecology 30: 147--201. Mulholland PJ, Hall ROJ, Sobota DJ, et al. (2009) Nitrate removal in stream ecosystems measured by 15N addition experiments: Denitrification. Limnology and Oceanography 54: 666--680. Mulholland PJ, Helton AM, Poole GC, et al. (2008) Excess nitrate from agricultural and urban areas reduces denitrification efficiency in streams. Nature 452: 202--205.

301

Mulholland PJ, Tank JL, Sanzone DM, et al. (2000) Nitrogen cycling in a forest stream determined by a N-15 tracer addition. Ecological Monographs 70(3): 471--493. Mulholland PJ, Valett HM, Webster JR, et al. (2004) Stream denitrification and total nitrate uptake rates measured suing a field 15N isotope tracer approach. Limnology and Oceanography 49: 809--820. Naiman RJ, Bilby RE, Schindler DE, and Helfield JM (2002) Pacific salmon, nutrients, and the dynamics of freshwater and riparian ecosystems. Ecosystems 5(4): 399--417. Newbold JD, Elwood JW, O’Neill RV, and Sheldon AL (1983) Phosphorus dynamics in a woodland stream ecosystem: A study of nutrient spiralling. Ecology 64: 1249--1265. Newbold JD, Elwood JW, O’Neill RV, and VanWinkle W (1981) Measuring nutrient spiralling in streams. Canadian Journal of Fisheries and Aquatic Sciences 38: 860--863. Nilsson C and Berggren K (2000) Alterations of riparian ecosystems caused by river regulation. BioScience 50(9): 783--792. Nilsson M, Sagerfors J, Buffman I, et al. (2008) Contemporary carbon accumulation in a boreal oligotrophic minerogenic mire – a significant sink after accounting for all C-fluxes. Global Change Biology 14(10): 2317--2332. Novotny V (2003) Water Quality: Diffuse Pollution and Watershed Management, 2nd edn. New York: Wiley. 864pp. O’Sullivan PE (1995) Eutrophication. International Journal of Environmental Studies 47: 173--195. Oliveira RS, Dawson TE, Burgess SSO, and Nepstad DC (2005) Hydraulic redistribution in three Amazonian trees. Oecologia 145: 354--363. Oyarzu´n CE, Godoy R, Schrijver AD, Staelens J, and Lust N (2004) Water chemistry and nutrient budgets in an undisturbed evergreen rainforest of southern chile. Biogeochemistry 71(1): 107--123. O¨berg G (2003) The biogeochemistry of chlorine in soil. In: Gribble GW (ed.) The Handbook of Environmental Chemistry, Volume 3: Natural Production of Organohalogen Compounds, pp. 43--62. Berlin: Springer. O¨berg G, Nordlund E, and Berg B (1996) In situ formation of organically bound halogens during decomposition of Norway spruce needles: Effects of fertilization. Canadian Journal of Forest Research 26(6): 1040--1048. O¨berg G and Sande´n P (2005) Retention of chloride in soil and cycling of organic matter-bound chlorine. Hydrological Processes 19(11): 2123--2136. Palmer MA, Reidy Liermann CA, Nilsson C, et al. (2008) Climate change and the world’s river basins: Anticipating management options. Frontiers in Ecology and the Environment 6: (doi:10.1890/060148). Paul MJ and Meyer JL (2001) Streams in the urban landscape. Annual Review Ecology and Systematics 32: 333--365. Penman HL (1961) Weather, plant and soil factors in hydrology. Weather 16: 207--219. Peters NE (1984) Comparison of air and soil temperatures at forested sites in the westcentral Adirondack Mountains. Northeastern Environmental Science 3(2): 67--72. Peters NE (1994) Water-quality variations in a forested Piedmont catchment, Georgia, USA. Journal of Hydrology 156: 73--90. Peters NE (2008) Water-quality monitoring and process understanding in support of environmental policy and management. In: Tchiguirinskaia I, Demuth S, and Hubert P (eds.) 9th Kovacs Colloquium: River Basins – From Hydrological Science to Water Management/Les bassins versant – de la science hydrologique a` la gestion des eaux, vol. 323, pp. 93--109. Paris: International Association of Hydrological Sciences. Peters NE, Cerny J, and Havel M (1998) Factors controlling streamwater nitrate concentrations in a forested headwater catchment, Krusne´ hory Mountains, Czech Republic. In: Haigh MJ, Krecek J, Rajwar GS, and Kilmartin MP (eds.) Headwaters: Water Resources and Soil Conservation, pp. 147--157. Rotterdam: AA Balkema. Peters NE and Driscoll CT (1987a) Hydrogeologic controls of surface-water chemistry in the Adirondack region of New York State. Biogeochemistry 3(1/3): 163--180. Peters NE and Driscoll CT (1987b) Sources of acidity during snowmelt at a forested site in the west-central Adirondack Mountains, New York. In: Swanson RH, Bernier PY, and Woodard PD (eds.) Forest Hydrology and Watershed Management, vol. 167, pp. 99--108. Wallingford: IAHS. Peters NE and Driscoll CT (1989) Temporal variations in solute concentrations of meltwater and forest floor leachate at a forested site in the Adirondacks, New York. In: Lewis JE (ed.) 46th Annual Eastern Snow Conference, Eastern Snow Conference, pp. 45–56. Quebec City, Canada. Peters NE and Leavesley GH (1995) Biotic and abiotic processes controlling water chemistry during snowmelt at Rabbit Ears Pass, Rocky Mountains, Colorado, USA. Water, Air, and Soil Pollution 79(1/4): 171--190. Peters NE and Meybeck M (2000) Water quality degradation effects on freshwater availability: Impacts of human activities. Water International 25(2): 185--193. Peters NE, Meybeck M and Chapman DV (2005) Effects of human activities on water quality. In: Anderson MG and McDonnell JJ (eds.) Encyclopedia of Hydrological Sciences, vol. 3, Part 8, ch. 93, pp. 1409–1425. Chichester: Wiley.

302

Hydrology and Biogeochemistry Linkages

Peters NE and Murdoch PS (1985) Hydrogeologic comparison of an acidic-lake basin with a neutral-lake basin in the west-central Adirondack Mountains. New York. Water, Air and Soil Pollution, 26: 387--402. Peters NE and Ratcliffe EB (1998) Tracing hydrologic pathways using chloride at the Panola Mountain Research Watershed, Georgia, USA. Water, Air, and Soil Pollution 105(1–2): 263--275. Peterson BJ, Wollheim WM, Mulholland PJ, et al. (2001) Control of nitrogen export from watersheds by headwater streams. Science 292(5514): 86--90. Pezeshki SR (2001) Wetland plant responses to soil flooding. Environmental and Experimental Botany 46(3): 299--312. Philip JR (1967) Sorption and infiltration in heterogeneous media. Australian Journal of Soil Research 5(1): 1--10. Phillips FM, Hogan JF, and Scanlon BR (2004) Section I: Introduction and overview. In: Hogan JF, Phillips FM, and Scanlon BR (eds.) Groundwater Recharge in a Desert Environment: The Southwestern United States, pp. 1--14. Washington, DC: American Geophysical Union. Pitelka LF (1994) Air pollution and terrestrial ecosystems. Ecological Applications 4(4): 627--628. Plummer LN, Busby JF, Lee RW, and Hanshaw BB (1990) Geochemical modeling of the Madison Aquifer in parts of Montana, Wyoming, and South Dakota. Water Resources Research 26(9): 1981–2014. Poff NL, Allan JD, Bain MB, et al. (1997) The natural flow regime – a paradigm for river conservation and restoration. BioScience 47(11): 769--784. Poff NL, Bledsoe BP, and Cuhaciyan CO (2006) Hydrologic variation with land use across the contiguous United States: Geomorphic and ecological consequences for stream ecosystems. Geomorphology 79: 264--285. Pomeroy JW, Davies TD, Jones HG, Marsh P, Peters NE, and Tranter M (1999) Transformations of snow chemistry in the boreal forest: Accumulation and volatilization. Hydrological Processes 13(14–15): 2257--2273. Pruppacher HR, Semonin RG, and Slinn WGN (1983) Precipitation Scavenging, Dry Deposition, and Resuspension, Volume1: Precipitation Scavenging London: Elsevier. 729pp. Pusch M, Fiebig D, Brettar I, et al. (1998) The role of micro-organisms in the ecological connectivity of running waters. Freshwater Biology 40(3): 453--495. Rademacher LK, Clark JF, Hudson GB, Erman DC, and Erman NA (2001) Chemical evolution of shallow groundwater as recorded by springs, Sagehen basin; Nevada County, California. Chemical Geology 179: 37--51. Rascher CM, Driscoll CT, and Peters NE (1987) Concentration and flux of solutes from snow and forest floor during snowmelt in the west-central Adirondack Region of New York. Biogeochemistry 3: 209--224. Rassam DW, Knight JH, Turner J, and Pagendam D (2006) Groundwater surface water interactions: Modelling denitrification and d15NNO3–d18ONO3 fractionation during bank storage. In: Institution of Engineers Australia (ed.) Proceedings of the 30th Hydrology and Water Resources Symposium, pp. 157–161. Launceston, TAS, Australia, 4–7 December 2006. Sandy Bay, TAS: Conference Design. Rassam DW, Pagendam DE, and Hunter HM (2008) Conceptualisation and application of models for groundwater–surface water interactions and nitrate attenuation potential in riparian zones. Environmental Modelling and Software 23(7): 859--875 (doi:10.1016/j.envsoft.2007.11.003). Reddy KR and DeLaune RD (2008) Biogeochemistry of Wetlands: Science and Applications. Boca Raton, FL: Taylor and Francis. Reeburgh WS (1983) Rates of biogeochemical processes in anoxic sediments. Annual Review of Earth and Planetary Sciences 11: 269--298. Reynolds B, Emmett BA, and Woods C (1992) Variations in streamwater nitrate concentrations and nitrogen budgets over 10 years in a headwater catchment in mid-Wales. Journal of Hydrology 136(1–4): 155--175. Reynolds B, Fowler D, and Thomas S (1996) Chemistry of cloud water at an upland site in mid-Wales. Science of the Total Environment 188: 115--125. Reynolds B, Neal C, and Norris DA (2001) Evaluation of regional acid sensitivity predictions using field data: Issues of scale and heterogeneity. Hydrology and Earth System Science 5(1): 75--81. Reynolds B, Ormerod SJ, and Gee AS (1994) Spatial patterns in stream nitrate concentrations in upland Wales in relation to catchment forest cover and forest age. Environmental Pollution 84(1): 27--33. Richard TL and Steenhuis TS (1988) Tile drain sampling of preferential flow on a field scale. Journal of Contaminant Hydrology 3: 307--325. Richards JH and Caldwell MM (1987) Hydraulic lift: Substantial nocturnal water transport between soil layers by Artemisia tridentata roots. Oecologia 73: 486--489. Richey JE, Melack JM, Aufdenkampe AK, Ballester VM, and Hess LL (2002) Outgassing from Amazonian rivers and wetlands as a large tropical source of atmospheric CO2. Nature 416(6881): 617--620.

Roberts BJ and Mulholland PJ (2007) In-stream biotic control on nutrient biogeochemistry in a forested headwater stream, West Fork of Walker Branch. Journal of Geophysical Research 112: G04002 (doi:10.1029/2007JG000422). Robertson K (2001) Emissions of N2O in Sweden – natural and anthropogenic sources. AMBIO 20(3–4): 151--155. Rose S and Peters NE (2001) Effects of urbanization on streamflow in the Atlanta area (Georgia, USA): A comparative hydrological approach. Hydrological Processes 15(8): 1441--1457. Rosenberry DO, Glaser PH, and Siegel DI (2006) The hydrology of northern peatlands as affected by biogenic gas: Current developments and research needs. Hydrological Processes 20(17): 3601--3610. Rosenberry DO and Labaugh JW (2008) Field techniques for estimating water fluxes between surface water and ground water. In: US Geological Survey (ed.) US Geological Survey Technical Water Resources Investigations, Book 4, ch. D2, 128pp. Washington, DC: USGS. Rosenberry DO and Winter TC (1997) Dynamics of water-table fluctuations in an upland between two prairie-pothole wetlands in North Dakota. Journal of Hydrology 191: 266--269. Roulet N, Lafleur P, Richard P, Moore T, Humphreys E, and Bubier J (2007) Contemporary carbon balance and late Holocene carbon accumulation in a northern peatland. Global Change Biology 13(2): 379--411. Roulet NT (1990) Hydrology of a headwater basin wetland: Groundwater discharge and wetland maintenance. Hydrological Processes 4(4): 387--400. Saunders DL and Kalff J (2001) Nitrogen retention in wetlands, lakes and rivers. Hydrobiologia 443: 205--212. Schimel J (2004) Playing scales in the methane cycle: From microbial ecology to the globe. Proceedings of the National Academy of Sciences of the United States of America 101(34): 12400--12401. Schindler DW (1988) Effects if acid rain on freshwater ecosystems. Science 239(4836): 149--157. Schindler DW (1999) From acid rain to toxic snow. AMBIO 28(4): 350--355. Schindler DW, Hecky RE, Findlay DL, et al. (2008) Eutrophication of lakes cannot be controlled by reducing nitrogen input: Results of a 37-year whole-ecosystem experiment. Proceedings of the National Academy of Sciences of the United States of America 105(32): 11254--11258. Schlesinger WH (1997) Biogeochemistry: An Analysis of Global Change, 2nd edn. San Diego, CA: Academic Press. Schnoor JL and Stumm W (1986) The role of chemical weathering in the neutralization of acidic deposition. Aquatic Sciences 48(2): 171--195. Scholl MA (2000) Effects of heterogeneity in aquifer permeability and biomass on biodegradation rate calculations – results from numerical simulations. Ground Water 38(5): 702–712. Scholl MA, Cozzarelli IM, and Christenson SC (2006) Recharge processes drive sulfate reduction in an alluvial aquifer contaminated with landfill leachate. Journal of Contaminant Hydrology 86(3–4): 239–261. Scholl MA, Giambelluca TW, Gingerich SB, Nullet MA, and Loope LL (2007) Cloud water in windward and leeward mountain forests: The stable isotope signature of orographic cloud water. Water Resources Research 43: W12411. Schulze E-D, Caldwell MM, Canadell J, et al. (1998) Downward flux of water through roots (i.e. inverse hydraulic lift) in dry Kalahari sands. Oecologia 115: 460--462. Schumm SA (1977) The Fluvial System. New York: Wiley. 338pp. Schuster PF, Reddy MM, LaBaugh JW, et al. (2003) Characterization of lake water and ground water movement in the littoral zone of Williams Lake, a closed-basin lake in north central Minnesota. Hydrological Processes 17(4): 823--838. Scott NA, Sagger S, and McIntosh PD (2001) Biogeochemical impact of Hieracium invasion in New Zealand’s grazed tussock grasslands: Sustainability implications. Ecological Applications 11(5): 1311--1322. Sebilo M, Billen G, Grably M, and Mariotti A (2003) Isotopic composition of nitrate– nitrogen as a marker of riparian and benthic denitrification at the scale of the whole Seine River system. Biogeochemistry 65: 35--51. Seitzinger S, Harrison JA, Bo¨hlke JK, et al. (2006) Denitrification across landscapes and waterscapes: A synthesis. Ecological Applications 16(5): 2064--2090. Shanley JB and Peters NE (1993) Variations in aqueous sulfate concentrations at Panola Mountain, Georgia. Journal of Hydrology 146: 361--382. Skjelkva˚le BL, Evans C, Larssen T, Hindar A, and Raddam GG (2003) Recovery from acidification in European surface waters: A view to the future. AMBIO 32(3): 170--175. Smith RL, Baumgartner LK, Miller DN, Repert DA, and Bo¨hlke JK (2006) Assessment of nitrification potential in ground water using short term, single-well injection experiments. Microbial Ecology 51: 22--35. Smith RL, Bo¨hlke JK, Garbedian SP, Revesz KM, and Yoshinari T (2004) Assessing denitrification in groundwater using natural gradient tracer tests with 15N: In situ measurement of a sequential multistep reaction. Water Resources Research 40: W07101 (doi:10.1029/2003WR002919).

Hydrology and Biogeochemistry Linkages

Smith RL, Garabedian SP, and Brooks MH (1996) Comparison of denitrification activity measurements in groundwater suing cores and natural-gradient tracer tests. Environmental Science and Technology 30(12): 3448--3456. Smith VH (1983) Low nitrogen to phosphorus ratios favor dominance by blue-green algae in lake phytoplankton. Science 221: 669--671. Sorrell BK, Mendelssohn IA, McKee KL, and Woods RA (2000) Ecophysiology of wetland plant roots: A modelling comparison of aeration in relation to species distribution. Annals of Botany 86(3): 675--685. Soulsby C, Tetzlaff D, Rodgers P, Dunn S, and Waldron S (2006) Runoff processes, stream water residence times and controlling landscape characteristics in a mesoscale catchment: An initial evaluation. Journal of Hydrology 325: 197--221. St. Louis VL, Rudd JWM, Kelly CA, Beaty KG, Flett RJ, and Roulet NT (1996) Production and loss of methylmercury and loss of total mercury from boreal forest catchments containing different types of wetlands. Environmental Science and Technology 30(9): 2719--2729. Stackelberg PE, Furlong ET, Meyer MT, Zaugg SD, Henderson AK, and Reissman DB (2004) Persistence of pharmaceutical compounds and other organic wastewater contaminants in a conventional drinking-water treatment plant. Science of the Total Environment 339: 99--113. Stanley EH, Fisher SG, and Grimm NB (1997) Ecosystem expansion and contraction in streams. BioScience 47(7): 427--435. Sterner RW and Elser JJ (2002) Ecological Stoichiometry: The Biology of Elements from Molecules to the Biosphere. Princeton, NJ: Princeton University Press. 466pp. Stevens PA, Norris DA, Sparks TH, and Hodgson AL (1994) The impacts of atmospheric N inputs on throughfall, soil and stream water interactions for different aged forest and moorland catchments in Wales. Water, Air and Soil Pollution 73(1): 297--317. Stockner JG and MacIsaac EA (1996) British Columbia lake enrichment programme: Two decades of habitat enhancement for sockeye salmon. Regulated Rivers: Research and Management 12: 547--561. Stoddard JL, Jeffries DS, Lukewille A, et al. (1999) Regional trends in aquatic recovery from acidification in North America and Europe. Nature (London) 401: 575--578. Strahler AN (1964) Quantitative geomorphology of drainage basins and channel networks. In: Chow VT (ed.) Handbook of Applied Hydrology, pp. 4--11. New York: McGraw-Hill. Stream Solute Workshop (1990) Concepts and methods for assessing solute dynamics in stream ecosystems. Journal of the North American Benthological Society 9: 95--119. Stumm W and Morgan JJ (1996) Aquatic Chemistry, 3rd edn. New York: Wiley. 1022pp. Svensson BH, Lantsheer JC, and Rodhe H (2001) Sources and sinks of methane in Sweden. AMBIO 20(3–4): 155--160. Szramek K and Walter LM (2004) Impact of carbonate precipitation on riverine inorganic carbon mass transport from a mid-continent, forested watershed. Aquatic Geochemistry 10(1–2): 99--137. Tabacchi E, Lambs L, Guilloy H, PlantyTabacchi AM, Muller E, and Decamps H (2000) Impacts of riparian vegetation on hydrological processes. Hydrological Processes 14(16–17): 2959--2976. Tank JL, Rosi-Marshall EJ, Baker MA, and Hall RO (2008) Are rivers just big streams? A pulse method to quantify nitrogen demand in a large river. Ecology 89(10): 2935--2945 (doi:10.1890/07-1315.1). Tetzlaff D, Seibert J, and Soulsby C (2009) Inter-catchment comparison to assess the influence of topography and soils on catchment transit times in a geomorphic province; the Cairngorm mountains. Scotland. Hydrological Processes, 23(13): 1874--1886. Tetzlaff D, Soulsby C, Waldron S, et al. (2007) Conceptualization of runoff processes using a geographical information system and tracers in a nested mesoscale catchment. Hydrological Processes 2(110): 1289–1307. Tilman D, Kilham SS, and Kilham P (1982) Phytoplankton community ecology: The role of limiting nutrients. Annual Review Ecology and Systematics 13: 349--372. Tranter M, Brimblecombe P, Davies TD, Vincent CE, and Abrahams PW (1986) The composition of snowfall, snowpack and meltwater in the Scottish Highlands – evidence for preferential elution. Atmospheric Environment 20(3): 517--525. Tranter M and Jones HG (2001) The chemistry of snow: Processes and nutrient cycling. In: Jones HG, Pomeroy JW, Walker DA, and Hoham RW (eds.) Snow Ecology: An Interdisciplinary Examination of Snow-Covered Ecosystems, pp. 127--165. Cambridge: Cambridge University Press. Treml V and Banas M (2008) The effect of exposure on alpine treeline position: A case study from the High Sudetes, Czech Republic. Arctic, Antarctic, and Alpine Research 40(4): 751--760. Triska FJ, Kennedy VC, Avanzino RJ, Zellweger GW, and Bencala KE (1989) Retention and transport of nutrients in a third-order stream in northwestern California: Hyporheic processes. Ecology 70(6): 1893--1905.

303

Troch PA, Paniconi C, and van Loon EE (2003) Hillslope-storage Boussinesq model for subsurface flow and variable source areas along complex hillslopes: 1. Formulation and characteristic response. Water Resources Research 39(11): 1316 (doi:10.1029/2002WR001728). Vannote RL, Minshall GW, Cummins KW, Sedell JR, and Cushing CE (1980) The river continuum concept. Canadian Journal of Fisheries and Aquatic Sciences 37(1): 130--137. Veatch W, Brooks PD, Gustafson JR, and Molotch NP (2009) Quantifying the effects of forest canopy cover on net snow accumulation at a continental, mid-latitude site. Ecohydrology 2(2): 115--128. Verry ES, Brooks KN, and Barten PK (1988) Streamflow response from an ombrotorhic mire. In: International Symposium on the Hydrology of Wetlands in Temperate and Cold Regions, Joensuu, Finland, pp. 52–59. Helsinki, Finland: Academy of Finland. Vitousek PM, Gosz JR, Grier CC, Melillo JM, and Reiners WA (1982) A comparative analysis of potential nitrification and nitrate mobility in forest ecosystems. Ecological Monographs 52(2): 155--177. Vogel JC, Talma AS, and Heaton THE (1981) Gaseous nitrogen as evidence for denitrification in groundwater. Journal of Hydrology 50: 191--200. Vollenweider RA (1969) Possibilities and limits of elementary models concerning the budget of substances in lakes. Archiv fu¨r Hydrobiologie 66: 1--36. Vollenweider RA and Kerekes J (1982) Eutrophication of Waters. Paris: OECD. 154pp. Walsh CJ, Roy AH, and Feminella JW (2005) The urban stream syndrome: Current knowledge and the search for a cure. Journal of the North American Benthological Society 24(3): 706--723. Walter BP and Heimann M (2000) A process-based, climate-sensitive model to derive methane emissions from natural wetlands: Application to five wetland sites, sensitivity to model parameters, and climate. Global Biogeochemical Cycles 14(3): 745--765. Walter BP, Heimann M, and Matthews E (2001) Modeling modern methane emissions from natural wetlands, 1. Model description and results. Geophysical Research Letters 106(D24): 34189--34206. Watmough SA, Aherne J, and Dillon PJ (2003) Potential impact of forest harvesting on lake chemistry in south-central Ontario at current levels of acid deposition. Canadian Journal of Fisheries and Aquatic Sciences 60(9): 1095--1103. Watmough SA and Dillon PJ (2003) Base cation and nitrogen budgets for seven forested catchments in central Ontario, 1983–1999. Forest Ecology and Management 177: 155--177. Weathers KC and Likens GE (1997) Clouds in southern Chile: An important source of nitrogen to nitrogen-limited ecosystems? Environmental Science and Technology 31: 210--213. Weathers KC, Lovett GM, and Likens GE (1995) Cloud deposition to a spruce forest edge. Atmospheric Environment 29(6): 65--672. Webster JR and Patten BC (1979) Effects of watershed perturbation on stream potassium and calcium dynamics. Ecological Monographs 49(1): 51--72. Weissmann GS, Zhang Y, LaBille EM, and Fogg GE (2002) Dispersion of groundwater age in an alluvial aquifer system. Water Resources Research 38(1). (doi:10.1029/ 2001WR000907). Wetzel RL (2001) Limnology: Lake and River Ecosystems, 3rd edn. San Diego, CA: Academic Press. 1006pp. White AF and Blum AE (1995) Effects of climate on chemical weathering in watersheds. Geochimica et Cosmochimica Acta 59(9): 1729--1747. White AF and Brantley SL (2003) The effect of time on the weathering of silicate minerals: Why do weathering rates differ in the laboratory and field? Chemical Geology 202: 479--506. White PJ and Broadley MR (2001) Chloride in soils and its uptake and movement within the plant: A review. Annals of Botany 88(6): 967--988. Wilcox DA, Shedlock RJ, and Hendrickson WH (1986) Hydrology, water chemistry and ecological relations in the raised mound of Cowles Bog. Journal of Ecology 74(4): 1103--1117. Wildenschild D, Jensen KH, Villholth K, and Illangasekare TH (1994) A laboratory analysis of the effect of macropores on solute transport. Ground Water 32(3): 381--389. Williams K, Caldwell MM, and Richards JH (1993) The influence of shade and clouds on soil water potential: The buffered behavior of hydraulic lift. Plant and Soil 157: 83--95. Williams MW, Bales RC, Brown AD, and Melack JM (1995) Fluxes and transformations of nitrogen in a high-elevation catchment, Sierra Nevada. Biogeochemistry 28(1): 1--31. Winter TC (1999) Relation of streams, lakes and wetlands to groundwater flow systems. Hydrogeology Journal 7: 28--45. Winter TC, Harvey JW, Franke OL, and Alley WM (1999) Ground water and surface water – a single resource. US Geological Survey Circular 1139: 1--79.

304

Hydrology and Biogeochemistry Linkages

Winter TC and Woo M-K (1990) Hydrology of lakes and wetlands. In: Wolman MG and Riggs HC (eds.) Surface Water Hydrology. The Geology of North America, vol. 0–1, pp. 159--187. Boulder, CO: Geological Society of America. Wollheim WM, Vo¨ro¨smarty CJ, Peterson BJ, Seitzinger SP, and Hopkinson CS (2006) Relationship between river size and nutrient removal. Geophysical Research Letters 33: L06410 (doi:10.1029/2006GL025845). Worrall F, Harriman R, Evans CD, et al. (2004) Trends in dissolved organic carbon in UK rivers and lakes. Biogeochemistry 70(3): 369--402. Wright RF, Aherne J, Bishop K, et al. (2006) Modelling the effect of climate change on recovery of acidified freshwaters: Relative sensitivity of individual processes in the MAGIC model. Science of the Total Environment 365: 154--166. Wright RF, Larssen T, Camarero L, et al. (2005) Recovery of European acidfied surface waters. Environmental Science and Technology 39(3): 64A--72A (doi: 10.1021/ es0531778).

Wright RF, Norton SA, Brakke DF, and Frogner T (1988) Experimental verification of episodic acidification of freshwaters by sea salts. Nature 334: 422--424. Xenopoulos MA, Lodge DM, Fretress J, et al. (2003) Regional comparisons of watershed determinants of dissolved organic carbon in temperate lakes from the Upper Great Lakes region and selected regions globally. Limnology and Oceanography 48(6): 2321--2334. Zhang Q, Zhang Y, Peng S, Yirdaw E, and Wu N (2009) Spatial structure of alpine trees in Mountain Baima Xueshan on the southeast Tibetan Plateau. Silva Fennica 42(2): 197--208.

2.12 Catchment Erosion, Sediment Delivery, and Sediment Quality DE Walling, University of Exeter, Exeter, UK SN Wilkinson, CSIRO Land and Water, Townsville, QLD, Australia AJ Horowitz, US Geological Survey, Atlanta, GA, USA & 2011 Elsevier B.V. All rights reserved.

2.12.1 2.12.2 2.12.2.1 2.12.2.2 2.12.2.3 2.12.3 2.12.3.1 2.12.3.2 2.12.3.3 2.12.3.4 2.12.4 2.12.4.1 2.12.4.2 2.12.4.2.1 2.12.4.2.2 2.12.4.2.3 2.12.4.2.4 2.12.4.3 2.12.4.3.1 2.12.4.3.2 2.12.4.4 2.12.4.4.1 2.12.4.4.2 2.12.4.4.3 2.12.5 2.12.5.1 2.12.5.2 2.12.5.3 2.12.5.3.1 2.12.5.3.2 2.12.5.3.3 2.12.5.3.4 2.12.5.4 References

A Changing Context Sediment Budgets The Sediment Budget as an Integrating Concept The Functioning of the Sediment Budget The Global Sediment Budget Documenting Catchment Sediment Budgets The Background The Use of Fallout Radionuclides Sediment Source Fingerprinting The Future Modeling the Catchment Sediment Budget The Requirement Model Development Modeling approaches and model complexity Empirical modeling of catchment sediment yield Conceptual process modeling of catchment sediment budgets Mechanistic, physically based modeling of hillslope processes SedNet – A Sediment Budget Model for River Networks Model outline Management applications Current Status and Future Directions Modeling across scales for planning and management Directions in modeling erosion and deposition processes Model uncertainty considerations The Quality Dimension Introduction Basic Sediment Geochemistry Major Issues Associated with Sediment Quality Background/baseline sediment-associated constituent concentrations The collection of representative sediment samples and the issues of spatial and temporal variability The chemical analysis of suspended and bed sediments Bioavailability and toxicity Future Directions

2.12.1 A Changing Context Although it has attracted the interest of fluvial geomorphologists, geologists, sedimentologists, and hydrologists, the study of erosion and sediment transport by rivers has traditionally been largely the preserve of the agricultural engineer and the hydraulic or civil engineer (e.g., ASCE, 1975; Schwab et al., 1981; Lal, 1994; Julien, 1995, 2002; Morgan, 1995; Yang, 1996; US Department of Agriculture, 1997; Chien and Wan, 1999; Fangmeier et al., 2006; US Bureau of Reclamation, 2006; Garcia, 2008). In the case of erosion, attention focused largely on soil erosion on agricultural land and emphasized the assessment of rates of soil loss and their implications for crop productivity and the sustainability of land use practices, as well as the design of soil conservation measures.

305 307 307 307 310 312 312 313 314 315 316 316 317 317 317 318 319 319 319 320 320 320 322 323 323 323 323 324 324 325 329 330 330 331

The well-known Universal Soil Loss Equation (USLE) developed in the USA (see Wischmeier and Smith, 1978) and modified for application elsewhere (e.g., Schwertmann et al., 1990; Larionov, 1993) was a product of this interest in erosion processes, providing a basis for predicting the spatial variation of rates of soil loss in response to their control by rainfall, topography, and land use practices, and for assessing the potential impact of improved management and cropping practices. Hydraulic engineers directed attention to the study of sediment transport by rivers and related problems associated with the management of river channels for navigation and flood control and reservoir sedimentation, as well as to the design of hydraulic structures able to cope with high sediment loads. Such work commonly emphasized the coarser fractions of the sediment load, as this was most important in terms of

305

306

Catchment Erosion, Sediment Delivery, and Sediment Quality

river morphology. It was also more readily predicted from a knowledge of sediment properties and flow conditions, than the finer washload, which was generally a noncapacity load and therefore a supply-controlled load. The washload was frequently viewed as being of limited importance, since it was readily transported through a river system and had limited morphological impact. The emphasis on the transport of coarser sediment either as bedload or as suspended bed material is well demonstrated by the large number of sediment transport formulas that were developed in the middle years of the twentieth century for predicting these components of the sediment load (see ASCE, 1975). As a noncapacity or supplycontrolled load, the washload of a river was seen as something that was not easy to predict using sediment transport formulas and, if it was of interest, it therefore needed to be measured. Against this background, attention traditionally focused on soil loss from upstream areas, with emphasis on on-site problems of soil degradation and loss of productivity and on downstream sediment transport and sediment yield. There was often only limited contact between those working on these two aspects. Stated very simply, soil loss from fields was lost from the farm and its ultimate fate was of limited interest to the agricultural engineer. Equally, eroded soil generally contributed primarily to the finer washload of a river, which was seen as being of limited importance in terms of river morphology and hydraulics, although it was an important cause of reservoir sedimentation and some knowledge of downstream suspended sediment yields was therefore needed. In general, the degree of attention given to the study of erosion and sediment transport was broadly proportional to the magnitude of erosion rates and sediment loads. In countries such as the USA, there was considerable activity in these areas, whereas in countries such as the UK, where erosion rates were low and soil erosion was not perceived to be a problem, and sediment yields were also low and rivers relatively small, activity and interest were limited. A major change in the above situation occurred in the latter part of the twentieth century. Changing perspectives on erosion and sediment transport promoted increased interest in this general field and emphasized the need for a more multidisciplinary perspective and a more integrated approach that directed attention to the functioning of the entire catchment system. Greater emphasis was therefore placed on the hydrological context. Several key drivers of these changes can be identified. One was the recognition of the importance of fine sediment, both as a key water-quality parameter in terms of its physical presence and also as an important control on river water quality more generally. Many pollutants and contaminants, including heavy metals, pesticides and other organic contaminants, as well as nutrients such as phosphorus, are transported primarily in association with sediment and interactions between the solid (sediment) and liquid (water) phases exert an important influence on water quality (e.g., Golterman, 1977; Shear and Watson, 1977; UNESCO, 1978; Allan, 1979; Fo¨rstner and Wittmann, 1981; Hart, 1982; Salomons and Fo¨rstner, 1984; Horowitz, 1991; Ongley et al., 1992; Santiago et al., 1994; US Environmental Protection Agency, 1997; House and Warwick, 1999; Warren et al., 2003). In addition to considering the amount of sediment transported by a river, there was also an increasing need to consider the

quality of that sediment. Another important driver was the increasing evidence of the detrimental effects of fine sediment in degrading aquatic habitats and ecosystems, through, for example, the siltation of fish spawning gravels, the smothering of aquatic vegetation and increased nutrient inputs to floodplains, riparian areas, and other water bodies, through sediment deposition (see Ritchie, 1972; Clark, 1985; Clark et al., 1985; Waters, 1995; Newcombe and Jensen, 1996; Wood and Armitage, 1997; Soulsby et al., 2001; Suttle et al., 2004; Cavalcanti and Lockaby, 2005). Both in terms of its physical presence and its quality, fine sediment is frequently an important cause of environmental degradation and it has been widely referred to as the world’s number one pollutant. Diffuse source pollution was increasingly recognized as an important cause of water pollution, and sediment, which can be mobilized from throughout a river basin, is a major component of such pollution. Around the Great Lakes of North America, for example, concern for the eutrophication and pollution of these water bodies and particularly Lake Erie, directed attention to the need to control diffuse source pollution and sediment assumed a central role in the Pollution from Land Use Activities Reference Group (PLUARG) program developed by the International Joint Commission on the Great Lakes (Coote et al., 1982; Ongley, 1982). In many ways, this program was ahead of its time in recognizing the importance of sediment and the role of land use in influencing sediment mobilization and transfer to water bodies. Sediment has also assumed considerable importance in the recent EU Water Framework and Habitats Directives (Fo¨rstner, 2002, 2003) aimed at improving land management practices, protecting aquatic habitats, and maintaining conditions of good ecological status in rivers. In addition, increasing interest in the functioning of the Earth’s system has highlighted the important role of land–ocean sediment transfer in global geochemical cycling, and particularly the carbon cycle (Ludwig et al., 1996; Lyons et al., 2002; Beusen et al., 2005; Seitzinger et al., 2005; Gislason et al., 2006; Van Oost et al., 2007; Saenger et al., 2008). River sediment loads have been shown to be very sensitive to the various drivers of global change (e.g., Walling and Fang, 2003; Walling, 2008) and to exert an important influence on the health of receiving waters in the coastal zone. Within the International Geosphere Biosphere Programme (IGBP), launched in 1987 by the International Council for Scientific Unions (ICSU) particular attention was directed to land–ocean sediment and material transfers through its Land–Ocean Interactions in the Coastal Zone (LOICZ) core project. Significant outcomes of the evolution of these new perspectives on erosion and sediment transport include the following. First there has been an increasing emphasis on fine sediment (see Owens et al., 2005). This is the most significant fraction of the sediment load in terms of pollution and sediment-associated transport of nutrients and contaminants, since contaminants are in most cases preferentially associated with the finer (o63 mm) particles (Horowitz, 1991). Equally, it is generally fine sediment which is of greatest importance in terms of the degradation of aquatic ecosystems and habitats. Since the fine sediment load of a river is commonly a noncapacity load and supply-limited, interest in fine sediment transport has necessarily shifted the emphasis of sediment

Catchment Erosion, Sediment Delivery, and Sediment Quality

transport studies. It has moved away from a hydraulic approach, that emphasized hydraulic controls and the transport conditions in the channel toward developing an improved understanding of sediment mobilization and transfer within the entire catchment and thus the supply of sediment to the river system. Clearly, this demands a hydrological approach. Second, with greater emphasis being directed to sediment quality, and increasing recognition that sediment quality is in most instances closely related to sediment source, there has been a need to develop an improved understanding of potential sediment sources and transfer pathways. This was also a key requirement for any attempt to develop sediment management or control programs and to implement mitigation measures. Resources need to be targeted to those sources or parts of a catchment that provide the main source of the sediment transported by a river and which need to be controlled. Again, therefore, a distributed hydrological approach for understanding and modeling the sediment dynamics of a catchment or river system has been increasingly required. Third, the shift of emphasis away from concern for sediment problems linked primarily to the amount of sediment and thus the magnitude of erosion rates and sediment yields to the more wide ranging environmental significance of fine sediment has broadened the relevance of the study of erosion and sediment transport to include most areas of the world. Paradoxically, it is often areas with low erosion rates and low sediment yields where the environmental impacts of sediment are potentially greatest and the need to develop an improved understanding of the processes of sediment mobilization and transfer is therefore strongest. Finally, as indicated above, these new perspectives and requirements have created the need to integrate studies of erosion and sediment transport. The offsite problems of soil erosion, which relate to the onward transfer of the mobilized sediment through a drainage basin and the impacts of this sediment, are frequently seen as being equally, if not more, important than the on-site problems of soil loss. Equally, the increased emphasis on fine, rather than coarse, sediment transport has focused attention on sediment supply to river channels and the need to look beyond the river channel and to consider the processes of sediment mobilization and transfer within the entire upstream catchment area. A hydrological perspective is a key requirement for the new perspectives on erosion and sediment transport outlined above. This has in turn strengthened the position of the study of erosion and sediment transport as an important branch of hydrology. This position has long been recognized by the International Association of Hydrological Sciences (IAHS), through the activities of its Commission on Continental Erosion which was established in the middle years of the twentieth century. Equally, the Hydrology section of the American Geophysical Union and the Hydrological Sciences division of the European Geosciences Union both include groups devoted to the study of erosion and sedimentation. The need for a broader multidisciplinary perspective is also demonstrated by the emergence of specialist groups focusing on sediment studies, such as the International Association for Sediment Water Science (IASWS), which was established in 1984, and the World Association for Sedimentation and Erosion Research (WASER), which was founded in 2004.

307

This contribution reviews some of the key developments associated with the changing focus of studies of erosion and sediment yield outlined above. Attention is directed first to the sediment budget concept, which provides a valuable framework for studying, modeling, and managing erosion and sediment yield in catchments. Second, approaches to modeling catchment sediment budgets are considered. Finally, several of the key contemporary issues associated with sediment quality are discussed.

2.12.2 Sediment Budgets 2.12.2.1 The Sediment Budget as an Integrating Concept Although there is undoubtedly still a place for a reductionist approach, which focuses attention on the dynamics of a particular process associated with erosion and sediment transport, recognition of the wide ranging environmental significance of fine sediment and the need to link information on sediment output from drainage basins with information on sediment sources and sediment mobilization, transfer, and storage has resulted in a general acceptance of the sediment budget as a central integrating concept for the study of erosion and sediment yield. In addition to integrating the various components of sediment mobilization, transfer, storage, and output and providing a valuable scientific framework for research investigations, the sediment budget concept also provides an essential management tool (Walling and Collins, 2008). It identifies the key sediment sources and transfer pathways within a catchment, which are likely to represent the focus of any management strategy. Furthermore, it emphasizes the sensitivity of the sediment response of a catchment to environmental change and the potentially complex links between changing erosion rates and changes in sediment yield, which must be recognized when planning and implementing sediment management and control strategies. Figure 1, which is based on the classic work of Trimble (1983) in the 360 km2 catchment of Coon Creek, Wisconsin, USA, provides a useful demonstration of the catchment sediment budget concept and the way in which it integrates consideration of sources, sinks, and output and thus sediment mobilization, transport, deposition and storage, as well as the dynamic interaction of these components. In the Coon Creek study, two separate budgets were developed. The first was for the period of poorly managed agriculture and severe erosion that followed land clearance and the expansion of agriculture in the latter half of the nineteenth century and the early part of the twentieth century. The second was for the subsequent period, when soil conservation measures were introduced to control erosion and soil degradation. An important feature of the budgets for both periods is that only a relatively small proportion of the total mass of sediment mobilized within the basin by erosion reaches the basin outlet (i.e., B5–7%). This emphasizes that information on the sediment yield at a basin outlet may provide a poor indication of the overall amount of sediment mobilized and moved through a basin and emphasizes that the key to understanding the system frequently lies in identifying and quantifying the sediment sinks or stores. Whereas attention has traditionally focused on erosion processes and sediment transport, the sinks and stores can

308

Catchment Erosion, Sediment Delivery, and Sediment Quality

dominate the functioning of a catchment sediment budget, with much of the sediment mobilized in catchments being deposited on hillslopes, in riverbeds, on floodplains, and in reservoirs, rather than contributing to catchment export (e.g., Dunne et al., 1998; Trimble and Crosson, 2000). Figure 1 demonstrates that during both periods, large amounts of sediment were being stored in the colluvial deposits associated with the hillslopes within the upland areas and in alluvial sinks within both the tributary valleys and the main valley of Coon Creek. Comparison of the sediment budgets for the two periods shows that although the implementation of soil conservation measures after 1938 greatly reduced upland erosion rates, producing a substantial (i.e., B25%) reduction in sediment mobilization from the slopes, the sediment yield at the basin outlet changed very little, due to the increased efficiency of sediment transfer through the channel system (i.e., reduced deposition) and the remobilization of sediment that had accumulated within the middle valley during the preceding period of accelerated erosion. From a management perspective, a sediment budget, such as that presented in Figure 1, provides valuable information for use in developing effective catchment-based sediment management and control strategies. It identifies the most important sediment sources that would need to be targeted in any attempt to reduce downstream sediment fluxes and thus facilitates the optimum use of the resources available for implementing sediment control measures. Equally, it also

Table 1

emphasizes that reduction of upstream erosion may not necessarily result in a significant reduction of the downstream sediment yield. Much of the sediment generated upstream may have previously been deposited and stored before reaching the catchment outlet, and reduction in upstream sediment mobilization could be offset by remobilization of sediment from intervening stores. An understanding of the sediment budget of a drainage basin is clearly also important for predicting the likely impact of future climate change on downstream sediment response. This could change significantly, if hydrological changes resulted in the remobilization of sediment from existing sediment sinks, for example, through changing channel morphology and increased channel migration and erosion.

2.12.2.2 The Functioning of the Sediment Budget As indicated above, the sediment budget concept provides a valuable integrating framework for studying the various processes of sediment mobilization and delivery operating within a catchment. Table 1 lists most of the key processes involved and some of the recent research aimed at developing an improved understanding of these processes. Although, as shown in Figure 1, the emphasis is commonly placed on the magnitude of the fluxes and stores, and thus on the quantities of sediment involved, it is also important to recognize that the properties of the sediment associated with different

Examples of recent research on the component processes of the sediment budget

Process Sediment mobilization Interrill or sheet erosion

References

Abrahams et al. (2001), Gime´nez and Govers (2002), Prosser and Rustomji (2000), Valmis et al. (2005), Wei et al. (2009), Zhang et al. (2009a, 2009b).

Rill erosion

Cerdan et al. (2006), Govers et al. (2007), Lei et al. (2006), Merten et al. (2001), Schiettecatte et al. (2008).

Gully erosion

Gomez et al. (2003), Gordon et al. (2008), Poesen et al. (2003), Rustomji (2006), Valentin et al. (2005).

Mass movements

Brayshaw and Hassan (2009), Chappell et al. (2004), Hassan et al. (2005), Heimsath et al. (2002), Lavigne and Suwa (2003),Wemple et al. (2001), Schuerch et al. (2006).

Channel bank erosion

Atkinson et al. (2003), Florsheim et al. (2008), Fox et al. (2007), Rinaldi et al., (2008), Hupp et al. (2009), Jeffries et al. (2003), Laubel et al. (2003), Wynn and Mostaghimi (2006).

Sediment transfer or delivery Slope to channel Rustomji and Prosser (2001), Croke et al. (2005), Deasy et al. (2009), Haygarth et al. (2006), Preston and Schmidt (2003), Smith and Dragovich (2008). Channel

Droppo et al. (2001, 2004), Forbes and Lamoreux (2005), Malmon et al. (2005), Petticrew et al. (2007), Rubin and Topping (2001), Simon et al. (2004), Stone et al. (2008). Sediment deposition and storage Colluvial Brardinoni et al. (2009), Cochrane and Flanagan (2006), Croke et al. (1999), de Moor and Verstraeten (2008), Macaire et al. (2002), Rommens et al. (2006).

Channel

Collins and Walling (2007), Hart (2002), Macnab et al. (2006), Petticrew et al. (2007), Smith et al. (2003), Steiger et al. (2003).

Alluvial fans

Field (2001), Harvey (2002), Harvey et al. (2005), Leeder and Mack (2001), Ritter et al. (2000), Staley et al. (2006).

Floodplain

Aalto et al. (2003, 2008), Hughes et al. (2009), Kronvang et al. (2009), Lauer and Parker (2008), Sweet et al. (2003), Swanson et al. (2008), Jeffries et al. (2003), Thonon (2006), Thonon et al. (2007).

Sediment yield Ali and de Boer (2008), Evans and Slaymaker (2004), Haregeweyn et al. (2008), Steegen et al. (2001), Syvitski and Milliman (2007), Molina et al. (2007), Tamene et al. (2006), Verstraeten and Poesen (2002), Verstraeten et al. (2003).

Catchment Erosion, Sediment Delivery, and Sediment Quality Coon Creek 1853−1938 Upland sheet and Sources (t × 103) rill erosion 630 Upland gullies Tributaries 80 46 Sediment discharge at mouth 42

Lower Middle Tributary valley valley 230 Upland valleys 78 valleys 96 Hillslopes 269 42 Sinks and stores (t × 103) Coon Creek 1938−1975 Upland sheet and rill erosion Sources (t × 103) 456 Upland gullies Tributaries Middle 71 39 valley 30 Sediment discharge at mouth 40

Lower Middle valley valley Upland 153 30 valleys 42 Hillslopes 332 Sinks and stores (t × 103)

Figure 1 The sediment budgets for Coon Creek, Wisconsin, USA for the periods 1853–1938 and 1938–1975 produced by Trimble (1983). The fluxes shown are mean annual values. From Trimble (1983) A sediment budget for Coon Creek basin in the Driftless Area, Wisconsin, 1853–1977. American Journal of Science 283: 454–474.

components of the budget may change, as sediment is transferred from source to sink. Mobilization, transfer, and deposition processes will frequently involve selectivity related to both particle size and particle density, resulting in contrasts between the composition of sources and sediment associated with different components of the budget (e.g., Fontaine et al., 2000; Stone and Droppo, 1996). In the case of variations in particle size, the contrast between the effective or in situ grain size and the ultimate or absolute grain size of the sediment, which reflects the existence of composite particles (i.e., aggregates and flocs), can introduce further complexity and exert an important influence on the properties of sediment associated with individual components of the sediment budget (see Stone and Saunderson, 1992; Stone and Walling, 1997; Walling et al., 2000; Blake et al., 2005; Woodward and Walling, 2007). In the case of sinks, for example, coarser particles may be preferentially deposited, but if these coarser particles comprise aggregates or flocs, they may contain a significant proportion of fine particles, the deposition of which might otherwise be unexpected (Nicholas and Walling, 1996). Droppo (2001) has called for a rethinking of conventional approaches to investigating suspended sediment dynamics to

309

reflect the existence and importance of such composite particles. Equally, sediment mobilized from different sources may be characterized by different properties, and the sediment associated with different components of the budget may change according to its source or the relative contribution from different sources. In this context, contrasts in the properties of sediment derived from different sources may reflect both the source type (e.g., sheet and rill erosion vs. gully and channel bank erosion) and the spatial variability of source material properties caused by variations in geology, soil type, or land use across a catchment. Key characteristics of the functioning of a sediment budget include its connectivity and thus the extent to which the slopes or upstream parts of a catchment are linked to the channel system or the catchment outlet. The connectivity of a system will depend on the incidence and efficiency of the transfer pathways and the magnitude of the stores. The connectivity of the system is clearly of fundamental importance when investigating and attempting to control sediment-induced diffuse source pollution. Detailed assessment of connectivity necessitates consideration of the transfer pathways involved and their efficiency in transferring sediment through the sediment delivery system. By providing a clearer representation of the links between erosion or sediment mobilization and sediment yield, the catchment sediment budget represents an important advance over the sediment delivery ratio concept. The latter simple blackbox concept (e.g., Roehl, 1962; Walling, 1983) recognized that the sediment output was likely to be less than the gross sediment mobilization and represented the ratio of the former to the latter. The magnitude of this ratio was in turn linked to the size of the catchment, with its magnitude commonly decreasing as the scale of the catchment increased. Many limitations of the sediment delivery ratio concept have been widely debated (e.g., Walling, 1983; Parsons et al., 2006; de Vente et al., 2007) and Beven et al. (2005) provide a useful overview of the problems to be faced in conceptualizing sediment delivery to stream channels. The precise form taken by the sediment budget of a catchment will reflect a wide range of controls, including the local topography and the hydrological regime, as well as the size of the catchment. Figure 2 provides an indication of the potential nature and extent of such variability by indicating the key characteristics of the sediment budgets of four small drainage basins on the Russian Plain documented by Golosov et al. (1992). These are all relatively small basins, heavily impacted by agricultural land use and associated soil erosion. The investigation aimed to establish the proportion of the sediment mobilized within the catchments by different erosion processes that reached the basin outlets. In this environment, three key sediment mobilization processes were identified. These comprised sheet erosion (i.e., widespread erosion of the surface by surface wash), rill erosion (i.e., erosion by concentrated flow in micro-channels developed on the slopes), and gully erosion (i.e., erosion within deeper ephemeral channels that dissect the landscape and where sediment is mobilized by mass movements on the gully sides as well as by the flow through the gully). In this environment, sheet and rill erosion are generally more important than gully erosion as a sediment source and there is little evidence of

310

Catchment Erosion, Sediment Delivery, and Sediment Quality

sediment storage on the lower parts of the slopes. Slopes are frequently convex, terminating at the margins of balkas (flat floored, gully-like features), that dissect the landscape. Even within this relatively homogeneous area, the proportion of the sediment mobilized by erosion within the individual catchments that reaches the basin outlet ranges from 0% to 89%. In most of the catchments, both the balka bottoms and the river floodplains constitute major sinks for sediment moving through the system and as with Coon Creek (see Figure 1), the sinks represent a very important component of their sediment budgets. As the scale of the drainage basin increases, deposition of sediment on the river floodplains in the lower parts of the basin will commonly assume increasing importance. Work within the catchments of the Rivers Ouse (3315 km2) and Wharfe (818 km2) in Yorkshire, UK, reported by Walling et al. (1998) has, for example, shown that as much as 30–40% of the sediment delivered to the main channel system is deposited on the adjacent floodplains during overbank flood events and does not reach the basin outlet. At a larger scale, Bobrovitskaya et al. (1996) provided information on the sediment budget of the lower River Ob which drains a vast catchment of 2 950 000 km2 in Siberia to the Arctic Ocean. The available information on the suspended load of this river provided by two gauging stations on its lower reaches separated by an 870-km reach indicates that in its lower reaches approximately 50% of its suspended sediment load is deposited on the well-developed floodplain bordering the river and fails to reach the lowest measuring station. At this larger scale, tectonic subsidence within the interior of a river basin can also promote the development of major sediment sinks which reduce the downstream sediment flux. This is well illustrated by the Rio Madeira, a major tributary of the Amazon in Bolivia and information reported by Baby et al. (2009). The upper basin of this river, which extends to B170 000 km2, drains the Andean Cordillera where erosion rates are high and consequently transports a very high annual suspended sediment load of B500–600 Mt. However, on leaving the Andes, the downstream course of the Rio Madeira passes through a subsiding foreland basin where of the order of 270 Mt of sediment is deposited each year. As a result, only about 45% of the upstream load of the Rio Madeira is transported downstream into the main Amazon river system. The operation of sediment budgets, such as those depicted in Figures 1–3, can be considered over several different timescales. Several recent sediment budget investigations that have taken a longer-term perspective have emphasized the importance of sediment storage, with the majority of the sediment mobilized being stored within the catchment over long periods. For example, Rommens et al. (2005) reported a Holocene sediment budget for a small 103 ha agricultural catchment in the Belgian loess belt that shows that 58–80% of the sediment mobilized within the catchment had been stored near its source and not delivered to downstream rivers. Similarly, Prosser et al. (2001a) estimated that as much as 80% of the sediment eroded from large coastal catchments in Eastern Australia in historical times remains stored in their channels and floodplains. Many of the sediment sinks associated with a sediment budget are likely to be long-term sinks. For example, the sediment deposited on the lower parts of a slope will

Veduga Creek (86.9 km2)

Sheet 51%

Sheet 87%

Gully 49%

Balka Rolzavets (181.5 km2)

Rill 6.5% Gully 6.5% Balkas 55% Floodplain 45%

Balkas 91% 9% Output

Little Kolysheley River (181.5 km2) Sheet 72.5% Rill 5% Gully 22.5%

Kijuchi Creek Slope 42% Rill 4.5% Gully 53.5%

(8 km2)

Balkas 11%

Slope18% Floodplain 54% 28% Output

89% Output

Figure 2 The sediment budgets for four small drainage basins on the Russian Plain, established by Golosov et al. (1992). From Golosov et al. (1992) Sediment budgets of river catchments and river channel aggradation on the Russian plain. Geomorphology (Moscow) 4: 62–71 (in Russian).

commonly remain in near-permanent storage, unless there is a significant change in the pattern of erosion. River floodplains will frequently also represent longer-term sinks, with bank erosion perhaps causing some loss, which is balanced by point bar formation and deposition elsewhere. Floodplain sinks could, however, be rapidly remobilized by changes in channel pattern and increased channel migration associated with changes in the flow regime caused by human activity in the upstream catchment or climate change. Some sinks will, however, operate as shorter-term stores. This was the case with the middle valley sink within the Coon Creek catchment depicted in Figure 1. Furthermore, at the annual timescale, sediment deposited within the channel system may accumulate during one period of the year, only to be remobilized and flushed out during a subsequent period (e.g., Collins and Walling, 2007). In this situation, storage is clearly temporary and in their study of three groundwater-dominated lowland catchments in the UK, Collins and Walling (2007) demonstrated that fine sediment accumulated within the channel during the winter period, when most sediment was transported through the system, and was subsequently remobilized during the summer period. Estimates of the average mass of sediment stored in the channel systems of the three catchments during the 2-year study period demonstrated that this was equivalent to between 21% and 38% of the mean annual

Catchment Erosion, Sediment Delivery, and Sediment Quality

311

Soil redistribution rate (t ha−1 year−1) 5 0 −5 −10

30

t (m)

h Heig

−15 20

10

350

30

He

igh

20

250

t (m )

300

10

ce

tan

Dis

200

0

25

)

(m

150 0

20

100

0

15 0 10

50

ce

n ista

)

(m

D

50 Figure 3 The pattern of soil redistribution within a 7.5 ha field at Butsford Barton near Colebrooke, Devon, UK established by Walling and his coworkers using 137Cs measurements. More than 200 bulk cores were collected from the field at the intersections of a 20 m grid. The soil redistribution rates depicted represent mean annual values for a B40 year period prior to the mid-1990s.

suspended sediment export from the three catchments. By documenting changes in storage through time, it was also possible to estimate the total amount of sediment entering and leaving channel storage within the three catchments over the study period. The amounts of sediment entering and leaving channel storage within the three catchments were equivalent to between B20% and 75% and between 25% and 70%, respectively, of the mean annual sediment yield, demonstrating that a significant proportion of the sediment flux passed through this short-term sink.

2.12.2.3 The Global Sediment Budget Although the application is somewhat different, in terms of both scale and the nature of the budget, a sediment budget approach has also been used to assess the impact of recent changes in the sediment loads of the world’s rivers on the global land–ocean sediment transfer. Such changes have important implications for global geochemical cycling. In this case the emphasis has been on the total land–ocean sediment flux and the magnitude of the changes in this flux that have

occurred as a result of human activity. Although reliable information on sediment loads is unavailable for many world rivers, there is a general consensus that the contemporary land–ocean sediment flux is of the order of 15 Gt yr1, and a recent study reported by Syvitski et al. (2005) suggests that the value may be somewhat lower around 12.6 Gt yr1. However, it is known that this flux is changing as a result of human impact (Walling, 2006a; Syvitski and Milliman, 2007). In some rivers it is increasing, due to land clearance and the expansion of agricultural land use and associated increases in erosion, whereas in others it is declining due to the trapping of sediment by dams. In some river basins both drivers may be operating and the net effect will depend on their relative importance. A key issue is the extent to which the global sediment budget has been perturbed by human influence. This involves establishing the likely natural land–ocean sediment flux and then assessing the extent to which it has been increased and reduced by land disturbance and dam construction respectively. Syvitski et al. (2005) have used a regression model incorporating the main controls on natural river loads to

312

Catchment Erosion, Sediment Delivery, and Sediment Quality

estimate the pre-human sediment loads of the world’s major rivers as being B14 Gt yr1. This is 1.4 Gt yr1 greater than their estimate of the contemporary land–ocean sediment flux (12.6 Gt yr1), which will have been influenced by both increases and decreases relative to the pre-human flux, as a result of land disturbance and sediment trapping by dams, respectively. Lack of sediment load data for many rivers in the developing world, where sediment loads are likely to have increased as a result of population growth, makes it difficult to estimate the magnitude of any increase, but more information is available on the impact of dams in reducing sediment fluxes. Although the impact of dams in reducing the sediment loads of the world’s rivers is widely recognized (see Milliman et al., 1984; Vo¨ro¨smarty et al., 1997, 2003; Walling and Fang, 2003; Walling, 2006a), there is currently considerable uncertainty associated with existing estimates of the likely amount of sediment sequestered behind dams on the world’s rivers and the resulting reduction in the global land–ocean sediment flux. Vo¨ro¨smarty et al. (2003) estimated that more than 40% of the global river discharge is currently intercepted by large (Z0.5 km3 maximum storage capacity) reservoirs, and by coupling this information with estimates of reservoir trap efficiency they estimated that reservoirs are currently sequestering B4–5 Gt yr1 of sediment, with the potential for this value to be considerably higher if the large number of smaller reservoirs are also taken into account. Using a similar approach, Syvitski et al. (2005) suggested that the contemporary land–ocean sediment flux is being reduced by B3.6 Gt yr1 as a result of sediment trapping by dams. These values are, however, an order of magnitude lower than the estimate of the current sedimentation behind the world’s major dams provided by a recent study involving the B33 000 dams included in the ICOLD World Register of Dams (ICOLD, 2006), undertaken by the ICOLD Reservoir Sedimentation Committee and reported by Basson (2008). The data provided by this study suggest that sedimentation behind the world’s major dams is currently equivalent to an annual sequestration of B60 Gt yr1 (see Walling, 2008). It is, however, important to recognize that the estimate of the current rate of sediment sequestration in the world’s reservoirs of B60 Gt yr1 presented above represents the mass of sediment sequestered behind the dams and this does not equate to the associated reduction in the land–ocean sediment flux. Much of this sediment would previously not have reached the oceans, due to deposition and storage within the river system, and particularly on river floodplains. The conveyance loss associated with sediment movement through a river system can clearly be expected to vary according to the magnitude of the sediment flux, the sediment transport and flood regime of the river, and the morphology of the channel system, and is likely to decrease in heavily managed channels, where the flow is constricted and flood inundation restricted. It is therefore difficult to propose a typical value for the conveyance loss likely to be associated with the B60 Gt yr1 of sediment currently being sequestered behind the dams constructed on the world’s rivers. However, Walling, (2008) has suggested a value of 60% as a first-order estimate. Use of this value would mean that 40% of the total B60 Gt yr1 might be expected to have previously reached the oceans and that dam construction is currently reducing the global land–ocean

Table 2 A comparison of the estimates of the major components of the global sediment budget and their modification by human activity provided by Syvitski et al. (2005) with those generated by Walling (2008), using a different estimate of the reduction in the contemporary sediment flux caused by sediment trapping Component

Syvitski et al. (2005)

Pre-human land–ocean flux (Gt yr1) Contemporary land–ocean sediment Flux (Gt yr1) Reduction in flux associated with reservoir trapping (Gt yr1) Contemporary flux in the absence of reservoir trapping (Gt yr1) Increase over pre-human flux due to human activity (%) Reduction in contemporary gross flux due to reservoir trapping (%)

14.0 12.6 3.6 16.2

Walling (2008) 14.0 12.6 24 36.6

22

160

16

66

sediment flux by about 24 Gt yr1, a value that is considerably in excess of the likely contemporary global land–ocean sediment flux. This value of 24 Gt yr1 is approaching an order of magnitude greater than the values of 3.6 Gt yr1 suggested by Syvitski et al. (2005) as representing the reduction in the contemporary global annual land–ocean flux resulting from sediment trapping by reservoirs. Taking the above information on the likely magnitude of the contemporary and pre-human land–ocean sediment flux and the potential impact of sediment trapping by dams, it is possible to speculate further on the possible nature of the global sediment budget and the extent to which it has been perturbed by human activity (see Table 2). If the contemporary land–ocean sediment flux is taken to be 12.6 Gt yr1, but it is assumed that this has been reduced by 3.6 Gt yr1 as a result of reservoir trapping, the contemporary flux in the absence of reservoir trapping would be 16.2 Gt yr1. This represents a B16% increase over the pre-human flux, with this contemporary flux being reduced by B22%. As such, the perturbation associated with human activity is fairly limited. If, however, the same value is used for the contemporary land–ocean flux (12.6 Gt yr1), but it is assumed that this has been reduced by 24 Gt yr1 as a result of reservoir trapping, the contemporary flux in the absence of reservoir trapping would be 36.6 Gt yr1. This represents a B169% increase over the pre-human flux and this has, in turn, been reduced by B66% as a result of reservoir trapping. Under this scenario, human activity must be seen to have had a major influence on the global sediment budget. Further research is clearly required to confirm the magnitude of human impact on the global sediment budget.

2.12.3 Documenting Catchment Sediment Budgets 2.12.3.1 The Background Traditionally, fine sediment monitoring programs in river basins have focused on measuring the sediment load at the outlet of the catchment or river basin under investigation. This

Catchment Erosion, Sediment Delivery, and Sediment Quality enables the sediment yield (t yr1) and specific sediment yield (t km2 yr1) to be quantified. Such traditional measuring programs are commonly based on manual suspended sediment sampling, using samplers designed to collect representative suspended sediment samples from the measuring cross section (e.g., Guy and Norman, 1970; Gray et al., 2008). These samples are subsequently filtered to determine the suspended sediment concentration (mg l1). The suspended sediment flux at the time of sampling is computed as the product of the water discharge and the sediment concentration, taking account of the variation of both sediment concentration and flow velocity in the cross section. If frequent samples are collected, it is possible to interpolate the record of sediment flux between the sampling times, to compute the total load for the study period. Where, as in many situations, fewer samples are collected, rating curves representing relationships between suspended sediment concentration or discharge and water discharge are established using the available samples and the rating curve is used in conjunction with the record of water discharge to estimate the load for the period of interest. The use of rating curves introduces the potential for significant errors in the estimate of sediment flux (see Walling, 1977; Ferguson, 1986; Walling and Webb, 1988). Recent technological advances have greatly expanded the potential for obtaining more reliable estimates of suspended sediment load. Programmable automatic samplers can be used to increase the sampling frequency and to ensure that samples are collected at key times during a flood event (e.g., Lewis and Eads, 2001; Alexandrov et al., 2003). In situ sensors can also be deployed to record surrogate information that can be used to provide a continuous or near continuous record of suspended sediment concentration. Turbidity measurements obtained using both optical backscatter and transmission sensors have proved to be particularly valuable for this purpose (e.g., Gippel, 1995; Glysson and Gray, 2002; Schoellhamer and Wright, 2003), although their use is generally limited to relatively low levels of suspended sediment concentration. Other principles, involving lasers and ultrasonic sensors, have also been successfully used to collect information on the variation of both suspended sediment concentration and its grain size composition through time (e.g., Melis and Topping, 2003; Thonon et al., 2005; Topping et al., 2005). The requirement for information on the overall sediment budget of a catchment, rather than simply an estimate of the sediment load at the catchment outlet, has, however, necessarily introduced a need to develop new approaches capable of documenting rates of sediment mobilization, quantifying the storage elements of sediment budgets and obtaining information on sediment sources and transfer pathways. Traditional techniques provide some scope for assembling such information, but in a recent paper Walling (2006b) suggested that there was a need for a new paradigm which focused on tracing rather than monitoring. Monitoring was seen as continuing to be important, and indeed to be an essential component of any comprehensive measurement program, but the use of tracing techniques was viewed as representing the only effective means of assembling much of the information required to establish a sediment budget. Two key advances in the application of tracer techniques for investigating catchment sediment budgets can usefully be

313

highlighted. The first is the use of fallout radionuclides to obtain information on soil and sediment redistribution rates within a catchment and the second is the use of sediment source fingerprinting techniques to provide information on the relative contribution of a range of potential sources to the sediment output from a catchment. Both applications are briefly considered below.

2.12.3.2 The Use of Fallout Radionuclides The use of fallout radionuclides to obtain information on soil and sediment redistribution rates within a catchment is founded on the existence of a number of natural and manmade radionuclides that reach the land surface as fallout, primarily as wet fallout in association with rainfall, and are rapidly and strongly fixed by the surface soil or sediment. The subsequent redistribution of these radionuclides within a catchment or river system is a direct reflection of the movement of the soil or sediment particles to which the radionuclides are attached. By studying the post-fallout redistribution and fate of the selected fallout radionuclide, it is possible to obtain information on soil and sediment redistribution and, therefore, on erosion and deposition rates. The fallout radionuclide most widely used for this purpose is cesium-137 (137Cs) (see Ritchie and Ritchie, 2008). Cesium137 is a man-made radionuclide that was produced by the testing of thermonuclear weapons in the 1950s and early 1960s. The 137Cs released by these bomb tests was carried up into the stratosphere and transported around the globe. Significant fallout occurred in most areas of the world during the period extending from the mid-1950s through to the 1970s, although the depositional fluxes were much greater in the northern than the southern hemisphere. In the absence of further bomb tests after the Nuclear Test Ban Treaty in 1963, fallout effectively ceased in the mid-1970s. However, in some areas of the world a further fallout input occurred in 1986 as a result of the Chernobyl accident. Fallout from that accident was short-lived, but in some neighboring regions the total fallout associated with the Chernobyl accident exceeded the earlier bomb fallout. Cesium-137 has a half-life of 30.2 years and much of the original fallout still remains within the upper horizons of the soils and sediments of a catchment. By investigating the current distribution of the radionuclide in the landscape, it is possible to obtain information on the net effect of soil and sediment redistribution processes operating over the past B50 years (see Zapata, 2002). When sampling the soils and sediments in a catchment, attention is usually directed to both the inventory or the total amount of 137Cs contained in the soil or sediment (Bq m2) and its depth distribution. However, emphasis is frequently placed on the collection of bulk soil cores and their use to determine the inventory at the sampling point, since the sectioning of a core to determine the depth distribution necessitates the analysis of a much larger number of samples, which can prove to be costly and time-consuming. Samples are analyzed by gamma spectrometry and count times of 12–24 h may be required when activities are low. Mean soil redistribution rates over the past B50 years, since the main period of fallout, are established by comparing the inventories measured at individual sampling points with the reference inventory for the study site.

314

Catchment Erosion, Sediment Delivery, and Sediment Quality

The latter is commonly based on cores collected from an adjacent undisturbed area with minimum slope that can be expected to have experienced neither erosion nor deposition over the past B50 years. Points with inventories less than the reference inventory are indicative of eroding areas, whereas those with inventories in excess of the reference value indicate deposition. A range of conversion models have been developed for use in estimating erosion and deposition rates, based on the degree of departure of the measured inventory from the reference inventory (e.g., Walling and He, 1999a; Walling et al., 2002; Li et al., 2009). Using a similar approach, 137 Cs measurements have also been successfully used to estimate deposition rates on river floodplains over the past B50 years (Walling and He, 1993, 1997; Terry et al., 2002; Ritchie et al., 2004). Cesium-137 has now been successfully used in many areas of the world to obtain hitherto essentially unavailable information on medium-term rates of soil and sediment redistribution (see Figures 3 and 4, Table 3; Ritchie and Ritchie, 2008) and its value as a tracer has been strongly promoted by the International Atomic Energy Agency (IAEA) (see Zapata, 2002). Most applications have involved relatively small areas, since this permits the collection of sufficient samples to obtain representative information on the spatial patterns of soil and sediment redistribution involved. There is, nevertheless, a need for further work to establish procedures for using the approach in a reconnaissance mode, in order to obtain information from larger areas without a major increase in the number of samples that need to be collected and analyzed. Key advantages of the approach include the ability to obtain retrospective information on medium-term soil redistribution rates, the need for only a single sampling campaign, the provision of spatially distributed information relating to the individual sampling points, and the ability to collect information from the natural landscape, without the need to install plots or to otherwise constrain the location of the measuring points. Although most studies employing fallout radionuclides have been based on 137 Cs, both excess lead-210 (210Pbex) and beryllium-7 (7Be) have also been used in a similar manner (see Mabit et al., 2008). Lead-210 is a natural geogenic radionuclide produced as a product of the uranium decay series. Radium-226 (226Ra) is found in most soils and rocks and this decays to produce gaseous radon-222 (222Rn), which in turn decays to 210Pb. Some of the 222Rn diffuses upward through the soil and escapes into the atmosphere where it decays to 210Pb and is deposited as fallout. As with 137Cs, the 210 Pb fallout reaching the land surface is rapidly fixed by the soil and its subsequent redistribution is governed by the movement of soil and sediment particles. The fallout 210Pb is termed excess or unsupported 210Pb, to distinguish it from the 210 Pb produced by in situ decay, which will be in equilibrium with, or supported by, the parent 226Ra. The use of 210Pbex to document soil and sediment redistribution within the landscape employs similar assumptions and procedures to those used with 137Cs. Walling and He (1999b) discuss its use in soil erosion studies and He and Walling (1996) provide examples of its application for estimating rates of overbank sedimentation on river floodplains. The half-life of 210Pb is 22.3 years and therefore similar to that of 137Cs. However, because 210Pb

is a natural geogenic radionuclide, the fallout has been essentially constant through time and the activity in the soil will reflect fallout receipt and subsequent decay over the past B100 years. In the case of soil redistribution on slopes, the influence of past erosion on the present inventory will increase toward the present. Measurements of 210Pbex activity can therefore provide information on longer-term soil and sediment redistribution rates over the past B100 years and use of both 137Cs and 210Pbex in combination can provide additional information on the erosional or depositional behavior of a study area (He and Walling, 1996; Walling et al., 2003a) (see also Figure 4). Beryllium-7 is a natural cosmogenic radionuclide formed in the upper atmosphere by its bombardment with cosmic rays. In contrast to 137Cs and 210Pb, 7Be has a very short half-life of only 53 days and, because of this, it can be used to provide information on soil and sediment redistribution rates associated with individual events or short periods of heavy rainfall extending over a few weeks (e.g., Walling et al., 1999; Blake et al., 2002; Wilson et al., 2003; Schuller et al., 2006; Sepulveda et al., 2007). The principles involved in applying 7Be measurements are similar to those for 137Cs and 210 Pbex, but for most approaches it is important to ensure that the period of interest conforms to a number of requirements, to avoid carry-over effects from previous periods of heavy rain, which could influence the magnitude and spatial distribution of 7Be inventories across the study area. Walling et al. (2009) have recently described a refined procedure for employing 7Be measurements, which largely overcomes this constraint and makes the approach more generally applicable.

2.12.3.3 Sediment Source Fingerprinting Sediment source fingerprinting techniques can provide important information on the source of the suspended sediment transported by a stream. In simple terms, the techniques attempt to match the properties of the sediment to those of potential sources within the catchment and to establish the relative contribution of those sources to a given sediment sample. Source can be interpreted in terms of both spatial sources, representing different parts of the catchment, perhaps different tributaries or areas underlain by different rock types, and source types, representing sediment mobilized by different processes or from areas with different land use. A set of potential source types could, for example, include surface erosion from cultivated areas and areas of permanent pasture or range, gully erosion, and channel erosion. A wide of sediment properties including color (e.g., Krein et al., 2003), geochemistry (e.g., Collins and Walling, 2002), mineral magnetic properties (e.g., Caitcheon, 1993; Hatfield and Maher, 2009), radionuclide content (e.g., Wallbrink et al., 1998; Matisoff et al., 2002), and stable isotopes (e.g., Fox and Papanicolaou, 2007) have been used to fingerprint potential sources and in most cases a composite fingerprint incorporating several properties is required to discriminate between potential sediment sources. A mixing (or unmixing) model is used to estimate the relative contribution of the potential sediment sources to the sediment sample under consideration. Walling (2005) provides an overview of the development of source tracing techniques and their potential, emphasizing many complexities that need to be taken into account in order

Catchment Erosion, Sediment Delivery, and Sediment Quality

137 Cs estimated sedimentation rate (g cm−2 yr−1)

R. Sev ern

0.45 0.37 0.3 0.22 0.15

)

)

6

ght (m

1.76 200

3.2 1.7

6

Hei

ht (m

Heig

3.26

315

200

150

Dis

100

ce

tan

100

0

) (m

50

0 −10

0

nce

(m)

ta Dis

B4380 Study area

200

−

(a)

Cockshot rough

Severn

Dismantled railway 500m

0

210

Pb estimated sedimentation rate (g cm−2 yr−1)

R. S

)

1.76 200

6

1.7

6

200

150

Dis

100

ce

tan

100 D

)

(m

0 (m) nce a t is

50 0 −10

0 (b)

ght

3.2

Hei

3.26

ht (m

Heig

0.45 0.37 0.3 0.22 0.15

(m)

ever n

0

−20

Figure 4 The patterns of overbank deposition of fine sediment on a portion of the floodplain of the River Severn near Buildwas, Shropshire, UK established by Walling and his co-workers using 137Cs and 210Pbex measurements. 124 bulk cores were collected from the floodplain at the intersections of a 25-m grid. The estimates of mean annual sedimentation rate estimated using the 137Cs measurements relate to the past B40 years, whereas those based on the 210Pbex measurements relate to the past B100 years.

to obtain meaningful and reliable results. Of particular importance are the need to verify statistically the discriminatory power of the fingerprints employed (e.g., Collins et al., 1997a), to recognize many sources of uncertainty incorporated into the approach (e.g., Rowan et al., 2000) and to express and interpret the results accordingly, and to take account of possible differences between source material and sediment samples in terms of grain size composition and organic matter content (e.g., Collins et al., 1997b). The success of source fingerprinting techniques depends heavily on identifying a range of elements and/or isotopes that are capable of

discriminating potential sources with a high degree of reliability. Fallout radionuclides have frequently been successfully incorporated into fingerprints used to discriminate between surface sources under different land use (e.g., cultivation, pasture, and forest) and channel/subsurface sources (e.g., Walling et al., 2008) but a new generation of fingerprint properties based on compound specific stable isotopes (CSSIs) associated with the fatty acids produced by plants appears to offer the potential to discriminate between source areas supporting different vegetation covers. Gibbs (2008) reports a study undertaken in North Island New Zealand where CSSIs

316

Catchment Erosion, Sediment Delivery, and Sediment Quality

were used to establish the relative importance of areas under sheep pasture, indigenous forest, and exotic forest plantations as sources of the sediment deposited in a downstream estuary. Most existing fingerprinting studies focus on establishing the source of the suspended sediment transported by a river. In early work this commonly involved collecting individual samples of sediment, but more recent studies have frequently made use of time-integrating sediment traps (e.g., Phillips et al., 2000), in order to provide a single sample representative of the sediment transported during the period of sample

Table 3 A spatially integrated assessment of soil redistribution within the 7.5 ha field at Butsford Barton near Colebrooke, Devon, UK, based on the estimates of soil redistribution rates provided by 137Cs measurements presented in Figure 3 Parameter

Value

Percentage area with erosion (%) Percentage area with deposition (%) Mean erosion rate for eroding area (t ha1 yr1) Mean deposition rate for deposition zones (t ha1 yr1) Net erosion rate for the field (t ha1 yr1) Sediment delivery ratio for the field (%)

79 20 10 7.5 6.5 81

collection. In addition to suspended sediment, the fingerprinting approach has also been applied to overbank sediment deposits on floodplains (e.g., Bottrill et al., 2000) and fine sediment recovered from salmon spawning gravels (e.g., Walling et al., 2003c) and it is possible to generate a temporal perspective and to investigate changes in sediment source through time by applying the same approach to a sediment core collected from a lake or floodplain and interpreting downcore changes in sediment properties in terms of source fingerprints (e.g., Collins et al., 1997b; Walling et al., 2003b; Pittam et al., 2009). Taken together and combined with more traditional monitoring techniques for obtaining information on the sediment flux at the catchment outlet, these two sets of sediment tracing techniques afford a valuable means of obtaining much of the information required to establish a catchment sediment budget (e.g., Walling et al., 2001, 2006). Thus, for example, it is possible to link information on the source of the sediment load at the catchment outlet provided by source fingerprinting techniques with information on rates of sediment redistribution within those source areas and rates of accretion in sediment sinks such as river floodplains provided by fallout radionuclides, to quantify the key sources and sinks within the sediment budget of a catchment. Figure 5 depicts

Pang Cultivated fields

Pasture fields

Cultivated fields Gross erosion 79997t

Gross erosion 55263t

Lambourn Pasture fields

Gross erosion 1425t

Gross erosion 1960t In-field deposition 26378t

In-field deposition 54900t

In-field deposition 1680t

Field-to-channel Field-to-channel deposition deposition 28551t 121t

Input to channel system (334t) (159t) Channel banks/ subsurface sources 5t

Field-to-channel deposition 24559t

In-field deposition 675t

Field-to-channel deposition 223t

Input to channel system (538t) (527t) Channel banks/ subsurface sources 11t

Channel bed sediment storage 191t

Suspended sediment yield 498t

Channel bed sediment storage 229t

Suspended sediment yield 1076t

(Sediment delivery ratio = 1%)

(Sediment delivery ratio = 1%)

Figure 5 The sediment budgets established for the Pang ( B166 km ) and Lambourn (B234 km2) catchments in southern England by Walling et al. (2006). 2

Catchment Erosion, Sediment Delivery, and Sediment Quality the sediment budgets of the Pang (B166 km2) and Lambourn (B234 km2) catchments (located on the chalk of southern England), which were established using this approach. In this case, the highly permeable strata underlying the catchments and the resulting dominance of groundwater in the runoff from the catchment mean that storm runoff is limited and that little sediment leaves the catchment. However, there is evidence of relatively high rates of sediment mobilization and redistribution within the catchments, and in this environment the functioning of their sediment budgets is dominated by the internal sediment sinks.

2.12.3.4 The Future Recent years have seen important advances in the development of improved methods for characterizing and establishing catchment sediment budgets. The methods now available are able to provide information and understanding to support the development of sediment management programs. However, further advances are required to meet future information requirements, which are likely to place increasing emphasis on the targeting of sediment control strategies, in order to maximize the benefits achieved by investment in such strategies. The use of new sediment source fingerprints, such as CSSIs, can be expected to provide significant improvements in tracing sediment from specific sources and assessing the importance of those sources. In the case of fallout radionuclide applications, most existing work has focused on small areas and there is an important need to upscale their use to larger areas. Because of the limitations on sample numbers commonly associated with sample counting facilities, this upscaling cannot be achieved by simply increasing the number of samples collected. Attention needs to be directed to the development of reconnaissance sampling strategies capable of maximizing the information supplied by a small number of samples. In turn, there is a need to integrate the use of fallout radionuclide techniques with numerical modeling and geographical information systems (GISs), in order to optimize the spatial extrapolation of the resulting information. Advances in sensor technology will undoubtedly bring new and improved methods for monitoring sediment fluxes at catchment outlets that increase the temporal resolution of the records obtained and extend the scope of the data obtained. Scope undoubtedly exists to obtain valuable information on the grain size composition of the sediment load, as well as its magnitude. Such developments are likely to prove to be important in compensating for the progressive reduction in field staff dedicated to sediment sampling and other related monitoring activities that have occurred in many countries in recent years.

2.12.4 Modeling the Catchment Sediment Budget 2.12.4.1 The Requirement Supporting national and international programs that aim to reduce the threat soil erosion poses to agricultural production and downstream aquatic environments require systematic assessments of sediment sources and their connectivity to downstream impacts (Phillips, 1986). Assessments can provide a technical basis to assist targeting of limited resources to

317

maximize benefits (NLWRA, 2001; Bohn and Kershner, 2002). At river basin scale, managers are often faced with a paucity of erosion data or uneven distribution of measurements across the assessment area. Measurements will have been made at different times and various scales for a range of purposes. Modeling can enable systematic assessment of erosion severity over much larger areas than can be practically covered by measurement alone (Reid and Dunne, 2003). Modeling can also enable assessment over longer time periods with a wider range of climatic conditions, including potential future climates, which is critical for long-term planning, given the high temporal variability of erosion and sediment delivery. Modeling periods of decades may be required to represent adequately the aggregate effects of climatic variability. Several requirements for modeling erosion and sediment delivery can be identified from a management perspective: 1. Models should explicitly represent the primary erosion processes occurring, so that priorities can be developed, effective interventions identified, and the effect of alternative management scenarios simulated. 2. Spatial patterns in erosion rates should be identified by representing the primary environmental drivers of erosion and sediment delivery. 3. The connectivity of upstream erosion sources to downstream sediment loads should be represented, which requires consideration of sediment sinks as well as sources, and the potential for source connectivity to vary spatially depending on the location of sediment sinks. 4. Assessing erosion and sediment delivery at national or continental scales requires models with modest data requirements. 5. Where sediment-associated pollutants, such as phosphorus and agro-chemicals, are a focus of management, then sediment particle size fractions should be explicitly represented. Pollutants preferentially attach to fine sediment fractions (Section 2.12.5.2), and erosion and transport process behavior will differ between fine and coarse fractions. These requirements, particularly the first three, suggest that a sediment budget is a suitable framework for modeling, because it accounts for the sources, transport and sinks of material, with a river basin being the confining domain (Trimble, 1993; Reid and Dunne, 2003). The following subsections describe three aspects of sediment budget modeling: (1) the evolution and development of models, (2) an example of integrated sediment budget modeling, and (3) the current status and future directions of model development and application.

2.12.4.2 Model Development 2.12.4.2.1 Modeling approaches and model complexity Erosion and sediment delivery modeling can be based upon interpolation and synthesis of measurements, or upon physical reasoning and identification of the environmental factors that control the key processes. Typically, a combination is used (DeRoo, 1996), with process models providing spatial resolution between measurement points and helping to identify the upstream erosion and delivery processes, and

318

Catchment Erosion, Sediment Delivery, and Sediment Quality

measurement points being used to constrain model predictions. Three approaches to erosion and sediment delivery modeling can be identified, representing different weightings between measurement and mathematical process description (Beck, 1987; Merritt et al., 2003): 1. Empirical modeling, usually based on a small number of causal variables, often spatially and temporally lumped, and calibrated to measurements. These include catchmentspecific relationships between catchment area and sediment yield for example. 2. Conceptual process modeling, which represents generation, routing and storage processes within landscape units or catchments using simple representation of their controlling parameters, and without process interactions. They may be semi-lumped into units or subcatchments with time-steps of days to decades (see Merritt et al., 2003). 3. Mechanistic physical-process modeling provides detailed representation of runoff-generation processes, and usually for application at point, field, or small watershed scale, at finer temporal and spatial resolution than employed in conceptual modeling.

2.12.4.2.2 Empirical modeling of catchment sediment yield

(large)

Empirical modeling of catchment sediment yield emphasizes available data rather than process representations, and this feature can represent both a strength and weakness, depending on the problem being addressed. On the one hand, sensitivity to data constrains model output and enhances the opportunity for new system understanding, making empirical modeling particularly useful where system understanding is weak. Model empiricism is also vital for investigating or calibrating models of sediment yield processes for which the fundamental physical constraints are not sufficiently well known or described. Empirical models are generally relatively simple, and consequently have modest data requirements.

Model unable to exploit data

Identifiability problems

(small)

Predictive performance

Model complexity generally increases from empirical through conceptual to mechanistic approaches, as the level of process representation and the level of spatial and temporal resolution required increase. There is interdependence between process complexity and the temporal and spatial resolution of a model; finer spatial and temporal resolution requires more complexity in process representation, and vice versa. For example, conceptual rainfall-runoff models require more storage terms and model parameters for predicting daily or monthly runoff than for long-term average runoff (Jothityangkoon et al., 2001). The interdependence between process complexity and model resolution scale means that most empirical models focus on lumped process representations of catchment sediment yield or individual erosion processes, and can be implemented over large areas and long time periods.

Most mechanistic models are spatially distributed and focus on hillslopes or small watersheds for individual events. Conceptual models are often semi-lumped, and focus on simple representation of catchment erosion and deposition processes. The most appropriate model design considers each component of the erosion and sediment system with a level of complexity that is appropriate for the problem at hand and for the data available (Reid and Dunne, 2003). This is important because, for a given data availability and process knowledge, there is a maximum model complexity which fully exploits the information provided by input data and above which predictive capacity is reduced (Grayson and Blo¨schl, 2000; Figure 6). Above this level of complexity, it also becomes difficult to identify appropriate parameter values (Beck, 1987). The optimization of model complexity to maximize predictive capacity has contributed to the number of models which have been developed in recent decades. The evolution of each modeling approach is briefly described below, where some of the more widely known models are used as examples and model-specific reviews are cited.

(la

rge

)

Da

ta

)

ava ila

(large

bili

ty

(sm

all)

(small)

Model

xity comple

Figure 6 Schematic diagram of the relationship between hydrologic model complexity, data availability and predictive performance. Reproduced from Grayson R and Blo¨schl G (2000) Spatial modelling of catchment dynamics. In: Grayson R and Blo¨schl G (eds.) Spatial Patterns in Catchment Hydrology, ch. 3, pp. 51–81. Cambridge: Cambridge University Press.

Catchment Erosion, Sediment Delivery, and Sediment Quality

Examples of empirical sediment yield modeling include:

•

Modeling the sediment load at a river station, using discharge and sediment concentration data. Commonly, frequent discharge measurements (L3 T1) are available, with occasional measurements of suspended sediment concentration (M L3). There is considerable short-term temporal variability in sediment concentration (Nistor and Church, 2005). The key modeling challenge is to explain and predict the variation in sediment concentrations between observations, with sediment load in a given time period (M T1) then being the product of discharge and concentration. A common approach is to model concentration using functions of discharge, or sediment rating curves (Asselman, 2000). Rating curves assume steady-state behavior, although records can be divided into multiple time windows to represent system changes. More recently, neural network and other modeling techniques have been employed, which consider the influence of antecedent as well as present discharge and concentration on sediment load (Kisi, 2005). Lumped models of sediment yield, based on upstream catchment area (Wasson, 1994), include other basin metrics such as relief, runoff, climate zone, lithology, and anthropogenic factors (Syvitski and Milliman, 2007).

319

constraints on process behavior. Thus, conceptual models evolve as the process understanding improves. Process component models are usually somewhat empirical in nature, requiring some calibration to match observed erosion and deposition rates, but providing predictive capacity where measurements are not available. There are several types of conceptual process models of erosion and sediment budgets, which employ different types of input data:

2.12.4.2.3 Conceptual process modeling of catchment sediment budgets

1. A catchment sediment budget developed from a combination of air-photo interpretation, mapping of geomorphic units or zones, field erosion measurements, and dating of sediment deposits (Trimble, 1983; Wasson et al., 1998; Curtis et al., 2005). As the amount of data increase, these budget models become increasingly complex in terms of the sources and sinks represented, and the spatial and temporal resolutions and extents. These models can help guide rehabilitation efforts (Trimble, 1993). Prediction outside the study area is limited where measurements dominate model inputs. 2. A reach sediment budget generated from load estimates based on sediment monitoring data, which can be used to identify reaches of net erosion or deposition (Singer and Dunne, 2001). 3. A source fingerprinting analysis based on sediment tracer properties, which assesses the relative contribution of erosion processes and/or source areas to river sediment without requiring direct measurement of erosion or deposition rates (Walling, 2005; Walling and Collins, 2008; Davis and Fox, 2009). 4. Semi-lumped spatial models of sediment budgets which commonly use GIS functions to divide river basins into subcatchments or watersheds, each draining to a river link, and route sediment through river networks (Benda and Dunne, 1997; Prosser et al., 2001b; Wilkinson et al., 2006, 2009). 5. Distributed spatial models which compute hillslope runoff, erosion, and sediment delivery at the resolution of input data sets, with sediment routed according to topography. Examples include AGNPS (Agricultural Nonpoint Source model; Young et al., 1989) and SEDEM (SEdiment DElivery Model; Van Rompaey et al., 2001). More precisely, these models should be seen as partially distributed, with some lumped elements remaining.

Conceptual process modeling generally represents sediment routing through a catchment, using a semi-lumped structure of subcatchments or management units. Source (erosion) and sink (deposition) processes are commonly represented for each defined spatial unit, although not all models calculate complete sediment budgets. Feedbacks between processes will not usually feature. Models that identify sediment sources and delivery through the river network are more consistent with field-based evidence of sediment fluxes (Trimble and Crosson, 2000), and are more suitable than lumped models of sediment yield for the modeling requirements identified in Section 2.12.4.1 (Phillips, 1986). Individual erosion and deposition components of conceptual process models are based on prior field measurements and studies, which provide the basis to identify and mathematically formulate the environmental controls and physical

Spatial lumping of model domains into morphological units or subcatchments, and application of lumped parameter values across units and timescales is common in conceptual models. However, the scale of lumping can strongly influence the model predictions. For example, spatial patterns of land use within units may be far from random, which can affect sediment delivery. For practical reasons related to the time and effort involved, more complex semi-lumped and distributed models were originally limited to small hillslope applications. By the late 1990s, the wide availability of GIS software and techniques enabled more complex models with spatially varying input data to be easily applied to larger areas. Today, data requirements and the ability to realistically describe processes are the dominant limitations on modeling.

•

Weaknesses of empirical modeling include: (1) the spatial and temporal resolution and extent are limited by the data available; (2) their lack of explicit process representations can limit predictive capacity outside the study area or measured range of environmental characteristics; (3) the heterogeneity of catchment characteristics such as rainfall, topography, lithology, and land use is not usually represented in spatially lumped models; this reduces predictive capacity, given the significant spatial correlations, and nonlinear dependencies, between slope gradient, runoff, and other driving variables of erosion (Van Rompaey et al., 2001); (4) the absence of source and sink process representations in empirical sediment yield models can limit the number of different types of data which can be meaningfully assembled. Data sets that may enhance the model may instead be used to interpret model results (e.g., Singer and Dunne, 2001).

320

Catchment Erosion, Sediment Delivery, and Sediment Quality

2.12.4.2.4 Mechanistic, physically based modeling of hillslope processes Available mechanistic models span a range of complexity in the process representations used to model runoff and erosion. Some models are more physically based, representing soil infiltration and runoff routing analytically, for example, using kinematic waves, while others use empirical approaches such as runoff curve numbers. Surface erosion is modeled using USLE-based detachment equations, or more complex shear stress and stream power functions (DeRoo et al., 1996; Srivastava et al., 2007). Some models predict net soil loss considering both detachment and deposition (Nearing et al., 1989). Indicating their focus toward hillslope or small watershed scales, mechanistic models do not commonly include gully and riverbank erosion and channel deposition processes. Several stand-alone models predict ephemeral gully erosion using flow shear stress and sediment transport capacity, although their predictive capacity has received little testing (Poesen et al., 2003). Several mechanistic models of permanent gullies have also been developed, describing the evolution of morphology during the early stages of gully development and the final morphological characteristics (Poesen et al., 2003). Again, however, their applicability has not been widely tested. Available physically based models cover a wide variety of spatial and temporal scales, with the former ranging from individual hillslopes to fields. Some models are pixel-based, and accommodate GIS data to facilitate modeling small watersheds (DeRoo et al., 1996; Srivastava et al., 2007). The majority of physically based models are designed to simulate individual events rather than long time periods (Aksoy and Kavvas, 2005). Many mechanistic and distributed models have input data requirements that are technically or financially unattainable over large river basins or continents or multiyear time periods (Van Rompaey et al., 2001). The distributed or hillslope unit structure of mechanistic models addresses problems associated with spatial lumping, such as spatial correlation between inputs. However, considerable spatial variability in surface roughness, slope gradient, and other variables often exist even within model spatial units. Distributed models are sensitive to errors in surface slope and topography. Consequently, it is common for mechanistic physics-based models to require calibration to reproduce observed behavior, which potentially introduces error in process modules.

2.12.4.3 SedNet – A Sediment Budget Model for River Networks 2.12.4.3.1 Model outline To illustrate considerations in design and application of sediment budget models, this section describes the SedNet (Sediment budget river Network) model. SedNet constructs budgets of the primary sources and sinks of sediment for each link in a river network (Prosser et al., 2001b; Wilkinson et al., 2004, 2009). This model structure may be considered generic to all semi-lumped sediment budget models. The structure enables spatial representation of the connectivity of

upstream sources to downstream yields, including the role of floodplains and impoundments within catchments (Prosser et al., 2001c). Predicted suspended sediment yield is supply limited in the long term, which is consistent with observations. The river network is defined from a digital elevation model (DEM). Separate budgets are constructed for sand and gravel bed material (Wilkinson et al., 2006), for suspended sediment (Wilkinson et al., 2009), and for particulate and dissolved phosphorus and nitrogen (Wilkinson et al., 2004). The sediment yield from the downstream end of each link accounts for material sourced from hillslope and gully erosion in the subcatchment which drains directly to the link, bank erosion along the link, and from upstream tributaries. Deposition is accounted for on floodplains, in impoundments or reservoirs, and accumulation of bed material in the river channel. The processes of land-sliding, debris flow, and hillslope soil creep are not significant sediment sources in the Australian environment (for which SedNet was developed), although they could be added for model application elsewhere. The net change in channel storage of suspended sediment over decades is assumed to be negligible relative to other terms, and so in-channel deposition and re-entrainment of suspended sediment is ignored. The budget is reported as mean annual values for a set of conditions. The effects of temporal variability in climate and hydrology on each source and sink are modeled by regionalizing statistics of daily discharge. The process representations are generally conceptual in nature, designed to show the primary physical controls to provide predictive capacity in low-data environments. For example, hillslope erosion is represented by the Revised Universal Soil Loss Equation (RUSLE; Renard et al., 1997), with a sediment delivery ratio accounting for deposition within hillslopes. Gully sediment yield is constrained by the estimated volume of gully networks and their period of development. Spatial variation in riverbank erosion is estimated as a function of stream power, and the extent of erodible soil and riparian vegetation (Wilkinson et al., 2009). Parameter values for a given environment are specified based on the knowledge of erosion and deposition processes developed through field measurement (Bartley et al., 2007), reconstruction of erosion histories (Wasson et al., 1998), sediment tracing (Wallbrink et al., 1998), and independent sediment yield estimates (Rustomji et al., 2008).

2.12.4.3.2 Management applications SedNet was developed for the Australian National Land and Water Resources Audit, which investigated the spatial patterns in erosion processes, and the offsite impacts of agriculture across the Australian continent (NLWRA, 2001). The modeling indicated marked differences in sediment supply between regions, with gully and river bank erosion dominating sediment supply in temperate regions, and hillslope erosion dominating in tropical regions, due to the higher rainfall intensity. Only 25% of fine sediment delivered to streams was predicted to be delivered to estuaries overall (Prosser et al., 2001a).

Catchment Erosion, Sediment Delivery, and Sediment Quality

hotspots. This approach provides improved predictive capacity over spatially lumped models (Wilkinson et al., 2006). River links predicted to have bed material accumulation have impaired biological health, with lower abundance Suspended sediment export (kt yr−1)

SedNet has since been applied at regional scale using higher-resolution datasets, to better support catchment planning. For example, increased riverine sediment exports from the catchments draining to the Great Barrier Reef (GBR) threaten to degrade near-shore coral reef and benthic ecosystems (De’ath and Fabricius, 2010). Modeling predicted that 70% of sediment export comes from just 20% of the total catchment area (Figure 7). The spatial pattern of contribution to export was highest in near-coastal areas with high rainfall intensity, steep slopes, and more intensive land management (McKergow et al., 2005). The model has also been used to compare scenarios of future management. Figure 8 demonstrates that targeting erosion control to areas and erosion sources in descending order of their contribution to sediment export can achieve reductions in export several times larger than would be achieved by spatial random changes in land management. Evaluation of SedNet against yield estimates from suspended sediment rating curves, and against sediment tracer data, indicates that the model can reliably differentiate between the areas contributing most and least to basin yield, provided that input data are of good quality (Wilkinson, 2008; Wilkinson et al., 2009). SedNet has also been used to assess the location and extent of accumulations of sand and gravel bed material within river networks, indicating that up to 25% of the river network is affected in some river basins, particularly downstream of gully and riverbank erosion

321

620 Targeted riparian revegetation

600

Random riparian revegetation

580 560 540 520 500 0

200 400 600 800 Riparian revegetation (km)

1000

Figure 8 Simulated reductions in suspended sediment export from the Murrumbidgee River catchment, showing that spatially-targeted control of gully and river bank erosion can achieve larger reductions in export than spatially-random control measures. Reproduced from Wilkinson SN, Prosser IP, Olley JM, and Read A (2005) Using sediment budgets to prioritise erosion control in large river systems. In: Batalla RJ and Garcia C (eds.) Geomorphological Processes and Human Impacts in River Basins, IAHS Publication 299, pp. 56–64. Solsona, Catalonia: IAHS Press, with permission from IAHS press.

Ratio of hillslope to channel Specific suspended Contribution of suspended Current minus natural (gully and bank) erosion sediment load (t ha−1 yr−1) sediment to the coast (t ha−1 yr−1) contribution of suspended sediment < 0.01 < 0.2 < 0.5 to the coast (t ha−1 yr−1) 0.2−0.5 0.5−2 0.01−0.05 < 0.01 0.05−0.1 0.5−1.4 >2 (a) (b) (d) (c) 0.01−0.05 > 1.4 0.1−05 0.05−0.1 0.5−1 0.1−0.5 >1 0.5−1 >1 N

0

200

400 km

Figure 7 SedNet results for the catchments draining to the Great Barrier Reef (a) estimated ratio of hillslope erosion to channel erosion (gully plus riverbank) in each subcatchment, (b) predicted specific suspended sediment load, (c) predicted contribution of suspended sediment to the coast under current conditions, and (d) the difference between estimated current and natural contribution to suspended sediment export. Reproduced from McKergow et al. (2005) Sources of sediment to the Great Barrier Reef World Heritage Area. Marine Pollution Bulletin 51: 200–211, with permission from Elsevier.

322

Catchment Erosion, Sediment Delivery, and Sediment Quality

of habitat-sensitive macro-invertebrate taxa (Harrison et al., 2008).

2.12.4.4 Current Status and Future Directions 2.12.4.4.1 Modeling across scales for planning and management There is an increasing demand for land management and environmental stewardship to be underpinned by technical assessments. Land management and planning is increasingly driven by, and administered through, national and international programs, rather than locally. The challenge for erosion and sediment modeling is to provide robust assessments of the effects of local and dynamic changes in land management on erosion and sediment yield outcomes over large areas (100–100 000 km2) and long time periods (10–100 years). Modelling and measurement are becoming more integrated in the study of catchment erosion and sediment delivery. Remote sensing has provided a rapid increase in the quality of spatial data of topography, soil, and vegetation cover (Vrieling, 2006). A range of data types are now being used to constrain the modeled sediment budget, including independent load estimates and sediment tracing (Rustomji et al., 2008; Wilkinson et al., 2009), and dating of sediment deposits (de Moor and Verstraeten, 2008). There are no models that simultaneously operate at all scales from point land management practices to basin sediment yields, and such a model would require large increases in input data resolution (Srivastava et al., 2007). The present response to the need for information at multiple scales is to apply erosion and sediment delivery models of different scales in a more closely coupled and integrated fashion. For example, conceptual river basin models can be used to identify priority source areas, where mechanistic models are applied at the field scale to optimize management practices, the outputs of which are then represented in the basin-scale model. Given the considerable resources and expertise required to operate models and improve data inputs, having a clear strategy for model integration will help achieve the most effective outcomes. The current status of empirical, conceptual, and mechanistic modeling approaches is summarized below. Empirical models of sediment yield can help to identify areas or catchments with more intense erosion or deposition (Singer and Dunne, 2001). Because they do not usually identify sediment source processes, their ability to simulate climate or land management scenarios is limited. Empirical sediment yield models are often used to validate conceptual and mechanistic sediment budget process models (Takken et al., 1999; Wilkinson et al., 2009). In this context, methods to quantify the uncertainty associated with empirical sediment load estimates are important (Rustomji and Wilkinson, 2008). Conceptual and mechanistic models provide frameworks for routing material from sources to sinks and downstream environments, to identify erosion hotspots and to estimate the effects of practice changes. The benefit of conceptual and mechanistic modeling approaches are realized only when all of the important erosion and deposition processes are represented. Many models provide in-depth treatment of surface erosion but omit gully and riverbank erosion and channel

deposition processes, despite their importance at basin scale (de Vente and Poesen, 2005). Inappropriate or omitted process representations can lead to a model predicting the right river basin sediment yield for the wrong reasons, jeopardizing investment priorities (Boomer et al., 2008). The application of mechanistic models across the large areas and long time-periods of interest for land management is often constrained by limited data with which to specify parameter values, and simplified versions of mechanistic models have emerged (Van Rompaey et al., 2001; Borah and Bera, 2002; Brasington and Richards, 2007). Stochastic or probabilistic descriptions of hydrology and soil properties have also been proposed (Aksoy and Kavvas, 2005). Spatial interfaces have been developed to facilitate application of hillslope profile models to broader areas with complex topography (e.g., Ascough et al., 1997; Renschler, 2003). Simulating the effects of global warming-induced climate change on erosion and sediment yield is likely to become more common, but requires careful consideration of all model inputs, including changes in variability and mean condition.

2.12.4.4.2 Directions in modeling erosion and deposition processes A fundamental principle guiding further developments in modeling erosion and sediment delivery processes is that the complexity and the spatial and temporal scales should be appropriate to the depth of process knowledge for the study area, the input data available, and the modeling objective. Four less-well developed areas can be identified as foci for further development of process modeling: Overland sediment transport capacity. It is now well recognized that models of catchment sediment delivery should separate surface erosion and overland sediment transport from erosion and deposition processes occurring at larger spatial scales, such as gully erosion and floodplain deposition (Trimble and Crosson, 2000; de Vente et al., 2007). Predicting the spatial variations in overland transport capacity is challenging due to local variability in terrain, soil properties, and vegetation cover, and consequently predictions are usually calibrated to match observed hillslope sediment yields (Verstraeten et al., 2007). Recent approaches include: (1) pixelbased sediment transport capacity estimation as a function of the erosion rate (Van Rompaey et al., 2001), (2) functions of stream power (Young et al., 1989; Prosser and Rustomji, 2000), and (3) subcatchment delivery ratio estimation using functions of storm duration and runoff travel time, considering the distance to stream, surface slope, and roughness (Ferro and Minacapilli, 1995; Lu et al., 2005). Hillslope mass failures are an important sediment source in many mountainous and hilly areas. Modeling the spatial controls on hillslope mass failure has received less attention to date than sediment mobilization processes that are more common in lowland areas. Rainfall intensity and duration, slope gradient, and soil properties are important determining factors (Chang and Chiang, 2009). Predicting the occurrence of mass failure at given locations is difficult over broad areas, and the temporal patterns of sediment delivery to streams can be described stochastically (Benda and Dunne, 1997).

Catchment Erosion, Sediment Delivery, and Sediment Quality

Channel network erosion. Available models of the extent and erosion rate of gully networks are empirical, which is an appropriate given the strong random component to the upstream extent of incised channel networks (Shreve, 1966). Modeling gully extent across large river basins can be based on manual mapping of sample areas. Soil type, slope, and climate variability provide useful, but not powerful, explanatory variables (Hughes et al., 2001; Kuhnert et al., 2010). Where gully extent estimates are available, gully volume and age provide constraints on long-term gully sediment yield (Prosser et al., 2001b; Wilkinson et al., 2006). Gully sediment yield also declines with gully age (Prosser and Winchester, 1996). There are no widely validated models available for assessing gully sediment yield dynamics over shorter time periods, although runoff, land use, drainage area, slope gradient, and soil properties are factors commonly applied to explain local variability in measured rates (Poesen et al., 2003; Valentin et al., 2005). The most common drivers used to predict spatial variation in bank erosion rates are bankfull discharge (Rutherfurd, 2000), stream power (Finlayson and Montgomery, 2003), riparian vegetation, and bank erodibility (Wilkinson et al., 2009). Recent work is improving ability to represent the mechanisms driving bank erosion, including subaerial weathering, scour, mass failure, and channel meandering (e.g., Lawler, 1995; Sun et al., 1996; Abernethy and Rutherfurd, 1998; Langendoen and Simon, 2008). Achieving robust predictions at finer temporal resolution is a particular challenge, but is not always required for management purposes. There have been limited data on the extent and rates of the above erosion processes, which have constrained model development. However, high-resolution DEMs from laser altimetry are now providing data over much larger spatial extents on gully dimensions (Ritchie, 1996) and on bank erosion rates (Notebaert et al., 2009). Pixel resolution is a key factor influencing the utility of laser altimetry DEMs for mapping erosion features and quantifying erosion rates (Notebaert et al., 2009). River channel and floodplain deposition. Representing deposition remains essential to modeling sediment delivery at river basin scale in most environments. However, hydraulic controls on sediment deposition cannot generally be resolved in basin-scale modeling (Nicholas et al., 2006). Conceptually, the primary controls on floodplain deposition are the sediment delivery to floodplains, controlled by the overbank discharge and concentration, and the residence time of overbank flow determined by floodplain size (Prosser et al., 2001c). More complex models consider shallow-water hydraulics more explicitly, and the uncertainty associated with model parameters (Nicholas et al., 2006). Dating of floodplain sediments is useful to verify model predictions (Nicholas et al., 2006; de Moor and Verstraeten, 2008). Long-term fine sediment deposition within river channels, such as lateral point-bar accretion, is controlled by similar fluid mechanics principles to vertical floodplain deposition. Accounting for temporary sediment storage within river channels, which is remobilized in subsequent flow events, can be important for modeling fine sediment yield in shorter time periods, and accounting for the progression of bed material through river networks (Viney and Sivapalan, 1999; Wilkinson et al., 2006).

323

However, few river basin models currently represent temporary channel storage, and further development is warranted. Analytic representations of channel storage and sediment wave routing are under active development, but data availability has prevented application or verification at basin scale (Cui et al., 2005; Lauer and Parker, 2008).

2.12.4.4.3 Model uncertainty considerations Uncertainty analysis methods are covered in the Chapter Chapter 2.17 Uncertainty of Hydrological Predictions. Many of these methods are applied in erosion and sediment modeling, especially with more complex models where parameter values may be less well defined. The importance of quantifying model uncertainty, relative to generating best estimates of catchment function, depends on whether the modeling purpose is to formulate hypotheses for further investigation, or for practical application such as developing and justifying investment priorities (Sivapalan, 2009). Understanding the relative contributions of sources of uncertainty is useful for guiding efforts to improve in model performance. This is determined by model sensitivity to changes in parameter values, but also to the levels of uncertainty in each parameter (Reid and Dunne, 2003). Calibration data on erosion and deposition rates are sparser than data on rainfall and runoff, and there is potential for calibration of erosion and sediment models to distort predicted spatial patterns and sediment source contributions. The method of model calibration should ideally align with the modeling purpose. For example, sediment yield data are useful for calibrating sediment yield predictions, but if the modeling purpose is to predict relative contributions of erosion processes to yield then sediment tracing data may be more useful. The requirement for evaluating model predictions is especially important when modeling environments in which the erosion and deposition processes are less well understood.

2.12.5 The Quality Dimension 2.12.5.1 Introduction Although, as indicated in Section 2.12.1, fluvial/lacustrine suspended and bed sediments have traditionally been treated as a physical issue (e.g., reservoir sedimentation, channel and harbor silting, bridge scour, and soil erosion and loss), they also can pose a significant chemical/toxicological (waterquality) problem (Vanoni, 1977; Walling, 1977; Baker, 1980; Fo¨rstner and Wittmann, 1981; Salomons and Fo¨rstner, 1984; Ferguson, 1986; Horowitz, 1991, 1995; de Vries and Klavers, 1994; Stumm and Morgan, 1996; US Environmental Protection Agency, 1997; Horowitz et al., 2001; Walling et al., 2003d; Blum et al., 2004; Cinque et al., 2004; Reed et al., 2004; de Vente and Poesen, 2005; Radaone and Radaone, 2005; Walling, 2005; de Arau´jo et al., 2006; Black et al., 2007; Domenici et al., 2007; Horowitz and Stephens, 2008). Chemical constituents that primarily are sediment-associated fall into a general class called hydrophobes or hydrophobic compounds (e.g., Fo¨rstner and Wittmann, 1981; Luthy et al., 1997; Warren et al., 2003). This group includes heavy metals/trace elements (e.g., Cu, Pb, Zn, As, and Hg), nutrients (e.g., N, P, Si, and C), and persistent organic compounds such

324

Catchment Erosion, Sediment Delivery, and Sediment Quality

as polycyclic aromatic hydrocarbons (PAHs), polychlorinated biphenyls (PCBs), dioxin, kepone, and chlorinated pesticides (e.g., Aldrin, Chlordane, Mirex, DDT and its breakdown products DDD and DDE; e.g., US Environmental Protection Agency, 1997; Simpson et al., 2005). Even in relatively pristine environments, barely detectable dissolved-constituent concentrations occurring in the water column can simultaneously be detected at levels 3–5 orders of magnitude higher in association with naturally occurring suspended and bed sediments (Fo¨rstner and Wittmann, 1981; Fo¨rstner, 1989; Horowitz, 1991; Chapman, 1992; Foster and Charlesworth, 1996; Horowitz and Stephens, 2008). Further, bed sediments can make substantial chemical contributions to interstitial water, usually as the result of changing redox conditions that trigger post-depositional mineralogical changes and subsequent chemical remobilization, often into a more bioavailable form. Numerous studies have demonstrated that sedimentassociated chemical constituents can affect aquatic organisms. The organisms can range from small zooplankton (near the base of the food chain), through benthic organisms that live in intimate contact with bed sediment and its surrounding interstitial water, to humans, who may be affected ultimately as constituent levels increase and bioaccumulate up the food chain (Fo¨rstner and Wittmann, 1981; Salomons and Fo¨rstner, 1984; Chapman, 1992; Fo¨rstner and Heise, 2006). Ever since the publication of the Hawkes and Webb (1962) treatise on geocemical exploration, as well as the subsequent publication of numerous geochemical atlases (e.g., Webb et al., 1978; Fauth et al., 1985; Ottesen et al., 2000), there is a widely accepted perception that suspended and bed sediments reflect local environmental inputs. This is the result of both physical (e.g., grain size, surface area) and chemical (e.g., unbalanced surface charges, presence of oxyhydroxide or organic coatings) factors that make aquatic sediments akin to chemical sponges (Fo¨rstner and Wittmann, 1981; Horowitz, 1991; US Environmental Protection Agency, 1997). Hence, sediment-associated chemical levels can increase or decrease in response to natural environmental processes (e.g., changes in Eh, pH, and grain-size distribution) and interactions (e.g., changes in local geology, volcanic activity). For similar reasons, sediment chemistry also tends to reflect anthropogenically derived contributions, for example, due to land use changes, from both point and nonpoint sources (e.g., Reimann and Garrett, 2005; Horowitz and Stephens, 2008).

2.12.5.2 Basic Sediment Geochemistry The majority of sediment-associated chemical constituents are found on or near the surface of sediment particles, and usually are held by sorption or complexation as a result of unbalanced surface charges (e.g., Fo¨rstner and Wittmann, 1981; Salomons and Fo¨rstner, 1984; Horowitz, 1991). As such, grain size and particle surface area play a significant role in controlling sediment chemical-associated concentrations (e.g., Horowitz, 1991). Generally, as particle size decreases, total surface area increases, as do the chemical levels; hence, elevated concentrations are more likely to be found associated with silt- and/

or clay-sized particles (r63 mm) than coarser sand-sized material (Z63 mm). Although some contaminants can attach directly to the surfaces of sediment particles (e.g., clay minerals), it is more typical to find them associated with particle coatings that are composed of either organic matter or Fe and/ or Mn oxides and oxyhydroxides (e.g., Fo¨rstner and Wittmann, 1981; Fo¨rstner, 1989; Horowitz, 1991; Foster and Charlesworth, 1996). Authigenic minerals that form in situ, and which exist as separate particles, may entrain chemical constituents within their crystalline or cryptocrystalline structure as a result of either chemical bonding or physical trapping (e.g., Fo¨stner and Wittmann, 1981; Horowitz, 1991). As with surface coatings, sorption/desorption processes may increase or reduce associated chemical levels depending on changing physicochemical conditions. Less commonly, substantial concentrations of inorganic constituents may be held within mineral lattices. This is most likely to occur in association with mining or mining related and some industrial activities, and occurs as a result of the discharge of ore minerals and/or mining/industrial waste (typically sulfides such as pyrite (Fe), arsenopyrite (As), galena (Pb), sphalerite (Zn), etc.), through the physical erosion of exposed mine tailings (e.g., Horowitz et al., 1988, 1993; Pope, 2005), or industrial discharges.

2.12.5.3 Major Issues Associated with Sediment Quality Despite the potential environmental impacts of sedimentassociated chemical constituents, only a very limited number of countries currently (e.g., Canada, The Netherlands, Australia, New Zealand, and Germany) have established sediment-chemical regulatory limits; however, many have established guidelines (e.g., Persaud et al., 1993; US Environmental Protection Agency, 2005; Simpson et al., 2005). This situation reflects a number of long-standing arguments associated with sediment chemical quality that have yet to be fully resolved. Four of the most significant ones are: (1) what are the background/baseline concentrations for a variety of sediment-associated constituents; (2) how best to collect representative suspended and bed sediment samples for subsequent chemical analysis; (3) how best to determine the concentrations of sediment-associated constituents; and (4) how to estimate/determine bioavailability?

2.12.5.3.1 Background/baseline sediment-associated constituent concentrations Unlike the vast majority of sediment-associated synthetic organic compounds that have no natural source(s), unless they are manufactured copies of natural substances, sedimentassociated inorganic constituents typically do occur in the environment. As a result, a background/baseline level must be established to determine the presence of contamination and/ or the impact of variations in land use (e.g., Goldschmidt, 1958; Hawkes and Webb, 1962; Plant et al., 1997; Reimann and Garrett, 2005). Background and baseline are concepts that tend to be used interchangeably, and often are qualified by terms such as geochemical, natural, or ambient. However, background concentrations usually refer to chemical levels that imply the exclusion of anthropogenic influence whereas baseline concentrations are typically determined at a

Catchment Erosion, Sediment Delivery, and Sediment Quality

particular point in space and/or time; albeit, it may imply limited anthropogenic effects (e.g., Gough, 1993; Reimann and Garrett, 2005). The occurrence of natural geological phenomena (e.g., volcanic eruptions, unworked ore deposits) and changing local geology, combined with the advent of the industrial revolution, and the concomitant eolian and fluvial distribution/redistribution of a variety of materials, and their associated chemical constituents means that it is unlikely that background concentrations can be determined from any current surficial material. It also means that background/baseline concentrations can change spatially and temporally. This leads to a major issue: how to define the natural inorganic chemical composition of sediments, and has led many geochemists/ environmental chemists to accept two precepts: (1) background chemical composition is neither spatially nor temporally static, and should be viewed as a range rather than as a single value; and (2) chemical changes induced by natural processes should not be viewed as contamination, even when the source may be an unmined mineralized zone. Hence, only sediment-chemical enhancements derived from anthropogenic activities/sources should be viewed as contamination. On the other hand, regulatory agencies tend to take a broader view, and define contaminated sediment as that which ‘‘contains chemical substances in excess of appropriate geochemical, toxicological, or sediment quality criteria or measures, or is otherwise considered to pose a threat to human health or the environment’’ (US Environmental Protection Agency, 1997). For many years, geochemists attempted to provide a single set of chemical values in an effort to define the natural background or baseline for the inorganic composition of sediments, and used average crustal chemical abundances, or constructed values for a so-called average shale for that purpose (e.g., Clarke and Washington, 1924; Poldervaart, 1955; Turekian and Wedepohl, 1961; Taylor, 1964; Krauskopf, 1967; Bowen, 1979; Wedepohl, 1995). In turn, these background/ baseline concentrations were then used to determine the presence and extent of sediment-chemical contamination. A recent US Geological Survey (USGS) National Water Quality Assessment (NAWQA) Program study has provided a fairly comprehensive and up-to-date (1990–2000) continental-scale assessment of baseline values for a wide variety of sedimentassociated inorganic constituents (Horowitz and Stephens, 2008). The results are based on the chemical analyses of the r63-mm fraction of nearly 450 bed-sediment samples obtained from undeveloped or agricultural sites within the conterminous US (Table 4). The baseline values associated with these samples do not appear markedly different from those generated from other studies performed in the US and globally (Bowen, 1979; Shacklette and Boerngen, 1984; Horowitz, et al., 1991; Gustavsson, et al., 2001; Manheim and Hayes, 2002). Hence, they probably represent a useful benchmark for identifying anthropogenically enhanced (contaminated) sediment-associated constituents and levels. The same study also indicates that with the exception of mining, urbanization, and population density (Horowitz and Stephens, 2008), land use does not exercise substantive controls on sediment-associated inorganic chemical concentrations (Table 4).

325

2.12.5.3.2 The collection of representative sediment samples and the issues of spatial and temporal variability It should be accepted as given in any sediment-associated chemical study that no amount of high-quality analytical work can overcome poor and/or nonrepresentative sampling (e.g., Horowitz, 1997). Consequently, the number, location, and density of sampling locations need to be carefully evaluated within the context of study objectives, acceptable levels of associated error, and some knowledge of the expected levels of chemical variation. In other words, what is the minimum level of acceptable error that still permits sound management decisions and or data interpretations (e.g., Keith, 1988; Mudroch and MacKnight, 1991; Horowitz, 2008). Both bed and suspended sediments have unique characteristics and distribution patterns that must be understood prior to designing an adequate sampling and analysis program (e.g., Keith, 1988; Horowitz, 1991, 2008; Mudroch and MacKnight, 1991). Studies have provided a clear picture regarding the distribution of suspended sediment in fluvial cross sections, as well as the impacts of those distributions on sediment-associated chemical concentrations (Horowitz et al., 1989, 1992; Horowitz, 1991, 1995, 2008). Suspended sediment and suspended sediment-associated chemical concentrations can exhibit marked short-term spatial and temporal variability (e.g., Horowitz, 1991, 1995, 2008). These distributions tend to support the use of equal-width increment (EWI) or equal-discharge increment (EDI), depth- and width-integrated isokinetic sampling. In other words, collecting a dip sample at a single location in a normal fluvial cross section is unlikely to generate a representative sample of suspended sediment or sediment-associated chemical constituents (Horowitz et al., 1989, 1990, 1992; Horowitz, 1991, 1995, 2008). There are exceptions, such as during low flow (baseflow) conditions when suspended sediment concentrations are very low (r10 mg l1), or at elevated energy levels in relatively narrow channels when suspended sediment concentrations can be very high, but such cases require careful sampling and subsequent analysis to justify the use of localized (e.g., a dip sample, an autosampler) rather than depth- and widthintegrated sampling (e.g., Horowitz, 2008). When both sand- (463 mm) and silt/clay-sized (o63 mm) particles are present in a stream, the concentration of suspended sediment tends to increase with increasing distance from the riverbanks (Figure 9). This results from increasing stream velocity (discharge) due to decreasing frictional resistance away from the riverbanks and the riverbed (in shallow water) (e.g., Vanoni, 1977). Note that the typical cause of an increase in suspended sediment concentration is an increase in the amount of sand-sized (463 mm) material (Figure 9). These concentration changes can occur over relatively short distances (e.g., o3 m). There is a concomitant decrease in the concentrations of most sediment-associated chemical constituents with increasing distance from the riverbanks (Figure 9). This decrease occurs as a direct result of the increase in sand-sized (463 mm) particles because the coarser material typically contains markedly lower chemical concentrations (e.g., trace elements, nutrients) than the finer silt/clay-sized material (Fo¨rstner and Wittmann, 1981; Salomons and Fo¨rstner, 1984; Horowitz, 1991). Note that Al does not follow

Table 4 A continental scale assessment of baseline sediment chemistry for the conterminous US. The table presents information on minimum, maximum, mean, median, and median absolute deviations (MAD) chemical concentrations for background, as well as various land-use categories Al (%)

Sb (mg kg1)

Be (mg kg1)

Cd (mg kg1)

Ca (%)

Ce (mg kg1)

Cr (mg kg1)

Co (mg kg1)

Cu (mg kg1)

Fe (%)

La (mg kg1)

Pb (mg kg1)

Li (mg kg1)

Mg (%)

Background chemical concentrations – all data Count 448 446 447 447 448 Min 0.2 0.1 0.1 7.0 0.1 Max 13.0 3.7 60 1300 7.0 Mean 6.0 0.8 8.1 470 1.6 Median 5.9 0.7 6.6 490 1.8 MAD 1.0 0.2 2.2 110 0.8

445 0.1 2.8 0.5 0.4 0.2

447 0.1 28 3.0 1.8 1.3

447 12.0 360 79 69 15

447 6.3 270 66 58 13

448 0.5 78 14 12 4.0

448 1.0 150 24 20 6.0

448 0.2 10 3.3 2.9 0.7

447 6.3 190 42 39 8.0

448 2.0 200 24 20 6.0

448 3.0 97 33 30 10

447 0.04 4.3 1.0 0.9 0.4

Agricultural sites (Z50%) Count 237 237 237 Min 3.2 0.1 2.4 Max 10.5 3.0 60 Mean 6.0 0.8 8.8 Median 5.8 0.7 7.2 MAD 0.9 0.2 1.9 Forest sites (Z50%) Count 286 284 Min 1.4 0.1 Max 13.0 3.7 Mean 6.4 0.8 Median 6.5 0.7 MAD 0.9 0.2 Rangeland Count Min Max Mean Median MAD

sites (Z50%) 59 59 2.1 0.1 11.0 24 5.9 1.1 5.7 0.6 0.6 0.1

As (mg kg1)

285 0.1 41 7.9 6.8 2.0

Ba (mg kg1)

Mn (mg kg1)

Hg (mg kg1)

Mo (mg kg1)

Ni (mg kg1)

P (%)

K (%)

Se (mg kg1)

Ag (mg kg1)

Na (%)

448 15 9000 1100 840 360

448 0.01 3.1 0.08 0.04 0.02

448 0.3 13 1.1 1.0 0.0

447 1.0 160 28 23 7.0

447 0.02 0.47 0.11 0.10 0.02

447 0.03 3.1 1.4 1.5 0.3

447 0.1 5.6 0.8 0.7 0.2

445 0.1 4.3 0.3 0.2 0.1

447 0.02 2.2 0.7 0.6 0.3

S (%)

Sn (mg kg1)

V (mg kg1)

448 17 970 160 150 60

439 0.03 1.5 0.12 0.08 0.04

433 1.2 54 2.8 2.5 o0.1

448 5.1 380 92 83 21

Zn (mg kg1)

Ti (%)

OC (%)

TC (%)

448 5.2 430 100 91 20

444 0.04 1.9 0.38 0.33 0.08

425 0.01 25 3.7 2.4 1.1

426 0.7 25 4.5 3.3 1.6

237 9.0 860 490 500 70

237 0.5 6.0 1.4 1.0 0.3

237 0.1 2.8 0.4 0.4 0.1

237 0.1 12.0 3.3 2.6 1.8

237 35 360 74 66 12

237 34 200 65 58 11

237 5.0 78 14 12 3.0

237 6.0 86 24 22 6.0

237 1.4 10.0 3.3 2.9 0.6

237 19 150 40 36 6.0

237 6.0 310 24 20 5.0

237 20 110 33 30 8.0

237 0.1 4.4 1.2 1.0 0.4

237 190 8400 1200 870 260

237 0.01 1.0 0.06 0.04 0.02

237 0.3 6.5 1.2 1.0 0.1

237 12 160 30 25 5.0

237 0.04 0.31 0.11 0.10 0.02

237 0.1 3.1 1.4 1.5 0.3

237 0.1 4.8 0.8 0.7 0.2

237 0.1 1.0 0.2 0.2 0.1

237 0.03 1.8 0.6 0.6 0.2

237 24 470 160 140 40

236 0.03 0.75 0.10 0.07 0.04

222 1.4 5.0 2.7 2.5 0.0

237 47 260 96 88 21

237 30 190 99 93 19

234 0.14 1.1 0.37 0.31 0.06

220 0.02 18 2.7 2.2 0.8

220 0.8 18 3.6 3.2 1.4

285 7.0 1300 420 420 80

286 0.5 7.0 1.9 2.0 0.2

283 0.1 4.2 0.6 0.4 0.1

285 0.1 26 1.9 1.0 1.0

285 13 350 89 81 14

285 13.0 270 71 65 12

286 2.0 64 18 17 4.0

286 1.0 250 29 24 8.0

286 0.7 8.5 3.7 3.6 0.8

285 9.0 190 47 43 7.0

286 2.0 200 34 28 7.0

286 6.0 97 38 39 14

285 0.1 4.2 0.8 0.6 0.2

286 20 20 000 1400 1000 400

286 0.01 3.1 0.12 0.07 0.02

286 0.3 13 1.2 1.0 0.0

285 6.0 140 30 29 6.5

285 0.02 0.39 0.13 0.11 0.04

285 0.1 2.7 1.3 1.4 0.3

285 0.1 8.6 0.9 0.7 0.2

283 0.1 4.3 0.3 0.2 0.1

285 0.0 2.2 0.6 0.4 0.2

286 17 660 130 100 38

278 0.03 1.5 0.13 0.09 0.03

275 1.2 54 3.1 2.5 0.0

286 14 380 91 86 21

285 21 440 130 110 31

282 0.05 1.9 0.43 0.41 0.06

269 0.01 25 4.8 3.3 0.9

270 0.7 25 5.2 3.6 1.0

59 10 80 33 30 8.0

59 0.4 3.8 1.2 1.1 0.3

59 170 3200 660 490 120

59 0.01 4.7 0.12 0.03 0.02

59 0.3 1.4 0.9 1.0 0.0

58 7.0 160 24 19 5.0

59 0.04 0.19 0.11 0.10 0.02

59 0.5 2.1 1.7 1.7 0.2

59 0.2 2.6 0.8 0.6 0.2

59 0.1 1.0 0.3 0.2 0.1

59 0.1 1.7 0.8 0.8 0.3

59 110 970 290 250 50

59 0.03 0.48 0.12 0.08 0.04

59 1.4 5.0 2.5 2.5 0.0

59 28 200 82 72 15

59 38 150 85 84 17

59 0.14 0.72 0.32 0.30 0.04

59 0.5 3.7 1.6 1.4 0.6

59 0.7 9.9 2.9 2.6 1.2

94 0.04 2.0 0.7 0.5 0.3

94 39 1000 180 120 40

94 0.03 1.0 0.20 0.15 0.08

88 2.0 69 7.4 2.5 0.1

94 52 180 99 94 18

94 45 1700 330 270 120

93 0.18 0.85 0.42 0.40 0.12

88 0.01 16 4.4 3.3 1.4

88 0.9 19 5.5 4.9 1.7

505 31 1600 160 120 40

504 0.03 1.7 0.16 0.11 0.06

471 1.1 92 5 2.5 0

505 24 240 91 86 16

504 2 1700 200 150 51

492 0.1 1.1 0.4 0.36 0.1

468 0.01 29 3.9 2.9 1.1

467 0.7 29 4.8 3.9 1.5

59 1.9 36 6.9 5.2 1.4

58 10 1100 590 590 85

59 0.5 3.0 1.5 1.6 0.6

59 0.1 1.6 0.4 0.3 0.1

59 1.0 25 5.0 3.7 2.0

59 28 130 69 69 10

58 18 220 59 48 11

59 4.0 32 11 9.0 2.0

59 4.0 79 24 20 5.0

59 1.0 6.2 2.7 2.4 0.4

59 17 100 40 40 5.0

59 6.0 330 24 18 4.0

94 1.3 140 13 9.1 3.0

93 33 920 430 450 70

94 0.5 4.3 1.8 1.7 0.4

94 0.2 7.3 1.4 1.0 0.5

94 0.2 19.0 2.9 1.7 1.0

94 27 270 85 78 22

94 45 700 97 81 18

94 5.0 64 18 16 4.0

94 9.0 420 76 53 24

94 1.6 11.0 4.2 3.9 0.8

94 12 120 45 41 12

94 8.0 590 110 76 35

94 10 100 33 30 10

94 0.2 4.7 1.2 0.9 0.4

94 130 12 000 1600 1100 570

94 0.03 2.2 0.25 0.13 0.07

94 0.9 11 2.2 1.0 0.0

94 11 130 39 36 7.0

94 0.04 1.3 0.20 0.14 0.04

94 0.2 2.5 1.4 1.4 0.3

94 0.2 4.1 1.0 0.7 0.3

94 0.1 17 1.1 0.5 0.3

Population density sites (Z50 Percentile; Z27 p km2) Count 505 504 504 501 505 503 Min 1.7 0.1 1.2 6 0.5 0.1 Max 14 24 160 920 12 18 Mean 6.4 1.2 10 460 1.8 0.9 Median 6 0.9 7.9 460 2 0.5 MAD 1 0.2 2.3 80 0.5 0.2

504 0.1 20 3 1.7 1.2

505 15 270 83 74 20

504 11 700 79 69 15

505 0.9 21 3.8 3.6 0.8

505 10 130 45 40 11

505 8 590 64 39 16

505 6 110 37 34 6

505 0.1 4.7 1.1 0.9 0.4

505 74 12 000 1400 1000 400

504 0.01 14.5 0.22 0.09 0.04

505 0.3 34 1.6 1 0

504 4.2 170 34 30 7

505 0.03 1.8 0.15 0.13 0.04

505 0.1 2.8 1.5 1.5 0.3

504 0.1 13 0.9 0.7 0.2

503 0.1 17 0.7 0.4 0.2

Urban sites (Z50%) Count 94 94 Min 3.9 0.2 Max 13.0 10 Mean 6.6 1.6 Median 5.8 1.1 MAD 0.8 0.4

Sr (mg kg1)

505 3 170 17 15 5

505 6 620 51 36 13

505 0 2.6 0.6 0.5 0.3

Catchment Erosion, Sediment Delivery, and Sediment Quality Arkansas River, CO: 11 May 1987

327

Cowlitz River, WA: 20 April 1987 600

Concentration (mg l−1)

800 600

400

400 200 200

0

0 D-6.1

D-12.2

D-15.2

D-22.9

D-27.4

< 63 µm Fraction

D-22.9

D-39.6

% Concentration (mg kg−1)

4

0

0

D-99.1

Al

Zn

Cu

200

20

0 40

mg kg−1

Fe

2

40

D-70.1

>63 µm Fraction

Arkansas River, CO: 11 May 1987 8

4

D-57.9

100

0 Pb

12

Co

8

20

4

0

0 D-6.1 D-12.6 D-15.2 D-22.9 D-27.4

D-6.1 D-12.6 D-15.2 D-22.9 D-27.4

Figure 9 Horizontal cross sectional changes in suspended sediment concentration for the Arkansas and Cowlitz rivers based on isokinetic depthintegrated vertical samples. The numbers following the D are distances, in meters, from the left bank (upper). Horizontal cross sectional variations in selected suspended sediment-associated trace element in depth-integrated isokinetic vertical samples from the Arkansas River on May 11, 1987. The numbers following the D are distances, in meters, from the left bank of the river (lower).

this pattern, because the majority of this element is lattice-held rather than sorbed to mineral surfaces, and both fractions contain substantial quantities of aluminosilicates (Horowitz, 2008). Vertical concentrations of suspended sediment in fluvial systems tend to increase with increasing depth; this also is due to an increase in sand-sized material (Figure 10). This occurs because the velocity (discharge) in most rivers, under normal

flow conditions, is insufficient to homogeneously distribute the coarser material. The majority of sand-sized particles tend to be transported on or near the riverbed. The increase in sand-sized particle concentration, from top to bottom, also leads to a concomitant decrease in sediment-associated chemical concentrations (Figure 10). As with the horizontal variations noted previously, these changes can occur over relatively short distances (o1.5 m).

328

Catchment Erosion, Sediment Delivery, and Sediment Quality Arkansas River, CO: 11 May 1987

Depth (%)

6.45 m from left bank

17.4 m from left bank

20

20

40

40

60

60

80

80 0

200 400 600 Concentration (mg l−1)

0

63 µm fraction

Arkansas River, CO: 11 May 1987 6.45 m from left bank

17.4 m from left bank

20 40 Fe (%) 60 80 0

2

4

0

2

4

20 40

Cu (mg kg−1)

Depth (%)

60 80 0

10

20

30

0

20

40

20 40

Zn (mg kg−1)

60 80 0

100

200

0

100

200

300

20 40

Pb (mg kg−1)

60 80 0

10

20 30 Concentration

40

0

20 40 Concentration

60

Figure 10 Vertical cross sectional changes in suspended sediment and selected sediment-associated trace element concentrations for the Arkansas River on May 11, 1987, based on isokinetic point samples collected at 20%, 40%, 60%, and 80% of depth. One vertical was 6.45 m and the other was 17.4 m from the left bank.

Sediment chemistry, especially suspended sediment chemistry is markedly affected by hydrology and can display substantial changes in concentration over relatively short as well as relatively longer timescales (e.g., Horowitz, 1995, 2008; Horowitz et al., 2008). This accrues for two reasons, and

means that in order to delimit the range of sediment-associated constituent concentrations at any particular location, samples must be collected over a range of flow conditions and temporal scales. The hydrologic linkage can be both direct and indirect. The direct linkage results from the changes in

Catchment Erosion, Sediment Delivery, and Sediment Quality

suspended sediment grain-size distribution mentioned earlier. As velocity (discharge) increases, the median grain size of suspended sediment tends to increase that normally produces a concomitant decrease in sediment-associated constituent concentrations (e.g., Horowitz, 1991, 1995, 2008; Horowitz, et al., 2008). Although sediment-associated chemical concentrations decline as discharge increases, the fluxes (loads) of these same constituents usually increase. This occurs because the increase in discharge, in conjunction with increasing amounts of suspended sediment, although coarser, typically more than compensates for the decline in actual chemical concentrations (e.g., Horowitz, 1995, 2008; Old et al., 2003, 2006; Lawler et al., 2006; Horowitz et al., 2008). The indirect linkage between hydrology and sedimentassociated chemical concentrations occurs as a result of changing sediment sources. Although it is something of an over-simplification, there is a generally accepted perception that under baseflow, sediment chemistry tends to be dominated by point sources, whereas during high flow (stormflow), sediment chemistry tends to be dominated by nonpoint (diffuse) sources (e.g., Horowitz, 1995, 2008; Old, et al., 2003, 2006; Horowitz, et al., 2008). Sediment from nonpoint sources in urban (e.g., trace elements, nutrients, PAHs, and pesticides), mining (e.g., trace/major elements), and agricultural areas (e.g., nutrients, agricultural chemicals) is particularly enriched in a wide variety of both organic and inorganic constituents (e.g., Horowitz, 1995; Horowitz, et al., 2008). At least in the US, baseflow point-source sedimentassociated chemical effects tend to be limited in terms of both amount and chemical concentration by controlling end-of-pipe discharges through permitting processes such as National Pollutant Discharge Elimination System (NPDES) under the Clean Water Act of 1972. On the other hand, nonpoint-source discharges are much harder to limit/control, and entail the application of a variety of best-management practices (BMPs) such as increasing the width of riparian zones, procurement of additional green space, reforestation, low or no tillage in agricultural areas, installation of highway runoff settling ponds, etc. As a result, of all the foregoing, accurate assessments of suspended sediment-associated chemical concentrations may require sampling over several different temporal scales (e.g., sampling over the course of a storm to determine concentration changes associated with the rising limb, the peak, and the falling limb of the hydrograph, sampling between storms to determine the impact of different lengths of antecedent dry conditions, and seasonal sampling to deal with the application of various agricultural chemicals (e.g., fertilizers, pesticides), or to evaluate the impact of, for example, deicing salts). On the other hand, such factors as changing demographics (e.g., population density) and land use factors (e.g., urbanization) can affect sediment-associated chemical concentrations over longer temporal scales, of the order of years or decades, depending on the size of the hydrologic system. Bed sediments can exhibit marked spatial variability, but rarely short-term temporal variability (e.g., Horowitz, 1991, 1995). In addition, unlike suspended sediment samples, bed sediments rarely pose a problem relative to collecting sufficient masses to meet any requisite analytical needs. As such, it usually is far easier to collect representative bed sediment

329

rather than suspended sediment samples. As long as the goal of a study is not intended to determine relative levels of localized spatial variability, the sampling issues associated with bed sediments usually can be addressed through the production of composite samples generated by combining a sufficient number of spatially separated equal-volume aliquots. The number of requisite aliquots typically is predicated on the level of local spatial variability in conjunction with acceptable levels of chemical variance. Normally, the more subsamples that are combined, the more representative the composite, and the smaller the level of associated chemical variance (e.g., Hakanson, 1984; Garner et al., 1988; Horowitz, 1991; Mudroch and MacKnight, 1991).

2.12.5.3.3 The chemical analysis of suspended and bed sediments As noted previously, sediment-associated constituent concentrations typically occur at levels three to five orders of magnitude higher than found in solution (e.g., Horowitz, 1991, 1995, 2008). As such, analytical sensitivity normally is not an issue except when samples masses are very small; this rarely is the case with bed sediment, but can be an issue with suspended sediment. However, even in the case of suspended sediment, there are a number of techniques available, such as flow-through centrifugation or filtration that permit the collection/concentration of sufficient amounts of suspended matter for subsequent chemical analyses (e.g., Horowitz, 1995, 2008). Although there are a number of direct techniques for the chemical analysis of solid-phase materials, the majority, with limited exceptions, are performed on either liquid extracts (organic constituents) or liquid digests (inorganic constituents). In the past, the method of choice for the solubilization of sediment-associated organic constituents was the soxhlet extraction (e.g., Amalric and Mouvet, 1997). However, modern production laboratories have begun to switch to either microwave extraction (e.g., Morozova et al., 2008; Forster et al., 2009) or accelerated solvent extraction (especially useful for the determination of hydrophobic persistent organic pollutants; e.g., Jacobsen et al., 2004; Silvia Diaz-Cruz et al., 2006; Berrada et al., 2008). The latter two procedures are markedly faster than the soxhlet extraction, and can be modified to be compound and/or compound-class specific through the selection of an appropriate solvent, and by programming the temperature and/or pressure, and the length of time for the extraction (e.g., Raynie, 2004, 2006). The analytical instruments of choice currently in use are: (1) gas chromatography using various detectors; (2) gas chromatography-mass spectroscopy; and (3) liquid chromatography-mass spectroscopy. Extraction efficiencies are normally determined by the concomitant analysis of reference materials, and analytical recoveries are typically determined by spiking extracts with known amounts of the analyte(s) of interest. The analytical instrumentation of choice for a wide variety of inorganic constituents including trace elements (e.g., Cu, Zn, Cd, and Pb), major elements (e.g., Fe, Al, Na, and K), and some nutrients (e.g., P, S) is some type of inductively coupled plasma (ICP)-based system. The two most common devices are ICP-atomic emission spectroscopy (ICP-AES), sometimes

330

Catchment Erosion, Sediment Delivery, and Sediment Quality

called ICP-optical emission spectroscopy (ICP-OES), and ICPmass spectroscopy (ICP-MS). The former is a purely optical system that measures emitted light at a specific (constituentspecific) wavelength whereas the latter is based on determining the concentrations of specific individual stable isotopes, and converting that value to a total concentration in the digestate based on fixed isotopic percentages. Both systems require sample solubilization before quantitation. Several common nutrient determinations (e.g., total carbon, total organic carbon, total nitrogen, and total sulfur) are determined directly on dried sediment by combusting the sample at high temperatures, in the presence of oxygen, and quantitating the evolved gases using a variety of different detectors. There are numerous choices for the solubilization of various sediment-associated inorganic constituents for subsequent chemical analysis (e.g., Johnson and Maxwell, 1981; van Loon, 1985; Batley, 1989; ASTM, 2008). Essentially, these methods fall into one of the three categories: total analyses, total recoverable analyses, or selective extractions (e.g., Horowitz, 1995, 2008). Geochemists normally define a total analysis as one in which Z95% of the analyte of concern is quantified. This approach usually entails the complete breakdown of mineral lattices, and requires a mixture of concentrated mineral acids and relatively high temperatures, and/or some type of fusion with various fluxes (e.g., sodium carbonate, lithium tetraborate/metaborate) at elevated temperatures, with the solubilization of the resulting bead (e.g., Johnson and Maxwell, 1981). Analytical precision and bias are normally determined by the concomitant analysis of appropriate reference materials. Analyses of this type are unambiguous because they are independent of sediment-associated mineralogical/ petrological variation; hence, they are comparable across spatial and/or temporal scales, and represent a known endmember because of the level of recovery (Z95%). Total recoverable digestions, typically favored by regulatory agencies at least in the US, are nonspecific partial extractions usually employing mineral acids and some level of heating. The levels of constituent solubilization and subsequent quantitation are highly dependent on variations in mineralogy/ petrology, as such, concentrations determined using this procedure may not be comparable across spatial and/or temporal scales, and can be very difficult to interpret. It sometimes has been claimed that this type of digestion produces a measure of bio- and/or environmental-availability but that interrelation has never been demonstrated (e.g., Horowitz, 1991, 2008). Another difficulty associated with this approach is a lack of certified reference materials, so while it is possible to determine analytical precision (reproducibility) by replicate analyses of the same material, it is difficult to evaluate analytical bias as well as percent recoveries. Spiking digestates is not a typical sediment-associated inorganic analytical procedure because the major source of analytical variance does not result from the quantitation step, but from the solubilization step. Selective extractions represent a special category of partial and/or recoverable digestions that are intended to provide specific information about sediment-chemical partitioning (e.g., Kersten and Fo¨rstner, 1987; Horowitz, 1991). Sedimentchemical partitioning entails all the various procedures that can be used to determine how (mechanistic approach; e.g.,

adsorption, complexation; within mineral lattices) and/or where (phase approach; e.g., iron oxides, manganese oxides, organic matter; carbonates) various chemical constituents are associated with sediment particles (e.g., Chao, 1984; Bately, 1989; Horowitz, 1991; Hall and Pelchat, 1999). Unfortunately, all these procedures can be categorized as operational definitions and there are numerous procedures that purport to provide similar information (e.g., bound to iron oxides), but which do not provide equivalent analytical results (e.g., Horowitz, 1991). These procedures usually also have additional limitations such as they can only be used on sediment collected in oxidized environments (e.g., Tessier et al., 1979).

2.12.5.3.4 Bioavailability and toxicity By far, the single biggest barrier to the widespread development and acceptance of sediment quality guidelines/ regulatory limits is the continuing controversy regarding the bioavailability, and potential toxicity of sediment-associated chemical constituents. Unlike dissolved constituents, where the total concentration is presumed to be bioavailable, it has always been assumed that only limited portions of sedimentassociated constituents (e.g., non-lattice held) fall into the same category (e.g., Allen et al., 1993; Hansen et al., 2005; Simpson et al., 2005). Additional concerns stem from disagreements over the interpretation of toxicity tests in terms of exposure rates, selection of appropriate test organisms, and relative measures of lethality, as well as an understanding of sediment-chemical partitioning and potential bioaccumulation (e.g., Lahr et al., 2003; Hansen et al., 2005). Current sediment quality guidelines (SQGs) are typically provided at two concentrations. The first level is the lower of the two, and normally reflects concentrations at which little or no biological/ecological effects are expected. This level has been given a variety of names such as lowest effect level (Persaud et al., 1993), threshold-effects concentration (TEC; MacDonald et al., 2000), and threshold-effects level (TEL; US Environmental Protection Agency, 1997). The second level is the higher of the two, and normally reflects concentrations at which biological/ecological effects are expected. This level has also been given a variety of names such as severe effect level (Persaud et al., 1993), probable-effects concentration (PEC; MacDonald, et al., 2000), and probable-effects level (PEL; US Environmental Protection Agency, 1997). Although these various concentrations tend to overlap, there currently appears to be no consensus on a single set of values for either category or for various sediment-associated constituents (e.g., Persaud et al., 1993; US Environmental Protection Agency, 1997; MacDonald et al., 2000; Hansen et al., 2005; Simpson et al., 2005). To further complicate this issue, other terms have been used to indicate the potential effects of sediment-associated chemical constituents. For example, in 1997, the US EPA evaluated sediment chemical data from over 21000 locations in the US and found that 26% had a higher probability and 49% had an intermediate probability of adverse effects on aquatic life and human health. The chemical constituents most often associated with these increased probabilities were PCBs, Hg, DDT, Cu, Ni, and Pb (US Environmental Protection Agency, 1997).

Catchment Erosion, Sediment Delivery, and Sediment Quality

Other than the results from various toxicity tests, there appear to be two basic approaches associated with establishing SQGs; one is based on equilibrium partitioning models whereas the other is based on some type of sediment-chemical partitioning (e.g., Hansen et al., 2005; Simpson et al., 2005). In either case, evaluations have to be made on a site-specific basis (e.g., Simpson et al., 2005). The equilibrium partitioning approach requires measuring chemical concentrations in interstitial water; the underlying assumption being that whatever concentrations in the porewater represent the bioavailable component of the equivalent sediment-associated constituent (e.g., Simpson et al., 2005). The actual guideline concentrations are then based on those established for dissolved constituents. There are three major issues with this approach. The first is that it obviously does not apply to suspended sediment. The second issue is associated with those organisms that actually ingest sediment. The physicochemical conditions in the gut of an organism are obviously not the same as those that exist in the interstitial water (e.g., they are likely to be more acidic and less oxygen will be present); hence, bioavailable concentrations derived from interstitial water may be inappropriate. The third issue is associated with specific size fractions from a bulk sediment sample. Examination of the gut contents of various aquatic organisms indicates that they tend to limit their intake to specific grain-size ranges. As such, determining equilibrium partitioning concentrations based on interstitial water derived from a bulk sediment sample may be inappropriate. The sediment partitioning approach is predicated on the view that some form of chemical extraction/digestion can be found that functions as a measure of, or surrogate for bioavailability (e.g., Di Toro et al., 1992; Allen et al., 1993). One such procedure that has received a good deal of attention is acid volatile sulfides-simultaneously extracted metals (AVSSEM). This extraction employs cold dilute HCl; those metals that exceed the concentration of available sulfide, on a molar basis, are considered bioavailable (Di Toro et al., 1992; Allen et al., 1993; Hansen et al., 2005; Simpson et al., 2005). However, a number of studies have indicated that the AVSSEM method can over- or under-predict bioavailability when compared to other approaches, and may not be appropriate for all environments and/or organisms (e.g., Chen and Mayer, 1999; Lahr et al., 2003; Meador et al., 2005; Simpson and Batley, 2007; Prica et al., 2008).

2.12.5.4 Future Directions While the foregoing summary, covering the current status of sediment quality, clearly indicates that a great deal is known about the subject, there still remains much to do with respect to achieving scientific consensus in a variety of areas. While the scientific community has begun to reach some level of consensus relative to background/baseline values for sedimentassociated constituents, more refinement is needed. It would also be appropriate to begin to delineate those constituents that are particularly sensitive to, or indicative of various types of land use or source material. The general lack of broad scale agreement on appropriate sampling and analytical procedures is likely to continue to make transboundary/multinational studies difficult as a result of potential data incompatibilities.

331

Lastly, it appears that a great deal of more work needs to be done in the areas of sediment-associated constituent bioavailability and toxicity before there can even be a modicum of consensus that could lead to fairly universal sediment-quality guidelines.

References Aalto R, Lauer JW, and Dietrich WE (2008) Spatial and temporal dynamics of sediment accumulation and exchange along Strickland River floodplains (Papue New Guinea) over decadal-to-centennial timescales. Journal of Geophysical Research Earth Surface 113: F01F04. Aalto R, Maurice-Bourgoin L, Dunne T, Montgomery DR, Nittrouer CA, and Guyot J-L (2003) Episodic sediment accumulation on Amazonian floodplains influenced by El Nino/Southern Oscillation. Nature 425: 493--497. Abernethy B and Rutherfurd I (1998) Where along a river’s length will vegetation most effectively stabilise stream banks? Geomorphology 23: 55--75. Abrahams AD, Li G, Krishnan C, and Atkinson JF (2001) A sediment transport equation for interrill overland flow. Earth Surface Processes and Landforms 26: 1443--1459. Aksoy H and Kavvas ML (2005) A review of hillslope and watershed scale erosion and sediment transport models. Catena 64: 247--271. Alexandrov Y, Laronne JB, and Reid I (2003) Suspended sediment concentration and its variation with water discharge in a dryland ephemeral channel, Northern Negev, Israel. Journal of Arid Environments 53: 73--84. Ali KF and De Boer DH (2008) Factors controlling sediment yield in the upper Indus River basin, northern Pakistan. Hydrological Processes 22: 3102--3114. Allan RJ (1979) Sediment-Related Fluvial Transmission of Contaminants: Some Advances by 1979, Inland Waters Directorate, Environment Canada, Scientific Series No. 107. Ottawa: Environment Canada. Allen HE, Fu G, and Deng B (1993) Analysis of acid-volatile sulfide (AVS) and simultaneously extracted metals (SEM) for the estimation of potential toxicity in aquatic sediments. Environmental Toxicology and Chemistry 12: 1441--1453. Amalric L and Mouvet C (1997) Comparison of SFE, soxhlet and sonication for the extraction of organic contaminants from sandstone. International Journal of Environmental Analytical Chemistry 68: 171--186. ASCE (1975) Sedimentation Engineering. American Society of Civil Engineers Manuals and Reports on Engineering Practice No. 54. New York: ASCE. Ascough JC, Baffaut C, Nearing MA, and Liu BY (1997) The WEPP watershed model. 1. Hydrology and erosion. Transactions of the ASAE 40: 921--933. Asselman NEM (2000) Fitting and interpretation of sediment rating curves. Journal of Hydrology 234: 228--248. ASTM (2008) Annual Book of ASTM Standards, vol. 11.02, pp. 384–387. West Conshohocken, PA: ASTM International. Atkinson PM, German SE, Sear DA, and Clark MJ (2003) Exploring the relations between riverbank erosion and geomorphological controls using geographically weighted logistic regression. Geographical Analysis 35: 58--82. Baby P, Guyot J-L, and He´rail G (2009) Tectonic control of erosion and sedimentation in the Amazon Basin of Bolivia. Hydrological Processes 23: 3225--3229 (doi:10.1002/hyp.7391). Baker RA (ed.) (1980) Contaminants and Sediments, Vol. 1: Fate and Transport, Case Studies, Modeling, Toxicity Ann Arbor, MI: Ann Arbor Science Publishers. Bartley R, Hawdon A, Post DA, and Roth CH (2007) A sediment budget for a grazed semi-arid catchment in the Burdekin Basin, Australia. Geomorphology 87: 302--321. Basson G (2008) Reservoir sedimentation – an overview of global sedimentation rates and predicted sediment deposition. In: Erosion, Transport and Deposition. Workshop Berne, Switzerland, April 2008, Abstracts, pp. 74–79. Batley GE (1989) Trace Element Speciation: Analytical Methods and Problems. Boca Raton, FL: CRC Press. Beck MB (1987) Water-quality modeling – a review of the analysis of uncertainty. Water Resources Research 23: 1393--1442. Benda L and Dunne T (1997) Stochastic forcing of sediment routing and storage in channel networks. Water Resources Research 33: 2865--2880. Berrada H, Borrull F, Guillermina F, and Marce RM (2008) Determination of macrolide antibiotics in meat and fish using pressurized liquid extraction and liquid chromatography-mass spectrometry. Journal of Chromatography, A 1208: 83--89. Beusen AHW, Dekkers ALM, Bouwman AF, Ludwig W, and Harrison J (2005) Global river export of suspended solids and particulate carbon and nutrients. Global Biochemical Cycles 19: GB4S05.

332

Catchment Erosion, Sediment Delivery, and Sediment Quality

Beven K, Heathwaite L, Haygarth P, Walling D, Brazier R, and Withers P (2005) On the concept of delivery of sediment and nutrients to stream channels. Hydrological Processes 19: 551--556. Black KS, Athey S, Wilson P, and Evans D (2007) The use of particle tracking in sediment transport studies: A review. Geological Society, Special Publication 274: 73--91. Blake WH, Droppo IG, Wallbrink PJ, Doerr SH, Shakesby RA, and Humphreys GS (2005) Impacts of wildfire on effective sediment particle size: Implications for postfire sediment budgets. In: Walling DE and Horowitz AJ (eds.) Sediment Budgets 1, IAHS Publication No. 236, pp. 115–123. Wallingford: IAHS Press. Blake WH, Walling DE, and He Q (2002) Using cosmogenic Beryllium-7 as a tracer is sediment budget investigations. Geografiska Annala 84A: 89--102. Blum WEH, Bu¨sing J, and Montanarella L (2004) Research needs in support of the European thematic strategy for soil protection. TrAC – Trends in Analytical Chemistry 23: 680--685. Bobrovitskaya NN, Zubkova C, and Meade RH (1996) Discharges and yields of suspended sediment in the Ob and Yenesey Rivers of Siberia. In: Walling DE and Webb BW (eds.) Erosion and Sediment Yield: Global and Regional Perspectives, IAHS Publication No. 325, pp. 323–338. Wallingford: IAHS Press. Bohn BA and Kershner JL (2002) Establishing aquatic restoration priorities using a watershed approach. Journal of Environmental Management 64: 355--363. Bottrill LJ, Walling DE, and Leeks GJL (2000) Using recent overbank deposits to investigate contemporary sediment sources in large river systems. In: Foster IDL (ed.) Tracers in Geomorphology, pp. 369--387. Chichester: Wiley. Bowen HJM (1979) Environmental Chemistry of the Elements. London: Academic Press. Brardinoni F, Hassan MA, Rollerson T, and Maynard D (2009) Colluvial sediment dynamics in mountain drainage basins. Earth and Planetary Science Letters 284: 310--319. Boomer KB, Weller DE, and Jordan TE (2008) Empirical models based on the universal soil loss equation fail to predict sediment discharges from Chesapeake Bay catchments. Journal of Environmental Quality 37: 79--89. Borah DK and Bera M (2002) Watershed-scale hydrologic and nonpoint-source pollution models: Review of mathematical bases. In: ASAE International Meeting, pp. 1553–1566. Chicago, IL, USA. Brasington J and Richards K (2007) Reduced-complexity, physically-based geomorphological modelling for catchment and river management. Geomorphology 90: 171--177. Brayshaw D and Hassan MA (2009) Debris flow initiation and sediment recharge in gullies. Geomorphology 109: 122--131. Caitcheon GG (1993) Sediment source tracing using environmental magnetism – a new approach with examples from Australia. Hydrological Processes 7(4): 349--358. Cavalcanti GG and Lockaby BG (2005) Effects of sediment deposition on fine root dynamics in riparian forests. Soil Science Society of America Journal 69: 729--737. Cerdan O, Poesen J, Govers G, et al. (2006) Sheet and rill erosion rates in Europe. In: Boardman J and Poesen J (eds.) Soil Erosion in Europe, pp. 501--514. Chichester: Wiley. Chang KT and Chiang SH SH (2009) An integrated model for predicting rainfallinduced landslides. Geomorphology 105: 366--373. Chao TT (1984) Use of partial dissolution techniques in geochemical exploration. Journal of Geochemical Exploration 20: 101--135. Chapman D (ed.) (1992) Water Quality Assessments, a Guide to the Use of Biota, Sediments, and Water in Environmental Monitoring. London: Chapman and Hall. Chappell NA, Douglas I, Hanapi JM, and Tych W (2004) Sources of suspended sediment within a tropical catchment recovering from selective logging. Hydrological Processes 18: 685--702. Chen Z and Mayer LM (1999) Assessment of sedimentary Cu availability: A comparison of biometric and AVS approaches. Environmental Science and Technology 33: 650--652. Chien N and Wan Z (1999) Mechanics of Sediment Transport. New York: ASCE. Cinque K, Stevens MA, Roser DJ, Ashbolt NJ, and Leeming R (2004) Assessing the health implications of turbidity and suspended particles in protected catchments. Water Science and Technology 50: 205--210. Clark EH (1985) The off-site costs of soil erosion. Journal of Soil and Water Conservation 40: 19--22. Clark EH, Haverkamp JA, and Chapman W (1985) Eroding Soils: the Off-Farm Impacts. Washington, DC: The Conservation Foundation. Clarke FW and Washington HS (1924) The composition of the earth’s crust. U.S. Geological Survey Professional Paper 127. Washington, DC: US Geological Survey.

Cochrane T and Flanagan D (2006) Sediment deposition in a simulated rill under shallow flow conditions. Transactions of the ASAE 49: 893--903. Collins AL and Walling DE (2002) Selecting fingerprint properties for discriminating potential suspended sediment sources in river basins. Journal of Hydrology 261: 218--244. Collins AL and Walling DE (2007) Fine-grained bed sediment storage within the main channel system of the Frome and Piddle catchments, Dorset, UK. Hydrological Processes 21: 1448--1459. Collins AL, Walling DE, and Leeks GJL (1997a) Source type ascription for fluvial suspended sediment based on a quantitative composite fingerprinting technique. Catena 29: 1--27. Collins AL, Walling DE, and Leeks GJL (1997b) Sediment sources in the Upper Severn catchment: A fingerprinting approach. Hydrology and Earth Systems Science 1: 509--521. Coote DR, MacDonald EM, Dickinson WT, Ostry RC, and Frank R (1982) Agriculture and water quality in the Canadian Great Lakes Basin: I. Representative agricultural watersheds. Journal of Environmental Quality 11: 473--481. Croke J, Hairsine P, and Fogarty P (1999) Sediment transport, redistribution and storage on logged forest hillslopes in south-eastern Australia. Hydrological Processes 13: 2705--2720. Croke J, Mockler S, Fogarty P, and Takken I (2005) Sediment concentration changes in runoff pathways from a forest road network and the resultant spatial pattern of catchment connectivity. Geomorphology 68: 257--268. Cui Y, Parker G, Lisle TE, Pizzuto JE, and Dodd AM (2005) More on the evolution of bed material waves in alluvial rivers. Earth Surface Processes and Landforms 30: 107--114. Curtis JA, Flint LE, Alpers CN, and Yarnell SM (2005) Conceptual model of sediment processes in the upper Yuba River watershed, Sierra Nevada, CA. Geomorphology 68: 149--166. Davis CM and Fox JF (2009) Sediment Fingerprinting: Review of the method and future improvements for allocating nonpoint source pollution. Journal of Environmental Engineering 135: 490--504. Deasy C, Brazier RE, Heathwaite AL, and Hodgkinson R (2009) Pathways of runoff and sediment transfer in small agricultural catchments. Hydrological Processes 23: 1349--1358. de Arau´jo JC, Gu¨ntner A, and Bronstert A (2006) Loss of reservoir volume by sediment deposition and its impact on water availability in semiarid Brazil. Hydrological Sciences Journal 51: 157--170. De’ath G and Fabricius K (2010) Water quality as a regional driver of coral biodiversity and macroalgae on the Great Barrier Reef. Ecological Applications 20(3): 840--850. de Moor JJW and Verstraeten G (2008) Alluvial and colluvial sediment storage in the Geul River catchment (The Netherlands) – combining field and modelling data to construct a Late Holocene sediment budget. Geomorphology 95: 487--503. de Vente J and Poesen J (2005) Predicting soil erosion and sediment yield at the basin scale: Scale issues and semi-quantitative models. Earth Science Reviews 71: 95--125. de Vente J, Poesen J, Arabkhedri J, and Verstraeten G (2007) The sediment delivery problem revisited. Progress in Physical Geography 31: 155--178. de Vries A and Klavers HC (1994) Riverine fluxes of pollutants: Monitoring strategy first, calculation methods second. European Journal of Water Pollution Control 4: 12--17. DeRoo APJ (1996) Soil erosion assessment using G.I.S. In: Singh VP and Fiorentino M (eds.) Geographical Information Systems in Hydrology, pp. 339--356. Dordrecht: Kluwer. DeRoo APJ, Wesseling CG, and Ritsema CJ (1996) LISEM: A single-event physically based hydrological and soil erosion model for drainage basins. 1: Theory, input and output. Hydrological Processes 10: 1107--1117. Di Toro DM, Mahony JD, Hansen DJ, Scott KJ, Carlson AR, and Ankley GT (1992) Acid volatile sulfide predicts the acute toxicity of cadmium and nickel in sediments. Environmental Science and Technology 26: 96--101. Domenici P, Claireaux G, and McKenzie DJ (2007) Environmental constraints upon locomotion and predator–prey interactions in aquatic organisms: An introduction. Philosophical Transactions of the Royal Society B: Biological Sciences 362: 1929--1936. Droppo IG (2001) Rethinking what constitutes suspended sediment. Hydrological Processes 15: 1551--1564. Droppo IG, Lau YL, and Mitchell C (2001) The effect of depositional history on contaminated bed sediment stability. Science of the Total Environment 266: 7--13. Droppo IG, Nackaerts K, Walling DE, and Williams N (2004) Can flocs and water stable soil aggregates be differentiated within fluvial systems. Catena 60: 1--18. Dunne T, Mertes LAK, Meade RH, Richey JE, and Forsberg BR (1998) Exchanges of sediment between the flood plain and channel of the Amazon River in Brazil. Geological Society of America Bulletin 110: 450--466.

Catchment Erosion, Sediment Delivery, and Sediment Quality

Edwards TK and Glysson GD (1988) Field methods for measurement of fluvial sediment. U.S. Geological Survey Open-file Report 86–531, 118pp. Evans M and Slaymaker O (2004) Spatial and temporal variation of sediment delivery from alpine lake basins: Cathedral Provincial Park, Southern British Columbia. Geomorphology 61: 209--224. Fangmeier DD, Elliot WJ, Workman SR, Huffman RL, and Schwab GO (2006) Soil and Water Conservation Engineering. New York: Delmar Thompson. Fauth H, Hindel R, Siewers U, and Zinner J (1985) Geochemischer Atlas Bundesrepublik Deutschland. Stuttgart: BGR Hannover; Schweizerbart’sche Verlagsbuchhandlung. Ferguson RI (1986) River loads underestimated by rating curves. Water Resources Research 22: 74--76. Ferro V and Minacapilli M (1995) Sediment delivery processes at basin scale. Hydrological Sciences Journal (Des Sciences Hydrologiques) 40: 703--717. Field J (2001) Channel avulsion on alluvial fans in southern Arizona. Geomorphology 37: 93--104. Finlayson DP and Montgomery DR (2003) Modeling large-scale fluvial erosion in geographic information systems. Geomorphology 53: 147--164. Florsheim JL, Mount JF, and Chin A (2008) Bank erosion as a desirable attribute of rivers. BioScience 58: 519--529. Fontaine TA, Moore TD, and Burgoa B (2000) Distributions of contaminant concentrations and particle size in fluvial sediment. Water Research 34: 3473--3477. Forbes AC and Lamoureux SF (2005) Climatic controls on streamflow and suspended sediment transport in three large middle Arctic catchments, Boothia Peninsula, Nunavut, Canada. Arctic, Antarctic and Alpine Research 37: 304--315. Forster M, Laabs V, Lamshoft M, et al. (2009) Sequestration of manure-applied sulfadizine residues in soils. Environmental Science and Technology 43: 1824--1830. Fo¨rstner U (1989) Contaminated Sediments. Lecture on Environmental Aspects of Particle-Associated Chemicals in Aquatic Systems. New York: Springer. Fo¨rstner U (2002) Sediments and the European water framework directive. Journal of Soils and Sediments 2: 54. Fo¨rstner U (2003) Sediments and the European water framework directive. Journal of Soils and Sediments 3: 138. Fo¨rstner U and Heise S (2006) Assessing and managing contaminated sediments: Requirements on data quality – from molecular to river basin scale. Croatica Chemica Acta 79: 5--14. Fo¨rstner U and Wittmann GTW (1981) Metal Pollution in the Aquatic Environment, 2nd edn. New York: Springer. Foster IDL and Charlesworth SM (1996) Heavy metals in the hydrocycle: Trends and explanation. Hydrological Processes 10: 227--261. Fox GA, Wilson GV, Simon A, Langendoen EJ, Akay O, and Fuchs JW (2007) Measuring streambank erosion due to groundwater seepage: Correlation to bank pore water pressure, precipitation and stream stage. Earth Surface Processes and Landforms 32: 1558--1573. Fox JF and Papanicolaou AN (2007) The use of carbon and nitrogen isotopes to study small watershed erosion processes. Journal of the American Water Resources Association 43: 1047--1064. Garcia MH (ed.) (2008) Sedimentation Engineering: Processes, Measurements, Modeling and Practice, American Society of Civil Engineers Manuals and Reports on Engineering Practice No. 110. New York: ASCE. Garner FC, Stapanian MA, and Williams LR (1988) Composite sampling for environmental monitoring. In: Keith LH (ed.) Principles of Environmental Sampling, pp. 363--374. Washington, DC: American Chemical Society. Gibbs MM (2008) Identifying source soils in contemporary estuarine sediments: A new compound-specific isotope method. Estuaries and Coasts 31: 344--359. Gime´nez R and Govers G (2002) Flow detachment by concentrated flow on smooth and irregular beds. Soil Science Society of America Journal 66: 1475--1483. Gippel CJ (1995) Potential of turbidity monitoring for measuring the transport of suspended solids in streams. Hydrological Processes 9: 83--97. Gislason SR, Oelkers EH, and Snorrason A (2006) Role of river-suspended material in the global carbon cycle. Geology 34: 49--52. Glysson GD and Gray JR (eds.) (2002) Proceedings of the Federal Interagency Workshop on Turbidity and Other Sediment Surrogates, U.S. Geological Survey Circular 1250. Washington, DC: US Geological Survey. Goldschmidt VM (1958) Geochemistry. Oxford: Clarendon. Golosov VN, Ivanova NN, Litvin LF, and Sidorchuk AYu (1992) Sediment budgets of river catchments and river channel aggradation on the Russian plain. Geomorphology (Moscow) 4: 62--71 (in Russian). Golterman HL (ed.) (1977) Interactions between Sediment and Fresh Water. The Hague: Dr. W Junk Publishers.

333

Gomez B, Banbury K, Marden M, Trustrum NA, Peacock DH, and Hoskin PJ (2003) Gully erosion and sediment production: Te Weraroa stream, New Zealand. Water Resources Research 39: 1187. Gordon LM, Bennett SJ, Alonso CV, and Bingner RL (2008) Modeling long-term soil losses on agricultural fields due to ephemeral gully erosion. Journal of Soil and Water Conservation 63: 173--189. Gough LP (1993) Understanding our Fragile Environment: Lessons from Geochemical Studies, U.S. Geological Survey Circular 1105. Washington, DC: US Geological Survey. Govers G, Gime´nez R, and van Oost K (2007) Rill erosion: Exploring the relationship between experiments, modelling and field observations. Earth-Science Reviews 84: 87--102. Gray J, Glysson D, and Edwards T (2008) Suspended-sediment samplers and sampling methods. In: Sedimentation Engineering: Processes, Measurements, Modeling and Practice, American Society of Civil Engineers Manuals and Reports on Engineering Practice No. 110. New York: ASCE. Grayson R and Blo¨schl G (2000) Spatial modelling of catchment dynamics. In: Grayson R and Blo¨schl G (eds.) Spatial Patterns in Catchment Hydrology, ch. 3, pp. 51–81. Cambridge: Cambridge University Press. (http://www.hydro.tuwien.ac.at/ downloads.html). Gustavsson N, Bolviken B, Smith DB, and Severson RC (2001) Geochemical landscapes of the conterminous United States – new map presentations for 22 elements. U.S. Geological Survey Professional Paper 1648. Washington, DC: US Geological Survey. Guy HP and Norman VW (1970) Field Methods for Measurement of Fluvial Sediment, U.S. Geological Survey Techniques of Water Resources Investigations, Book 3, ch. C2. Washington, DC: US Geological Survey. Hakanson L (1984) Sediment sampling in different aquatic environments – statistical aspects. Water Resources Research 20: 41--46. Hall GEM and Pelchat P (1999) Comparability of results obtained by the use of different selective extraction schemes for the determination of element forms in soil. Water, Air, and Soil Pollution 112: 41--53. Hansen DJ, Di Toro DM, Berry WJ, et al. (2005) Procedures for the Derivation Equilibrium Partitioning Sediment Benchmarks (ESBs) for the Protection of Benthic Organisms: Metal Mixtures (Cadmium, Copper, Lead, Nickel, Silver and Zinc). U.S. EPA/600/R-02/011. Washington, DC: US Environmental Protection Agency, Office of Research and Development. Haregeweyn N, Poesen J, Nyssen J, et al. (2008) Sediment yield variability in Northern Ethiopia: A quantitative analysis of its controlling factors. Catena 75: 65--78. Harrison ET, Norris RH, and Wilkinson SN (2008) Can a catchment-scale sediment model be related to an indicator of river health? Hydrobiologia 600: 49--64. Hart BT (1982) Uptake of trace metals by sediments and suspended particulates: A review. Hydrobiologia 91–92: 299--313. Hart EA (2002) Effects of woody debris on channel morphology and sediment storage in headwater streams in the Great Smoky Mountains, Tennessee-North Carolina. Physical Geography 23: 492--510. Harvey AM (2002) The role of base-level change in the dissection of alluvial fans: Case studies from southeast Spain and Nevada. Geomorphology 45: 67--87. Harvey AM, Mather AE, and Stokes M (2005) Alluvial fans: Geomorphology, sedimentology, dynamics – introduction. A review of alluvial fan research. Geological Society London Special Publication 252: 1--7. Hassan MA, Church M, Lisle TE, Brardinoni F, Benda L, and Grant GE (2005) Sediment transport and channel morphology of small forested streams. Journal of the American Water Resources Association 41: 853--876. Hatfield RG and Maher BA (2009) Fingerprinting upland sediment sources: Particle size-specific magnetic linkages between soils, lake sediments and suspended sediments. Earth Surface Processes and Landforms 34(10): 1359--1373 (doi:10.1002/esp.1824). Hawkes HE and Webb JS (1962) Geochemistry in Mineral Exploration. New York, NY: Harper and Row. Haygarth PM, Bilotta GS, Bol R, et al. (2006) Processes affecting transfer of sediment and colloids with associated phosphorus, from intensively farmed grasslands: An overview of key issues. Hydrological Processes 20: 4407--4413. He Q and Walling DE (1996) Use of fallout Pb-210 measurements to investigate longer-term rates and patterns of overbank sediment deposition on the floodplains of lowland rivers. Earth Surface Processes and Landforms 21: 141--154. Heimsath AM, Chappell J, Spooner NA, and Questiaux DG (2002) Creeping soil. Geology 30: 111--114. House WA and Warwick MS (1999) Interactions of phosphorus with sediments in the River Swale, Yorkshire, UK. Hydrological Processes 13: 1103--1115. Horowitz AJ (1991) A Primer on Sediment-Trace Element Chemistry, 2nd edn. Chelsea, MI: Lewis.

334

Catchment Erosion, Sediment Delivery, and Sediment Quality

Horowitz AJ (1995) The Use of Suspended Sediment and Associated Trace Elements in Water Quality Studies. IAHS Special Publication No. 4. Wallingford: IAHS Press. Horowitz AJ (1997) Some thoughts on problems associated with various sampling media used for environmental monitoring. Analyst 122: 1193--1200. Horowitz AJ (2008) Contaminated sediments: Inorganic constituents. In: Anderson MG (ed.) Encyclopedia of Hydrological Sciences, Wiley Interscience, published online at http://mrw.interscience.wiley.com/emrw/9780470848944/ehs/article/hsa317/ current/html, DOI: 10.1002/0470848944.hsa317, 27pp. Horowitz AJ (2008) Determining annual suspended sediment and sediment-associated trace element and nutrient fluxes. Science of the Total Environment 400: 315--343. Horowitz AJ, Elrick KA, and Callender E (1988) The effect of mining on the sedimenttrace element geochemistry of cores from the Cheyenne River arm of Lake Oahe, South Dakota, U.S.A. Chemical Geology 67: 17--33. Horowitz AJ, Elrick KA, and Cook RB (1993) The effect of mining and related activities on the sediment-trace element geochemistry of Lake Coeur d’Alene, Idaho, U.S.A. – Part I: Surface sediments. Hydrological Processes 7: 403--423. Horowitz AJ, Elrick KA, Demas CR, and Demcheck DK (1991) The use of sedimenttrace element geochemical models for the identification of local fluvial baseline concentrations. In: Peters NE and Walling DE (eds.) Sediment and Stream Water Quality in a Changing Environment: Trends and Explanation, IAHS Publication 203, pp. 339–348. Wallingford: IAHS Press. Horowitz AJ, Elrick KA, and Hooper RC (1989) A comparison of instrumental dewatering methods for the separation and concentration of suspended sediment for subsequent trace element analysis. Hydrological Processes 3: 163--184. Horowitz AJ, Elrick KA, and Smith JJ (2001) Annual suspended sediment and trace element fluxes in the Mississippi, Columbia, Colorado, and Rio Grande drainage basins. Hydrological Processes 15: 1169--1207. Horowitz AJ, Elrick KA, and Smith JJ (2008) Monitoring urban impacts on suspended sediment, trace element, and nutrient fluxes within the City of Atlanta, Georgia, U.S.A.: Program design, methodological considerations, and initial results. Hydrological Processes 22: 1473--1496. Horowitz AJ, Elrick KA, von Guerard PB, Young NO, Buell GR, and Miller TL (1992) The use of automatically collected point samples to estimate suspended sediment and associated trace element concentrations for determining annual mass transport. In: Bogen J, Walling D, and Day T (eds.) Proceedings of the IAHS Symposium Erosion and Sediment Transport Monitoring Programmes in River Basins, IAHS Publication No. 210, pp. 209–218. Oslo: IAHS Press. Horowitz AJ, Rinella F, Lamothe P, et al. (1990) Variations in suspended sediment and associated trace element concentrations in selected riverine cross sections. Environmental Science and Technology 24: 1313--1320. Horowitz AJ and Stephens VC (2008) The effects of land use on fluvial sediment chemistry for the conterminous U.S. – results from the first cycle of the NAWQA Program: Trace and major elements, phosphorus, carbon, and sulfur. Science of the Total Environment 400: 290--314. Hughes AO, Olley JM, Croke JC, and Webster IT (2009) Determining floodplain sedimentation rates using 137Cs in a low fallout environment dominated by channel- and cultivation-derived sediment inputs, central Queensland, Australia. Journal of Environmental Radioactivity 100(10): 858--865 (doi:10.1016/ j.jenrad.2009.06.011). Hughes AO, Prosser IP, Stevenson J, et al. (2001) Gully erosion mapping for the National Land and Water Resources Audit. Technical Report 26/01. Canberra: CSIRO Land and Water. Hupp CR, Schenk ER, Richter JM, Peet RK, and Townsend PA (2009) Bank erosion along the dam-regulated lower Roanoke River, North Carolina. In: Management and Restoration of Fluvial Systems with Broad Historical Changes and Human Impacts, The Geological Society of America Special Paper 451, pp. 97–108. Boulder, CO: Geological Society of America. ICOLD (2006) World Register of Dams. ICOLD http://www.icold-cigb.net (accessed April 2010). Jacobsen AM, Halling-Sorenson B, Ingerslev F, and Hansen SH (2004) Simultaneous extraction of tetracycline, macrolide and sulfonamide antibiotics from agricultural soils using pressurized liquid extraction, followed by solid-phase extraction and liquid chromatography-tandem mass spectrometry. Journal of Chromatography A 1038: 157--170. Jeffries R, Darby SE, and Sear DA (2003) The influence of vegetation and organic debris on flood-plain sediment dynamics: Case study of a low-order stream in the New Forest England. Geomorphology 51: 61--80. Johnson WM and Maxwell JA (1981) Rock and Mineral Analysis. New York, NY: Wiley. Jothityangkoon C, Sivapalan M, and Farmer DL (2001) Process controls on water balance variability in a large semi-arid catchment: Downward approach to hydrological model development. Journal of Hydrology 254: 174--198.

Julien P (1995) Erosion and Sedimentation. Cambridge: Cambridge University Press. Julien P (2002) River Mechanics. Cambridge: Cambridge University Press. Keith LH (ed.) (1988) Principles of Environmental Sampling, American Chemical Society. York, PA: Maple Press. Kersten M and Fo¨rstner U (1987) Effect of sample pretreatment on the reliability of solid speciation data of heavy metals: Implications for the study of early diagenetic processes. Marine Chemistry 22: 299--312. Kisi O (2005) Suspended sediment estimation using neuro-fuzzy and neural network approaches. Hydrological Sciences Journal (Des Sciences Hydrologiques) 50: 683--696. Krauskopf KB (1967) Introduction to Geochemistry. New York, NY: McGraw-Hill. Krein A, Petticrew E, and Udelhoven T (2003) The use of fine sediment fractal dimensions and color to determine sediment sources in a small watershed. Catena 53: 165--179. Kronvang B, Hoffmann CC, and Dro¨ge R (2009) Sediment deposition and net phosphorus retention in a hydraulically restored lowland river floodplain in Denmark: Combining field and laboratory experiments. Marine and Freshwater Research 60: 638--646. Kuhnert PM, Henderson A, Bartley R, and Herr A (2010) Incorporating uncertainty in gully erosion calculations using the Random Forests modelling approach. Environmetrics 21(5): 493--509. doi:10.1002/env.999. Lahr J, Maas-Diepeveen JL, Stuijfzand SC, et al. (2003) Responses in sediment bioassays used in the Netherlands: Can observed toxicity be explained by routinely monitoring priority pollutants? Water Research 37: 1691--1710. Lal R (ed.) (1994) Soil Erosion Research Methods. Ankeny, IA: Soil and Water Conservation Society. Langendoen EJ and Simon A (2008) Modeling the evolution of incised streams. II: Streambank erosion. Journal of Hydraulic Engineering-ASCE 134: 905--915. Larionov GA (1993) Erosion and Wind Blown Soil. Moscow: Moscow State University Press. Laubel A, Kronvang B, Hald AB, and Jensen C (2003) Hydromorphological and biological factors influencing sediment and phosphorus loss via bank erosion in small lowland rural streams in Denmark. Hydrological Processes 17: 3443--3463. Lauer JW and Parker G (2008) Modeling framework for sediment deposition, storage, and evacuation in the floodplain of a meandering river: Theory. Water Resources Research 44: W04425. Lavigne F and Suwa H (2003) Contrasts between debris flows, hyperconcentrated flows and stream flows at a channel of Mount Semeru, East Java, Indonesia. Geomorphology 61: 41--58. Lawler DM (1995) The impact of scale on the processes of channel-side sediment supply: A conceptual model. In: Effects of Scale on Interpretation and Management of Sediment and Water Quality, IAHS Publication 226, pp. 175–184. Boulder, NC: IAHS Press. Lawler DM, Petts GE, Foster IDL, and Harper S (2006) Turbidity dynamics during spring storm events in an urban headwater river system: The Upper Tame, West Midlands, UK. Science of the Total Environment 360: 109--126. Leeder MR and Mack GH (2001) Lateral erosion (‘toe-cutting’) of alluvial fans by axial rivers: Implications for basin analysis and architecture. Journal of the Geological Society 158: 885--893. Lei TW, Zhang QW, Zhao J, and Nearing MS (2006) Tracing sediment dynamics and sources with rare earth elements. European Journal of Soil Science 57: 287--294. Lewis J and Eads RE (2001) Turbidity threshold sampling for suspended sediment load estimation. In: Proceedings of the 7th Federal Interagency Sedimentation Conference, pp. III-110–III-117. Reno, Nevada. Federal Interagency Project, Technical Committee of the Subcommittee on Sedimentation. Li S, Lobb DA, Tiessen KHD, and McConkey BG (2009) Selecting and applying cesium-137 conversion models to estimate soil erosion rates in cultivated fields. Journal of Environmental Quality 39: 204--219. Lu H, Moran CJ, and Sivapalan M (2005) A theoretical exploration of catchment-scale sediment delivery. Water Resources Research 41: W09415 (doi:10.1029/ 2005WR004018). Ludwig W, Probst J-L, and Kempe S (1996) Predicting the oceanic input of organic carbon by continental erosion. Global Biogeochemical Cycles 10: 23--41. Luthy RG, Aiken GR, Brusseau ML, et al. (1997) Sequestration of hydrophobic organic contaminants by geosorbents. Environmental Science and Technology 31: 3341--3347. Lyons WB, Nezat CA, Carey AE, and Hicks DM (2002) Organic carbon fluxes to the ocean from high-standing islands. Geology 30: 443--446. Mabit L, Benmansour M, and Walling DE (2008) Comparative advantages and limitations of fallout radionuclides (137Cs, 210Pb and 7Be) to assess soil erosion and sedimentation. Journal of Environmental Radioactivity 99: 1799--1807.

Catchment Erosion, Sediment Delivery, and Sediment Quality

Macaire J-J, Bellemlih S, Di-Giovanni C, De Luca P, Visset L, and Bernard J (2002) Sediment yield and storage variations in the Negron River catchment (south western Parisian Basin, France) during the Holocene period. Earth Surface Processes and Landforms 27: 991--1009. MacDonald DD, Ingersoll CG, and Berger TA (2000) Development and evaluation of consensus-based sediment quality guidelines for freshwater ecosystems. Archives of Environmental Contamination and Toxicology 39: 20--31. Macnab K, Jacobson C, and Brierley G (2006) Spatial variability of controls on downstream patterns of sediment storage: A case study in the Lane Cove Catchment, New South Wales, Australia. Geographical Research: Journal of the Institute of Australian Geographers 44: 255--271. Malmon DV, Reneau SL, Dunne T, Katzman D, and Drakos PG (2005) Influence of sediment storage on downstream delivery of contaminated sediment. Water Resources Research 41: W05008. Manheim FT and Hayes LH (2002) Lake Ponchartrain basin, bottom sediments and related resources. U.S. Geological Survey Professional Paper 1634, CD-ROM. Matisoff G, Bonniwell EC, and Whiting PJ (2002) Soil erosion and sediment sources in an Ohio watershed using beryllium-7, cesium-137, and lead-210. Journal of Environmental Quality 31: 54--61. McKergow L, Prosser I, Hughes A, and Brodie J (2005) Sources of sediment to the Great Barrier Reef World Heritage Area. Marine Pollution Bulletin 51: 200--211. Meador JP, Ernest DW, and Kagley AN (2005) A comparison of the non-essential elements cadmium, mercury, and lead found in fish and sediment from Alaska and California. Science of the Total Environment 339: 189--205. Melis TS and Topping DG (2003) Testing laser based sensors for continuous in situ monitoring of suspended sediment in the Colorado River, Arizona. In: Bogen J, Fergus T, and Walling DE (eds.) Erosion and Sediment Transport Measurement in Rivers: Technological and Methodological Advances, IAHS Publication No. 283, pp. 21–27. Wallingford: IAHS Press. Merritt WS, Letcher RA, and Jakeman AJ (2003) A review of erosion and sediment transport models. Environmental Modelling and Software 18: 761--799. Merten GHG, Nearing MA, and Borges ALO (2001) Effect of sediment load on soil deteachment and deposition in rills. Soil Science Society of America Journal 65: 861--868. Milliman JD, Quirashee GS, and Beg MAA (1984) Sediment discharge from the Indus River to the ocean: Past, present and future. In: Haq BU and Milliman JD (eds.) Marine Geology and Oceanography of Arabian Sea and Coastal Pakistan, pp. 65--70. New York: Van Nostran Rheinhold. Molina A, Govers G, Poesen J, Van Hemelryck H, De Bie`vre B, and Vanecker V (2007) Environmental factors controlling spatial variation in sediment yield in a Central Andean mountain area. Geomorphology 98: 178--186. Morgan RPC (1995) Soil Erosion and Conservation. Harlow: Longman Scientific and Technical. Morozova VS, Eremin SA, Nesterenko PN, Klyuev NA, Shelepchikov AA, and Kubrakova IA (2008) Microwave and ultrasonic extraction of chlorophenoxy acids from soil and their determination by fluorescence polarization immunoassay. Journal of Analytical Chemistry 63: 127--134. Mudroch A and MacKnight SD (1991) Handbook of Techniques for Aquatic Sediment Sampling. Boca Raton, FL: CRC Press. Nearing MA, Foster GR, Lane LJ, and Finkner SC (1989) A process-based soil-erosion model for USDA-Water Erosion Prediction Project technology. Transactions of the ASAE 32(5): 1587--1593. Newcombe CP and Jensen JOT (1996) Channel suspended sediment and fisheries: A synthesis for quantitative assessment of risk and impact. North American Journal of Fish Management 16: 693--727. Nicholas AP and Walling DE (1996) The significance of particle aggregation in the overbank deposition of suspended sediment on river floodplains. Journal of Hydrology 186: 275--293. Nicholas AP, Walling DE, Sweet RJ, and Fang X (2006) Development and evaluation of a new catchment-scale model of floodplain sedimentation. Water Resources Research 42: W10426 (doi:10.1029/2005WR004579). Nistor CJ and Church M (2005) Suspended sediment transport regime in a debris-flow gully on Vancouver Island, British Columbia. Hydrological Processes 19: 861--885. NLWRA (2001) Australian Agriculture Assessment 2001. National Land and Water Resources Audit, Canberra. http://www.anra.gov.au/topics/publications (accessed April 2010). Notebaert B, Verstraeten G, Govers G, and Poesen J (2009) Qualitative and quantitative applications of LiDAR imagery in fluvial geomorphology. Earth Surface Processes and Landforms 34: 217--231. Old GH, Leeks GJL, Packman JC, Smith BPG, Lewis S, and Hewitt EJ (2006) River flow and associated transport of sediments and solutes through a highly urbanised catchment, Bradford, West Yorkshire. Science of the Total Environment 360: 98--108.

335

Old GH, Leeks GJL, Packman JC, et al. (2003) The impact of a convectional summer rainfall event on river flow and fine sediment transport in a highly urbanized catchment: Bradford, West Yorkshire. Science of the Total Environment 314/316: 495--512. Ongley ED (1982) The PLUARG experience: Scientific implications for diffuse source management. In: Hart BT (ed.) Water Quality Management: Monitoring Programs and Diffuse Runoff, pp. 87--101. Melbourne: Water Studies Centre, Chisholm Institute of Technology and Australian Society for Limnology. Ongley ED, Krishnappan BG, Droppo I, Rao SS, and Maguire RJ (1992) Cohesive sediment transport: Emerging issues for toxic chemical management. Hydrobiologia 235–236: 177--187. Ottesen RT, Bogen J, Bolviken B, Volden T, and Haugland T (2000) Geokjemisk Atlas for Norge, del. 1. Trondheim, Norway: Norges geologiske undersokelse, Norges vassdrags–og energidirektorat. Owens PN, Batalla RJ, Collins AJ, et al. (2005) Fine-grained sediment in river systems: Environmental significance and management issues. River Research and Applications 21: 693--717. Parsons AJ, Wainwright J, Brazier RE, and Powell DM (2006) Is sediment delivery a fallacy? Earth Surface Processes and Landforms 31: 1325--1328. Persaud D, Jaagumagi R, and Hayton A (1993) Guidelines for the Protection and Management of Aquatic Sediment Quality in Ontario. Ontario: Water Resources Branch, Ontario Ministry of the Environment and Energy. Petticrew EL, Krein A, and Walling DE (2007) Evaluating fine sediment mobilisatioin and storage in a gravel-bed river using controlled reservoir releases. Hydrological Processes 21: 198--210. Phillips JD (1986) The utility of the sediment budget concept in sediment pollutioncontrol. Professional Geographer 38: 246--252. Phillips JM, Russell MA, and Walling DE (2000) Time-integrated sampling of fluvial suspended sediment: A simple methodology for small catchments. Hydrological Processes 14: 2589--2602. Pittam NJ, Foster IDL, and Mighall TM (2009) An integrated lake-catchment approach for determining sediment source changes at Aqualate Mere, Central England. Journal of Paleolimnology 42(2): 215--232 (doi:10.1007/s10933-008-9272-9). Plant JA, Klaver G, Locutura J, Salminen R, Vrana K, and Fordyce FM (1997) The forum of European geological surveys geochemistry task group inventory 1994– 1996. Journal of Geochemical Exploration 59: 123--146. Poesen J, Nachtergaele J, Verstraeten G, and Valentin C (2003) Gully erosion and environmental change: Importance and research needs. Catena 50: 91--133. Poldervaart A (ed.) (1955) Crust of the earth. Geological Society of America Special Paper 62. Denver, CO: Geological Society of America. Pope LM (2005) Assessment of contaminated streambed sediment in the Kansas part of the historic Tri-State Lead and Zinc Mining District, Cherokee County, 2004. U.S. Geological Survey Scientific Investigations Report 2005–5251. Lawrence, KS: US Geological Survey. Preston N and Schmidt J (2003) Modelling sediment fluxes at large spatial and temporal scales. Lecture Notes in Earth Sciences 101: 53--72. Prica M, Dalmacija B, Roncevic S, Krcmar D, and Becelic M (2008) A comparison of sediment quality results with acid volatile sulfides (AVS) and simultaneously extracted metals (SEM) ratio in Vojvodina (Serbia) sediments. Science of the Total Environment 389: 235--244. Prosser IP, Hughes A, Rustomji P, Young WJ, and Moran CJ (2001a) Predictions of the sediment regime of Australian rivers. In: Rutherfurd ID, Sheldon F, Brierley G, and Kenyon C (eds.) Third Australian Stream Management Conference, pp. 529--533. Brisbane: Cooperative Research Centre for Catchment Hydrology. Prosser IP and Rustomji P (2000) Sediment capacity relationships for overland flow. Progress in Physical Geography 24: 179--193. Prosser IP, Rustomji P, Young B, Moran C, and Hughes A (2001b) Constructing river basin sediment budgets for the National Land and Water Resources Audit. Technical Report 15/01. Canberra: CSIRO Land and Water. Prosser IP, Rutherfurd ID, Olley JM, Young WJ, Wallbrink PJ, and Moran CJ (2001c) Large-scale patterns of erosion and sediment transport in river networks, with examples from Australia. Marine and Freshwater Research 52: 81--99. Prosser IP and Winchester SJ (1996) History and processes of gully initiation and development in eastern Australia. Zeitschrift fur Geomorphologie, Supplement Bund 105: 91--109. Radaone M and Radaone N (2005) Dams, sediment sources and reservoir silting in Romania. Geomorphology 71: 112--125. Raynie DE (2004) Modern extraction techniques. Analytical Chemistry 76: 4659--4664. Raynie DE (2006) Modern extraction techniques. Analytical Chemistry 78: 3997--4004. Reed T, Wielgus SJ, Barnes AK, Schiefelbein JJ, and Fettes AL (2004) Refugia and local controls: Benthic invertebrate dynamics in Lower Green Bay, Lake Michigan following zebra mussel invasion. Journal of Great lakes Research 30: 390--396.

336

Catchment Erosion, Sediment Delivery, and Sediment Quality

Reid LM and Dunne T (2003) Sediment budgets as an organizing framework in fluvial geomorphology. In: Kondolf MG and Pie´gay H (eds.) Tools in Fluvial Geomorphology, pp. 463--500. New York: Wiley. Reimann C and Garrett RG (2005) Geochemical background – concept and reality. Science of the Total Environment 350: 12--27. Renard KG, Foster GA, Weesies DK, McCool DK, and Yoder DC (1997) Predicting Soil Erosion by Water: A Guide to Conservation Planning with the Revised Universal Soil Loss Equation. Agriculture Handbook 703. Washington, DC: United States Department of Agriculture. Renschler CS (2003) Designing geo-spatial interfaces to scale process models: the GeoWEPP approach. Hydrological Processes 17: 1005--1017. Rinaldi M, Mengoni B, Luppi L, Darby SE, and Mosselman E (2008) Numerical simulation of hydrodynamics and bank erosion in a river bend. Water Resources Research 44: W09428. Ritchie JC (1996) Remote sensing applications to hydrology: Airborne laser altimeters. Hydrological Sciences Journal (Des Sciences Hydrologiques) 41: 625--636. Ritchie JC (1972) Sediment, fish, and fish habitat. Journal of Soil and Water Conservation 27: 124--125. Ritchie JC, Finney VL, Oster KJ, and Ritchie CA (2004) Sediment deposition in the flood plain of Stemple Creek watershed, northern California. Geomorphology 61: 347--360. Ritchie JC and Ritchie CA (2008) Bibliography of publications of 137cesium studies related to erosion and sediment deposition. USDA-ARS Hydrology and Remote Sensing Laboratory Occasional Paper HRSL-2008-02. http://www.ars.usda.gov/ Main/docs.htm?docid ¼ 17939 (accessed April 2010). Ritter JB, Miller JR, and Husek-Wulforst J (2000) Environmental controls on the evolution of alluvial fans in Buena Vista Valley, North Central Nevada, during late quaternary time. Geomorphology 36: 63--87. Roehl JE (1962) Sediment source areas, delivery ratios and influencing morphological factors. In: Commission of Land Erosion (Proceedings Symposium of Bari, October, 1962), IAHS Publication No. 59, pp. 202–213. Wallingford: IAHS Press. Rommens T, Verstraeten G, Bogman P, et al. (2006) Holocene alluvial sediment storage in a small river catchment in the loess area of Central Belgium. Geomorphology 77: 187--201. Rommens T, Verstraeten G, Lang A, et al. (2005) Soil erosion and sediment deposition in the Belgian loess belt during the Holocene: Establishing a sediment budget for a small agricultural catchment. Holocene 15: 1032--1043. Rowan JS, Goodwill P, and Franks SW (2000) Uncertainty estimation in fingerprinting suspended sediment sources. In: Foster IDL (ed.) Tracers in Geomorphology, pp. 279--290. Chichester: Wiley. Rubin DM and Topping DJ (2001) Quantifying the relative importance of flow regulation and grain size regulation of suspended sediment transport and tracking changes in grain size of bed sediment. Water Resources Research 37: 133--146. Rustomji P (2006) Analysis of gully dimensions and sediment texture for southeast Australia for catchment sediment budgeting. Catena 67: 119--127. Rustomji P, Caitcheon G, and Hairsine P (2008) Combining a spatial model with geochemical tracers and river station data to construct a catchment sediment budget. Water Resources Research 44: W01422. Rustomji P and Prosser IP (2001) Spatial patterns of sediment delivery to valley floors: Sensitivity to sediment transport capacity and hillslope hydrology relations. Hydrological Processes 15: 1003--1018. Rustomji P and Wilkinson SN (2008) Applying bootstrap resampling to quantify uncertainty in fluvial suspended sediment loads estimated using rating curves. Water Resources Research 44: W09434. Rutherfurd ID (2000) Some human impacts on Australian stream channel morphology. In: Brizga S and Finlayson B (eds.) River Management: The Australasian Experience, pp. 11--49. London: Wiley. Saenger C, Cronin TM, Willard D, Halka J, and Kerhin R (2008) Increased terrestrial to ocean sediment and carbon fluxes in the northern Chesapeake Bay associated with twentieth century land alteration. Estuaries and Coasts 31: 492--500. Salomons W and Fo¨rstner U (1984) Metals in the Hydrocycle. New York, NY: Springer. Santiago S, Thomas RL, Larbaigt G, et al. (1994) Nutrient, heavy metal and organic pollutant composition of suspended and bed sediments in the Rhone River. Aquatic Sciences 56: 220--242. Schiettecatte W, Gabriels D, Cornelis WM, and Hofman G (2008) Enrichment of organic carbon in sediment transport by interrill and rill erosion processes. Soil Science Society of America Journal 72: 50--55. Schoellhamer DH and Wright SA (2003) Continuous measurement of suspendedsediment discharge in rivers by use of optical backscatterance sensors. In: Bogen J, Fergus T, and Walling DE (eds.) Erosion and Sediment Transport Measurement in Rivers: Technological and Methodological Advances, IAHS Publication No. 283, pp. 28–36. Wallingford: IAHS Press.

Schuerch P, Densmore AL, McArdell BW, and Molnar P (2006) The influence of landsliding on sediment supply and channel change in a steep mountain catchment. Geomorphology 78: 222--235. Schuller P, Iroume A, Walling DE, Mancilla HB, Castillo A, and Trumper RE (2006) Use of beryllium-7 to document soil redistribution following forest harvest operations. Journal of Environmental Quality 35: 1756--1763. Schwab GO, Frevert RK, Edminster TW, and Barnes KK (1981) Soil and Water Conservation Engineering, 3rd edn. New York: Wiley. Schwertmann U, Vogel W, and Kainz M (1990) Bodenerosion Durch Wasser. Stuttgart: Eugen Ulmer GmbH. Seitzinger SP, Harrison JA, Dumont E, Beusen AHW, and Bouwman AF (2005) Sources and delivery of carbon, nitrogen and phosphorus to the coastal zone: An overview of global nutrient export from watersheds (NEWS) models and their application. Global Biogeochemical Cycles 19: GB4S01. Sepulveda A, Schuller P, Walling DE, and Castillo A (2007) Use of 7Be to document soil erosion associated with a short period of extreme rainfall. Journal of Environmental Radioactivity 99: 35--49. Shacklette HT and Boerngen JG (1984) Elemental concentrations in soils and other surficial materials of the conterminous United States. U.S. Geological Survey Professional Paper 1270. Washington, DC: US Geological Survey. Shear H and Watson AEP (eds.) (1977) The Fluvial Transport of Sediment-Associated Nutrients and Contaminants. Windsor, ON: International Joint Commission. Shreve RL (1966) Statistical law of stream numbers. Journal of Geology 74: 17--37. Silvia Diaz-Cruz M, Lopex de M, and Barcelo D (2006) Determination of antimicrobials in sludge from infiltration basins at two artificial recharge plants by pressurized liquid extraction–liquid chromatography–tandem mass spectrometry. Journal of Chromatography A 1130: 72--82. Simon A, Dickerson W, and Heine A (2004) Suspended-sediment transport rates at the 1.5-year recurrence interval for ecoregions of the United States: Transport conditions at the bankfull and effective discharge? Geomorphology 58: 243--262. Simpson SL and Batley GE (2007) Predicting metal toxicity in sediments: A critique of current approaches. Integrated Environmental Assessment and Management 3: 18--31. Simpson SL, Bately GE, Charlton AA, et al. (2005) Handbook for Sediment Quality Assessment. Lucas Heights, NSW: CSIRO, Centre for Environmental Contaminants Research. Singer MB and Dunne T (2001) Identifying eroding and depositional reaches of valley by analysis of suspended sediment transport in the Sacramento River, California. Water Resources Research 37: 3371--3381. Sivapalan M (2009) The secret to doing better hydrological science?: Change the question!. Hydrological Processes 23: 1391--1396. Smith BPG, Naden PS, Leeks GJL, and Wass PD (2003) The influence of storm events on fine sediment transport, erosion and deposition in a reach of the River Swale, Yorkshire, UK. Science of the Total Environment 314–316: 451--474. Smith HG and Dragovich D (2008) Sediment budget analysis of slope-channel coupling and in-channel sediment storage in an upland catchment, southeastern Australia. Geomorphology 101: 643--654. Soulsby C, Youngson AF, Moira HJ, and Malcolm IA (2001) Fine sediment influence on salmonid spawning habitat in a lowland agricultural stream: A preliminary assessment. Science of the Total Environment 265: 295--307. Srivastava P, Migliaccio KW, and Simunek J (2007) Landscape models for simulating water quality at point, field, and watershed scales. Transactions of the ASABE 50: 1683--1693. Staley DM, Wasklewicz TA, and Blaszczynski JS (2006) Surficial patterns of debris flow deposition on alluvial fans in Death Valley, CA using airborne laser swath mapping data. Geomorphology 74: 152--163. Steegen A, Govers G, Takken I, Nachtergaele J, Posen J, and Merckx R (2001) Factors controlling sediment and phosphorus export from two Belgian agricultural catchments. Journal of Environmental Quality 30: 1249--1258. Steiger J, Gurnell AM, and Goodson JM (2003) Quantifying and characterizing contemporary riparian sedimentation. River Research and Applications 19: 335--352. Stone M and Droppo IG (1996) Distribution of lead, copper and zinc in sizefractionated river bed sediment in two agricultural catchments of southern Ontario, Canada. Environmental Pollution 93: 353--362. Stone M, Krishnappan BG, and Emelko MB (2008) The effect of bed age and shear stress on the particle morphology of eroded cohesive river sediment in an annular flume. Water Research 42: 4179--4187. Stone M and Saunderson H (1992) Particle size characteristics of suspended sediment in Southern Ontario rivers tributary to the Great Lakes. Hydrological Processes 6: 189--198.

Catchment Erosion, Sediment Delivery, and Sediment Quality

Stone PM and Walling DE (1997) Particle size selectivity considerations in suspended sediment budget investigations. Water, Air and Soil Pollution 99: 63--70. Stumm W and Morgan JJ (1996) Aquatic Chemistry: Chemical Equilibria and Rates in Natural Waters. New York, NY: Wiley. Sun T, Meakin P, Jossang T, and Schwarz K (1996) A simulation model for meandering rivers. Water Resources Research 32: 2937--2954. Suttle KB, Powers ME, Levine JM, and McNeely C (2004) How fine sediment in riverbeds impairs growth and survival of juvenile salmonids. Ecological Applications 14: 969--974. Swanson KM, Watson E, Aalto R, et al. (2008) Sediment load and floodplain deposition rates: Comparison of the Fly and Strickland Rivers, Papue New Guinea. Journal of Geophysical Research, Earth Surface 113: F01S03. Sweet RJ, Nicholas AP, Walling DE, and Fang X (2003) Morphological controls on medium-term sedimentation rates on British lowland river floodplains. Hydrobiologia 494: 177--183. Syvitski JPM and Milliman JD (2007) Geology, geography and humans battle for dominance over delivery of fluvial sediment to the coastal ocean. Journal of Geology 115: 1--19. Syvitski JPM, Vo¨ro¨smarty CJ, Kettner AJ, and Green P (2005) Impact of humans on the flux of terrestrial sediment to the global coastal ocean. Science 308: 376--380. Takken I, Beuselinck L, Nachtergaele J, Govers G, Poesen J, and Degraer G (1999) Spatial evaluation of a physically-based distributed erosion model (LISEM). Catena 37: 431--447. Tamene L, Park SJ, Dikau R, and Vlek PLJ (2006) Analysis of factors determining sediment yield variability in the highlands of northern Ethiopia. Geomorphology 76: 76--91. Taylor SR (1964) Abundance of chemical elements in the continental crust: A new table. Geochimica et Cosmochimica Acta 28: 1273--1285. Terry JP, Garimella S, and Kostaschuk RA (2002) Rates of floodplain accretion in a tropical island river system impacted by cyclones and large floods. Geomorphology 42: 171--182. Tessier A, Campbell PGC, and Bisson M (1979) Sequential extraction procedure for the speciation of particulate trace metals. Analytical Chemistry 51: 844--851. Thonon I (2006) Deposition of Sediment and Associated Heavy Metals on Floodplains. PhD Thesis, University of Utrecht. Thonon I, de Jong K, van der Perk M, and Middelkoop H (2007) Modelling floodplain sedimentation using particle tracking. Hydrological Processes 21: 1402--1412. Thonon I, Roberti JR, Middelkoop H, van der Perk M, and Burrough P (2005) In situ measurement of sediment settling characteristics in floodplains using a LISST-ST. Earth Surface Processes and Landforms 30: 1327--1343. Topping D, Wright S, Melis T, and Rubin D (2005) High-resolution monitoring of suspended-sediment concentration and grain size in the Colorado River using laser-diffraction instruments and a three-frequency acoustic system. In: Proceedings of the Federal Interagency Sediment Monitoring Instrument and Analysis Research Workshop. Flagstaff, AZ, USA, 9–11 September 2003. US Geological Survey Circular 1276. Trimble SW (1983) A sediment budget for Coon Creek basin in the Driftless Area, Wisconsin, 1853–1977. American Journal of Science 283: 454--474. Trimble SW (1993) The distributed sediment budget model and watershed management in the Paleozoic plateau of the upper midwestern United States. Physical Geography 14: 285--303. Trimble SW and Crosson P (2000) U.S. soil erosion rates – myth and reality. Science 289: 248--250. Turekian KK and Wedepohl KH (1961) Distribution of the elements in some major units of the earth’s crust. Bulletin of the Geological Society of America 72: 175--192. UNESCO (1978) Monitoring of Particulate Matter Quality in Rivers and Lakes. Paris: UNESCO. US Bureau of Reclamation (2006) Erosion and Sedimentation Manual. Denver: US Bureau of Reclamation. US Department of Agriculture (1997) Predicting Soil Erosion by Water, a Guide to Conservation Planning with the Revised Universal Soil Loss Equation (RUSLE), USDA Agricultural Research Service, Agricultural Handbook No. 703. Washington, DC: USDA. US Environmental Protection Agency (1997) The Incidence and Severity of Sediment Contamination in Surface Waters of the United States, vol. 1, EPA-823-R-97-006. Washington, DC: National Sediment Quality Survey, Office of Science and Technology. US Environmental Protection Agency (2005) Procedures for the Derivation of Equilibrium Partitioning Sediment Benchmarks (ESBs) for the Protection of Benthic Organisms: Metal Mixtures (Cadmium, Copper, Lead, Nickel, Silver, and Zinc), EPA-600-R-02-011. Washington, DC: US Environmental Protection Agency, Office of Research and Development.

337

VA (ed.) (1977) Sedimentation Engineering, pp. 17--316. New York, NY: American Society of Civil Engineers, Task Committee for the Preparation of the Manual on Sedimentation of the Hydraulics Division. Valentin C, Poesen J, and Li Y (2005) Gully erosion: Impacts, factors and control. Catena 63: 132--153. Valmis S, Dimoylannis D, and Danalatos NG (2005) Assessing interrill erosion rate from soil aggregate instability index, rainfall intensity and slope angle on cultivated soils in central Greece. Soil and Tillage Research 80: 139--147. van Loon JC (1985) Selected Methods of Trace Metal Analysis: Biological and Environmental Samples. New York, NY: Wiley. Van Oost K, Quine TA, Govers G, et al. (2007) The impact of agricultural soil erosion on the global carbon cycle. Science 318: 626--629. Van Rompaey AJJ, Verstraeten G, Van Oost K, Govers G, and Poesen J (2001) Modelling mean annual sediment yield using a distributed approach. Earth Surface Processes and Landforms 26: 1221--1236. Verstraeten G and Poesen J (2002) Regional scale variability in sediment and nutrient delivery from small agricultural watersheds. Journal of Environmental Quality 31: 870--879. Verstraeten G, Poesen J, de Vente J, and Koninckx X (2003) Sediment yield variability in Spain: A quantitative and semiqualitative analysis using reservoir sedimentation rates. Geomorphology 50: 327--348. Verstraeten G, Prosser IP, and Fogarty P (2007) Predicting the spatial patterns of hillslope sediment delivery to river channels in the Murrumbidgee catchment, Australia. Journal of Hydrology 334: 440--454. Viney NR and Sivapalan M (1999) A conceptual model of sediment transport: Application to the Avon River Basin in Western Australia. Hydrological Processes 13: 727--743. Vo¨ro¨smarty CJ, Meybeck M, Fekete B, and Sharma K (1997) The potential impact of neo-Castorization on sediment transport by the global network of rivers. In: Human Impact on Erosion and Sedimentation, IAHS Publication No. 245, pp. 261–273. Wallingford: IAHS Press. Vo¨ro¨smarty CJ, Meybeck M, Fekete B, Sharma K, Green P, and Syvitski JPM (2003) Anthropogenic sediment retention: Major global impact from registered river impoundments. Global and Planetary Change 39: 169--190. Vrieling A (2006) Satellite remote sensing for water erosion assessment: A review. Catena 65: 2--18. Wallbrink PJ, Murray AS, Olley JM, and Olive LJ (1998) Determining the sources and transit times of suspended sediment in the Murrumbidgee River, New South Wales, Australia, using fallout 137 Cs and 210Pb. Water Resources Research 34: 879--887. Walling DE (1977) Assessing the accuracy of suspended sediment rating curves for a small basin. Water Resources Research 12: 1869--1894. Walling DE (1983) The sediment delivery problem. Journal of Hydrology 65: 209--237. Walling DE (2005) Tracing suspended sediment sources in catchments and river systems. Science of the Total Environment 344: 159--184. Walling DE (2006a) Human impact on land–ocean sediment transfer by the world’s rivers. Geomorphology 79: 192--216. Walling DE (2006b) Tracing versus monitoring: New challenges and opportunities in erosion and sediment delivery research. In: Owens PN and Collins AJ (eds.) Soil Erosion and Sediment Redistribution in River Catchments, pp. 13--27. Wallingford: CABI. Walling DE (2008) The changing sediment loads of the world’s rivers. In: Schmidt J, Cochrane T, Phillips C, Elliott S, Davies T, and Basher L (eds.) Sediment Dynamics in Changing Environments, IAHS Publication No. 325, pp. 323–338. Wallingford: IAHS Press. Walling DE and Collins AL (2008) The catchment sediment budget as a management tool. Environmental Science and Policy 11: 136--143. Walling DE, Collins AL, and McMellin GK (2003c) A reconnaissance survey of the source of interstitial fine sediment recovered from salmonid spawning gravels in England and Wales. Hydrobiologia 497: 91--108. Walling DE, Collins AL, and Sichingabula HM (2003a) Using unsupported lead-210 measurements to investigate soil erosion and sediment delivery in a small Zambian catchment. Geomorphology 52: 193--213. Walling DE, Collins AL, Jones PA, Leeks GJL, and Old G (2006) Establishing finegrained sediment budgets for the Pang and Lambourn LOCAR catchments, UK. Journal of Hydrology 330: 126--141. Walling DE, Collins AL, Sichingabula HW, and Leeks GJL (2001) Integrated assessment of catchment sediment budgets. Land Degradation and Development 12: 387--415. Walling DE, Collins AL, and Stroud RW (2008) Tracing suspended sediment and particulate phosphorus sources in catchments. Journal of Hydrology 350: 274--289. Walling DE and Fang D (2003) Recent trends in the suspended sediment loads of the world’s rivers. Global and Planetary Change 39: 111--126.

338

Catchment Erosion, Sediment Delivery, and Sediment Quality

Walling DE and He Q (1993) Use of cesium-137 as a tracer in the study of rates and patterns of floodplain sedimentation. In: Peters NE, Hoehn E, Leibundgut Ch, Tase N, and Walling DE (eds.) Tracers in Hydrology, IAHS Publication No. 215, pp. 319–328. Wallingford: IAHS Press. Walling DE and He Q (1997) Use of fallout 137Cs in investigations of overbank deposition on floodplains. Catena 29: 263--282. Walling DE and He Q (1999a) Improved models for estimating spoil erosion rates from cesium-137 measurements. Journal of Environmental Quality 28: 611--622. Walling DE and He Q (1999b) Using fallout lead-210 measurements to estimate soil erosion on cultivated land. Soil Science Society of America Journal 63: 1404--1412. Walling DE, He Q, and Appleby PG (2002) Conversion models for use in soil-erosion, soil redistribution and sedimentation investigations. In: Zapata F (ed.) Handbook for the Assessment of Soil Erosion and Sedimentation Using Environmental Radionuclides, pp. 111--164. Dordrecht: Kluwer. Walling DE, He Q, and Blake WH (1999) Use of 7Be and 137Cs measurements to document short- and medium-term rates of water-induced soil erosion on agricultural land. Water Resources Research 35: 3865--3874. Walling DE, Owens PN, Carter J, et al. (2003d) Storage of sediment-associated nutrients and contaminants in river channel and floodplain systems. Applied Geochemistry 18: 195--220. Walling DE, Owens PN, Foster IDL, and Lees JA (2003b) Changes in the fine sediment dynamics of the Ouse and Tweed basins over the last 100–150 years. Hydrological Processes 17: 3245--3269. Walling DE, Owens PN, and Leeks GJL (1998) The role of channel and floodplain storage in the suspended sediment budget of the River Ouse, Yorkshire, UK. Geomorphology 22: 225--242. Walling DE, Owens PN, Waterfall BD, Leeks GJL, and Wass PD (2000) The particle size characteristics of fluvial suspended sediment in the Humber and Tweed catchments, UK. Science of the Total Environment 251–252: 205--222. Walling DE, Schuller P, Zhang Y, and Iroume´ A (2009) Extending the timescale for using beryllium 7 measurements to document soil redistribution by erosion. Water Resources Research 45: W02418 (doi:10.1029/2008WR007143). Walling DE and Webb BW (1988) The reliability of rating curve estimates of suspended sediment yield. Some further comments. In: Sediment Budgets, Proceedings of the Porto Alegre Symposium. IAHS Publication No. 174, pp. 337–350. December 1988. Wallingford: IAHS Press. Warren N, Allan IJ, Carter JE, House WA, and Parker A (2003) Pesticides and other micro-organic contaminants in freshwater sedimentary environments – a review. Applied Geochemistry 18: 159--194. Wasson RJ (1994) Annual and decadal variation of sediment yield in Australia, and some global comparisons. In: Olive LJ, Loughran RJ, and Kesby JA (eds.) Variability in Stream Erosion and Sediment Transport, IAHS Publication 224, pp. 269–279. Canberra: IAHS Press. Wasson RJ, Mazari RK, Starr B, and Clifton G (1998) The recent history of erosion and sedimentation on the Southern Tablelands of southeastern Australia: sediment flux dominated by channel incision. Geomorphology 24: 291--308. Waters TF (1995) Sediment in Streams: Sources, Biological Effects and Control. Bethesda: American Fisheries Society. Webb JS, Thornton I, Thompson M, Howarth RJ, and Lowenstein PL (1978) The Wolfson Geochemical Atlas of England and Wales. Oxford: Oxford University Press. Wedepohl KH (1995) The composition of the continental crust. Geochimica et Cosmochimica Acta 59: 1217.

Wei H, Nearing MA, Stone JJ, et al. (2009) A new splash and sheet erosion equation for rangelands. Soil Science Society of America Journal 73: 1386--1392. Wemple BC, Swanson FJ, and Jones JA (2001) Forest roads and geomorphic process interactions, Cascade Range Oregon. Earth Surface Processes and Landforms 26: 191--204. Wilkinson S, Henderson A, Chen Y, and Sherman B (2004) SedNet User Guide, Version 2. Client Report. Canberra: CSIRO Land and Water. http:// www.toolkit.net.au/sednet (accessed April 2010). Wilkinson SN, Prosser IP, Olley JM, and Read A (2005) Using sediment budgets to prioritise erosion control in large river systems. In: Batalla RJ and Garcia C (eds.) Geomorphological Processes and Human Impacts in River Basins, IAHS Publication 299, pp. 56–64. Solsona, Catalonia: IAHS Press. Wilkinson SN, Prosser IP, and Hughes AO (2006) Predicting the distribution of bed material accumulation using river network sediment budgets. Water Resources Research 42: W10419. Wilkinson SN (2008) Testing the capability of a sediment budget model to target remediation measures to reduce suspended-sediment yield. In: Schmidt J, Cochrane T, Phillips C, et al. (eds.) Sediment Dynamics in Changing Environments, IAHS Publication 325, pp. 559–566. Christchurch: IAHS Press. Wilkinson SN, Prosser IP, Rustomji P, and Read AM (2009) Modelling and testing spatially distributed sediment budgets to relate erosion processes to suspended sediment yields. Environmental Modelling and Software 24: 489--501. Wilson CG, Matisoff G, and Whiting PJ (2003) Short-term erosion rates from a 7Be inventory balance. Earth Surface Processes and Landforms 28: 967--977. Wischmeier WH and Smith DD (1978) Predicting Rainfall Erosion Losses. A Guide to Conservation Planning. Agriculture Handbook No. 537. Washington, DC: US Department of Agriculture. Wood PJ and Armitage PD (1997) Biological effects of fine sediment in the lotic environment. Environmental Management 21: 203--217. Woodward JC and Walling DE (2007) Composite suspended sediment particles in river systems: Their incidence, dynamics and physical characteristics. Hydrological Processes 21: 3601--3614. Wynn T and Mostaghimi S (2006) The effects of vegetation and soil type on streambank erosion, southwestern Virginia, USA. Journal of the American Water Resources Association 42: 69--82. Yang CT (1996) Sediment Transport: Theory and Practice. New York: McGraw-Hill. Young RA, Onstad CA, Bosch DD, and Anderson WP (1989) AGNPS – a nonpointsource pollution model for evaluating agricultural watersheds. Journal of Soil and Water Conservation 44: 168--173. Zapata F (2002) Handbook for the Assessment of Soil Erosion and Sedimentation Using Environmental Radionuclides. Dordrecht: Kluwer. Zhang G, Liu Y, Han Y, and Zhang XC (2009a) Sediment transport and soil detachment on steep slopes: I. Transport capacity estimation. Soil Science Society of America Journal 73: 1291--1297. Zhang G, Liu M, Han Y, and Zhang XC (2009b) Sediment transport and soil detachment on steep slopes: II. Sediment feedback relationship. Soil Science Society of America Journal 73: 1298--1304.

Relevant Websites http://www.loicz.org Land Ocean Interactions in the Coastal Zone.

2.13 Field-Based Observation of Hydrological Processes M Weiler, Albert-Ludwigs University of Freiburg, Freiburg, Germany & 2011 Elsevier B.V. All rights reserved.

2.13.1 2.13.1.1 2.13.1.2 2.13.1.3 2.13.2 2.13.2.1 2.13.2.2 2.13.2.3 2.13.3 References

Runoff Generation Processes Early Research Defining the Pathways of Storm Runoff Current Directions Quantifying the Processes Field-Based Observations Quantifying the Processes: Hydrometric Observations Quantifying the Processes: Tracers Conclusion

2.13.1 Runoff Generation Processes 2.13.1.1 Early Research Infiltration was the first process recognized as being significant to runoff generation during a precipitation event. In the early part of the twentieth century, Robert Elymer Horton first described quantitatively the process of infiltration into the soil surface and introduced terminology still used by hydrologists today (Horton, 1933). Following Horton, others recognized that surface runoff was often not the dominant process responsible for increased stream discharge observed during precipitation events. Beginning with Hursh and Brater’s (1941) work at Coweeta (North Carolina, USA), subsurface flow became recognized as a potentially important component of storm flow. Later, studies identified the concepts of variable runoff source areas and the importance of subsurface flow as a contributor to event stream flow response (Betson, 1964; Hewlett and Hibbert, 1963, 1967). Shortly after these developments, old water (pre-event water stored in the watershed as soil water or/and groundwater) was identified as being a significant contributor to runoff (e.g., Pinder and Jones, 1969; Sklash and Farvolden, 1979). Indeed, it is now widely accepted that old water constitutes the majority of stormflow in humid watersheds (e.g., Pearce et al., 1986). However, new water may still be an important contribution to storm runoff in urbanized watersheds or many arid and mountainous watersheds. Horton (1933) defined infiltration as a result of the need to describe the physical process by which water moves into the soil, distinct from other terms sometimes used such as percolation or absorption. Horton defined infiltration capacity as ‘‘the maximum rate at which rain can be absorbed by a given soil at a given condition’’ (Horton, 1933: 453). Horton attributed surface runoff to rainfall intensities that exceeded the infiltration capacity of the soil. This is widely known as Hortonian overland flow or infiltration excess overland flow. However, Horton was not working in forested environments and therefore probably concluded incorrectly that runoff for an individual storm event was mainly or wholly surface runoff. The storm hydrograph response in a forested watershed was shown to consist of subsurface flow and channel

339 339 340 341 342 342 343 345 347 348

precipitation by Hursh and Brater (1941). Engler (1919) already recognized the importance of subsurface stormflow after making detailed measurements of infiltration and physical properties of soil, including porosity, water content, soil texture, and hydraulic conductivity. Subsequently, soil depth, topography, and hydrologic characteristics associated with different elevations were shown to influence peak discharge (Hoover and Hursh, 1943). Hewlett and Hibbert (1963) first recognized the importance of unsaturated flow and concluded that unsaturated flow could not be ignored in hydrograph analysis. Utilizing a concrete trough to observe unsaturated flow at the Coweeta experimental watershed, they coined the term ‘translatory flow’ to describe unsaturated flow and attributed it to the thickening of water films surrounding soil particles, which results in a pulse of water. Substantial amounts of runoff can be generated on areas which have been saturated with water (Dunne and Black, 1970). Furthermore, not only water quantity but also water chemistry and quality are affected by runoff from saturated soils (Mole´nat et al., 2002). Cappus (1960) recognized that saturated areas often occur at specific locations in a watershed which led to the development of the partialcontribution area concept (Betson, 1964). Runoff-generating areas are frequently located in valley floors and on particularly shaped slopes (Amerman, 1965). Even though extent and location of runoff generation areas can vary notably, it has been demonstrated that generally only a small part of a watershed contributes to storm runoff from saturated areas (Ragan, 1968). Cappus (1960) characterized a catchment in terms of its runoff-generating areas. His research showed that it is parts of the watershed, not the whole area, that contribute to runoff (partial contribution area concept). Concerning the involved parts of the catchment, he differentiated between infiltration excess (roads and compacted soil) and saturation areas (valley bottoms). The variable source area concept was developed in the early 1960s and is largely attributed to Betson (1964). He found out that contributions made by different parts of the catchment depend on the precipitation intensity, but the variability is so small that the contributing areas remain almost constant for the duration of one event in his study area. Betson (1964) demonstrated that contributing areas were

339

340

Field-Based Observation of Hydrological Processes

almost constant during heavy rainfalls. For such events, infiltration excess overland flow was observed. Research by Ragan (1968) supports the partial contribution area concept, when he states that only a small part of the basin, less than 3% of the total watershed, contributes an appreciable amount to the storm hydrograph. Amerman (1965) found out that these areas are often located on ridge tops, in valley floors, and on valley slopes. Dunne and Black (1970) collected subsurface flow and saturation overland flow (SOF) with a large trench and could demonstrate that the partial area concept can be extended from infiltration excess overland flow (Betson, 1964) to saturation excess overland flow. Furthermore, Weyman (1970) described that the concept of partial contributing areas, approved by his experiments, can be extended to subsurface flow. He observed that subsurface throughflow and saturation excess surface runoff mainly occurred in specific parts of his watershed. Subsurface stormflow was finally recognized as being an important contributor to event-based stream discharge. In addition, it was previously observed that preferential subsurface stormflow could occur in forest soils (i.e., water moving faster than the soil matrix should allow, typically through some form of soil pipe) (Whipkey, 1965). Whipkey

was the father of trench studies, in which trenches are commonly excavated along the base of a hillslope down to the impermeable layer and flow from the soil horizons is collected and measured. Figure 1 shows an example of how permanent trenches can be built to collect runoff at the soil surface and in the subsurface.

2.13.1.2 Defining the Pathways of Storm Runoff Development of runoff theory proceeded rapidly in the 1960s and 1970s and the studies conducted by Dunne and Black (1970) set a precedent that was rarely exceeded during the next two decades. Dunne and Black used intensive instrumentation across various hillslope types to observe subsurface processes in a wet, mountainous area of Vermont, USA. Three hillslopes consisting of well-drained sandy loams over glacial till, with convex, concave, and straight contours, were instrumented with wells and piezometers to measure water-table elevation and pressure potential, and a nuclear depth probe was used to measure soil moisture along a transect up the middle of each slope. A trench was excavated along the base to the hillslopes to measure runoff at various levels and weirs were installed above and below the reach of river channel running at the

Cross section

Geomembrane (AMERDRAIN 200)

Pipes (runoff collection)

PVC-foil (0.5 mm)

Drainage pipe PVC (perforated)

Geotextile Backfill

Collection sheet (steel) Concrete

Undisturbed soil Figure 1 Cross section and the actual picture (trench is in construction and refilled to the upper observation depth) of a permanent trench for measuring surface runoff and subsurface runoff in two different depths.

Field-Based Observation of Hydrological Processes

base of the study site. Subsurface stormflow was found to occur only during large events and SOF occurred in significant quantities only on the concave (hollow) hillslope. Overland flow occurring on the concave hillslope during large precipitation events was the only flow measured in large enough quantities to account for the measured stream flow. Other important contributions during this decade include Weyman’s (1973) study, which advocated the theory of a saturated wedge developing from riparian margins and moving upslope with increasing precipitation. Groundwater hydrologists, such as Alan Freeze, were developing theories on regional groundwater flow in the early 1970s (e.g., Freeze and Witherspoon, 1967). Freeze et al. (1972) suggested that the majority of event hydrograph response could be attributed to subsurface stormflow. Near the end of the 1970s, a series of studies focused on searching for the mechanism that could explain this process (e.g., Sklash and Farvolden, 1979). Until this time, subsurface flow was considered to be a function of measurable physical properties of soil, namely hydraulic conductivity. However, measurement of soil hydraulic conductivity could not account for the rates of flow necessary to deliver water, via the subsurface, to the stream channel in order to affect the observed stream response. This quandary is resolved by separately considering the flow in the soil matrix described by Darcy’s law (where flow is dependent on soil hydraulic conductivity) and preferential flow pathways via soil pipes and macropores (e.g., Harr, 1977). Studies such as Mosley (1979) showed that rates of subsurface flow could be large enough to account for the observed hydrograph response in a steep headwater catchment with very moist conditions (M8 catchment, Maimai, New Zealand). Large peak flow rates observed at concentrated locations of soil pit faces were found to coincide with stream hydrograph peaks and dye-tracing experiments were used to quantify the rate of water movement through the profile. Mosley’s dye experiments led him to conclude that the majority of the stream flow response was from the contribution of event or new water (water contributed by the current precipitation event). Significant debate over the source of water that generates the storm hydrograph response followed Mosley’s (1979) published work. Mosley believed that the new water was entering macropores and the soil surface, and flowing by lateral macropores without interacting with the soil matrix. Around the time Mosley was working in Maimai, a forested headwater catchment in New Zealand, a new method of examining the source of stormflow was conceived. Pinder and Jones (1969) were the first to use the two-component mixing model to separate event water on the basis of chemical signatures by measuring various ions in rainwater, storm discharge, and stream baseflow. However, it would be almost 20 years before hydrochemical observations were combined with hydrometric observations. Pinder and Jones (1969) concluded that up to 42% of event stream flow might be old water in the Nova Scotia catchment studies. Later, Sklash and Farvolden (1979) measured tritium, oxygen-18 (d18O), and deuterium isotope ratios across various watersheds and concluded that groundwater was the main contributor to the event hydrograph. The process responsible for the transfer of old water to the stream in sufficient quantities to explain their observations was attributed to groundwater ridging near the riparian

341

margins via the rapid conversion of the tension saturated capillary fringe to phreatic water (i.e., saturation occurred soon after the commencement of an event). Gillham (1984) studied these groundwater ridging processes and further realized the importance for stream flow generation in watersheds with extending riparian zones. Following Mosley (1979), Pearce et al. (1986) and Sklash et al. (1986) published the results of studies in which they examined the relative concentrations of chloride, deuterium, and d18O in addition to the electrical conductivity of samples of rainfall, streamflow, and soil water flowing from pit faces in the Maimai catchment, New Zealand. Generally, old (preevent) water and new (event) water were thought to be mixing in the soil profile and then discharging to the stream in a fairly uniform mixture in terms of isotopic and chemical composition (Pearce et al., 1986, Sklash et al., 1986). Groundwater ridging and saturated wedge development from the rapid conversion of tension-saturated zones to positive potentials were cited as the mechanisms responsible for the delivery of stormflow, although hydrometric data were not available to augment these findings. If conversion of tension-saturated zones to positive potentials was occurring, rapid transmission of new water was not needed to explain stormflow. The majority of stormflow would be contributed by old water already stored in the soil and only a small amount of new water would be needed as input (Pearce et al., 1986; Sklash et al., 1986). To solve the old-water new-water dichotomy, a unification of hydrochemical and hydrometric measurements was necessary and McDonnell (1990) did just that in the same catchment (Maimai-M8) as studied by Mosley (1979), Pearce et al. (1986), and Sklash et al. (1986). Using the same soil pits excavated in the previous studies, a combination of isotope and chemical tracing, and an extensive tensiometer network, McDonnell (1990) observed that water tables arising at the soil bedrock interface were not maintained but correlated well with throughflow rates. Soil piping (connection of macropores) was suggested to explain the rapid dissipation of the water table and pore water pressures (McDonnell, 1990). To explain his observations, McDonnell (1990) suggested that rapidly infiltrating new water perched at the impermeable layer and mixed with larger volumes of old water and subsequently drained as the saturated areas in the hillslope expanded creating continuous saturated areas thus affecting rapid stormflow. The formation of these saturated areas, then, largely depends on topography. McGlynn et al. (2002) provided a thorough review of the experiments to date at the Maimai research area. Another explanation of the old-water dominance was provided by the transmissivity feedback mechanism (Rodhe, 1989; Seibert et al., 2003). The process, which was mainly observed in glacial till soils in Scandinavia and Canada, describes the rapid rise of the water table into the more transmissive (permeable) topsoil and the resulting higher subsurface flow. As water is stored below field capacity in the soils, the new water mixes with the old water and the resulting runoff is characterized by dominance in old water (Laudon et al., 2004). Beven (2006) compiled many relevant original papers about runoff generation that are also introduced in this chapter and provided the historical context in detail.

342

Field-Based Observation of Hydrological Processes

2.13.1.3 Current Directions A range of mechanisms facilitates subsurface stormflow and it is useful to separate these into various subareas in order to examine the controls on subsurface stormflow. From the above discussion, we know that increases in subsurface flow are a result of increasing hydraulic gradient, cross-sectional area, rise in the water table into more transmissive soil layers, and the linking of isolated saturated areas across the hillslope (variable source area). To what degree each of these phenomena influences subsurface stormflow appears to depend on various conditions such as antecedent moisture conditions and the morphology of the defined basin or hillslope. Subsurface stormflow initiation and the mechanisms for preferential flow are still debatable issues. Topographic control is being examined in greater detail as it becomes apparent how complex the influence of morphology is on subsurface flow. The aim to better understand the subsurface structures focused on using different geophysical methods. It is also recognized that there are thresholds to the occurrence of subsurface storm flow. Again, these thresholds seem to vary with individual site conditions and it is recognized that the response is nonlinear (Weiler et al., 2005). Research in the last decade focuses on any one or on a number of these issues, either directly or indirectly. Preferential flow has been shown to be important for both flow initiation and rapid lateral transport of water downslope (Mosley, 1979; McDonnell, 1990). In the first case, preferential infiltration has been identified as being significant enough for rapid development of saturated areas and water tables in more permeable soils (e.g., De Vries and Chow, 1978; McDonnell, 1990; Weiler and Naef 2003). In order for rapid stream flow response to be facilitated by subsurface storm flow, water must infiltrate and move down slope at rates greater than the estimates based on soil matrix properties would predict. Preferential flow can occur via macropores, cracks and soil pipes, and in areas of higher permeability in the soil, including highly permeable layers (Bonell, 1998; McGlynn et al., 2002). It has also been recognized that the permeability of macropore and crack walls may be lower than that of the soil matrix, which would allow for rapid unimpeded flow once water fills these conduits (Calver and Cammeraat, 1993; McDonnell, 1990). Rates of pipe flow are largely determined by their diameter, and it has been recognized that there are certain precipitation thresholds that must be exceeded before pipe flow will dominate the subsurface flow (Weiler and McDonnell, 2007). In Bonell’s (1998) review of runoff generation, he states ‘‘reconciling their [soil pipes] hydrochemistry coupled with the need for more sophisticated hydrometric studies to address the pipeflow issue, stands out as one of the principle research challenges connected with storm hydrograph separations.’’ We are still lacking the knowledge to address these pipe flow issues; however, there have been several attempts to perform more sophisticated tracer-based and hydrometric studies (Anderson et al. 2009a, 2009b; Anderson et al., 2010) or to use other approaches to understand the flow pathways along the soil bedrock interface (Graham, 2009). Geophysical methods offer the opportunity to rapidly collect subsurface information in a noninvasive or minimally

invasive manner, which may be a key information to see flow pathways and hence understand runoff generation processes. These techniques are sensitive to different physical properties (e.g., magnetic, elastic, and electrical properties) of subsurface materials. In near-surface environments, techniques such as ground-penetrating radar (GPR), electrical resistivity tomography (ERT), electromagnetic induction surveying, or different seismic methods have been proven to provide valuable data for a variety of applications (Butler, 2005). Much of the related work and progress made in the field of hydrology within the past decade is documented in Rubin and Hubbard (2005) and Vereecken et al. (2006). Especially, GPR and seismic methods may provide structural information to characterize the subsurface. In sandy sediments, GPR allows imaging of subsurface geometries up to depths of B10 m with a resolution at the dm-to-m scale (Beres et al., 1999). However, translation of geophysical observations into the relevant subsurface state and properties, such as moisture content, or hydraulic conductivity, remains a challenging task. The relations between geophysical and the hydrological target variables are usually complex, nonunique, and site specific (Scho¨n, 1998). Uncertainty in analyzing and interpreting geophysical data may be reduced by multimethod approaches. For example, time-lapse imaging of subsurface flow processes is possible by combining geophysical techniques with artificial tracers, as for instance, salt tracers in the unsaturated zone imaged by ERT or cross-hole radar attenuation tomography (Johnson et al., 2007). Kienzler (2007) developed a nice example of the potential combining artificial salt tracer injection to visualize preferential lateral flow pathways (Figure 2). Another interesting idea relies on extracting structural information such as correlation lengths from geophysical images. The basic assumption is that the statistical properties of geophysical structures and parameter variations, respectively, can be used as a proxy for the statistical characteristics of the target hydrological parameter field. Until now, this idea has primarily been used with GPR reflection data (e.g., Knight et al., 2007); more experience, using realistic synthetic scenarios and other geophysical data, is clearly needed. Nevertheless, such concepts can be extremely useful to investigate and characterize geophysical data and hydrological systems to understand runoff generation mechanisms. Until today, experimental studies to understand runoff generation processes and flow pathways have been conducted basically worldwide in various climatic and geological settings, leading to an advanced and refined perception of rainfall runoff processes (Graham, 2009; Kienzler, 2007; Scherrer et al., 2006). Despite the large variety of observed flow processes, one commonality is the strong nonlinearity of hillslope response to rainfall (Tromp-van Meerveld and Weiler, 2008). One possibility to explain this behavior is the sudden connection of different areas in the watershed that are locally generating runoff (either at the soil surface or as subsurface runoff) by different flow pathways (macropores, pipes, and channels). Tromp-van Meerveld and McDonnell (2006) proposed the so-called fill-and-spill mechanisms to explain the threshold behavior at the Panola watershed experiment, USA. Bachmair and Weiler (2010) extended this concept to the connect-and-react hypothesis to generally describe the sudden connection of runoff generation processes in the watershed

Field-Based Observation of Hydrological Processes

343

Tracer line source 15% 5

15

m

15 m

25 m m

2.5 m

30

m

10 m

40 m Figure 2 Subsurface flow paths detected with ground-penetrating radar (GPR) at the Koblenz experimental slope, Switzerland. The profiles have been taken before as well as 4 h after the start of a salt tracer injection. Displayed are the differences between these two measurements, large differences (yellow–red colors) indicate flow path locations. From Kienzler P (2007) Experimental Study of Subsurface Stormflow Formation. Combining Tracer, Hydrometric and Geophysical Techniques. Diss. ETH Nr. 17330, Eidgeno¨ssische Technische Hochschule (ETH), Zu¨rich, Switzerland.

and the resulting nonlinearity or threshold behavior in the runoff response. However, explaining and predicating the threshold behavior continue to be a challenging task (Zehe and Sivapalan, 2009).

2.13.2 Quantifying the Processes 2.13.2.1 Field-Based Observations Many different techniques and methods have been developed in the last 100 years to directly or indirectly observe runoff generation processes in the field. Many of these methods require only simple instruments or observations, but their power can be increased if enough spatial and temporal explicit observations (e.g., Trubilowicz et al., 2009) are being taken to understand the complex spatial–temporal processes during storm runoff generation. In Table 1, measurement methods are listed together with the spatial scale and the processes which can be observed. Many of these methods are explained in more detail in the following chapter by showing the potential and issues within sample application and past studies. The references listed in Table 1 are only possible sources of information.

2.13.2.2 Quantifying the Processes: Hydrometric Observations The contribution of SOF to storm runoff has been repeatedly studied since the first work of Dunne and Black (1970). A central aspect of the SOF estimation is the delineation of the contributing saturated areas. Soil saturation can be detected with remote sensing (e.g., Mohanty and Skaggs, 2001). However, saturation patterns cannot be captured efficiently with remote sensing in dense forests (e.g., Kite and Pietroniro, 1996). Soil saturation under forests was therefore mapped

based on soil and vegetation characteristics in order to evaluate the models for saturation predictions (Gu¨ntner et al., 2004). However, only few studies have evaluated the mapping criteria with direct saturation measurements (Rosin et al., 2009). Moreover, saturated areas have been mapped and modeled in a single climate (e.g., Blazkova et al., 2002), but only few investigations have been made to compare mapping and modeling in different climates, mostly using different data sources (Me´rot et al., 2003). It still seems to be important to improve our methods to better monitor the spatial dynamics and connectivity of saturated areas in particular in watersheds that are dominated by SOF. The observations in the Miniflet catchment in Sweden (Myrabo, 1997) are a nice example of the space–time variations of water-table response and saturation areas due to subsurface flow and flow accumulation (Figure 3). Observing subsurface runoff is still a challenge. Woods and Rowe (1996) excavated a trench in the Maimai M8 catchment and measured flow rate and quantity in a series of troughs along a trench face coupled with tensiometer and piezometer measurements. They found that bedrock topography was responsible for flow routing as saturated areas developed in hollows and converged as ribbons of concentrated flow. Flow volumes were highly variable and were not well predicted by the surface topography and flow accumulation. In a follow-up study, Woods et al. (1997) concluded that variability in runoff depends on both topography and soil moisture conditions. Freer et al. (1997) excavated another similar trench in the Panola watershed close to Atlanta, USA. They determined that bedrock topography improved predictions as subsurface flow routing is dependent on the morphology of the impeding layer or bedrock. Freer et al. (1997) also noted that antecedent soil moisture conditions were a significant control on the occurrence of saturation.

344

Field-Based Observation of Hydrological Processes

Table 1

Measurement methods to observe runoff generation processes at different scales

Measurement

Spatial scales

Processess

Issues

References

Infiltrometer Sprinkling experiments

Plot, hillslope Plot, hillslope

Infiltration Infiltration, overland, and subsurface flow

Mertens et al. (2002) Weiler and Naef (2003)

Soil moisture measurement

Point

Percolation, saturation, evapotranspiration

Lysimeter

Plot

Soil water potential

Point

Percolation, evapotranspiration, groundwater recharge Flow direction, saturation

Boundary conditions Drop size distribution and intensity. Trench is necessary for subsurface runoff Point measurement, high spatial, and temporal variability High costs, disturbed soil monolith

Anderson and Burt (1978)

Dye staining experiments

Plot, hillslope

Flow pathways, preferential flow

Artificial tracer experiments (1-D)

Plot, hillslope

Flow velocity, preferential flow

Artificial tracer experiments (2-D)

Plot

Trenching

Hillslope

Flow velocity, flow direction, preferential flow Subsurface flow

Point measurement, high spatial, and temporal variability Destructive but very visually informative, artificial experiment Sampling method influences results (destructive, flow, suction cups, etc.) Stationary conditions

Woods and Rowe (1996)

Chemical hydrograph separation Mapping of saturated areas Saturation collectors

Hillslope, catchment

Interruption of low pathways, artificial drainage Concentration of end members Depends on climate Point observations

Hillslope, catchment Plot, hillslopes

Water sources (spatially or temporally) Potential saturation Saturation

Utilizing smaller troughs to collect discharge from the soil profile of a hillslope and a network of piezometers, Hutchinson and Moore (2000) examined how throughflow is related to surface topography and basal till/confining layer. The troughs were oriented such that the hillslope was divided up into units so they could be compared. They found that at the lowest flow, the subsurface flow distribution was correlated well with upslope contributing area as calculated from the basal till layer topography. However, at the highest flows, subsurface flow was more closely related to the contributing area of the surface topography. In other words, the saturated area, or water table, shifted from being parallel to the confining layer to being parallel to the surface. Moreover, they observed macropores which can deliver enough discharge to negate the topography as a control on subsurface flow. It was suggested that macropores can route water laterally which questions the validity of models which use topographic controls to predict subsurface flow (Hutchinson and Moore, 2000). Finally, it is of great relevance to this study that topographic models usually assume a quasi-steady-state for throughflow, which Hutchinson and Moore (2000) did not find appropriate at their site. Scherrer et al. (2006) surpassed, in number of trenches, all other experimental trench studies to understand runoff generation processes. They performed sprinkling experiments at 60 m2 hillslopes and also measured, in addition to surface runoff, subsurface runoff in a 6-m-wide trench (Figure 4).

Jost et al. (2005)

Scanlon et al. (2002), Aboukhaled et al. (1982)

Weiler and Naef (2003), Anderson et al. (2005) Weiler et al. (1998), McGuire et al. (2005)

Roth et al. (1991)

Hooper and Shoemaker (1986) Merot et al. (1995) Rosin et al. (2009)

In order to quantify the internal processes, they also instrumented the slope with many time-domain reflectometer (TDR) probes, tensiometers, and piezometers. This combined hydrometric observation setup allowed a detailed description of flow processes within the hillslope, resulting in runoff generation processes. Tromp-van Meerveld and Weiler (2008) also instrumented a hillslope in detail, similar as Scherrer et al. (2006) did, but with the focus of a longer-term observation to explore the wet and dry season and its processes. In particular, the combination of soil moisture measurements, maximum water-table observations, and sap-flow measurements enabled them to observe the spatial–temporal patterns of flow processes in the soil and transpiration processes. Jost et al. (2005) presented another approach, focusing on soil moisture measurements with TDRs to observe the spatial– temporal patterns of transpiration, recharge, and soil moisture storage. They measured soil moisture dynamic at 198 locations in a 0.5-ha forest site and were able to use these observations together with geostatistical methods to predict the patterns of water fluxes at the soil vegetation atmosphere. In future, the potential to use a large number of sensors to better observe the spatial–temporal patterns of fluxes and storage changes in the saturated and unsaturated zone will increase when using wireless sensor techniques. The wireless sensors will only provide a more efficient way to collect the data, but the development of cost-effective sensors to observe the relevant fluxes states the need to go hand in hand. Trubilowicz

Field-Based Observation of Hydrological Processes Q = 0.1 mm h−1

345

Q = 0.54 mm h−1

Q = 0.61 mm h−1

Q = 4.9 mm h−1

Q = 6.8 mm h−1 cm −10 0 10 20 30 40 50 60 m 0

50

100

Figure 3 Water table variations and saturation (blue areas) at the Minifelt watershed in Sweden for different stream runoff conditions. From Myrabo S (1997) Temporal and spatial scale of response area and groundwater variation in till. Hydrological Processes 11: 1861–1880.

et al. (2009) tested the potential of low-cost, low-power wireless sensor networks (mote networks) for monitoring throughfall, temperature, humidity, soil water content, and water-table dynamics using 41 motes in a small forested watershed (Figure 5). They found that while motes gave the ability to monitor a catchment at resolution levels that were previously impossible, they still need to evolve into an easierto-use, more reliable platform before they can replace traditional data collection methods.

2.13.2.3 Quantifying the Processes: Tracers Hydrologists have used tracers to study water movement for several decades and there are a number of different tracers available some being better suited to specific applications (see also Chapter 2.09 Tracer Hydrology). There are two basic types of tracers, the one considered natural and the other

artificial. Natural tracers are ones that can be found in the natural environment such as oxygen isotope 18, tritium, or weathered materials like silicates. These can be measured from water samples in the soil, precipitation, groundwater, and the stream. Artificial tracers are applied to the system; this includes various types of dyes and anions (e.g., chloride). Of course, chloride is naturally occurring but it is often applied in much larger quantities so that it overrides the natural background concentrations. Natural and artificial tracers both have advantages and disadvantages and neither are necessarily better, even of these two types, one tracer may be completely inappropriate where another is very useful. For example, Rhodamine WT is not a very useful soil–water tracer in lab application, whereas Lissamine FF is (Trudgill, 1987). The first tracers to be used in runoff generation studies were naturally occurring. By measuring the relative concentrations in the different sources (soil water, precipitation, and

346

Field-Based Observation of Hydrological Processes 4m

Tensiometer

Sprinkler

15

m

TDR−probes (installed vertically) Humus

Piezometer A

Collector tray for overland flows Collection of subsurface flow via ditch

Mineral soil

B1 B2/Cv Trench

TDR probes (installed horizontly)

Rock

α

Slope angle: 15−55%

Figure 4 Experimental setup to study runoff generation processes at the hillslope scale. From Scherrer S (1997) Abflussbildung bei Starkniederschla¨gen, Identifikationvon Abflussprozessen mittels ku¨nstlicher Niederschla¨ge. Versuchsanstalt fu¨rWasserbau, Hydrologie und Glaziologie der ETH Zu¨rich, Zu¨rich, 147pp.

stream), researchers could separate the storm hydrograph into different component sources (e.g., Pinder and Jones, 1969). For example, McGuire et al. (2005) used d18O to measure the residence time of water falling on eight different catchments in the HJ Andrews experimental forest, Coastal Range Oregon. They measured d18O in the various input sources and in the stream water to determine the source of the stormflow and how long it has been in the catchment. By accounting for variations in isotope ratios with changes in elevation and also values for snowpack melt water, they were able to determine residence times for the catchments and compare how it varied across scale. Interestingly, residence time was not dependent on scale but was more closely related to simple topographic measures such as median flow path length and gradient. Other types of natural tracing include measurements of silica content and alkalinity. For example, Soulsby et al. (2004) used alkalinity and silica content measurements to determine the main sources of runoff in the Scottish highlands. Natural tracers lend themselves well to catchment scale and larger studies while artificial tracers are convenient for hillslope and plot scale applications. The main drawback of natural tracers is the uncertainty associated with characterizing the sources (e.g., Didszun and Uhlenbrook, 2008). We know that chemical signatures are variable in both space and time. The chemical signature of soil moisture varies spatially and temporally depending on the length of time that moisture has resided in the soil. In turn, residence time of soil water is dependent on the length of time since the last precipitation event and the size of that event. This dependence makes it difficult to account for the soil water signature. In addition, the influence of interception and the

spatial variation in chemical composition of precipitation at the catchment scale has received little attention to date. Most hydrochemical studies have focused on the hillslope scale and often use only a single rain gauge to determine the chemical signature of rainwater. As McGuire et al. (2005) pointed out, the isotope signature of rainfall varies with elevation. In addition, deposition of minerals and soil physical properties can vary over small spatial scales, which will alter the chemical signature of water flowing through various areas. It would be nearly impossible, or at least very labor intensive, to account for such variations. Nonetheless, naturally occurring tracers continue to be used and do provide certain advantages over artificial tracers, mainly that they can be used on a larger scale and are ubiquitous. Artificial tracers overcome the uncertainty associated with characterizing the naturally occurring tracers, in that we can control when, where, and how much to apply. Artificial tracers have been used around for some time as early as the late 1960s (e.g., radioactive tracers used by Pilgrim (1966)). Exploration of the utility of tracers in hydrological study continued through the 1970s (e.g., Pilgrim and Huff, 1978; Smart and Laidlaw, 1977). Pilgrim and Huff (1978) demonstrated the usefulness of artificial tracers for monitoring water movement in the subsurface and observed irregular patterns of movement despite a uniform surface. More recently, dye experiments have been used to examine infiltration in greater detail (Weiler and Flu¨hler, 2004; Weiler and Naef, 2003). Weiler and Flu¨hler (2004) used simulated rainfall with Brilliant Blue dye followed by soil pit excavation and image analysis to examine infiltration (example images in Figure 6). Extended vertically stained sections of macropore

Field-Based Observation of Hydrological Processes

347

2 1

PE

3

6 5

4

7

Figure 5 Wireless sensor network with measurement station ready for deployment: (1) pressure transducer; (2) mote with power supply; (3) tipping bucket; (4) soil moisture probe; (5) air temperature/humidity probe with radiation shield; (6) ground temperature thermistor; (7) overland flow weir. From Trubilowicz J, Cai K, and Weiler M (2009) Viability of motes for hydrological measurement. Water Resources Research 45: W00D22 (doi:10.1029/2008WR007046).

flow were observed which initiated close to the soil surface. This is in accordance with the conclusions of McDonnell (1990). Weiler and Naef (2003) concluded that although macropores make up a much smaller fraction of the total porosity (o1%), they account for the majority of saturated flow and preclude the use of Darcy’s law or the Richards’ equation to predict flow rates. However, dye-staining experiments at the hillslope scale seem to be possible. Anderson et al. (2009a) were able to reconstruct lateral preferential flow networks by staining a 30-m-long hillslope and excavating the pathways. The experiment revealed that larger contributing areas coincided with highly developed and hydraulically connected preferential flow paths that had flow with little interaction with the surrounding soil matrix. They found evidence of subsurface erosion and deposition of soil and organic material laterally and vertically within the soil (see detailed information about the experimental setup, results, and interpretation in the electronic supplements). These dyestaining results are important because they add to the understanding of the runoff generation, infiltration, solute transport, and slope stability of preferential flow-dominated soils. Artificial tracers have also been used at the hillslope scale often injected through piezometers at specific depth (e.g., Talamba et al. (2000) or Weiler et al. (1998)) or done as a line application at the top of a hillslope or hillslope plot (Weiler et al., 1998). It seems that the time has come for artificial tracers to be tested at the small catchment scale. Application is

the biggest limitation with artificial tracers. It is either labor intensive or expensive in the case of sprinkler systems. Rodhe et al. (1996) and Lange et al. (1996) conducted studies in catchment which have been covered below the canopy so that chemical signature of the input water could be controlled, an impressive undertaking. Rodhe et al. (1996) used d18O ratios while Lange et al. (1996) used LiBr as their tracers. Of greatest interest are the results of Lange et al. (1996) who had low recovery and concluded that residence time was long enough to permit equilibrium exchange between the soil water and soil matrix. They believed that hydrochemical processes related to catchment runoff are underestimated because they are often based on soil column studies that do not account for lateral movement. While artificial tracers have been used in hydrology for some time, their usefulness as a tool for studying runoff generation has not been extensively explored. It could be possible in future to use artificial tracer more extensively if instrumentation techniques to detect the tracers are becoming better and smaller amounts need to be applied to observe the movement of tracers in the watershed.

2.13.3 Conclusion In the 1960s and 1970s, the focus was on observing hydrological processes in the field. There have been many groundbreaking studies and experiments observing the

348

Field-Based Observation of Hydrological Processes

Figure 6 Examples of four dye patterns in forest soils after sprinkling 60 mm dyed water in 3 h. The soil types range from sandy soils to loamy soils; however, macropores, root channels, and hydrophobicity at the soil surface are more relevant for generating different infiltration patterns than soil type.

spatial–temporal dynamics of water and solute fluxes on the plot, hillslope, and catchment to understand the interplay of different hydrological processes. Unfortunately, the focus on hydrology shifted toward modeling and computer simulations in the last 20 years. Field-based observations are demanding and time consuming and the rewards are often not as pronounced compared to developing or applying a new hydrological model at the watershed or even at the continental scale. However, we have forgotten many important lessons that we have learned about the functioning of watersheds. Our current hydrological models are all very similar and most of them do not incorporate the hydrological processes and flow pathways that have been observed in the field. As pleaded, for example, by Weiler and McDonnell (2004), a new area of more connections and discussions between field hydrologists and

hydrological modeler is needed to overcome the current deficit in hydrological model development. It is also believed that new techniques and new sensors need to be developed, tested, and implemented into field-based observation to enhance the possibility and understanding of processes, in particular flow processes in the subsurface and surface–groundwater interaction.

References Aboukhaled A, Alfaro A, and Smith M (1982) Lysimeters, Irrigation and Drainage, Paper 39, 68pp. Rome: FAO. Amerman CR (1965) The use of unit-source watershed data for runoff prediction. Water Resources Research 1(4): 499--507.

Field-Based Observation of Hydrological Processes Anderson AE, Weiler M, Alila Y, and Hudson RO (2009a) Dye staining and excavation of a lateral preferential flow network. Hydrology and Earth System Sciences 13: 935--944. Anderson AE, Weiler M, Alila Y, and Hudson RO (2009b) Subsurface flow velocities in a hillslope with lateral preferential flow. Water Resources Research 45: W11407 (doi:10.1029/2008WR007121). Anderson AE, Weiler M, Alila Y, and Hudson RO (2010) Water table response in zones of a watershed with lateral preferential flow as a first order control on subsurface flow. Hydrological Processes (in press). Anderson MG and Burt TP (1978) The role of topography in controlling throughflow generation. Earth Surfaces Processes and Landforms 3: 331--334. Bachmair S and Weiler M (2010) New dimensions of hillslope hydrology. In: Levia D, Carlyle-Moses D, and Tanaka T (eds.) Forest Hydrology and Biogeochemistry: Synthesis of Research and Future Directions. New York, NY: Springer. Beres M, Huggenberger P, Green AG, and Horstmeyer H (1999) A study of glaciofluvial architectures using two- and three-dimensional georadar methods. Sedimentary Geology 129: 1--24. Betson RP (1964) What is watershed runoff? Journal of Geophysical Research 69(8): 1541--1551. Beven K (2006) Streamflow generation processes. In: McDonnell JJ (ed.) IAHS Benchmark Papers in Hydrology Series, 432pp. Wallingford: IAHS. Blazkova S, Beven KJ, and Kulasova A (2002) On constraining TOPMODEL hydrograph simulations using partial saturated area information. Hydrological Processes 16(2): 441--458. Bonell M (1998) Selected challenges in runoff generation research in forests from the hillslope to headwater drainage basin scale. Journal of the American Water Resources Association 34(4): 765--786. Butler DK (ed.) (2005) Near-Surface Geophysics. Tulsa, OK: Society of Exploration Geophysicists. Calver A and Cammeraat LH (1993) Testing a physically based runoff model against field observations on a Luxembourg hillslope. Catena 20: 273--288. Cappus P (1960) Etude des lois de l’eAˆ coulement, application au calcul et a la prevision des de bits. La Houille Blanche A 493--520. De Vries J and Chow TL (1978) Hydrologic behavior in a forested mountain soil in coastal British Columbia. Water Resources Research 14(5): 935--942. Didszun J and Uhlenbrook S (2008) Scaling of dominant runoff generation processes: Nested catchments approach using multiple tracers. Water Resources Research 44: W02410 (doi:101029/2006WR005242). Dunne T and Black RD (1970) An experimental investigation of runoff production in permeable soils. Water Resources Research 6(2): 478--490. Engler A (1919) Untersuchungen u¨ber den Einfluss des Waldes auf den Stand der Gewa¨sser. Mitteilung der Schweizerischen Anstalt fu¨r fortsliches Versuchswesen 12: 1--626. Freer J, McDonnell JJ, Beven KJ, et al. (1997) Topographic controls on subsurface storm flow at the hillslope scale for two hydrologically distinct small catchments. Hydrological Processes 11(9): 1347--1352. Freeze AR, McDonnell JJ, Beven KJ, et al. (1972) Role of subsurface flow in generating surface runoff: 2 Upstream source areas. Water Resources Research 8(5): 1272--1283. Freeze AR and Witherspoon PA (1967) Theoretical analysis of regional groundwater flow: 2. Effect of water-table configuration and subsurface permeability variation. Water Resources Research 3: 623--634. Gillham RW (1984) The capillary fringe and its effect on water-table response. Journal of Hydrology 67: 307--324. Graham C (2009) A Macroscale Measurement and Modeling Approach to Improve Understanding of the Hydrology of Steep, Forested Hillslopes. PhD Thesis, Oregon State University, USA. Gu¨ntner A, Seibert J, and Uhlenbrook S (2004) Modeling spatial patterns of saturated areas: An evaluation of different terrain indices. Water Resources Research 40: W05114 (doi:10.1029/2003wr002864). Harr RD (1977) Water flux in soil and subsoil on a steep forested slope. Journal of Hydrology 33: 37--58. Hewlett JD and Hibbert AR (1963) Moisture and energy conditions within a sloping soil mass during drainage. Journal of Geophysical Research 68: 1081--1087. Hewlett JD and Hibbert AR (1967) Factors affecting the response of small watersheds to precipitation in humid areas. In: Sopper WE and Lull HW (eds.) Forest Hydrology, pp. 275--291. New York, NY: Pergamon. Hooper RP and Shoemaker CA (1986) A comparison of chemical and isotopic hydrograph separation. Water Resources Research 22(10): 1444--1454. Hoover MD and Hursh CR (1943) Influence of topography and soil depth on runoff from forest land. Transactions of the American Geophysical Union 2: 693--698.

349

Horton RE (1933) The role of infiltration in the hydrological cycle. In: Transactions of the American Geophysical Union, Fourteenth Annual Meeting, pp. 445–460. Washington, DC. Hursh CR and Brater EF (1941) Separating storm-hydrographs from small drainageareas into surface- and subsurface-flow. Transactions of the American Geophysical Union 3: 863--871. Hutchinson DG and Moore RD (2000) Throughflow variability on a forested slope underlain by compacted glacial till. Hydrological Processes 14(10): 1751--1766. Johnson TC, Routh PS, Barrash W, and Knoll MD (2007) A field comparison of Fresnel zone and ray-based GPR attenuation-difference tomography for time-lapse imaging of electrically anomalous tracer or contaminant plumes. Geophysics 72: G21--G29. Jost G, Heuvelink GBM, and Papritz A (2005) Analysing the space-time distribution of soil water storage of a forest ecosystem using spatio-temporal kriging. Geoderma 128(3–4): 258--273. Kienzler P (2007) Experimental Study of Subsurface Stormflow Formation. Combining Tracer, Hydrometric and Geophysical Techniques. Diss. ETH Nr. 17330, Eidgeno¨ssische Technische Hochschule (ETH), Zu¨rich, Switzerland. Kite GW and Pietroniro A (1996) Remote sensing applications in hydrological modelling. Hydrological Sciences 563--591. Knight R, Irving J, Tercier P, Freeman G, Murray C, and Rockhold M (2007) A comparison of the use of radar images and neutron probe data to determine the horizontal correlation length of water content. In: Hyndman DW, Day-Lewis FD, and Singha K (eds.) Subsurface Hydrology: Data Integration for Properties and Processes, Geophysical Monograph Series, vol. 171, pp. 31–44. Washington, DC: AGU. Lange H, Lischeid G, Hoch R, and Hauhs M (1996) Water flow paths and residence times in a small headwater catchment in Ga˚rdsjo¨n, Sweden, during steady state storm flow conditions. Water Resources Research 32: 1689--1698. Laudon H, Seibert J, Kohler S, and Bishop K (2004) Hydrological flow paths during snowmelt: Congruence between hydrometric measurements and oxygen 18 in meltwater, soil water, and runoff. Water Resources Research 40: W03102 (doi:10.1029/2003WR002455). McDonnell JJ (1990) A rationale for old water discharge through macropores in a steep, humid catchment. Water Resources Research 26(11): 2821--2832. McGlynn BL, McDonnell JJ, and Brammer DD (2002) A review of the evolving perceptual model of hillslope flowpaths at the Maimai catchments, New Zealand. Journal of Hydrology 257: 1--26. McGuire KJ, McDonnell M, Weiler M, et al. (2005) The role of topography on catchment-scale water residence time. Water Resources Research 41: W05002. Merot Ph, Ezzehar B, Walter C, and Aurousseau P (1995) Mapping waterlogging of soils using digital terrain models. Hydological Processes 9: 27--34. Me´rot P, Squividant H, Aurousseau P, et al. (2003) Testing a climato-topographic index for predicting wetlands distribution along an European climate gradient. Ecological Modelling 163(1–2): 51--71. Mertens J, Jacques D, Vanderborght J, and Feyen J (2002) Characterisation of the field-saturated hydraulic conductivity on a hillslope: In situ single ring pressure infiltrometer measurements. Journal of Hydrology 263(1–4): 217--229. Mohanty BP and Skaggs TH (2001) Spatio-temporal evolution and time-stable characteristics of soil moisture within remote sensing footprints with varying soil, slope, and vegetation. Advances in Water Resources 24(9–10): 1051--1067. Mole´nat J, Durand P, Gascuel-Odoux C, Davy P, and Gruau G (2002) Mechanisms of nitrate transfer from soil to stream in an agricultural watershed of French Brittany. Water, Air, and Soil Pollution 133(1–4): 161--183. Mosley MP (1979) Streamflow generation in a forested watershed. Water Resources Research 15: 795--806. Myrabo S (1997) Temporal and spatial scale of response area and groundwater variation in till. Hydrological Processes 11: 1861--1880. Pearce AJ, Stewart MK, and Sklash MG (1986) Storm runoff generation in humid headwater catchments: 1. Where does the water come from? Water Resources Research 22: 1263--1272. Pilgrim DH (1966) Radioactive tracing of storm runoff on a small catchment. Journal of Hydrology 4: 289--326. Pilgrim DH and Huff DD (1978) A field evaluation of subsurface and surface runoff: I. Tracer studies. Journal of Hydrology 38: 299--318. Pinder GF and Jones JF (1969) Determination of the groundwater component of peak discharge for the chemistry of total runoff. Water Resources Research 5(2): 438--445. Ragan RM (1968) An experimental investigation of partial area contributions. In: Proceedings of the Berne Symposium, IAHS Publ. 76, pp. 241–249. Rodhe A (1989) On the generation of stream runoff in till soils. Nordic Hydrology 20: 1--8.

350

Field-Based Observation of Hydrological Processes

Rodhe A, Nyberg L, and Bishop K (1996) Transit times for water in a small till catchment from a step shift in the oxygen 18 content of the water input. Water Resources Research 32: 3497--3511. Rosin K, Weiler M, and Smith R (2009) Evaluating soil saturation models in forests in different climates. Journal of Hydrology (in review). Roth K, Jury WA, Flu¨hler H, and Attinger W (1991) Transport of chloride through an unsaturated field soil. Water Resources Research 27(10): 2533--2541. Rubin Y and Hubbard SS (2005) Hydrogeophysics. Dordrecht: Springer. Scanlon BR, Healy RW, and Cook PG (2002) Choosing appropriate techniques for quantifying groundwater recharge. Hydrogeology Journal 10(1): 18--39. Scherrer S (1997) Abflussbildung bei Starkniederschla¨gen, Identifikationvon Abflussprozessen mittels ku¨nstlicher Niederschla¨ge. Versuchsanstalt fu¨rWasserbau, Hydrologie und Glaziologie der ETH Zu¨rich, Zu¨rich, 147pp. Scherrer S, Naef F, Faeh AO, and Cordery I (2006) Formation of runoff at the hill-slope scale during intense precipitation. Hydrology and Earth System Sciences 11(2): 907--922. Scho¨n JH (1998) Physical Properties of Rocks: Fundamentals and Principles of Petrophysics. Oxford: Pergamon. Seibert J, Bishop K, Rodhe A, and McDonnell JJ (2003) Groundwater dynamics along a hillslope: A test of the steady state hypothesis. Water Resources Research 39(1). 2-1–2-9 (doi:1029/2002WR001404 2003). Sklash MG, Beven KJ, Gilman K, and Darling WG (1996) Isotope studies of pipeflow at Plynlimon, Wales, UK. Hydrological Processes 10(7): 921--944. Sklash MG and Farvolden RN (1979) The role of groundwater in storm runoff. Journal of Hydrology 43: 45--65. Sklash MG, Stewart MK, and Pearce AJ (1986) Storm runoff generation in humid headwater catchments: 2. A case study of hillslope and low-order stream response. Water Resources Research 22(8): 1273--1282. Smart PL and Laidlaw IMS (1977) An evaluation of some fluorescent dyes for water tracing. Water Resources Research 13: 15--33. Soulsby C, Rodgers PJ, Petry J, Hannah DM, Malcolm IA, and Dunn SM (2004) Using tracers to upscale flow path understanding in mesoscale mountainous catchments: Two examples from Scotland. Journal of Hydrology 291: 174--196. Talamba D, Joerin C, and Musy A (2000) Study of subsurface flow using environmental and artificial tracers: The Haute-Mentue case, Switzerland. In: Tracers and Modelling in Hydrogeology, IAHS Publication No. 262, pp. 559–264. Tromp-van Meerveld HJ and McDonnell JJ (2006) Threshold relations in subsurface stormflow: 2. The fill and spill hypothesis. Water Resources Research 42: W02411 (doi:10.1029/2004WR003800).

Tromp-van Meerveld I and Weiler M (2008) Hillslope dynamics modeled with increasing complexity. Journal of Hydrology 361(1–2): 24--40. Trubilowicz J, Cai K, and Weiler M (2009) Viability of motes for hydrological measurement. Water Resources Research 45: W00D22 (doi:10.1029/ 2008WR007046). Trudgill ST (1987) Soil water dye tracing, with special reference to the use of rhodamine WT, Lissamine FF and amino G acid. Hydrological Processes 1: 149--170. Vereecken H, Binley A, Cassiani G, Revil A, and Titov K (2006) Applied Hydrogeophysics. Dordrecht: Springer. Weiler M and Flu¨hler H (2004) Inferring flow types from dye patterns in macroporous soils. Geoderma 120(1–2): 137--153. Weiler M and McDonnell J (2007) Conceptualizing lateral preferential flow and flow networks and simulating the effects on gauged and ungauged hillslopes. Water Resources Research 43: W03403 (doi:10.1029/2006WR004867). Weiler M, McDonnell J, Tromp-van Meerveld HJ, and Uchida T (2005) Subsurface stormflow. In: Anderson MG and Jeffrey JJ (eds.) Encyclopedia of Hydrological Sciences, vol. 3, ch. 112, pp. 1719–1732. Chichester: Wiley. Weiler M and Naef F (2003) An experimental tracer study of the role of macropores in infiltration in grassland soils. Hydrological Processes 17(2): 477--493. Weiler M, Naef F, Leibundgut C (1998) Study of runoff generation on hillslopes using tracer experiments and physically based numerical model. IAHS Publication No. 248, pp. 353–360 Weyman DR (1970) Throughfall on hillslopes and its relation to the streamhydrograph. Bulletin of the International Association of the Scientific Hydrology 15: 23--25. Weyman DR (1973) Measurements of the downslope flow of water in a soil. Journal of Hydrology 20: 267--288. Whipkey RZ (1965) Subsurface storm flow from forested slopes. Bulletin of the International Association of the Scientific Hydrology 2: 74--85. Woods R and Rowe L (1996) The changing spatial variability of subsurface flow across a hillside. Journal of Hydrology (NZ) 35(1): 51--86. Woods RA, Sivapalan M, and Robinson JS (1997) Modelling the spatial variability of subsurface runoff using a topographic index. Water Resources Research 31: 2097--2110. Zehe E and Sivapalan M (2009) Threshold behavior in hydrological systems as (human) geo-ecosystems: Manifestations, controls, implications. Hydrology and Earth System Sciences 13: 1273--1297.

2.14 Observation of Hydrological Processes Using Remote Sensing Z Su, University of Twente, Enschede, The Netherlands RA Roebeling, Royal Netherlands Meteorological Institute, De Bilt, The Netherlands J Schulz, Deutscher Wetterdienst, Offenbach, Germany I Holleman, Royal Netherlands Meteorological Institute, De Bilt, The Netherlands V Levizzani, ISAC-CNR, Bologna, Italy WJ Timmermans, University of Twente, Enschede, The Netherlands H Rott, University of Innsbruck, Innsbruck, Austria N Mognard-Campbell, OMP/LEGOS, Toulouse, France R de Jeu, VU University Amsterdam, Amsterdam, The Netherlands W Wagner, Vienna University of Technology, Vienna, Austria M Rodell, NASA/GSFC, Greenbelt, MD, USA MS Salama, GN Parodi, and L Wang, University of Twente, Enschede, The Netherlands & 2011 Elsevier B.V. All rights reserved.

2.14.1 2.14.1.1 2.14.1.2 2.14.1.3 2.14.2 2.14.2.1 2.14.2.2 2.14.2.2.1 2.14.2.2.2 2.14.2.3 2.14.2.3.1 2.14.2.3.2 2.14.2.3.3 2.14.2.3.4 2.14.2.4 2.14.2.4.1 2.14.2.5 2.14.2.5.1 2.14.2.5.2 2.14.2.6 2.14.2.6.1 2.14.2.6.2 2.14.3 2.14.3.1 2.14.3.2 2.14.3.3 2.14.3.3.1 2.14.3.3.2 2.14.3.4 2.14.3.4.1 2.14.3.4.2 2.14.3.5 2.14.4 2.14.4.1 2.14.4.2 2.14.4.2.1 2.14.4.2.2 2.14.4.2.3 2.14.4.3 2.14.4.3.1 2.14.4.3.2 2.14.4.3.3 2.14.5

General introduction Water Cycle and Water Resources Management Water and Energy Balance of the Earth From Radiometric Observations to Object Properties Water in the Atmosphere: Clouds and Water Vapor Introduction Satellite RS Observing water vapor Observing clouds Retrieval Algorithms Water vapor Total column water vapor Water vapor profiles Upper tropospheric humidity Cloud Detection Cloud property retrievals Validation Water vapor Cloud properties Data Sets Water vapor products Cloud products Water from the Atmosphere: Precipitation Introduction Precipitation Measurements from Weather Radars Precipitation Measurements from Satellite Retrievals from VIS–IR sensors Retrievals from passive MW sensors Validation Weather radar retrievals Satellite retrievals Applications Water to the Atmosphere – Evaporation Introduction and Historic Development Current State of Science Statistical approaches Variability approaches Physical approaches Future Research Needs Scaling Feedbacks Validation Water on the Land – Snow and Ice

351 351 352 353 353 353 354 354 357 357 357 357 358 359 359 360 360 360 361 361 361 362 362 362 362 362 362 363 363 363 364 364 364 364 366 366 366 367 368 368 368 368 369

351

352

Observation of Hydrological Processes Using Remote Sensing

2.14.5.1 Introduction 2.14.5.2 Techniques for Retrieval of Extent and Physical Properties of Snow and Ice 2.14.5.3 Examples of Products and Applications 2.14.5.4 Future Research Needs 2.14.6 Water on the Land – Surface Water, River Flows, and Wetlands (Altimetry) 2.14.6.1 Introduction 2.14.6.2 In Situ Measurements 2.14.6.3 RS Techniques 2.14.6.4 Validation and Synergy of RS Techniques 2.14.6.5 Availability of the Satellite Data Sets 2.14.6.6 SWOT: The Future Satellite Mission Dedicated to Surface Hydrology 2.14.7 Water in the Ground – Soil Moisture 2.14.7.1 Introduction 2.14.7.2 State of the Art 2.14.7.3 Data Sets BBB 2.14.7.3.1 Active MW data sets 2.14.7.3.2 Passive MW data sets 2.14.7.4 Validation 2.14.8 Water in the Ground – Groundwater (Gravity Observations) 2.14.8.1 Introduction 2.14.8.2 GRACE Data Processing 2.14.8.3 Retrievals of Groundwater Storage with GRACE Data 2.14.8.4 GRACE Data Access 2.14.8.5 Concluding Remarks and Future Perspective 2.14.9 Optical RS of Water Quality in Inland and Coastal Waters 2.14.9.1 Introduction 2.14.9.2 Atmospheric Correction 2.14.9.3 Retrieval Algorithms 2.14.9.4 Uncertainty Estimates 2.14.9.5 Concluding Remarks and Future Perspective 2.14.10 Water Use in Agro- and Ecosystems 2.14.10.1 Introduction 2.14.10.2 Continuous Evaluation of Crop Water Use with Support from RS 2.14.10.3 Drought Indices and Soil Moisture Monitoring 2.14.10.4 Algorithm Retrievals and Operability 2.14.10.5 SEBS Algorithm 2.14.10.6 Evaluation Example Acknowledgment References

2.14.1 General introduction 2.14.1.1 Water Cycle and Water Resources Management The United Nations (UN) Millennium Declaration called on all members ‘‘to stop the unsustainable exploitation of water resources by developing water management strategies at the regional, national and local levels which promote both equitable access and adequate supplies.’’ Improving water management can make a significant contribution to achieving most of the Millennium Development Goals established by the UN General Assembly in 2000, especially those related to poverty, hunger, and major diseases. The World Summit on Sustainable Development (WSSD) in 2002 recognized this need. Water and sanitation in particular received great attention from the Summit. The Johannesburg Plan of Implementation recommended to improve water resources management and scientific understanding of the water cycle through joint cooperation and research. For this purpose, it is

369 369 370 371 371 371 371 372 372 373 373 374 374 375 377 377 377 378 379 379 379 379 380 380 380 380 381 381 381 382 382 382 383 385 386 386 387 388 388

recommended to promote knowledge sharing, provide capacity building, and facilitate the transfer of technology including remote-sensing (RS) and satellite technologies, especially to developing countries and countries with economies in transition, and to support these countries in their efforts to monitor and assess the quantity and quality of water resources, for example, by establishing and/or further developing national monitoring networks and water resources databases and by developing relevant national indicators. The Johannesburg Plan also adopted integrated water resources management as the overarching concept in addressing and solving water-related issues. As a result of the commitments made in the Johannesburg Plan of Implementation, several global and regional initiatives have emerged. Current international initiatives such as the Global Monitoring for Environment and Security (GMES) program of the European Commission and the European Space Agency (ESA), and the Global Earth Observation System of Systems (GEOSS)

Observation of Hydrological Processes Using Remote Sensing

10-Year Implementation Plan (GEO, 2005), have all identified Earth observation (EO) of the water cycle as the key in helping to solve the world’s water problems. More specifically, the 10-year implementation plan states that ‘‘Enhanced prediction of the global water cycle variation based on improved understanding of hydrological processes and its close linkage with the energy cycle and its sustained monitoring capability is a key contribution to mitigation of water-related damages and sustainable human development. Improved monitoring and forecast information, whether of national or global origin, if used intelligently, can provide large benefits in terms of reduced human suffering, improved economic productivity, and the protection of life and property. In many cases, the combination of space-based data and high-resolution in-situ data provides a powerful combination for effectively addressing water management issues. Information on water quantity and quality and their variation is urgently needed for national policies and management strategies, as well as for UN conventions on climate and sustainable development, and the achievement of the Millennium Goals’’ (GEO, 2005). The availability of spatial information on water quantity and quality will also enable closure of the water budget at river basin and continental scales to the point where effective water management is essential (e.g., as requested by the European Union’s Water Framework Directive (WFD), as well as national policies). Geo-information science and EO are vital in achieving a better understanding of the water cycle and better monitoring, analysis, prediction, and management of the world’s water resources. Subject to climate change, the security of freshwater resources has emerged as one of the key societal problems. According to a report prepared under the auspices of the Intergovernmental Panel on Climate Change (IPCC, 2008), ‘‘Observational records and climate projections provide abundant evidence that freshwater resources are vulnerable and have the potential to be strongly impacted by climate change, with wide-ranging consequences on human societies and ecosystems.’’ Floods, droughts, water scarcity, water usage, water quality, water and ecosystem interactions, and water and climate interactions are all issues of direct importance to our human society. The only key to safeguard the security of water resources is better water resources management. This in turn requires better understanding of the water cycle, water climate interactions, and water ecosystem interactions in the Earth’s climate system. To achieve such an understanding, it is essential to be able to measure hydroclimatic variables at different spatial and temporal scales, such as radiation, precipitation, evaporation and transpiration (or evapotranspiration (ET)), soil moisture, clouds, water vapor, surface water and runoff, vegetation state, albedo and surface temperature, etc. The major components of the water cycle of the Earth system and their possible observations are presented in Figure 1. Such observations are essential to understand the global water cycle and its variability, both spatially and temporally, and can only be achieved consistently by means of EOs. Additionally, such observations are essential to advance our understanding of coupling between the terrestrial, atmospheric, and oceanic branches of the water cycle, and how this coupling may influence climate variability and predictability. Figure 1 also shows the proportion of the water-cycle flux

353

components in the ocean (including evaporation of the ocean water into the atmosphere and condensation of the water vapor falling as precipitation into the ocean again), the proportion of the terrestrial water-cycle components (including precipitation as a consequence of condensation of water vapor generated by evaporation and transpiration from land and water vapor transported from the ocean, the river discharge, and groundwater discharge returning water into the ocean), water and ocean ice sheets in the ocean, permafrost and snow, soil moisture and groundwater on land, and atmospheric water vapor. Water resources management directly interferes with the natural water cycle in the forms of building dams, reservoirs, water transfer systems, and irrigation systems that divert and redistribute part of the water storages and fluxes on land. The water cycle is mainly driven and coupled to the energy cycle in terms of phase changes of water (changes among liquid, water vapor, and solid phases) and transport of water by winds in addition to gravity and diffusion processes. The water-cycle components can be observed with in situ sensors as well as airborne and satellite sensors in terms of radiative quantities. Processing and conversion of these radiative signals are necessary to retrieve the water-cycle components. To enhance prediction of the global water-cycle variation, based on improved understanding of hydrological processes and its close linkage with the energy cycle and its sustained monitoring capability, is a key contribution to mitigation of water-related damages and sustainable human development. In many cases, the combination of space-based data and high-resolution in situ data in a modeling system using data assimilation provides a powerful tool for effectively addressing water management issues.

2.14.1.2 Water and Energy Balance of the Earth The Sun is the primary source of energy of Earth’s climate system and its five major components: the atmosphere, the biosphere, the cryosphere, the hydrosphere, and the land surface (ESA, 2006). In Earth’s energy balance, the shortwave (solar) radiation is redistributed by different radiative climate forcing components. In the long term, the amount of incoming solar radiation absorbed by land, ocean, and atmosphere is balanced by releasing the same amount of outgoing longwave (terrestrial) radiation from Earth to space. About half of the incoming solar radiation is absorbed by the Earth’s surface. This energy is transferred to the atmosphere by warming the air in contact with the surface (thermals), by ET and by longwave radiation that is absorbed by clouds and greenhouse gases. The atmosphere in turn radiates longwave radiation back to the Earth’s surface as well as out to space. Changes in greenhouse gas concentrations cause altering the longwave radiation from the Earth out to space. The climate system will respond directly to such changes, as well as indirectly, through a variety of feedback mechanisms. For example, an increased concentration of water vapor enhances the amount of thermal radiation absorbed by the atmosphere and consequently leads to an increase of the surface temperature, but will also likely lead to an increase of cloud amount and precipitation. Simplified schematic representations of the annual mean energy flux budgets for the Earth,

354

Observation of Hydrological Processes Using Remote Sensing

Radiation

Vapor transport 10%

Precipitation (27%) Permafrost storage (23 k km3) (0.043 m)

Infiltration

River and lakes storage (178 k km3) (0.349 m) Evaporation/ transpiration (17%)

Soil moisture storage (122 k km3) (0.239 m)

River and groundwater discharge (10%) Groundwater storage (15 300 k km3) (29.996 m)

Atmosphere water storage (12.7 k km3) (0.025 m)

Condensation (90%)

Evaporation (100%, 413 k km3 yr−1)

Ocean ice storage (26 350 k km3) (51.659 m)

Ocean water storage (1 335 040 k km3) (2627 m)

Figure 1 Global water cycle of the Earth system and their possible observations with in situ, airborne instruments (low altitude and high altitude), and satellites. The flux components (condensation, water vapor transport, precipitation, evaporation and transpiration, and river and groundwater discharge) are normalized with the total ocean evaporation of 413 000 km3yr1 (100%). The storage components are also converted to water depth using the total surface area of earth 510 072 000 km2. Data from Trenberth et al. (2007).

land, and ocean are presented in Figure 2, using data reported by Trenberth et al. (2009). For the Earth energy budget, the incoming solar radiation at the top of atmosphere (TOA) is 341.3 W m2, equivalent to one-quarter of the solar constant 1365.2 W m2, of which 101.9 W m2 is reflected (79 W m2 by clouds and 23 Wm2 by the Earth’s surface) to space resulting in a planetary albedo (or TOA albedo) of 29.8%. The surface albedo 14.3% (at the bottom of atmosphere, BOA) is the ratio of the reflected solar radiation (23 W m2) to the absorbed solar radiation (161 W m2). In addition, 78 W m2 of the incoming solar radiation is absorbed by the atmosphere. The atmosphere emits 333 W m2 downward to the surface, while the surface emits 396 W m2 upward to the atmosphere, resulting in a net upward surface longwave of 93 W m2. Part of the emitted surface longwave radiation passing through the atmospheric window (40 W m2), together with the upward longwave radiation from the atmosphere (169 W m2) and that from clouds (30 W m2), makes up the outgoing longwave radiation to space (238.5 W m2). The sum of the net radiation at the surface 98 W m2 (net downward solar radiation 161 W m2 less net surface upward longwave radiation 63 W m2) is balanced by the thermals (i.e., sensible heat flux 17 W m2) and latent heat flux for evaporation/transpiration (80 W m2), with 0.9 W m2 absorbed by the surface. At the TOA, the radiation balance is

also 0.9 W m2 (incoming solar radiation 341.3 W m2 less reflected solar radiation 101.9 W m2 and outgoing longwave radiation 238.5 W m2), indicating a net gain of 0.9 W m2 in energy, which may be conceived as a possible warming of the Earth system. However, this quantity is derived only for the Clouds and the Earth’s Radiant Energy System (CERES) (Wielicki et al., 1996) period from March 2000 to May 2004 and cannot be taken as long-term evidence. Similar explanations can be made for the energy budgets for land and ocean separately. The differences in land and ocean energy budget components are caused mainly by different albedo over land and water as well as the different thermodynamic properties of land and water. EO of water cycle primarily uses information in the optical, thermal, and microwave (MW) regions of the electromagnetic spectrum to retrieve water-cycle components, though other measurement using, for example, gravity measurement has also shown great promise for monitoring mass changes. One example of EO of water cycle is the Water Cycle Multi-mission Observation Strategy (WACMOS) project initiated by the European Space Agency (ESA) and the Global Energy and Water Cycle Experiment (GEWEX) of the World Climate Research Programme (WCRP). The WACMOS project aims to develop and validate novel and improved multimission-based global water-cycle data sets using multimission satellite data.

Observation of Hydrological Processes Using Remote Sensing

355

Global energy budget, TOA albedo 29.8% 102

Reflected solar radiation 101.9 W m−2

Incoming solar radiation 341.3 W m−2

341

Reflected by clouds and atmosphere 79

Outgoing longwave radiation 238.5 W m−2

239

79 Emitted by atmosphere

40 169

30 Greenhouse gases

Absorbed by 78 atmosphere Latent 17 80 heat 40

Reflected by surface 23

356

333 Back radiation

396

161 Absorbed by surface

17 Thermals

80 Evapotranspiration

Atmospheric window

333

Surface radiation

Absorbed by surface

BOA albedo 14.3% (a)

Land energy budget, TOA albedo 34.4% 113

Reflected solar radiation 113.4 W m−2

Incoming solar radiation 330.2 W m−2

330

Reflected by clouds and atmosphere 74

Outgoing longwave radiation 232.4 W m−2

232

74 Emitted by atmosphere

40 163

30 Greenhouse gases

Absorbed by 78 atmosphere Latent 17 27 heat 40

Reflected by surface 40

343

Absorbed by surface

(b)

304 Back radiation

383

145 17 Thermals

27 Evapotranspiration

Net absorbed 0.0

Atmospheric window

Surface radiation

304 Absorbed by surface

BOA albedo 27.3%

Figure 2 Schematic representation of the mean annual energy budgets for the earth, land, and ocean. The surface emissivity is assumed to be 1.0. Scheme and primary data from Trenberth KE, Fasullo JT, and Kiehl J (2009) Earth’s global energy balance. Bulletin of the American Meteorological Society 311–323: doi:10.1175/2008BAMS2634.1.

356

Observation of Hydrological Processes Using Remote Sensing Ocean energy budget, TOA albedo 28.3% 98

Reflected solar radiation 97.8 W m−2

Incoming solar radiation 345.4 W m−2

345

Reflected by clouds and atmosphere 81

Outgoing longwave radiation 240.8 W m−2

241

81 Emitted by atmosphere

40 165

36 Greenhouse gases

Absorbed by 78 atmosphere Latent 97 heat

12

40 Reflected by surface 17

361

401

168 Absorbed by surface

12 Thermals

(c)

Atmospheric window

97 Evapotranspiration

Net absorbed 1.3

Surface radiation

343 Back radiation 343

Absorbed by surface

BOA albedo 14.3%

Figure 2 Continued.

2.14.1.3 From Radiometric Observations to Object Properties The scientific challenge in EO of water cycle is to determine turbulent, thermodynamic, and fluid dynamic properties of the whole water cycle by using radiometric observations. As illustrated in Figure 3, a sensor with a certain response function measuring radiation reflected or emitted from an object can have different geometric arrangements with respect to the object, each with always atmosphere between the sensor and the object. In order to retrieve the properties of the object in concern using data in terms of its range to the sensor, its combined temperature and emissivity (or reflectivity) at different times, at different spatial resolution, at different wavelengths, at different direction, and at different polarization, detailed data processing is needed (see Section 2.14.10 for a detailed example). In terms of the sensor response, we need to ask two types of questions: (A) How much radiation is detected at the sensor? (B) When and how does it arrive? If the answers are only relevant to question (A), then we have a passive sensor system, otherwise if the answers are relevant to both questions (A) and (B), then we have an active sensor system. Many excellent textbooks exist that deal with the theoretical aspects of the sensor–object relationships and the practical issues related to retrievals of object properties (e.g., Rees, 2001; and Liang, 2004). Applications of RS in hydrology and climate studies can be found in related chapters in Anderson and McDonnell (2005) and in Bolle (2003); the current chapter is a continuation of these earlier efforts on RS in hydrology and water resources management. Many Internet

sites also provide very useful resources, data, and examples of EO of water-cycle variables, some of the most relevant ones are provided in Table 1. There have been excellent field campaigns in hydrology – HAPEX Sahel, HAPEX Mobily (Goutorbe et al., 1997), International Satellite Land Surface Climatology Project (ISLSCP) Field Experiment (FIFE; Sellers et al., 1992), MONSOON’90 (Kustas and Goodrich, 1994), the Southern Great Plains Hydrology Experiment (SGP), and the Soil Moisture Experiment (SMEX; Jackson et al., 1999) – to study the complex hydrological processes and land–atmosphere interactions at local to regional scales. The FIFE project was a large-scale climatology project set in the prairies of central Kansas from 1987 to 1989. This project was designed to improve understanding of carbon and water cycles; to coordinate data collected by satellites, aircraft, and ground instruments; and to use satellites to measure these cycles. More information on FIFE can be found on the Internet. The MONSOON’90 large-scale interdisciplinary field experiment was conducted in the summer of 1990 in southeastern Arizona to investigate the utility of RS coupled with energy and water-balance modeling for providing large-area estimates of fluxes in semiarid rangelands. Large-scale field experiments related to soil moisture are the series of the SGP and the SMEX series, focusing on validation of retrieval algorithms and demonstration of technological feasibilities of RS of soil moisture. Some examples of both sensor systems and the retrievals of the relevant geo-biophysical parameters can be found in some recent large-scale field experiments, the SPAR 2004 and SEN2FLEX 2005 campaigns (Sobrino et al., 2008, 2009;

Observation of Hydrological Processes Using Remote Sensing

357

The fundamental of Earth observations (link sensor response to object properties in a thermodynamic environment)

Sensor response Sensor

A. How much radiation is detected? B. When and how does it arrive?

Object properties

23°

52°

Nadir Atmosphere

Object

Its range, its combined temperature, and emissivity (or reflectivity) at different times, or different spatial resolution, at different wavelengths, at different direction, at different polarization.

A : A passive sensor system A + B: an active senor system

(e.g., SEN2FLEX campaign) Figure 3 Schematic representation of links between radiometric observations and object properties.

Su et al., 2008) as well as the EAGLE 2006 campaign (Su et al., 2009). The data collected from these field experiments are open to the scientific community for collaborative investigations and are accessible at the European Space Agency’s Principle Investigators portal or by contacting the authors directly. As an example, the spectra of bright sand and of a young pine tree collected during the EAGLE 2006 campaign are shown in Figure 4, indicating the sensor responses to different properties of the object. Some most relevant web links to data portals, software tools, and training courses related to water and energy balance of the Earth are given Table 1. In the following sections, we discuss details of the different components of the water cycles from the perspectives of EO.

2.14.2 Water in the Atmosphere: Clouds and Water Vapor 2.14.2.1 Introduction Accurate information on the distribution of water vapor and clouds in the atmosphere is essential for water and energy balance studies. The atmosphere acts as a medium for the transport of water around the globe. Water vapor is brought into the atmosphere through evaporation from liquid water bodies (B90%) and transpiration from plants (B10%). Clouds are formed in lifting air parcels, in which water vapor condensates into cloud particles due to the cooler temperatures. Once in the atmosphere, clouds are moved around the globe by strong winds and either evaporate back into water vapor or disappear as precipitation to replenish the earthbound parts of the water cycle. The presence of water vapor and clouds in the atmosphere warms the Earth’s troposphere and surface, and acts as a partial blanket for the longwave radiation coming from the surface. Water vapor and clouds absorb and emit infrared (IR)

radiation and thus contribute to warming the Earth’s surface. For clouds, this effect is counterbalanced by the reflection of visible (VIS) radiation, which reduces the amount of shortwave (solar) incoming radiation at the Earth’s surface and has a cooling effect on the climate system. The net average effect of the Earth’s cloud cover in the present climate is a slight cooling because the reflection of radiation more than compensates for the blanketing effect of clouds. Information on the distribution of water vapor and clouds in the atmosphere is also relevant for studying the hydrological cycle. The shortwave and longwave radiation that reach the Earth’s surface directly affect the evaporation (latent) and sensible heat fluxes. The part of the radiation that is used to evaporate soil moisture (evaporation) or crop moisture (transpiration) is released to the atmosphere as water vapor. The evaporated water vapor, in turn, is carried upward where it condenses into cloud droplets, ice crystals, or precipitation. Ground-based measurements are inadequate to observe the large spatial and temporal variations in water vapor and cloud properties (Rossow and Schiffer, 1999). The advent of satellite RS has changed this situation. Satellites can provide the required information at adequate temporal and spatial scales. Satellite observations can be used to retrieve integrated water vapor amounts and cloud physical properties from passive MW radiometers or spectral radiances, respectively. The accuracies and precisions of these satellite retrievals have been well determined within various validation studies.

2.14.2.2 Satellite RS Since the 1960s, many satellites have been providing continuous observations of the state of the atmosphere over very large regions or even for the entire globe. The satellite instruments of most interest for observing water vapor and clouds are MW radiometers that measure emitted MW radiation of the Earth’s surface, atmosphere, and clouds, and the

358

Observation of Hydrological Processes Using Remote Sensing

Table 1 cycle

Some resources for Earth observation program, satellites and data, software, and training courses related to earth observation of water

Organization/program

Web links

Comments

Group on Earth Observations (GEO)

http://www.earthobservations.org/

GEO portal

http://www.geoportal.org/

GEO applications

http://www.earthobservations.org/ documents/the_full_picture.pdf http://www.esa.int/esaeo/ http://www.esa.int/esalp/

GEO coordinates international efforts to build a Global Earth Observation System of Systems (GEOSS). In its water societal benefit area it aims at improving water resource management through better understanding of the water cycle. The GEO portal provides an entry point to access remote sensing, geospatial static, and in situ data, information and services. Water is one of the nine societal benefit areas. ‘The Full Picture’ provides an overview of the progresses in applications of GEOSS in the nine societal benefit areas till 2007. ESA’s Earth observation programs include the Global Monitoring for Environment and Security (GMES) and the Living Planet Programme. Observation of the hydrosphere is one of the foci of this program. The ESA-MOST (Ministry of Science and Technology, China) Dragon program includes a dedicated training program to provide training in data processing, algorithm and product development from ESA Earth Observation (EO) data in land, ocean, and atmospheric applications. NASA’s Hydrological Sciences focuses on the interpretation of remotely sensed data and land surface hydrological, meteorological, and climate modeling. GLDAS generates optimal fields of land surface states and fluxes (Rodell et al., 2004) by assimilating satellite- and ground-based observational data products into advanced land surface models. The high-quality, global land surface fields provided by GLDAS support several current and proposed weather and climate prediction, water resources applications, and water-cycle investigations. GLDAS has resulted in a massive archive of modeled and observed, global, surface meteorological data, parameter maps, and output which include 11 and 0.251 resolution 1979–present simulations of the Noah, CLM, and Mosaic land surface models. The main purpose of the Land SAF is to increase the benefits from Meteosat Second Generation (MSG) and European Polar Satellite (EPS) data related to land, land–atmosphere interactions and biophysical applications by developing techniques, products and algorithms that will allow a more effective use of data from the two satellites of the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT). The Earth Observation Handbook presents the main capabilities of satellite Earth observations, their applications, and a systematic overview of present and planned CEOS agency Earth observation satellite missions and their instruments. It also explores society’s increasing need for information on our planet.

European space agency (esa)

European Space Agency (ESA) Dragon programme

http://earth.esa.int/dragon/

National Aeronautics and Space Administration (NASA)

http://neptune.gsfc.nasa.gov/hsb/

Global land data assimilation system (gldas)

http://ldas.gsfc.nasa.gov/

Land surface analysis satellite applications facility (lsa saf)

http://landsaf.meteo.pt/

Committee on Earth Observation Satellites (CEOS)

http://www.eohandbook.com

passive imagers that measure VIS, near-IR, and IR radiances. The importance of satellite observations is determined by the spatial and temporal sampling resolution of their instruments. Frequent sampling is especially required for parameters that are highly variable in space and time, such as clouds.

2.14.2.2.1 Observing water vapor The strong variations of water vapor in space and time lead to the necessity of monitoring this quantity globally from satellites. Absorption lines of water vapor are present in almost every part of the electromagnetic spectrum. A great variety of space-borne sensors are used to retrieve atmospheric profiles of humidity or the column amount, even if they were not designed for it. These sensors observe the interaction of radiation with water vapor in the different parts of the spectrum

(MW, IR, optical, and ultraviolet (UV)). The number of available instruments is further increased due to the need for measurements at different observation geometries (nadir view, limb scanning, occultation, and day or night) and at different orbit orientations. Here, only some of the observation systems for tropospheric water vapor are described. MW radiometers observe the radiation close to the 22-GHz water vapor absorption line that is closely related to the total column content of water vapor. These observations can be used over oceans in clear sky and cloudy conditions. The conically scanning special sensor microwave/imager (SSM/I) on the DMSP satellites is available since 1987 and is continued with the special sensor microwave imager sounder (SSMIS) instrument into the future. Among others, this type of radiometers is flown on the US TRMM satellite (TRMM Microwave Imager (TMI)) and on the US Aqua mission

Observation of Hydrological Processes Using Remote Sensing Sand − location #7

0.6

Reflectance (−)

359

0.4

0.2

0 350

550 750 950 1150 1350 1550 1750 1950 2150 2350 Wavelength (nm) Pine tree vertical − location #25

Reflectance (−)

0.6

0.4

0.2

0 350

550 750 950 1150 1350 1550 1750 1950 2150 2350 Wavelength (nm)

Figure 4 Example spectra of bright sand (upper panels) and of a young pine tree (lower panels). The frame of the photos represents an area of 1  1 m2. On the graphs, the gray lines show the measured spectra, the thick black line is the spectrum accepted as the characteristic spectrum of the site under consideration (Su et al., 2009).

(Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E)). In addition, the advanced microwave sounding unit (AMSU) makes its observations at this frequency but is a cross-track scanner. Imaging spectrometers, such as ESA’s medium-resolution imaging spectrometer (MERIS), are used to retrieve the total column content of water vapor at a very high spatial resolution (B300 m) from near-IR observations during daytime (Rast et al., 1999). MERIS is especially useful over land surfaces. Such observations are also available from the moderateresolution imaging spectroradiometer (MODIS) flown on the NASA Aqua and Terra satellites (NASA, National Aeronautics and Space Administration). Since 1977, the Meteosat Visible and Infrared Radiation Imager (MVIRI) and Spinning-Enhanced Visible and Infrared Imager (SEVIRI) instruments in geostationary orbit observe radiation at 6.3 and 7.2 mm (only SEVIRI), and allow the retrieval of upper tropospheric humidity (UTH) with a very high temporal resolution (up to 15 min) that allows for studies of atmospheric dynamics (Schmetz et al., 2002). Also in geostationary orbit, humidity sounders similar to the HIRS instrument are flown on the US Geostationary Operational Environmental Satellites (GOES).

UV/VIS spectrometers, such as the Global Ozone Monitoring Experiment (GOME) and Scanning Imaging Absorption Spectrometer for Atmospheric Cartography (Sciamachy), are used for the retrieval of total column water vapor over land and ocean surfaces with approximately the same accuracy as the SSM/I but only under daylight and clear sky conditions at much coarser spatial resolution (Burrows et al., 1999). Since 1978, the observations of the Advanced Television and Infrared Observation Satellite (TIROS) Operational Vertical Sounder (ATOVS) on NOAA and MetOp satellites, with its IR spectrometer (High resolution Infrared Radiation Sounder (HIRS)), and MW radiometers (AMSU-A/B and Microwave Humidity Sounder (MHS)), have been combined to derive atmospheric temperature and humidity profiles. Since 2007, EUMETSATs MetOp satellite has been carrying the Infrared Atmospheric Sounding Interferometer (IASI) instrument. This new generation of IR sounding instruments is capable of observing about 15 independent pieces of information on the vertical profile by performing observations over a large part of the IR spectrum (4–50 mm) (Simeoni et al., 1997). A similar instrument, called Atmospheric Infrared Sounder (AIRS), is flown since 2002 on NASAs Aqua mission (Aumann et al., 2003).

360

Observation of Hydrological Processes Using Remote Sensing

Finally, temperature and humidity profiles can also be retrieved from Radio Occultation measurements, performed by, for example, the GRAS (Loiselet et al., 2000) instrument onboard the MetOp or the COSMIC fleet (Anthes et al., 2008).

2.14.2.2.2 Observing clouds During the last 25 years, observations from passive imaging satellites have been successfully used for the retrieval of cloud cover and cloud physical properties. Rossow and Garder (1993) used observations from the Advanced Very High Resolution Radiometer (AVHRR) instrument onboard the NOAA series of polar-orbiting satellites to derive global cloud climatology since 1982. Recently, several more sophisticated instruments for EOs have been launched. These include the instruments that are flown onboard the NASA Earth Observing System (EOS) geosynchronous orbiting satellites, which were launched in 1999 (Terra) and in 2002 (Aqua). The MODIS instruments on both satellites operate the required spectral channels for the retrieval of cloud properties at high spatial resolutions (better than 1 1 km2) globally, but at low temporal resolutions (revisit time 1 day or more). The unprecedented sampling frequency of geostationary satellites (better than 30 min) allows for monitoring the diurnal variations in cloud properties. The SEVIRI instrument on board METEOSAT-8, which was launched in 2002, is the first instrument that can be used for the retrieval of these properties from a geostationary orbit (Figure 5). SEVIRI constitutes a valuable source of data for water and energy balance studies. Another types of instruments for cloud observations are passive MW radiometers. These instruments measure emitted MW radiances from the Earth surface and the overlaying atmosphere. Greenwald et al. (1993) showed that the radiances observed by these instruments can be used for a simultaneous retrieval of atmospheric water vapor and cloud liquid water.

Figure 5 METEOSAT-8/SEVIRI image of the visible channel (0.6 mm) for the SEVIRI field of view for 17 January 2006 at 11:45 UTC.

Recently, even more advanced satellite systems are available for observing clouds. The most advanced cloud measurements are provided by the radar on the Cloudsat satellites and lidar on the Calipso satellites, which were launched in 2006 and fly in the A-train constellation. These instruments measure vertical profiles of cloud reflectivity of large particles (radar) and small particles (lidar).

2.14.2.3 Retrieval Algorithms 2.14.2.3.1 Water vapor Retrieval methods have to correspond to the instrument spectral range and observation geometry. Processes in the atmosphere complicate the retrieval task, for example, the coexistence of the three thermodynamic phases of water on the Earth, interaction with aerosols, and varying surface emissivity. The number of retrieval algorithms is much larger than the number of sensors. Retrieval methods generally depend on a priori information, which could be the coefficients of a regression, the constraints for retrieval based on inversion, or the training set of a neural network. The quality of a retrieval scheme depends on the applicability of the a priori information to the prevailing environmental conditions, that is, surface properties, clouds, etc.

2.14.2.3.2 Total column water vapor Major instruments utilized for the retrieval of total column water vapor are MW radiometers (SSM/I), UV/VIS spectrometers (GOME), and VIS and near-IR imaging spectrometers (MERIS). Retrieval schemes for MW radiometer can be distinguished in semiphysical and physical schemes. In both cases, observations of the instrument are simulated using a radiative transfer model. Input to the model is the atmospheric state vector and instrument parameters. The semiphysical schemes then retrieve the water vapor content by applying a statistical scheme (linear regression or neural networks) based on the training data (Schlu¨ssel and Emery, 1990). The physical schemes mostly use a first guess, often coming from a numerical weather forecast model (NWP), as the basis for the forward computation and then vary the first guess until the used set of observed radiances is best matched (e.g., Wentz, 1997). The latter requires much more computer power but has generally replaced statistical methods in the past 10 years. The basic principle in retrieval applied to GOME is the Differential Optical Absorption Spectroscopy (DOAS) method to calculate the difference between the Sun normalized measured Earthshine radiance and absorption cross sections at wavelengths where water vapor absorbs radiation and relate this absorption depth to the water vapor column concentration (e.g., Noel et al., 1999). The DOAS method provides a global (land and ocean) completely independent data set, because it does not rely on any additional external information. Near-IR MERIS algorithms are based on radiative transfer simulations, where the radiance ratio between the MERIS channels 15 (900 nm) and 14 (885 nm) are used in an inversion procedure based on regression (Bennartz and Fischer, 2001). Near-IR and IR algorithms were also developed for the MODIS instrument by Huang et al. (2004).

Observation of Hydrological Processes Using Remote Sensing

2.14.2.3.4 Upper tropospheric humidity The relative humidity (RH) of the upper troposphere has a strong influence on the amount of outgoing longwave radiation. It is often derived employing IR and MW instruments as HIRS and MVIRI/SEVIRI as well as AMSU-B/MHS. The brightness temperature of one channel, 6.3 mm for IR and 183 GHz for MW, is related to the RH of the upper troposphere. A typical physical retrieval method for Meteosat is described in Schmetz and Turpeinen (1988). The retrieval is confined to areas with neither medium- nor high-level clouds. Similar schemes, using Jacobian vertical weighting, have been developed by Buehler and John (2005) for AMSU-B, Brogniez et al. (2007) for Meteosat, and Jackson and Bates (2001) for HIRS.

2.14.2.4 Cloud Detection In general, cloud detection methods are based on the fact that clouds have a higher reflectance and a lower temperature than the underlying Earth surface. In addition, cloudy scenes have a higher spatial and temporal variability than clear sky scenes. However, difficulties in cloud detection appear when the contrast between the cloud and underlying surface is small. At VIS wavelengths, it is difficult to detect clouds over high reflecting surfaces such as snow or desert. At IR wavelengths, it is difficult to discriminate low clouds from clear sky land surfaces during the night, when surface temperatures may drop below cloud-top temperatures. In these cases, testing the spatial coherence in IR radiances in cloudy and clear skies is an effective manner to identify cloudy areas (Coakley and Bretherton, 1982). Moreover, cloud edges are difficult to detect, as the satellite pixels at these edges are only partly cloudy. Part of the difficulties touched on above may be alleviated by the combined use of the multi-spectral observations from satellite (Saunders and Kriebel, 1988; Ackerman et al., 1998).

is based on the principle that the reflectance of clouds at a nonabsorbing wavelength in the VIS region is strongly related to the optical thickness and has very little dependence on particle size, whereas the reflectance of clouds at an absorbing wavelength in the near-IR region is primarily related to particle size (Nakajima and King, 1990; Han et al., 1995). The example simulated TOA solar reflectance spectra presented in Figure 6 shows the substantial differences in the absorption properties of water and ice in the near-IR solar region (0.7 mmol o4 mm). Especially at 1.6, 2.2, and 3.9 mm, ice exhibits stronger absorption than water. Due to differences in cloud optical thickness, the ice cloud is somewhat brighter than the water cloud in the VIS region (lo0.7 mm). An inversion procedure is used to relate observed radiances to cloud thermodynamic phase, optical thickness, and particle size. A radiative transfer model is used to prepare lookup tables of simulated reflectances for clouds with different optical thicknesses, thermodynamic phases, and particle sizes for a wide variety of solar and satellite viewing geometries. Liquid and ice water path are computed from retrieved cloud optical thickness and particle size. Note that the retrieval of particle size from near-IR reflectances is weighted toward the upper part of the cloud (Platnick, 2001). At thermal IR wavelengths, the retrieval of cloud microand macro-physical properties is based on the interpretation spectral variations in emitted radiances at the cloud top. Absorption and emission dominate cloud radiative transfer at these wavelengths, because cloud particles have a low single scattering albedo and a large asymmetry parameter. For clouds with an optical thickness smaller than 4, the amount of observed upwelling radiance at the top of the atmosphere will be affected by cloud properties, such as optical thickness, particle size, and thermodynamic phase (Baum et al., 1994). By selecting two (or more) appropriate wavelengths, it is feasible to infer the emissivity and temperature at the cloud top, and deduce information on cloud optical thickness, particle size, and thermodynamic phase. A major drawback of IR retrievals techniques is that cloud emissivities saturate at relatively low optical thicknesses. Passive MW radiometers measure radiances, expressed as brightness temperatures, at various frequencies between 10 and 100 GHz. These radiances have distinct atmospheric

0.6

Water cloud Ice cloud

Reflectance

2.14.2.3.3 Water vapor profiles In general, the so-called 1 D-VAR technique is employed for water profile retrievals. 1 D-VAR schemes use the variational principle to solve the retrieval problem (Eyre, 1989), and invert the radiances to simultaneously retrieve the temperature and humidity profile, the surface temperature and MW emissivity, as well as cloud amount and cloud-top pressure. It employs an iterative method, which finds the maximum probability solution to a nonlinear retrieval/analysis problem. Li et al. (2000) applied a 1 D-VAR scheme to TOVS/ATOVS observations. Moreover, 1 D-VAR schemes are also applied to atmospheric sounders, such as AIRS and IASI and to Radio Occultation instruments. Semiphysical schemes, such as neural networks, can also be used to simultaneously retrieve temperature and water vapor profiles (e.g., Kuligowski and Barros, 2001).

361

0.4

0.2

2.14.2.4.1 Cloud property retrievals The spectral variations in cloud reflectances provide a powerful diagnostic tool to identify differences in cloud micro- and macro-physical properties, such as cloud optical thickness, cloud thermodynamic phase, and cloud particle size. The retrieval of cloud physical properties from reflectances at VIS (0.6 or 0.8 mm) and near-IR (1.6, 2.2, or 3.9 mm) wavelengths

0.0 0.50

1.00

1.50

2.00

2.50

3.00

3.50

Wavelength (µm) Figure 6 Radiative transfer model simulated top of atmosphere reflectance spectra for water and ice clouds.

4.00

362

Observation of Hydrological Processes Using Remote Sensing

absorption characteristics, and can be used for a simultaneous retrieval of atmospheric water vapor and cloud liquid water (Greenwald et al., 1993). The liquid water path (LWP) retrievals from MW radiometers provide a measurement of the integrated LWP, and only represent the liquid droplets volume in the cloud. Numerous algorithms have been developed to estimate the incoming shortwave radiation from satellite radiances (Pinker et al., 1995). Some methods calculate the incoming shortwave radiance by directly interpreting the TOA albedo in terms of atmospheric transmission (Mueller et al., 2009), while others calculate the transmission for a clear and cloudy atmosphere separately, using atmospheric water vapor and cloud microphysical properties (Deneke et al., 2008).

2.14.2.5 Validation Validation is prerequisite to generate accurate data sets of water vapor and cloud properties for water and energy balance studies.

2.14.2.5.1 Water vapor The validation of atmospheric water vapor retrieval schemes is very difficult because classical observations are only sparsely available; for example, radiosondes are mostly available over land surfaces and their observation time does not match overpass times of polar orbiting satellites. Ground-based global positioning system (GPS) observations are available more often over land surfaces, but they are only suitable to validate total column water vapor estimates. As aircraft observations are only available along major flight paths, and the accuracy of their instruments often insufficient for validation, the upper troposphere and lower stratosphere are hard to validate. Instead, satellite systems are compared among themselves or to atmospheric reanalysis. Such comparison can also help to uncover specific instrumental and retrieval problems. For instance, the comparison of the passive MW AMSR-E and IR AIRS estimates of total water vapor content revealed some systematic differences due to the treatment of clouds in the AIRS retrievals (Fetzer et al., 2006). The most comprehensive comparison of SSM/I-based retrievals among themselves and to radiosondes, performed by Sohn and Smith (2003), revealed that differences in statistical retrievals are mostly caused by differences in the training data that were used. It was also found that statistical algorithms outperform physical ones because of simplifying assumptions on tangential factors, such as near-surface wind speed, sea-surface temperature, and residual cloud liquid water. On a seasonal scale (3 months means), the differences between satellite and radiosondes are B1 kg m2 bias and B2.5 kg m2 rms. Sensitivity of instruments influences the satellite comparisons. Fetzer et al. (2008) compared AIRS UTH with the Microwave Limb Sounder (MLS) data. The mean values agree well within 10% and standard deviations of their differences are 30% or less. Differences in wet and dry regimes were found to be caused by different sensitivities of the two instruments. Milz et al. (2009) compared monthly mean distributions of UTH products from AMSU-B, Humidity Sounder Brazil (HSB), and AIRS for January 2003. The UTH, based on simulated AMSU-B brightness temperatures from AIRS profiles, has a

slight moist bias of up to 4% in RH. This bias is small compared to the differences in UTH observations from radiosondes and nadir-looking IR sounders, which were between 10% and 15%, depending on the type of radiosondes (Soden and Lanzante, 1996). It is also small compared to the large differences in UTH between different climate models (John and Soden, 2007). Thus, most of the existing UTH data sets are suitable as benchmark for improving climate model representations of humidity. Li et al. (2000) reported for temperature profile retrievals from ATOVS, an accuracy of 2 K for temperatures at 1-km resolution and 3–6 K for dew-point temperatures. IASI profile retrievals have recently been evaluated by Pougatchev et al. (2009). Besides the very much improved temperature retrieval, they found that the instantaneous RH retrievals have a bias of about 710%, and a standard error lower than 10% in the 800–300-hPa range.

2.14.2.5.2 Cloud properties The validation data of cloud properties are obtained from flight measurements or special observatory sites. During flight measurement campaigns, heavily instrumented aircrafts collect very detailed measurements of cloud micro- and macrophysical properties over a limited period of time, providing valuable information to obtain a better understanding of cloud microphysics (e.g., EUCAARI over Europe, AMMA over Africa, and RICO over the Caribbean). Special observatory sites aim to measure the physical state of the (cloudy) atmosphere over longer periods of time (years). These sites are equipped with a suite of RS instruments to measure radiation, water vapor, and cloud properties. The number of these sites is limited, and comprises the three America Atmospheric Radiation Measurement (ARM) sites and the four Cloudnet sites in Northern Europe. The measurements of the above-described observatory sites play a key role in the continuous validation of cloud properties. Recently, measurements from Cloudsat (radar) and Calipso (lidar) can be used for validation as well. The combined use of radar and lidar observations allows the retrieval of vertical profiles of cloud optical thickness, cloud phase, particle size, and cloud water content (Delanoe¨ and Hogan, 2008). These retrievals are of great value for the validation of cloud property retrievals or for deriving global cloud climatology. Validation studies confirmed that LWP can be retrieved with high accuracy from both AVHRR (Han et al., 1995; Jolivet and Feijt, 2005) and SEVIRI (Roebeling et al., 2006). Although some retrieval algorithms use the 0.6-, 3.8-, and 10.5-mm radiances (Han et al., 1995), while others use the 0.6- and 1.6mm radiances (Jolivet and Feijt 2005; Roebeling et al. 2008), similar accuracies (B15 g m2) and precisions (B30 g m2 for thin clouds and up to 100 g m2 for thick clouds) were found. The above-mentioned accuracies suggest that LWP retrievals from satellite could be an appropriate source of information for the evaluation of climate-model-predicted LWP values. For nonprecipitating water clouds, Van Meijgaard and Crewell (2005) found differences up to 50 g m2 between climatemodel-predicted and MWR-inferred LWP values. During the FIRE Arctic cloud experiment, Curry et al. (2000) compared large-scale model LWP values to MWR-inferred LWP values.

Observation of Hydrological Processes Using Remote Sensing

They found that all models underestimate the mean LWP by 20–30 g m2, which corresponded to a relative accuracy worse than 60%.

2.14.2.6 Data Sets 2.14.2.6.1 Water vapor products In the framework of the GEWEX Water Vapor Project, the NVAP total column water vapor product (Randel et al., 1996) was derived from a combination of SSM/I, TOVS, and radiosonde data for the years 1988–2001. This product was partly renewed by the additional use of AMSU and TRMM data, but this covers only the years 2000–01. Over ocean, the total column water vapor derived from SSM/I (Figure 7) is available from the EUMETSAT Satellite Application Facility on Climate Monitoring (CM-SAF) (Schulz et al., 2009) and from RS systems (Wentz, 1997). These data sets have been successfully used for climate analysis, the evaluation of climate models, model-based reanalysis, trend studies (Trenberth et al., 2005), and investigations of the human impact on the water vapor distribution (Santer et al., 2007). Moreover, GOME/SCIAMACHY data sets have also been used to compute trends of total column water vapor (Mieruch et al., 2008). High-quality data sets of atmospheric profiles for climate studies, based on TOVS, have been derived by Scott et al. (1999). These profiles are highly correlated to corresponding ATOVS profiles. Operationally processed data from ATOVS, AIRS, and IASI exist at various places, such as NOAA, NASA, EUMETSAT, and the CM-SAF. UTH data sets are derived from AMSU-B and described in Buehler et al. (2008). UTH data sets derived from geostationary satellites have been used for the evaluation of climate models (Brogniez et al., 2005).

2.14.2.6.2 Cloud products The International Satellite Cloud Climatology Project (ISCCP) provided the first global climatology of cloud cover at an acceptable spatial resolution of 30  30 km (Rossow and Garder, 1993). For a limited area, Karlsson (2003) presented a cloud climatology from AVHRR observations for Scandinavia. With more advanced retrieval algorithms, MODIS continues the survey of cloud cover (Ackerman et al., 1998).

363

The ISCCP data have been successfully used to derive information on cloud physical properties, such as cloud phase, cloud optical depth, or cloud particle size (Rossow and Schiffer, 1999). In turn, these properties have been applied to derive parameters, such as the shortwave radiation budget (Gupta et al., 1999). Other global cloud climatologies are derived from AVHRR, such as the PATMOS climatology (Jacobowitz et al., 2003), or MODIS (Minnis et al., 2003; Platnick et al., 2003).

2.14.3 Water from the Atmosphere: Precipitation 2.14.3.1 Introduction Precipitation can be considered the most crucial link between the atmosphere and the surface in weather and climate processes. Quantitative precipitation estimates (QPEs) on high spatial and temporal resolutions are of increasing importance for water resources management, for improving the precipitation prediction scores in numerical weather prediction (NWP) models, and for monitoring seasonal to interannual climate variability. Although operational networks of weather radars are expanding over Europe and North America, large areas remain where information on the occurrence and intensity of rainfall is missing. Rain rate estimates from passive MW or VIS and IR imaging sensors on polar and/or geostationary orbiting satellites may bridge this gap, and provide quasi-global information on the spatial extent and intensity of rain.

2.14.3.2 Precipitation Measurements from Weather Radars Weather radars employ scattering of radio-frequency waves (5.6 GHz for C-band) to measure precipitation and other particles in the atmosphere (Rinehart, 2004). The intensity of the atmospheric echoes is converted to the so-called radar reflectivity factor Z using the Rayleigh scattering approximation (Probert-Jones, 1962):

Z¼

X

D6i

ð1Þ

i

where Di is the diameter of raindrop i and the summation is over all drops in a unit volume. Marshall and Palmer (1948) proposed a simple exponential form of the drop size distribution N(D) which is widely accepted:

60

NðDÞ ¼ N0 expðLDÞ

ð2Þ

WVPA (kg m−2)

50 40 30 20 10 0 Figure 7 EUMETSAT CM-SAF SSM/I-derived total column water vapor.

where the drop density N0 ¼ 8 103 mm1 m3 and L ¼ 4.1R0.21 mm1 depends on the rain rate R in mm h1. The radar reflectivity factor can be estimated from the sixth moment of the drop size distribution:

Z¼

Z

NðDÞD 6 dD ¼ 720 N0 =L7 ¼ 296R1:47

ð3Þ

with Z in mm6 m3. Many different Z–R power laws are used as the appropriate power law depends on climatic and actual meteorological circumstances (e.g., stratiform vs. convective precipitation). Apart from variations in the drop size

364

Observation of Hydrological Processes Using Remote Sensing

distribution, several other factors impact the quality of radarbased QPE (Rossa et al., 2005). The vertical profile of reflectivity (VPR) is, especially at higher latitudes, the major source of error in QPE deduced from weather radar observations (Joss and Waldvogel, 1990; Koistinen, 1991). At longer ranges, the height of observation will increase and in the presence of a significant gradient in the VPR this will typically generate an underestimation of the accumulated precipitation. Many different techniques have been developed to estimate the VPR and to subsequently correct the radar QPEs for this profile. The VPR can be estimated from weather radar data using climatological profiles, mean reflectivity profiles, or local profiles obtained at short ranges (Vignal and Krajewski, 2001). On the other hand, gauge adjustment techniques have been developed which correct the radar precipitation estimates using a second-order polynomial in range (Michelson et al., 2000). The radio frequency radiation transmitted and received by weather radar is scattered by precipitation. During very intense precipitation, scattering can become so strong that the radar beam is attenuated causing underestimation of precipitation intensity or even disappearance of the rain cells behind very strong cells. The observed radar reflectivity may be corrected for the attenuation when the one-way attenuation due to rainfall is approximated by a power law. However, the correction algorithm for attenuation is potentially highly unstable (Hitschfeld and Bordan, 1954). For the (near) future, dual-polarization weather radars offer promising new possibilities to correct for attenuation during intense rainfall (Bringi and Chandrasekar, 2001).

2.14.3.3 Precipitation Measurements from Satellite The reader can find an up-to-date review of satellite rainfall retrieval methods in Kidd et al. (2009) and Levizzani et al. (2007).

2.14.3.3.1 Retrievals from VIS–IR sensors Over the past decades, several rain rate retrieval methods based on observations from VIS and IR sensors were developed. The methods based on geostationary (GEO) satellites often use thermal IR observations and relate daily minimum cloud-top temperatures (Adler and Negri, 1988; Anagnostou et al., 1999) or cold cloud durations (CCD) to rain rates (Todd et al., 1995). These methods tend to perform reasonably well over areas where rainfall is governed by deep convection, but are less effective at higher latitudes, where precipitation originates from both convective and stratiform systems. A major limitation of the CCD methods is that rain rates are proportional to cloud duration, which is an assumption that fails in case high rain intensities occur over a short time period (Alemseged and Rientjes, 2007). Several methods have been developed that relate cloud physical properties, retrieved from passive imaging sensors, to precipitation. The GOES Multi-Spectral Rainfall Algorithm (GMSRA; Ba and Gruber, 2001) utilizes data from five channels, covering the VIS, near IR, water vapor, and two thermal channels, to extract information on the cloud and rain extent. Nauss and Kokhanovsky (2007) showed that cloud LWP retrievals from MODIS daytime observations are directly

proportional to the probability of rainfall. On the other hand, Rosenfeld and Gutman (1994) and Rosenfeld and Lensky (1998) found that clouds require droplets with effective radii 414 mm for the onset of precipitation. This is consistent with the findings of Twomey (1977), who reported that the precipitation efficiency of a given cloud depends on the size of the cloud droplets and the amount of aerosols in the air. Roebeling and Holleman (2009) present a novel approach, which uses information on cloud condensed water path, particle effective radius, and cloud thermodynamic phase to detect precipitating clouds, while information on condensed water path and cloud top height is used to estimate rain rates. The fact that their approach can be applied to GEO observations from the SEVIRI potentially allows for the provision of precipitation observations over large parts of the globe every 15 min. Figure 8 shows the effect of increasing threshold condensed water path and droplet effective radius values on the spatial extent of precipitation over the Netherlands as retrieved from SEVIRI.

2.14.3.3.2 Retrievals from passive MW sensors A more direct measurement of precipitation from satellite is made possible by the use of the MW frequencies as in this part of the spectrum precipitation-sized particles are the main source of atmospheric attenuation. Over ocean, the signal is mainly due to the increased emission of radiation from rain droplets so that rain areas appear warmer over the radiometrically ‘cold’ water background. Over land, rainfall is associated with scattering of the upwelling surface radiation by precipitation-related ice particles. The main problem of the passive-MW-based techniques is that the instruments are currently only available on low-Earth orbiting (LEO) satellites, and thus observations are available only twice per day per satellite (at best). Moreover, the resolutions of the measurements are for ocean rainfall products of the order of 50  50 km2, while over land they are typically no better than 10 10 km2. MW-based estimation techniques belong to two main groups: empirical techniques that calibrate the observations with surface data sets and physical techniques that minimize the difference between a modeled atmospheric rainfall event and the observation. An example of the physical techniques is the Goddard profiling (GPROF) technique (Kummerow et al., 2001) that uses a database of model-generated atmospheric profiles to which the observed satellite measurements are compared, and the best profile match is selected. The advantage of such a technique, first conceived for the TRMM TMI, is that it provides more information about the precipitation system than techniques that just provide information on surface rainfall. With the launch of sensors such as the Advanced Microwave Sounding Unit-B (AMSU-B, cross-track scanner) or the Special Sensor Microwave Imager/Sounder (SSMIS, conical scanner), higher frequency channels in the strong water vapor absorption lines at 183 GHz became available. The response functions of these channels peak at altitudes higher than 2 km and thus are much less influenced by ground emissivity features that represent a large portion of the errors in precipitation estimation over land. Several

Observation of Hydrological Processes Using Remote Sensing

365

30

% Precipitation

% Precipitation

15

r eT : 15.0 (µm)

25 20 15 10

Radar SEVIRI (PP)

5 0

0

200

400 CWPT (g

600

800

CWPT : 160.0 (gm−2)

10

5

0

Radar SEVIRI (PP)

0

6

12

m−2)

18

24

r eT (µm)

Figure 8 Relationship between spatial extents of precipitation and threshold condensed water path values (CWPT) for clouds with particle sizes larger than 15 mm (left), and threshold particle sizes (reT) for clouds with condensed water path values larger than 160 g m2 (right). The horizontal gray line indicates the spatial extent of precipitation derived from weather radar observations that were collocated and synchronized with the SEVIRI retrievals. Note that the optimum thresholds for the detection of precipitation are 160 gm2 for CWP and 15 mm for re.

56 52 48

52 48

0

4

4

−8

−4

0

4

20

15

10

48

48

−4

0

52

52

−8

−4

56

56

56

−8

5

0

Figure 9 22 November 2008. Precipitation retrieval (in mm h1) over NW France, the Channel and the UK using high-frequency AMSU-B MW channels for a mixed-type precipitating system (right) and radar retrieval from the NIMROD network (left). The circle delimits the area where both radar and satellite sense snowfall. Image courtesy of S. Laviola, ISAC-CNR.

algorithms are now available for operational applications, including detection of cloud droplets, snowfall, and snow on the ground (e.g., Ferraro et al., 2005; Laviola and Levizzani, 2008; Surussavadee and Staelin, 2008; Weng et al., 2003). An example of mixed-phase precipitation retrieval is shown in Figure 9.

2.14.3.4 Validation 2.14.3.4.1 Weather radar retrievals Surface networks of rain gauges can be used for both reduction of the gross errors and validation of quantitative precipitation estimates. Wilson (1970) pioneered with the integration of radar and rain gauge data and showed that this can improve the area rainfall measurements. A real-time calibration of radar-based surface rainfall estimates by telemetering rain

gauges was performed by Collier (1983) and an improved accuracy was seen on most locations. Nowadays, mean-field bias adjustment of radar-based quantitative precipitation estimates is widely used. At the Royal Netherlands Meteorological Institute (KNMI), mean-field bias adjustment with gauges is used operationally for an hourly updated QPE product (Holleman, 2007). An extensive spatial and temporal verification of the bias-adjusted radar composites over a 6-year period (2000–05) using dependent and independent gauge data is performed. It is found that the real-time adjustment scheme effectively removes the mean-field bias from the raw accumulations over a large area and that it substantially reduces the daily standard deviation. The adjustment method cannot correct for a rangedependent bias and it is recommended to use a simple VPR adjustment procedure for that.

366

Observation of Hydrological Processes Using Remote Sensing

2.14.3.4.2 Satellite retrievals Weather radar observation can be used to validate retrievals of the occurrence and intensity of precipitation from passive imaging satellites. Roebeling and Holleman (2009) compared 15-min SEVIRI retrievals of spatial extent of precipitation and rain rates against weather radar observations (Figure 10). The instantaneous rain rates from SEVIRI are retrieved with a high accuracy of about 0.1 mm h1, and a satisfactory precision of about 0.8 mm h1. SEVIRI is very accurate in detecting percentages of precipitation over larger domains (the Netherlands), which is shown by the high correlation of about 0.90 between spatial extents of precipitation from SEVIRI and weather radar. Similarly, the rain rates retrievals from SEVIRI correlate reasonably well with the weather radar observations (corr. ¼ 0.63). An international effort is being conducted by the International Precipitation Working Group (IPWG) to obtain reasonably homogeneous validation figures for the various satellite rainfall estimation algorithms over the various continents. Ebert et al. (2007) argued that the results of such a validation exercise so far confirm that the performance of satellite precipitation estimates is highly dependent on the rainfall regime and generally opposed to those of the NWP model Quantitative Precipitation Forecasts (QPF). Satellite estimates of rainfall occurrence and amount are most accurate during summer and at lower latitudes, whereas the NWP models show greatest skill during winter and at higher latitudes. In general, the more the precipitation regime tends toward deep convection, the more (less) accurate the satellite (model) estimates are.

2.14.3.5 Applications The third phase of the Network of European Meteorological Services (EUMETNET) Operational Program on the Exchange of Weather Radar Information (OPERA) is a joint effort of 30 European countries, which runs from 2007 till 2011, and is managed by KNMI. OPERA-3 is designed to firmly establish the Program as the host of the European Weather Radar Network. Currently, OPERA’s operational network consists of

more than 175 weather radars, of which roughly 100 systems have Doppler processing and about 15 systems have dualpolarization capability. In the coming years, the number of dual-polarization systems is expected to increase dramatically, thus offering new opportunities for quantitative precipitation estimation (Bringi and Chandrasekar, 2001). During this program phase, an OPERA Data Center (ODC) for the weather radar network should be specified, developed, and operated. This data center is crucial for reaching the main objective of OPERA-3, that is, establishing the weather radar networking as a solid element of the European infrastructure. The ODC will enhance and monitor availability of radar data, facilitate quality control of single-site radar data, stimulate exchange of volume radar data and quality information, and produce a homogeneous European radar composite. Furthermore, the ODC will deliver radar data to users inside and outside the National Meteorological Services. In May 2009, the EUMETNET Council has approved further development of the ODC. Start of operation of the ODC is planned for early 2011. More information on OPERA can be found in Holleman et al. (2008) and on the website. Satellite rainfall products, in spite of their intrinsic problems and not completely defined quality figures, are characterized by a global perspective that no other measurement method has. Because of this, their applications are numerous and cover very different fields such as meteorology, hydrology, civil protection, and climate. We will only mention a few examples without pretending of being complete. As is the case of radar precipitation measurement, the first field of application is in nowcasting, when a larger coverage than the one ensured by the radar is necessary. The satellite, in fact, ensures a mesoscale perspective, which becomes synoptic when LEO and GEO orbits are used. Another very important meteorological application is in data assimilation for NWP. Several physical (nudging) and variational methods have been developed in time at all scales. It is generally accepted that precipitation assimilation (e.g., Davolio and Buzzi, 2003) is more suited at the mesoscale, where current models start to incorporate the appropriate cloud parametrizations that general circulation models often lack.

Figure 10 9 June 2009, 12:30 UTC. An example of Opera rain rate composite (left) and MSG-SEVIRI rain rate retrievals (right) for Europe, presented in the projection of MSG.

Observation of Hydrological Processes Using Remote Sensing

Hydrological applications of satellite rainfall products span from the assimilation into hydrological models for basin management to global hydrological predictions. In all cases, uncertainty definition is the key to successfully use satellite data in this field (e.g., Voisin et al., 2008). Another important problem of current global products is the effect of orography on the retrieval (Adam et al., 2006). An upcoming very interesting application that merges a hydrological and a civil protection perspective is the one that uses satellite global rainfall data for landslide prediction (Hong and Adler, 2008); their methodology identifies landslide-prone areas on the basis of morphological information and rainfall for providing a hazard map. The number of climatological applications is expected to increase substantially over the next few years, given the global character of satellite data. The Global Precipitation Climatology Project (GPCP; Adler et al., 2003) has gathered global satellite rainfall estimations since 1979. GPCP products are now used to evaluate models and verify scenarios on the impacts of the various phenomena (ENSO, volcanic eruptions, etc.) on climate. The most important scenario is to verify whether global warming produces an acceleration of the global water cycle with more extremes (droughts on one side and extreme floods on the other) or not (e.g., Curtis et al., 2007). Finally, regional studies are conducted to examine the structure of propagating convective episodes in the warm season for their better forecasting and their modification in a climate perspective (e.g., Carbone and Tuttle, 2008; Laing et al., 2008).

2.14.4 Water to the Atmosphere – Evaporation 2.14.4.1 Introduction and Historic Development In the middle of the last century, ET from well-watered land surfaces was thought to be controlled by meteorological conditions, and only in the 1970s it was recognized that spatially and temporally dynamic feedback mechanisms between ET and land surface (e.g., albedo, rooting depth, and temperature) play an important role. This invoked the first applications of RS-based approaches, which mainly made use of airborne scanners (Bartholic et al., 1972; Idso et al., 1975; Jackson et al., 1977; Stone and Horton, 1974). It was only in the following decade that the first use of thermal data obtained from satellites to estimate ET was seen (Price, 1982; Seguin and Itier, 1983). They comprised of statistical approaches using linear relationships between daily totals of ET and net radiation and the difference between near-midday observations of radiant temperature and near-surface air temperature. Naturally, these linear relations needed local calibration, and the influences such as wind velocity, thermal stratification, and surface roughness were incorporated in later work (Riou et al., 1988), as such trying to bridge the gap already recognized by Seguin and Itier (1983) between sophisticated models useful for understanding basic processes and for performing informative simulations on the one hand and real estimation of ET on the other. These developments led to the general acceptance of the idea that spatial variability in ET is important, which in turn stimulated the development of effective methods for determining landscape-scale ET. However, still a need was noticed

367

for models that can realistically simulate the distributed nature of land surface processes and for techniques capable of upscaling estimates that are based on point-scale observations (Shuttleworth, 1988; Kalma and Calder, 1994), mainly used for validation. As observed by Kalma et al. (2008), this viewpoint developed more or less simultaneously with the use of airborne eddy correlation measurements (Schuepp et al., 1992; Mann and Lenschow, 1994), the development of scintillometry (de Bruin et al., 1995; Green et al., 2001), and an increased use of RS techniques. The main attraction of the last technique probably is the possibility of integration over a heterogeneous area at different resolutions and of routinely generating operational ET estimates. From 1990 onward until now, a vast amount of models have been developed and tested in a large number of multidisciplinary large-scale field experiments (Kustas and Goodrich, 1994; Shuttleworth et al., 1989; Kabat et al., 1997; Hollinger and Daughtry, 1999; Su et al., 2008, 2009). The increased understanding of the observed processes resulted in a number of excellent overview papers on both these processes and their typical impediments (Moran and Jackson, 1991; Becker and Li, 1995) as well as on the methodologies to estimate ET themselves by Kustas and Norman (1996) and Quattrochi and Luvall (1999) and more recently by Kalma et al. (2008). The models that have evolved mainly differ in type or purpose of the application which basically determines the type of RS data used and to which extent ancillary data are needed. What they all have in common is that the main input originating from RS is thermal information. It is obvious that no method or algorithm will outperform all other methods under all conditions and that a selection has to be based on the scale and purpose of the application as well as on the availability of the required data.

2.14.4.2 Current State of Science There are currently several methods being used, which can roughly be divided into three categories: they are either based on statistics and empirics, on spatial variability using either within image hydrological contrasts or some kind of index, or they are physically based, more specific on the energy balance at the Earth’s surface. As this chapter deals with the observation of hydrological processes, we will focus on the last category. For the sake of completeness, first, we briefly discuss the statistical and spatial variability methodologies followed by a description of current frequently used physical, or analytical, approaches.

2.14.4.2.1 Statistical approaches The methods using mainly statistical and empirical relations, also the first that were developed, make use of quasi-linear relationships between difference in daily amounts of ET and net radiation on one side and observed instantaneous differences between radiometric temperature and near-surface air temperature on the other. A prerequisite here is the use of near-midday temperature differences on clear days as these are representative of the entire day due to the regular course of climatic parameters during cloud-free days. These methods all originate from the work of Jackson et al. (1977), who derived a single statistical regression constant for the relation between

368

Observation of Hydrological Processes Using Remote Sensing

inverting an energy balance model by Boegh et al. (2002) and McVicar and Jupp (2002) to overcome this shortcoming.

the instantaneous temperature differences and daily ET and net radiation. Although later work incorporated aerodynamic surface properties on atmospheric stability effects (Seguin and Itier, 1983; Riou et al., 1988) and, as such, moved into the direction of a physically based energy balance approach (Nieuwenhuis et al., 1985; Soer, 1980; Lagouarde and McAneney, 1992), these types of approaches still require local calibration. Therefore, they are currently more often used in combination with scaled indices derived from scatterplots of midday temperature versus normalized difference vegetation index (NDVI; Carlson et al., 1995) as such reducing the need for local calibration and ancillary data input, making them more suitable for operational monitoring of ET, reaching accuracies of around 1 mm on a daily basis (Kustas and Norman, 1996).

2.14.4.2.3 Physical approaches This brings us to the physically based RS algorithms to derive ET estimates. They are all based on the idea that ET is a change of the liquid state of water to the gaseous state, hereby using available energy in the environment for vaporization. The available energy is the net radiation, which is the budget of all shortwave and longwave incoming and outgoing radiation at the Earth–atmosphere interface, less the heat used for heating up that interface, that is, the Earth’s surface, commonly referred to as the soil heat. The available energy is then thought to be used either for heating up the atmosphere, the so-called sensible heat, or for changing the state of water, the latent heat. Soil heat is generally considered a fraction of net radiation (Su, 2002; Norman et al., 1995; Anderson et al., 1997) depending on vegetation characteristics, and several studies have indicated that net radiation can be accurately determined from RS data (Timmermans et al., 2007; Boegh et al., 1999; Kustas and Norman, 1999; Su et al., 2001); the main remaining task is the division of the available energy between sensible and latent heat. The most widespread approach, also commonly used in land-surface modeling (Overgaard et al., 2006), is to consider the Earth–atmosphere interface, the Earth’s surface, as an electrical analog. Basically, this means that the rate of exchange (i.e., flux) of a quantity (e.g., temperature or vapor pressure) between two media (e.g., the Earth and the atmosphere) is driven by a difference in potential of that quantity, and controlled by a number of resistances that depend on the local climate as well as on the internal properties of the two media. The remote determination of vapor pressure is not feasible with the current state of technology. Therefore, the approach is to determine the rate of exchange of temperature between the Earth and the atmosphere, that is, the sensible heat flux, and determine the latent heat flux as a rest term. Dividing the latent heat flux by the latent heat of vaporization then yields ET. This means that current research efforts aim at the proper determination of the sensible heat flux. The different approaches to this problem are sketched in Figure 11. Basically, they differ in whether or not they discriminate between soil and canopy components. In Figure 11(a), the

2.14.4.2.2 Variability approaches

Soil

Canopy

Atmosphere

This brings us to the methods using the spatial variability within the image. They use either two-dimensional scatterplots of surface radiant temperature versus NDVI, the so-called triangle methods (Nemani and Running, 1989; Price, 1990) or the within-image relation between surface temperature and surface reflection (Bastiaanssen et al., 1998a; Roerink et al., 2000; Su et al., 1999), both based on the original work of Menenti and Choudhury (1993) to determine the hydrological wet and dry extremes. In the triangle approaches basically the observed position of a pixel within the scatterplot determines the ratio between actual and potential ET. The methods using temperature and reflective properties use a scaling between the observed wet and dry edges along the surface temperature, using either solely temperature such as in the S-SEBI approach (Roerink et al., 2000) or in combination with the use of a local surface roughness estimate as in the SEBAL approach (Bastiaanssen et al., 1998a). Whereas these methods do not need very accurate atmospheric correction techniques nor detailed meteorological inputs, they are limited by the fact that hydrological contrast needs to be present within the observed scene. In the case of the triangle methods, this is circumvented by comparison with a theoretically derived scatter triangle (Jiang and Islam, 2001; Carlson et al., 1994; Gillies et al., 1997; Venturini et al., 2004), whereas a temperature scaling was coupled to surface resistance by

(a)

Figure 11 Sketch of different resistance schemes.

(b)

(c)

Observation of Hydrological Processes Using Remote Sensing

sensible heat flux is driven by the difference between the aerodynamic surface temperature at the canopy source/sink height and the near-surface air temperature and controlled by a single aerodynamic resistance to sensible heat transfer between the canopy source/sink height and the air/atmosphere at a reference height above the canopy. The aerodynamic resistance is generally calculated from local wind speed, surface roughness length, and atmospheric stability (Brutsaert, 1982, 1992). Although these single-source models are known to give good results under a variety of conditions and environments (Kustas, 1990; Kustas et al., 1996; Bastiaanssen et al., 1998; Su, 2002; Jia et al., 2003), their main problem is that the necessary aerodynamic surface temperature at the mean canopy air stream is different from the radiometric surface temperature obtained from RS observations. This is usually corrected for by introducing an extra resistance that mainly depends on the inverse Stanton number, a dimensionless heat transfer coefficient that originates from the difference in source/sink heights for momentum and for heat transport. Although robust models exist to estimate this parameter (Massman, 1999; Su et al., 2001), it is known to vary widely, especially over sparse vegetation. This has led to the so-called dual-source models that treat the soil and canopy separately. Two different approaches are noticed. When the surface, or pixel, is divided into different fractions of bare soil and vegetation, the soil and vegetation components do not interact. These models, first introduced by Avissar and Pielke (1989), are known as patch models, or parallel resistance network as they are more frequently named in the RS community (see Figure 11(b)). However, when sparse vegetation is present, the soil and vegetation components are known to interact and a so-called series resistance network is more appropriate. In this case, the canopy consists of a semitransparent layer located above the soil surface such that heat and moisture have to enter or leave the surface layer through the canopy layer, whereby the component fluxes are allowed to interact (see Figure 11(c)). The structure proposed by Shuttleworth and Wallace (1985) is most widely used and incorporates a bulk stomata resistance for the vegetation as well as a resistance controlling the soil fluxes. It is assumed that aerodynamic mixing within the canopy invokes a mean canopy-airflow where fluxes from the components are allowed to interact after which they are exchanged with the atmosphere, controlled by a third aerodynamic resistance. Both structures require component temperatures, whereas a remote sensor only observes the effective radiometric surface temperature, which is a combination of the component temperatures, depending on viewing angle and fractional vegetation cover. To derive the component temperatures from the effective temperature, additional information is required. Several methods have been developed, ranging from empirical relationships (Lhomme et al., 1994), via coupling to a crop growth model (Chehbouni et al., 1996; Chehbouni et al., 1997) and the NDVI–surface temperature relationship (Boegh et al., 1999), to dual viewing angle approaches (Francois et al., 1997; Kustas and Norman, 1997; Merlin and Chehbouni, 2004). A different approach was developed by Norman et al. (1995) where transpiration initially is estimated through the Priestley–Taylor equation, whereby they were able to relate the canopy temperature to air temperature. This allowed initial

369

guesses of component temperatures, which were then used in an iterative procedure to derive soil evaporation and canopy transpiration that satisfied the energy balance. In their original paper, Norman et al. (1995) described an operational procedure both for the series as well as for the parallel approach that, in different improved versions (Kustas et al., 2001; Anderson et al., 1997, 2005; Kustas and Norman 1999, 2000), is nowadays widely used (Sanchez et al., 2008; French et al., 2002; Li et al., 2005). By now it may be clear that using RS observations to derive latent heat fluxes, or ET, requires a certain amount of assumptions depending on the model and data used as well as on the purpose of the application. As such, it is no surprise that different techniques lead to deviating estimates (Zhan et al., 1996; French et al., 2005; Timmermans et al., 2007) and works still need to be undertaken to minimize those deviations.

2.14.4.3 Future Research Needs From the previous some residual challenges and thus future directions of research follow. They relate to spatial and temporal scaling issues, coupling and feedback issues, and, last but not least, validation issues. Although they may be categorized, they are discussed here in a coherent manner, as most of them are interrelated.

2.14.4.3.1 Scaling Depending on the application purpose, models describing the land–atmosphere interaction assume that both processes and variables are scale invariant (Menenti et al., 2004), which means that it is assumed that the relation of observations with model variables is the same at all spatial scales. Intermodel variability of predicted fluxes is therefore often large and causes are difficult to pinpoint (Menenti et al., 2004), which is probably the reason why only a few pixel-by-pixel flux comparisons (Timmermans et al., 2007; French et al., 2005; Boegh et al., 2004; Timmermans et al., 2009; De Lathauwer et al., 2009) are made (Overgaard et al., 2006). An in-depth analysis is needed of the nature of feasible observations in the soil– vegetation–atmosphere system at different (Tol et al., 2009; Timmermans et al., 2009) and multiple scales (McCabe et al., 2006) to detect and understand inconsistencies in model variables and parametrizations. There is also a need for improved temporal scaling procedures to extrapolate instantaneous estimates of ET derived from RS platforms to hourly, daily or longer periods (Kalma et al., 2008). Concepts most widely used so far (Shuttleworth et al., 1989; Batra et al., 2006) to extrapolate to daily values yield unsatisfying results, especially over drying surfaces (Chehbouni et al., 2008; Gentine et al., 2007). Given the fact that cloudy conditions hamper the frequent remote observation of ET, alternative approaches have to be explored, especially on timescales longer than 1 day.

2.14.4.3.2 Feedbacks Describing transfer of energy into the atmosphere using the energy balance methods generally invokes assuming homogeneous atmospheric properties. This requires neglecting fast changes in air temperature and humidity, and thus in fluxes

370

Observation of Hydrological Processes Using Remote Sensing

thereof, due to turbulence. These changes occur at small spatial scales implying that spatial variability of the atmospheric properties at a given time is significant. A more realistic description of the structure and dynamics of the atmosphere is obtained by large eddy simulation (Albertson, 1996; Albertson et al., 2001). Opportunities to improve current RS-based ET estimates therefore include testing the spatial validity of the meteorological data used (Gowda et al., 2007). In addition, there is a critical need to understand the feedbacks between the land surface and the atmosphere at various scales (Wood, 1998). Feedback between land and atmosphere arises from the fact that the fluxes of heat and water from the land surface to the atmosphere will change the properties of the atmosphere, which in turn will change the fluxes. Therefore, more work is required in the line of Bertoldi et al. (2007) and Timmermans et al. (2008), who examined feedback effects at multiple scales using an RS-based energy balance model dynamically coupled to a large eddy simulation model.

2.14.4.3.3 Validation Apart from the ongoing discussion on the mismatch between available energy observations and turbulent flux measurements from eddy correlation (Foken, 2008) resulting in uncertainties of up to 30% in validation data, there is also considerable doubt on the applicability of scintillometry over very heterogeneous terrain (Timmermans et al., 2009; Ezzahar et al., 2007). Moreover, RS-based energy balance models tend to be validated versus a handful of tower-based measurements, which does not ensure a reliable performance over the broader landscape. To address this uncertainty, intercomparisons of flux model output need to be performed as reported by Timmermans et al. (2007) and French et al. (2005). In addition, a dynamic coupling of distributed hydrological and atmospheric models through an RS-based surface energy balance model, such as Timmermans et al. (2008), Velde et al. (2009), and Bertoldi et al. (2007), is vital for future applications and probably improves possibilities for making a more spatially detailed evaluation (Overgaard et al., 2006). To summarize, advances in improving parametrization and validation of physically based ET models will rely heavily on the understanding of physical processes at different scales as well as on the ability to obtain distributed physical information. In order to achieve this, satellite EO will prove to be of paramount importance in the future.

2.14.5 Water on the Land – Snow and Ice 2.14.5.1 Introduction The seasonal and perennial snow and ice masses (the cryosphere) cover a major part of the land surfaces. They are essential or dominating elements of the hydrological cycle in mid- and high latitudes, as well as in many mountain areas. The terrestrial cryosphere comprises the seasonal snow cover, lake and river ice, permafrost, seasonally frozen ground, glaciers, ice caps, and the large ice sheets of Greenland and Antarctica. Of these, seasonal snow cover and frozen ground on land dominate in spatial extent and temporal variability, covering at maximum about 50% of the land area in the Northern Hemisphere. Due to feedbacks with the atmosphere

and other elements of the hydrosphere, the cryosphere responds very sensitively to climate warming, as reports on past and ongoing changes of the snow and ice masses confirm (Lemke et al., 2007). Due to the large spatial extent and temporal variability of snow and ice coverage, RS techniques provide the only feasible means for timely and comprehensive observation of these elements of the Earth system. The potential of RS for monitoring snow and ice has been recognized already in the 1960s, applying optical imaging sensors of the NOAA satellites to mapping the global snow cover on a weekly basis (Robinson et al., 1993). Thanks to advancements in sensor technology, the 1970s brought in a big step forward in satellite-borne RS, including observations of the cryosphere. Optical sensors of improved spatial and spectral resolution and new active and passive MW sensors opened up the opportunity to monitor all the individual elements of the global cryosphere. Already at that time, RS became an indispensable tool for snow and ice monitoring and research that further evolved over the years, thanks to advancements in sensor technology and data processing (Key et al. 2007). Airborne sensors play an important role in the development of techniques for data processing and analysis, as well as in local to regional surveys of snow and ice. However, due to the near-global coverage and the regular repeat capabilities, satellite-borne sensors are the main tool for snow and ice monitoring. Therefore, in this chapter, we focus on applications of satellite sensors.

2.14.5.2 Techniques for Retrieval of Extent and Physical Properties of Snow and Ice Sensors in the VIS, infrared, and MW part of the electromagnetic spectrum are employed to monitor the extent and physical properties of snow and ice. In order to explain the information content of the various sensor types, the main features affecting the radiance reflected or emitted by snow and ice are summarized below. Electromagnetic waves, incident on a snow or ice medium, are subject to scattering at volume inhomogeneities (snow grains and air bubbles in ice) and absorption along the propagation path. In the case of melting snow, water adds as a third component of the mixture. The absorption and scattering characteristics are determined by the dielectric and structural properties of the medium and the sensor wavelength. At VIS wavelengths, the dielectric losses of ice and water are small, but increase considerably in the near- and mid-IR. In the thermal IR, snow is almost a black body (emissivity 0.99). Consequently, clean fresh snow has a high reflectance in the visible part of the spectrum (0.9oRVISo0.99), dropping to Ro0.1 in the shortwave IR at wavelengths X1.5 mm. The spectral reflectance in the visible decreases significantly with aging of snow due to pollution. In the near IR, between 0.9 and 1.3 mm, the reflectance decreases with increasing size of the snow grains, which is used to estimate this parameter from satellite measurements (Dozier and Painter, 2004). The direct effect of liquid water in a snow pack on near-IR reflectance is small, although the reflectivity decreases because melt metamorphosis causes snow grains to grow.

Observation of Hydrological Processes Using Remote Sensing

The decrease of reflectance in the IR is employed by the normalized difference snow index (NDSI) for discriminating snow cover and snow-free surfaces:

NDSI ¼

RVIS  RSWIR RVIS þ RSWIR

ð4Þ

The automated MODIS snow-mapping algorithm uses at satellite reflectances in MODIS bands 4 (0.545–0.565 mm) and 6 (1.628–1.652 mm) to calculate the NDSI (Hall et al., 2002). Different thresholds of the NSDI are used to detect snow in forested areas and open land (Salminen et al., 2009). For excluding cloud-covered pixels, the quality flag from the MODIS cloud-masking algorithm is applied which uses visible, SWIR, and thermal IR channels to detect clouds (Ackerman et al., 1998). In the case of patchy snow cover, the binary classification shows a trend of overestimating the total snow area. To account for these effects, spectral unmixing techniques, using VIS and near-IR channels to map snow cover fraction at subpixel scale, are applied (Vikhamar and Solberg, 2003; Dozier and Painter, 2004; Sirguey et al., 2009). Active MW sensors (synthetic aperture radar, SAR) and passive MW sensors (radiometers) are widely applied for mapping the extent and physical properties of the snow cover. For interpreting and analyzing MW measurements of snow, it is essential to consider the layers contributing to the observed signal. The penetration depth, dp, can be computed from the complex permittivity (e ¼ e0 – ie00 ) by

dp ¼

pffiffiffiffi l e0 2p e00

ð5Þ

The imaginary part of the permittivity of snow, e00 , and, therefore, also dp, shows a strong dependence on the liquid water content (Ma¨tzler, 1987). The dielectric losses in dry snow are small, and the penetration depth is of the order of several hundred wavelengths (e.g., about 20 m for C-band SAR with l ¼ 5.6 cm). On the other hand, the penetration depth in wet snow is only about one wavelength or less due to the high dielectric losses of water. In the C-band, for example, the penetration dp in snow with 5% by volume of liquid water is only 3 cm. This has an important impact on the signal observed by MW sensors. For wet snow, the MW signal reflected or emitted from a melting snow pack originates from a thin top snow layer and the snow surface, whereas for dry snow both the snow volume and the medium below the snow pack contribute to the observed signal (Rott, 1997). These properties cause distinct differences in the information provided by the various MW sensors. Imaging radars for snow mapping typically operate in the C- and X-band (l ¼ 5.6 cm, l ¼ 3 cm). At these wavelengths, the scattering contribution of a dry winter snow pack is small, and the backscatter contribution of the ground surface dominates. On the other hand, due to the high absorption losses, the radar reflectivity of melting snow is rather low. These characteristics enable to map melting snow areas by means of C- and X-band SAR, applying a change detection algorithm using SAR image time series (Nagler and Rott, 2000). Combining optical and SAR sensors for snow area mapping helps to overcome the cloud handicap of optical sensors which is a particular

371

problem for updating snow extent in the ephemeral melting snow zones (Solberg et al., 2008). For observing snow water equivalent (SWE), a critical parameter for snow hydrology, shorter MW wavelengths need to be employed to obtain a distinct signal of the snow volume. The radiance emitted by the ground is attenuated in the dry snow pack by scattering at the snow grains. The attenuation due to volume scattering increases inversely to the third power of the wavelength (Bl3) (Hallikainen et al., 1987). The scattering losses depend on snow depth, density, and grain size. Currently, no satellite-borne imaging radar systems are available at short wavelengths, but MW radiometers are applied to map the depth and water equivalent of the snow pack. Retrieval of SWE from passive MW data is conventionally based on empirically determined relationships between SWE and emitted brightness temperature (TB). Standard procedures apply the difference in TB at 37 GHz (l ¼ 0.8 cm) and 19 GHz (l ¼ 1.6 cm) to estimate SWE (Foster et al., 2005). In order to compensate for effects of grain size, the parameters of the retrieval algorithms need to be tuned to regional snow conditions (Derksen et al., 2003). Another option for compensating grain size effects is the assimilation of in situ snow measurements in the SWE processing line (Pulliainen, 2006). Due to the coarse resolution of the sensors and the saturation of the signal in deep snow, radiometric SWE retrievals are subject to major errors in mountain areas and forests. Satellite-borne RS is widely applied for mapping the extent, surface topography, and motion of glaciers. For glacier mapping, spectral ratios in optical imagery are applied, similar to the techniques for snow mapping (Kargel et al., 2005; Paul et al., 2002). Manual post-processing is required to correct for debris-covered glacier surfaces. Stereo-optical satellite imagery (ASTER, SPOT-5) is applied to map surface topography, but the limited radiometric contrast reduces the accuracy in the snow areas (Berthier and Toutin, 2008). This problem can be overcome by radar interferometry (InSAR). Single-pass interferometry with two antennas on a platform, as on the Shuttle Radar Topography Mission (SRTM), avoids the problem of temporal decorrelation of the radar signal. The SRTM data set, acquired in February 2000, is the basis of a freely available DEM (90 m grid) covering the land surfaces between 601 N and 561 S (Rodriguez et al., 2005). Repeat-pass SAR images enable the mapping of ice motion at high accuracy by means of differential processing techniques. Differential InSAR processing techniques are applied to separate the phase contributions of surface motion and topography (Hanssen, 2001). However, decorrelation of the radar phase due to snowfall, wind drift, or melt in the time interval between the image acquisitions severely limits the application of repeat-pass interferometry over snow and ice. A unique InSAR data set for glacier studies, less affected by decorrelation, was acquired during the concurrent (‘tandem’) operation of the satellites ERS-1 and ERS-2 in the years 1995– 99 (Weydahl, 2001). During the tandem phase, the two satellites imaged the same swath on the Earth’s surface at a time difference of 24 h. If stable features are apparent on a glacier surface, image correlation techniques can be applied to map glacier motion from repeat-pass images of high-resolution optical sensors and SAR. This technique is less sensitive to motion than InSAR, but

372

Observation of Hydrological Processes Using Remote Sensing

does not require phase coherence (Strozzi et al., 2002). The new very high resolution SAR systems of TerraSAR-X and COSMO-SkyMed are very useful for this application (Floricioiu et al., 2008).

2.14.5.3 Examples of Products and Applications Satellite-derived products on snow and land ice have been widely used in research and are increasingly applied also for operational applications in hydrology and water management. Table 2 provides an overview on some key snow and land ice products, including a few links to sample data sets. Medium-resolution optical sensors (e.g., AVHRR on NOAA and MODIS on the Terra and Aqua platforms) are the main data sources for snow mapping at national to global scales. These data are widely applied for studies in climate research and hydrology of snow-covered regions (e.g., Brown et al., 2008; Pu et al., 2007; Rodell and Houser, 2004; Shamir and Georgakakos, 2006). An example for a snow map derived ¨ tztal basin from MODIS data is shown in Figure 12 for the O in the Austrian Alps. The inset shows the area-altitude distribution of the snow cover, which is used as input to a semidistributed model for simulating and forecasting snowmelt runoff (Nagler et al., 2008). Daily snow maps are often rather fragmentary due to cloud cover, so that for some applications (e.g., climate studies) weekly composites are preferred. MODIS daily snow maps, 8-day composites, and monthly fractional snow cover can be found on the Internet. Sensors at higher spatial and/or spectral resolution are used for regional studies of snow physical properties and snow distribution (Dozier and Painter, 2004; Molotoch, 2009), but usually lack the temporal sequence required for real-time runoff forecasting applications. SAR data are used for regional snow mapping, with emphasis on snow depletion during the melt period, exploiting the sensitivity of the sensors for detecting melting snow. Preferably, SAR data of the wide swath mode (ScanSAR) are Table 2

used, providing a swath width of 400 km (Envisat ASAR) and 500 km (Radarsat) (Luojus et al., 2007; Nagler and Rott, 2005). SAR-derived snow maps are applied for snowmelt runoff modeling and forecasting (Nagler et al., 2008), for snow cover modeling linked to regional meteorological models (Longe´pe´ et al., 2009), and for climate studies. Global maps of snow depth and water equivalent, derived from satellite-borne multichannel MW radiometer data reaching back to 1979, are available for climate studies (Foster and Chang, 1993). However, in many regions, the data show systematic differences to in situ measurements, requiring further improvement of retrieval algorithms (Foster et al., 2005). In western Canada, weekly SWE maps retrieved from satellite MW radiometer data are produced on an operational basis since the 1980s (Derksen et al., 2003). An example of such a product for the Canadian Prairies is shown in Figure 13. Because the retrieval parameters are tuned for regional snow morphology, these SWE maps provide better accuracy than the global products. EO satellite data are widely applied for compiling and updating glacier inventories and provide key input data for models of glacier mass balance, hydrology, and ice dynamics. The main satellite products for glacier research and monitoring applications are maps of glacier area, topography, surface velocity, diagenetic facies, and albedo. The Global Land Ice Measurements from Space (GLIMS) initiative is aimed at compiling a global data base of glacier outlines in digital format from optical satellite data (Raup et al., 2007). The database, available to the public, includes satellite image glacier maps, vector outlines and related metadata, and, optionally, also snow lines, center flow lines, hypsometry data, and surface velocity fields. Observations of the temporal evolution of the snowline during the ablation period are used as input for modeling glacier mass balance and runoff (Rott et al., 2008). Satellite observations of changes in ice surface elevation and ice fluxes are also very relevant to mass balance studies (Bamber and Rivera, 2007). SAR interferometry is an

Overview on selected snow and land ice products derived from satellite observations

Product type

Sensor type

Spatial resolution (typical range)

Sensors (examples)

Selected data sets

Snow area (total)

Multispectral optical imager SAR, scatterometer Multispectral optical imager Imaging microwave radiometer

30 m–1 km

Modis, avhrr, landsat

Global snow area: http://modis-snowice.gsfc.nasa.gov/

30–100 m 250 m–1 km

Asar, radarsat Modis, meris

25 km

Ssm/i, amsr

Multispectral optical imager, SAR

30–250 m

Modis, asar, radarsat

Multispectral optical imager Interferometric SAR, optical stereo imager SAR, optical imager

5–30 m

Spot hrv, aster, landsat

10–100 m

Spot hrv, aster, srtm, asar ASAR, radarsat, terrasar-X, optical

Snow area (melting) Snow albedo Snow water equivalent

Lake and river ice extent and concentration Glacier outlines Glacier surface topography Glacier motion

3–30 m

http://www-modis.bu.edu/brdf/userguide/ albedo.html Amsr-e/aqua daily l3 global snow water equivalent: http://www.nsidc.org/data/ ae_dysno.html http://www.polarview.org/services/lim.htm http://www.polarview.org/services/rim.htm Glims glacier data base http://nsidc.org/glims/ SRTM data products: http://www2.jpl.nasa.gov/ srtm/cbanddataproducts.html

Observation of Hydrological Processes Using Remote Sensing

373

Figure 12 Snow map of the O¨tzal Alps, Austria, derived from MODIS image data, 25 April 2007. Superimposed to Google Earth. (Inset) Area-altitude distribution of snow-covered and snow-free surfaces in the sub-basin Vent.

important tool for studying glacier hydraulics (Magnu´sson et al., 2007) and provides detailed maps of ice flow, which can be used for estimating the ice export of calving glaciers (Stuefer et al., 2007).

2.14.5.4 Future Research Needs Currently available satellite missions and sensors are providing important information on the extent and physical properties of snow and ice from local to regional and global scales. This potential has been utilized so far mainly for dedicated research studies in the fields of water balance and hydrology, surface energy fluxes, land surface processes, and Earth surface/ atmosphere interactions. However, the potential of remotely sensed cryosphere data for process modeling has so far been rarely exploited. Fostering the use of spatially distributed snow data requires further advancements of data assimilation techniques, a topic that has gained in importance over the last years (e.g., Clark et al., 2006; Kolberg et al., 2006; Nagler et al., 2008; Rodell and Houser, 2004; Slater and Clark, 2006). Regarding sensors and satellites, a large variety of imaging sensors in the optical and MW spectral range is available, many of which can be employed for snow and ice observations. However, many sensors lack continuity, which represents an obstacle for operational use in hydrology and water management. New initiatives will provide better continuity of observations, such as the Sentinel satellites within the GMES initiative of the ESA and the European Union. A major observational deficit is the lack of a sensor for spatially detailed

observations of the snow mass (SWE), a key parameter of the water balance. The feasibility of a satellite mission for SWE mapping with dual frequency (X-band and Ku-band) SAR is presently studied by ESA (Kern et al., 2008).

2.14.6 Water on the Land – Surface Water, River Flows, and Wetlands (Altimetry) 2.14.6.1 Introduction Terrestrial surface water is absolutely essential to life, economies, environment, climate, and weather. Both national and local economies rely on flowing rivers to transport storm waters, sewage, and other effluents away from cities besides offering major shipping lanes to inland areas. The ecologies of wetlands and floodplains depend on surface water flows to deliver nutrients and to exchange carbon and sediments. Surface waters play a role in global climate through energy and water mass exchange with the lower atmosphere. Moreover, local weather is strongly affected by the surface area of nearby water bodies. Runoff is a strong indicator of accumulated precipitation throughout a watershed, and large, periodically flooded wetlands provide vast surfaces for evaporation as well as water storage. Earth’s 6 billion people critically rely upon surface water availability for domestic use, agriculture, and industry, while human health is impacted by waterborne diseases (e.g., disease-vector-related malaria). National defense issues are related to surface water, particularly via politically charged water impoundment projects. The global

374

Observation of Hydrological Processes Using Remote Sensing

Snow water equivalent (mm) 0

10

20 18 50

01/03/03

30

40

50

60

70

80

90

100 110 120

No data Treeline

43

27 45

Wet snow

55 35

56 56

40 25

19 45

31

27 56 47 46

55 28 44 23

25

27 21 8

41

45 25 30

26

24 13 20

33 15 26

13 3

29

4 3

35 27 39

12 22 17 14 19

15 4 5

41 52

35 10

60 50

35 47

29 33

22 22

16 16 13

40 58 18 47 19

23

32 8 19 28 Depth of snow on ground (cm) 01/03/03 1971-00 Normal

MSC Climate Research Branch SMC Climate processes and earth observation division

Figure 13 Map of snow water equivalent (color coded) over the Canadian Prairies, derived from satellite microwave radiometer measurements. The numbers refer to SWE measured at snow stations. Courtesy: Meteorological Service of Canada.

water issues will have large effects on many of the world’s major decisions in the next decades and will require operational monitoring tools to support water policies. The following sections provide a review of the surface-water measurements available on the ground, onboard satellites, and a description of the future satellite mission Surface Water and Ocean Topography (SWOT), first satellite mission dedicated to the hydrology of continental surface water.

2.14.6.2 In Situ Measurements In situ gauging networks have been installed for several decades in many river basins, distributed nonuniformly throughout the world. In situ measurements provide time series of water levels and discharge rates, which are used for studies of regional climate variability as well as for socioeconomic applications (e.g., water resources allocation, navigation, land use, infrastructures, hydroelectric energy, and flood hazards) and environmental studies (rivers, lakes, wetlands, and floodplain ecohydrology). In situ methods are essentially a one-dimensional, point-based sampling of the water surface that relies on well-defined channel boundaries to confine the flow. Yet, water flow and storage changes across wetlands and floodplains are spatially complex with both vast

diffusive flows and narrow confined hydraulics. This complexity is fundamentally a three-dimensional process varying in space and time, which cannot be adequately sampled with one-dimensional approaches. In addition, gauging stations are scarce or even absent in parts of large river basins due to geographical, political, or economic limitations. For example, over 20% of the freshwater discharge to the Arctic Ocean is ungauged and surface water across much of Africa and portions of the Arctic either is not measured or has experienced the loss of over two-thirds of the gauges (Stokstad, 1999). Therefore, our ability to measure, monitor, and forecast global supplies of freshwater using in situ methods is essentially impossible because of (1) the decline in the numbers of gauges worldwide (Vo¨ro¨smarty et al., 2001), (2) the poor economic and infrastructure problems that exist for nonindustrialized nations, and (3) the physics of water flow across vast lowlands.

2.14.6.3 RS Techniques During the past decade, RS techniques (satellite altimetry, radar and optical imagery, active and passive MW techniques, InSAR, and space gravimetry) have been used to monitor some components of the water cycle in large river basins (Cazenave et al., 2004). Radar altimetry, in particular, has been used

Observation of Hydrological Processes Using Remote Sensing

2.14.6.4 Validation and Synergy of RS Techniques Surface water levels estimated from conventional nadir altimetry have been compared to those obtained from in situ

gauges located along the satellite tracks and in the proximity of the altimeter swath over many of the largest river basins. The rms differences between in situ and altimetry-derived water levels have been computed and are usually in the order of a few to several tens of centimeters (Kouraev et al., 2004). Combining nadir altimetry-derived water levels with satellite imagery provides a new method for remotely measuring surface water volumes over large floodplains. Figure 15 shows an example of the interannual surface water volume signal variability obtained with a combination of altimetry and NDVI data from the SPOT-4/Vegetation instrument over the lower Mekong River Basin compared with the GRACE signal (black) that integrates surface and underground water

Mekong Basin 40 Water volume (km3 month−1)

extensively in the recent years to monitor water levels of lakes, rivers, inland seas, floodplains, and wetlands (e.g., Birkett, 1995, 1998; Birkett et al., 2002; Mercier et al., 2002, Maheu et al., 2003; Kouraev et al., 2004). A few examples of altimetry-derived water level time series over rivers are presented in Figure 14. Nadir-viewing altimetry has a number of limitations over land because radar waveforms (e.g., raw radar altimetry echoes after reflection on the land surface) are complex and multipeaked due to interfering reflections from water, vegetation canopy, and rough topography. These effects result in less valid data than over oceans. Systematic reprocessing of raw radar waveforms with optimized algorithms provides decade-long time series of terrestrial water levels, at least over large (41 km width) rivers. Repeat-pass SAR interferometry has been shown to offer important information about floodplains in measuring small water-level changes (Alsdorf et al., 2000). Poor temporal resolutions are associated with repeat-pass interferometric SAR. Off-nadir single-pass interferometric SAR does not work over open water; instead, it requires special hydrogeomorphologies of flooded vegetation (Alsdorf et al., 2000; Lu et al., 2005; Kim et al., 2005). Optical sensors are used to provide estimates of surface water extent under favorable conditions when there are few or no clouds. The GRACE gravimetry mission provides estimates of water volume but its resolution, on the order of 400 km, is poor (Tapley et al., 2004). Although the Shuttle Radar Topography Mission (SRTM) produced a high spatial resolution image of heights, the errors over water surfaces are quite large and the mission was active for a sampling period of only 11 days in February 2000, preventing temporal change studies (e.g., 75.5 m; LeFavour and Alsdorf, 2005).

375

30 20 10 0 −10 −20 −30 −40 1998

1999

2000

2001

2002

2003

2004

2005

Year Figure 15 Variation of surface-water volume from ERS-2/ENVISAT altimetry and SPOT-VGT imagery (green) and total land water volume from GRACE (black) (Frappart et al., 2006).

Figure 14 Water-level time series over the Niger (upper panel; left), Yangtze (upper panel; right), Indus (lower panel; left), and Danube (lower panel; right) based on Topex–Poseidon altimetry. From http://www.legos.obs-mip.fr/en/soa/hydrologie/hydroweb/.

376

Observation of Hydrological Processes Using Remote Sensing

(Frappart et al., 2006). In the near future, when GRACE observations improve in terms of geographical resolution, it will be possible to estimate change in water volumes stored in soil and underground reservoirs by using in synergy GRACE, altimetry, and imagery data.

2.14.6.5 Availability of the Satellite Data Sets A recently developed water-level database for major rivers, lakes, and wetlands using altimetry measurements from Topex/ Poseidon, Jason-1, ERS-2, ENVISAT, and GFO satellites can be accessed through the Internet. The database includes water levels for over 130 lakes and man-made reservoirs, 250 virtual stations on rivers, and about 100 sites on flooded areas. The time series are regularly updated and the number of sites increases regularly. Users have access to associated errors. For optical sensors, several databases are available through the web. For instance, the SPOT VEGETATION products can be found on the website where the NDVI products are available; for the MERIS instrument, the data can be found on the Internet.

2.14.6.6 SWOT: The Future Satellite Mission Dedicated to Surface Hydrology The currently operating radar altimeters built to sample the surface of the open ocean miss numerous water bodies

between orbital tracks. Optical sensors cannot penetrate the canopy of inundated vegetation and typically fail to image water surfaces when clouds or smoke is present (e.g., Smith, 1997). The prevalent vegetation and atmospheric conditions in the tropics lead to much reduced performances for technologies operating in and near the optical spectrum. Hydraulic measurements with repetitive global coverage of the continental surface water are needed to accurately model the water cycle and to guide water management (Alsdorf et al., 2003; Alsdorf and Lettenmaier, 2003). The future satellite mission SWOT dedicated to continental surface hydrology in cooperation between NASA and CNES will contribute to a fundamental understanding of the global water cycle by providing for the first time global measurements of terrestrial surface water storage changes and discharge, which are critical for present and future climate modeling (Mognard and Alsdorf, 2006; Alsdorf et al., 2007; Mognard et al., 2007; Fu et al., 2009). The Ka-band Radar Interferometer (KaRIN) (Figure 16) is the technology capable of supplying the required imaging capability of water level (h) with global coverage at least twice every 21 days. KaRIN has two Ka-band SAR antennae at opposite ends of a 10-m boom with both antennae transmitting and receiving the emitted radar pulses along both sides of the orbital track. Look angles are limited to less than 4.51 providing a 120-km-wide swath. The 200-MHz bandwidth achieves cross-track ground

ay r arr

Sola

Interferometer antenna 1

Interferomete antenna 2

10 m baseline

950 km

Nadir altimeter

Surface water topography

Ocean topography H-pol interometer swath 10−70 km

Nadir altimeter path

V-pol inteferometer swath 10−70 km

Figure 16 Artistic view of the satellite SWOT with the Ka-band Radar Interferometer (KaRIN) instrument.

Intrinsic resolution from 2 m × 70 m to 2 m × 10 m

Observation of Hydrological Processes Using Remote Sensing

resolutions varying from about 10 m in the far swath to about 60 m in the near swath. A resolution of about 2 m in the along track direction is derived by means of synthetic aperture processing. SWOT will contribute to a fundamental understanding of the global water cycle by providing global measurements of terrestrial surface-water storage changes and discharge, which are critical for present and future climate modeling. SWOT will facilitate societal needs by (1) improving our understanding of flood hazards by measuring flood waves and water elevations, which are critical for hydrodynamic models; (2) freely providing water volume information to countries that critically rely on rivers that cross political borders; and (3) mapping the variations in water bodies that contribute to disease vectors (e.g., malaria).

2.14.7 Water in the Ground – Soil Moisture 2.14.7.1 Introduction Soil moisture is defined as the amount of water in the rooting zone, or any other depth in the unsaturated zone and is usually expressed in volumetric percentage (Hillel, 1998). It is a variable that has always been required in many disciplinary and cross-cutting scientific and operational applications such as numerical weather prediction, ecology, biogeochemical cycles, flood forecasting, etc. (Jackson et al., 1999). With increasing evidence of climate change, it becomes even more urgent to be able to elucidate the critical role of soil moisture. Unfortunately, soil moisture is notoriously difficult to observe at large (landscape to global) scale due to its high spatial and temporal variability. Most of our limited understanding of the role of soil moisture in meteorology, hydrology, ecology, and biogeochemistry has been developed from point to field-scale studies, where the emphasis was typically on the variation of soil moisture with depth. Our failure to translate this smallscale understanding to natural landscapes can be attributed largely to our lack of understanding of soil moisture variability at larger spatial scales. As a parallel consequence, most models have been designed around the available point data and do not reflect spatial variability (Leese et al., 2000). The potential to use MW RS for measuring soil moisture has been recognized early (Eagleman and Ulaby, 1975). The theoretical basis for measuring soil moisture at MW frequencies lies in the large contrast between the dielectric properties of liquid water and dry soil material. The large dielectric constant of water is the result of the water molecule’s alignment of its permanent electric dipole in response to an applied electromagnetic field. Therefore, when water is added to the soil matrix, the effective dielectric constant of the soil increases strongly (Hipp, 1974). As the emission and scattering properties of the soil are strongly influenced by the soil dielectric constant, both active and passive MW measurements are highly sensitive to soil moisture (Ulaby, 1976; Schmugge et al., 1974). Methodological problems, lack of validation, and limitations in computing have frequently delayed the research process to retrieve soil moisture from space observations (Wagner et al., 2007). But research in these fields evolved, resulting in several global-and continental-scale soil moisture

377

data sets (e.g., (Wagner et al., 2003; Owe et al., 2008; Njoku et al., 2003). This section gives a brief overview of the state of science on satellite soil moisture.

2.14.7.2 State of the Art Since the early days, satellite RS was seen as a potential tool to provide spatial and temporal continuous soil moisture measurements (Engman and Chaunhan, 1995). In particular, MW sensors are attractive because they can acquire imagery day and night unimpeded by cloud cover. However, even more important is the fact that many MW sensors are operated at frequencies below the relaxation frequency of water (9–17 GHz, depending on temperature) where the dielectric constant of soil changes strongly with the soil moisture content. For example, at 1.4 GHz, the dielectric constant of dry soil is around 3, while it is around 20–25 for a wet soil depending on soil texture (Wang and Schmugge, 1980). Given the strong effect of the soil dielectric properties on the emission and scattering of electromagnetic waves, both passive and active MW sensors provide a relatively direct means for assessing soil moisture when the soil is not frozen or snow covered. Further, sensors operating in the VIS and IR parts of the electromagnetic spectrum have been used for mapping soil moisture (Verstraeten et al., 2008). These methods use remotely sensed surface variables such as surface temperature or vegetation to constrain the surface energy and water balances to indirectly infer soil moisture. These methods essentially belong to the group of methods used for estimating evaporation and are hence discussed elsewhere. Active MW sensors used for soil moisture retrieval include synthetic aperture radars (SARs) for local- to regional-scale mapping and scatterometers for global monitoring (GM). These instruments transmit an electromagnetic pulse and measure the energy scattered back from the Earth’s surface. On the other hand, passive MW sensors (radiometers) merely record the radiation emitted by the Earth surface itself, which is related to the physical temperature of the emitting layer and the emissivity of the surface (Ulaby et al., 1981). Even though one might expect that active and passive sensors observe very different surface properties due to their different measurements principles, several land surface parameters, such as soil moisture, surface roughness, or vegetation biomass, have a comparable impact on both active and passive measurements. The fundamental reason for this is Kirchhoff’s law which, applied to the problem of RS of the Earth’s surface, states that the emissivity is one minus the hemisphere integrated reflectivity (Schanda, 1986). Therefore, soil moisture observed by active or passive sensors can be directly compared, particularly when the sensors are operated at the same frequency. The basic challenge for both active and passive soil moisture retrieval methods is that other surface variables, such as vegetation water content, vegetation structure, and surface roughness, also have a strong impact on the MW signal. Therefore, successful retrieval methods must be able to account for all these confounding land surface parameters. This might suggest that one should use models that describe the interaction of the MWs with the Earth’s surface in as much details as possible. Yet, such models become very complex and it is in general not possible to invert them. Even more

378

Observation of Hydrological Processes Using Remote Sensing

problematic is that one generally does not have enough experimental observations to falsify complex models, simply because different model structures and parameter sets may explain the observations equally well (Wagner et al., 2009). This so-called equifinality problem (Beven, 2001) is possibly the major reason why it is often not possible to transfer algorithms calibrated over one region to another. These considerations show that it is essential to develop models that capture the main physical phenomena, yet be simple enough to allow falsification and inversion. This implies that models may differ depending on the spatial resolution of the satellite system, because the dominant processes often change with scale. Equally important as the retrieval algorithm is the selection of MW instruments. A suitable sensor exhibits a high sensitivity to soil moisture while minimizing instrument noise and the perturbing impacts of other surface variables on the measured signal (Wagner et al., 2007a). Many RS studies conducted in the 1970s, 1980s, and 1990s indicated that lowfrequency MW radiometers should offer the best performance because of the minimal influence of surface roughness and vegetation on these measurements (Jackson et al., 1999). Therefore, the first satellite mission dedicated to measuring soil moisture on a global scale uses an MW radiometer operated at a frequency of 1.4 GHz (L-band), that is, the Soil Moisture and Ocean Salinity (SMOS) launched on November 2009 (Figure 17). To improve its spatial resolution, SMOS uses a passive interferometric design inspired from the very large baseline antenna concept in radio astronomy (Kerr et al., 2001). Yet, its spatial resolution will only be in the order of about 40 km, which limits its use to large-scale applications such as numerical weather prediction or climate change. To enlarge the number of potential applications, the Soil Moisture Active/Passive (SMAP) mission foreseen for launch in 2014 uses both active radar and passive radiometer instruments at L-band. It will use a 6-m large rotating mesh antenna shared by the radar and radiometer to cover a 1000-km-wide swath (Figure 17). Thus, SMAP will offer a 40-km soil moisture product derived from its passive observations and a 10km product derived from the combined active and passive observations (Entekhabi et al., 2010).

SMOS and SMAP employ novel measurement concepts with the goal to measure soil moisture with unprecedented accuracy, and also existing MW sensors operated at frequencies below about 10 GHz can provide valuable soil moisture information. Particularly, in recent years, several soil moisture data sets derived from both active and passive MW sensors have become freely available, which demonstrate the advances made in algorithmic research. Wagner et al. (2007) suggested that this initially less visible revolution became possible, thanks to the increasing availability of computer power, disk space, and powerful programming languages at affordable costs. This has allowed more students and researchers to develop and test algorithms on regional to global scales, which lead to a greater diversity of methods and, consequently, to more successful algorithms.

2.14.7.3 Data Sets BBB The soil moisture data sets described in this section are all available for user download via file transfer protocol (FTP) or web portals. Table 3 gives an overview of the different products. Most of these data sets have a rather coarse spatial resolution in the order of 20–50 km because they are derived from MW radiometers or scatterometers. In addition, a first continental-scale 1-km soil moisture data set derived from ENVISAT Advanced Synthetic Aperture Radar (ASAR) operated in GM mode has recently been published.

2.14.7.3.1 Active MW data sets Investigations into the potential of active MW sensors for soil moisture retrieval began already in the 1960s and gained momentum in the 1990s due to the launch of several satellites that carried a synthetic aperture radar (SAR) on board. Unfortunately, there is still no widely accepted method that delivers SAR-derived soil moisture data at fine spatial scales (10–100 m). This is to some extent surprising given that a large number of backscatter models and retrieval approaches were proposed and successfully applied within pilot studies (Dubois et al., 1995; Zribi et al., 2005). Unfortunately, independent testing and transferring of the methods to other

Figure 17 Artist impressions of ESAs Soil Moisture and Ocean Salinity Mission (left) and NASAs Soil Moisture Active Passive Mission (right). Both satellites will be used for global soil moisture mapping. (Left) Image courtesy: ESA. (Right) image courtesy NASA.

Observation of Hydrological Processes Using Remote Sensing Table 3

379

A short list of accessible satellite derived soil moisture products using active and passive microwave instruments

Product name

Satellite

Spatial resolution

Temporal resolution

Period

Url

Reference

NSIDC L3 soil moisture LPRM soil moisture

Amsr-e

0.251

Sub daily

2002–now

http://nsidc.org/

Amsr-e, trmm-tmi, ssm/i, smmr Windsat

0.251

Sub daily

1978–now

http://geoservices.falw.vu.nl

Njoku et al. (2003) Owe et al. (2008)

0.251

Daily

2003–now

http://www.nrl.navy.mil/windsat/

Li et al. (2009)

Ers-1, ers-2

50 km

B 6 days

1991–now

http://www.ipf.tuwien.ac.at/radar/

Ascat

Metop

25/50 km

Daily

2006–now

http://www.eumetsat.int/

Asar

Envisat

1–5 km

Weekly

2005–now

http://www.ipf.tuwien.ac.at/radar/

Wagner et al. (2003) Bartalis et al. (2007) Pathe et al. (2009)

Windsat soil moisture Scat

regions or data sets often did not yield the hoped-for results (Walker et al., 2004). The major problem appears to be the failure to accurately model surface roughness and vegetation effects at fine spatial scales (Verhoest et al., 2008), besides the technical characteristics of most SARs (revisit time, frequency, etc.) are not well suited for the task of soil moisture monitoring. Parallel to the work on SAR, some research groups started to investigate the potential of the ERS scatterometer for land applications in the 1990s (Pulliainen et al., 1998; Woodhouse and Hoekman, 2000). Despite scatterometers were designed for monitoring winds over the oceans, these studies quickly demonstrated the potential of the ERS scatterometer for soil moisture monitoring at a 50-km scale (Wagner et al., 1999; Wen and Su, 2003). From an algorithmic point of view, the advantage of working at a scale to 50 km is that surface roughness and land cover can reasonably be assumed to be constant. The major technical benefits of the ERS scatterometer are its short revisit time and its high radiometric accuracy. In addition, its three antennas acquire three quasiinstantaneous backscatter measurements from different azimuth and incidence angles, which is important for separating vegetation and soil moisture effects on the signal. The first global soil moisture data set was derived from ERS scatterometer data using a change detection algorithm (Wagner et al., 2003). It was released in 2003 and has since then been used in several validation and application studies (e.g. Scipal et al., 2008). Using the same algorithm, the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) has developed the first operational, near-real-time soil moisture monitoring system based upon the Advanced Scatterometer (ASCAT) flown on board of the Meteorological Operational (METOP) satellite series. ASCAT is the successor instrument of the ERS scatterometer and offers a twofold improved temporal and spatial resolution (Bartalis et al., 2007). The change-detection algorithm developed for the scatterometer has been adapted to 1-km GM mode data as acquired by the Advanced Synthetic Aperture (ASAR) on board of ENVISAT (Pathe et al., 2009). This particular SAR mode has a rather poor radiometric resolution, but requires less energy as high-resolution SAR modes. Thus, it offers a good temporal coverage suitable for studying soil moisture dynamics.

2.14.7.3.2 Passive MW data sets In the passive domain, soil moisture research already started in the 1970s and one of the first soil moisture retrieval algorithms was developed by Njoku and Kong (1977). This algorithm used a simple regression technique on multifrequency MW observations to obtain soil moisture from a controlled bare soil site. In time, this modeling approach started to become more complex with the addition of a surface roughness module (Choudhury et al., 1979; Wang and Choudhury, 1981; Wigneron et al., 2001), a vegetation module (Kirdiashev et al., 1979; Meesters et al., 2005), and a dielectric mixing module to convert the soil dielectric properties to soil moisture (Wang and Schmugge, 1980; Dobson et al., 1985; Mironov et al., 2004). On a later stage, an atmosphere module (Pellarin et al., 2003; Liebe, 2004) and snow module (Pulliainen et al., 1999) were introduced to obtain a better description of the MW emission as measured by the satellite. Most of the global soil moisture data sets from passive MW observations are based on a selection of the given modules and the differences between the different products vary on the choice of modules. In this section, we describe the two most commonly used global soil moisture data sets. The first global soil moisture product was developed by Njoku et al. (2003) and uses X-band AMSR-E MW observations to retrieve soil moisture. This model uses a multichannel iterative forward-model optimization method to solve simultaneously for surface temperature, soil moisture, and vegetation water content (Njoku et al., 2003). In the forward mode, the retrieval algorithm iteratively adjusts values of the retrieval parameters using Fresnel relations adjusted for surface roughness and attenuation by vegetation cover using time-invariant parameters based on land cover type (Njoku and Chan, 2006). The modeled brightness temperature is then compared to the observed at-sensor brightness temperature until an iterative least-squared minimized solution is obtained. Polarization ratios are used instead of absolute brightness temperature because these minimize the effects of surface temperature (Sahoo et al., 2008). The model uses the X-band frequency to minimize the effects of radio-frequency interference (RFI) on the at-sensor brightness temperature (Njoku et al., 2005). The final soil moisture data set is screened to remove data over large water bodies, dense vegetation, snow, and permanent ice. This product is distributed by the

380

Observation of Hydrological Processes Using Remote Sensing

National Snow and Ice Data Center (NSIDC) in EASE–GRID format with a nominal grid spacing of 25 km. The second global soil moisture product was obtained from different satellites sensors, including Nimbus SMMR, TRMM TMI, SSM/I, and AMSR-E (Owe et al., 2008). It used the Land Parameter Retrieval Model (LPRM) to retrieve soil moisture from passive MW observations. The soil moisture retrievals from LPRM were based on the solution of an MW radiative transfer model and solved simultaneously for the surface soil moisture and vegetation optical depth without a priori information of land surface characteristics (Meesters et al., 2005). The flexible approach created the possibility to retrieve soil moisture from a variety of frequencies. LPRM produced volumetric (approximately in m3 m3) soil moisture of approximately the first 12 cm for C-band MW observations. For X band, the penetration depth is a bit smaller, resulting in soil moisture values of the first centimeter. The data are distributed in a rectangular grid with a pixel spacing of 0.251

2.14.7.4 Validation Soil moisture products from passive and active MW satellite observations were extensively validated. In the absence of a homogeneous global soil moisture station network, the data sets were validated over regional networks (Ceballos et al., 2005; Wagner et al., 2007; Draper et al., 2009), intercompared (De Jeu et al., 2008; Rudiger et al., 2009; Mladenova et al., 2009), and evaluated against model data (Wagner et al., 2003). These studies found high correlations with in situ observations in semi-arid regions and somewhat lower correlations in agricultural areas. On average, the current active and passive MW soil moisture products have an accuracy of about 0.06 m3 m3 for sparse-tomoderate vegetated regions (De Jeu et al., 2008). For denser vegetation classes such as forests, the soil moisture retrievals start to become less accurate and at an LAI of about 4, no reliable soil moisture can be retrieved from the current passive MW sensors (De Jeu, 2003). Nevertheless, recent assimilation studies have demonstrated the potential use of these existing data sets for the regions where they can obtain reliable soil moisture. The assimilation of soil moisture observations from operational satellite systems was found to improve the model performance in agro-meteorology (de Wit and van Diepen, 2007), hydrology (Parajka et al., 2006), meteorology (Drusch et al., 2009; Zhao et al., 2006; Scipal et al., 2008), and climate (Liu et al., 2007; Loew et al., 2009). With the anticipated launch of the new satellites with more innovative sensors and the continuous scientific movement in algorithm development, an improvement on the quality of satellite soil moisture is expected. Furthermore, the use of satellite soil moisture in environmental research is not yet fully exploited, and further research is necessary to fully demonstrate the potential of these new data sets.

2.14.8 Water in the Ground – Groundwater (Gravity Observations) 2.14.8.1 Introduction Groundwater is vital for meeting agricultural, domestic, and industrial water needs, particularly in parts of the world where

the climate or topography does not allow for a reliable supply of surface water. It is also by far the most abundant form of fresh, unfrozen water on the Earth (Shiklomanov, 1993). Groundwater storage does not vary as rapidly as soil moisture or surface water, but it does exhibit significant seasonal and interannual variability (Rodell and Famiglietti, 2001) and it is susceptible to overexploitation (Alley et al., 2002). The slow process of groundwater recharge acts like a low-pass filter on transient weather conditions, so that multiannual water-table fluctuations in a natural setting may be a useful indicator of climate variations. Hence, groundwater storage observations are valuable for both practical and scientific applications. As with other water-cycle variables, monitoring groundwater storage at regional scales using in situ measurements is expensive and problematic, and at the global scale it is simply not feasible. RS has propelled global hydrology forward in the past 30 years, but because groundwater is hidden deep beneath the surface, it was the last component of the terrestrial water cycle to benefit from the technology. Near-surface stocks and fluxes of the water cycle can be inferred based on electromagnetic radiation (various wavelengths of light) emitted or reflected from the land surface and atmosphere. Satellites can only sense groundwater by the effect it has on Earth’s time-varying gravity field. Redistributions of water and other forms of mass cause changes in gravitational potential, which is imperceptible to human beings yet strong enough to perturb satellite orbits. This is the concept behind one of the most innovative Earth-observing satellite systems yet launched, the Gravity Recovery and Climate Experiment (GRACE).

2.14.8.2 GRACE Data Processing The primary goal of GRACE is to map the static and timevarying components of the Earth’s gravity field with better spatial resolution and accuracy than ever before (Tapley et al., 2004). GRACE comprises two satellites in a tandem, nearpolar orbit, approximately 200 km apart and 500 km above the Earth. As they orbit, a K-band MW ranging system continuously measures the distance between the two satellites, which is affected by heterogeneities in the Earth’s gravity field. These measurements, along with precise location information, can be used to construct a mathematical model of the shape of the gravity field, nominally on a monthly basis. Each gravity field solution is delivered as a set of spherical harmonic coefficients, rather than a gridded map. Wahr et al. (1998) and Rowlands et al. (2005) described two of the techniques available for converting the GRACE gravity data to mass anomalies (deviations from the long-term mean field). Further, in order to isolate changes in terrestrial water storage mass (groundwater, soil moisture, snow and ice, surface water, and biomass) one must remove the effects of atmosphere and ocean circulations using atmospheric analysis and ocean model outputs. Glacial isostatic adjustment must also be considered in certain regions, and a major earthquake can produce a gravitational anomaly, but the timescales of most solid earth processes are too long to be an issue (Dickey et al., 1997). Because of the nature of the measurements, GRACE has no ‘footprint’ or pixel resolution. Rather, there is a tradeoff between resolution and accuracy, so that the effective limit of resolution for estimating changes in terrestrial water storage is

Observation of Hydrological Processes Using Remote Sensing approximately 160 000 km2 (Rodell and Famiglietti, 1999; Rowlands et al., 2005; Swenson et al., 2006).

2.14.8.3 Retrievals of Groundwater Storage with GRACE Data Despite its origins in the field of geodesy, GRACE’s greatest contributions have been in the cryospheric and hydrologic sciences. GRACE has monitored the melting of the Greenland and Antarctic ice sheets as never before possible (Luthcke et al., 2006; Velicogna and Wahr, 2006) and quantified glacier melt in the Gulf of Alaska (Chen et al., 2006). GRACE terrestrial water-storage data have been used to constrain regional ET rates (Rodell et al., 2004), river discharge (Syed et al., 2005), soil moisture variations (Swenson et al., 2008), and surface-water-storage variations (Han et al., 2009), and to describe intercontinental teleconnections (Crowley et al., 2006). GRACE is also the first satellite system to observe regional scale variations in aquifer storage. Isolating groundwater from GRACE-derived terrestrial water-storage data requires knowledge of the other waterstorage variables, because gravimeters provide no indication of the sources or stratification of the mass changes affecting the time-variable gravity field. In polar and alpine regions, terrestrial water-storage variability is often dominated by changes in snow and ice (Niu et al., 2007). In humid tropical regions, such as the Amazon, surface water can be the major variable (Han et al., 2009). In the rest of the world, soil water typically exhibits the largest fluctuations on daily-to-seasonal timescales, whereas groundwater storage amplitudes can be as large or larger on seasonal and longer timescales (Rodell and Famiglietti, 2001). Biomass variations are near or below GRACE’s limit of detectability (Rodell et al., 2005). Following the approach suggested by Rodell and Famiglietti (2001), Yeh et al. (2006) and Rodell et al. (2007) demonstrated that groundwater storage variations can be isolated from GRACE terrestrial water-storage data using in situ root zone soil moisture observations or numerically modeled soil moisture fields. They verified their results using data from groundwater monitoring networks in Illinois and the Mississippi River Basin. Strassberg et al. (2007) achieved good results using the model-supported technique to estimate groundwater storage changes in the High Plains aquifer, likewise verified by monitoring well observations. Rodell et al. (2009) applied the technique to determine that groundwater beneath the Indian states of Rajasthan, Punjab, and Haryana (including Delhi) is being depleted at a rate of 17.7 km3 yr1 due to withdrawals for irrigation. Zaitchik et al. (2008) presented a more sophisticated approach for disaggregating GRACE-derived terrestrial water storage into its components, whereby an ensemble Kalman smoother is used to assimilate the GRACE data into a numerical land surface model. This approach has several advantages. First, physical equations of hydrologic and energetic processes, integrated within the model, provide a basis for synthesizing GRACE and other relevant observations such as precipitation in a physically consistent manner. Second, the model fills spatial and temporal data gaps, while observations anchor the results in reality. Third, in addition to separating groundwater, soil moisture, and other component

381

contributions, the assimilated output has much higher spatial and temporal resolutions than the original GRACE data. Zaitchik et al. validated the technique in the Mississippi River Basin using groundwater data from a network of wells, and showed significant improvement in both the timing and amplitude of modeled groundwater variations.

2.14.8.4 GRACE Data Access GRACE gravity data are produced and distributed by three centers that support the mission: the University of Texas Center for Space Research, NASA’s Jet Propulsion Laboratory (JPL), and the German Research Centre for Geosciences (GFZ). GRACE terrestrial water-storage products have been developed by many groups. Visualization and data portals include those provided by NASA/JPL, NASA/Goddard Space Flight Center (GSFC) and Stinger Ghaffarian Technologies, and the University of Colorado.

2.14.8.5 Concluding Remarks and Future Perspective Although other RS data can provide clues as to the location and characteristics of aquifers (Becker, 2006), satellite gravimetry is the only technology currently available for measuring regional-scale groundwater storage changes from space. In addition to GRACE, two other advanced gravity-monitoring satellites have been launched: GFZ’s Challenging Minisatellite Payload (CHAMP) in 2000 and the European Space Agency’s Gravity Field and Steady-State Ocean Circulation Explorer (GOCE) in 2009. CHAMP was a major advance in gravimetry at the time of launch, but it was not accurate enough to infer water-storage changes, and it was quickly made obsolete by GRACE. GOCE will map the static gravity field with significantly higher spatial resolution than GRACE, but it is not well suited for monitoring the time-variable gravity field and inferring changes in groundwater storage (Han and Ditmar, 2008). GRACE is in its extended mission phase, beyond its initial 5-year goal. It could potentially continue through 2012. NASA, ESA, and many independent reports (e.g., NRC, 2007) have recognized the importance of the data provided by GRACE and the need for a follow-on mission to enable continued monitoring of terrestrial water and ice as only satellite gravimetry can. Technology upgrades, such as a laser ranging system, a lower Earth orbit with drag-free propulsion, or more satellites and different orbital configurations, could increase the accuracy and spatial resolution of the products. However, at the time of writing, a next-generation time-variable gravity mission had not yet been approved.

2.14.9 Optical RS of Water Quality in Inland and Coastal Waters 2.14.9.1 Introduction Inland and coastal waters are important natural resources yet they are seriously threatened by eutrophication, salinization, and heavy metal contamination. Excessive concentrations of suspended particulate matter (SPM) influence the productivity and thermodynamic stability of inland and coastal waters (Muller-Krager, 2005: 348). Traditional measurements of

382

Observation of Hydrological Processes Using Remote Sensing

water quality are costly, time consuming, and are limited in their spatial and temporal coverage. EO data, on the other hand, provide rapid and repeated information over large and often inaccessible areas. EO, in conjunction with modeling and strategic in situ sampling, can play a crucial role in determining the current status of water-quality conditions and helps anticipate, mitigate, and even avoid future water catastrophes (DiGiacomo et al., 2007). The primary measurement of EO data over water is the visible light leaving the water column, hereafter called the water-leaving reflectance. In inland and coastal waters, this water-leaving reflectance is strongly affected by different materials, for example, terrigenous particulate and dissolved materials, resuspended sediment, or highly concentrated phytoplankton bloom. The majority of inland and coastal waters can therefore be classified as case 2 waters (Gordon and Morel, 1983). In case 2 waters, the constituents are independent of each other and do not covary with chlorophyll a as in case 1 waters. RS of inland and coastal waters is quite challenging due to the complicated signals from turbid water, substrate reflectance, and adjacent land surfaces (Figure 18). Consistent EO estimates of water-quality parameters in inland and coastal waters require three components: (1) a reliable atmospheric correction method; (2) an accurate retrieval algorithm; and (3) an objective method to estimate the uncertainty budget based on their sources. Because of limitation in length, the scope of this chapter has been narrowed to confine some of the recent developments in

each of the above-mentioned areas. Knowledge of the basic concepts of aquatic optics is assumed available.

2.14.9.2 Atmospheric Correction Most of the atmospheric correction procedures fail over inland and coastal waters, that is, case 2 waters. The failure of atmospheric correction might be attributed to the complexity of the recorded reflectance. The standard approach by Gordon and Wang (1994), for example, assumes a zero water-leaving reflectance in the near-infrared (NIR). In case 2 waters, this water-leaving reflectance has distinctive values at the NIR part of the spectrum (Siegel et al., 2000). The non-negligible value of water-leaving reflectance at the NIR was accounted by many authors (Carder et al., 1999; Gould et al., 1999; Ruddick et al., 2000; Hu et al., 2000; Salama et al., 2004). Coupled approaches are increasingly used to retrieve the optical properties of both water and atmosphere simultaneously. For each atmosphere–water setup, a TOA reflectance is simulated at variable viewing-illumination conditions. The parameters that define each media are tuned until the best convergence to the recorded reflectance is found (Chomko et al., 2003; Stamnes et al., 2003; Gordon et al., 1997; Zhao and Nakajima, 1997). However, most of these algorithms were developed for case 1 waters, that is, assuming known and spatially homogeneous water-leaving reflectance at the NIR. Newly developed algorithms are emerging for case 2 waters (Kuchinke et al., 2009a, 2009b). The spectral optimization method (Kuchinke et al.,

Scattering of direct and diffuse incident light

Observed reflectance by the sensor

Direct and diffuse incident sun light

ce e rfa Su ctanc le ref

Land

Water

Scattering, absorption and remittance by water constituents

A ref djac lec en tan t ce

Reflectance from mixed land water pixel

Bidirectional substrate reflectance Figure 18 Schematic diagram of the different processes that contribute to the observed remote-sensing reflectance at a pixel size in inland and coastal waters.

Observation of Hydrological Processes Using Remote Sensing 2009b) was constrained to 0.1 m m1 as a maximum value of backscattering coefficient of SPM at 0.443 mm. This value of backscattering is equivalent to 12 g m3 concentration of suspended particles using the specific backscattering coefficient of Albert and Gege (2006). On the other hand, artificial neural network techniques (Doerffer and Schiller, 2007) are usually limited to the range of their training sets. Most of estuarine and coastal waters have high loads of SPM, exceeding 12 g m3. For instance, the Yangtze estuarine water is extremely turbid with SPM concentration ranging between 80 and 500 g m3 (Shen et al., 2010). The spatial variabilities of the aerosol and water signals at the NIR part of the spectrum are characteristic features of turbid inland and coastal waters. These variabilities can be attributed to the different aerosol types that may coexist in this transaction zone as well as to the distinctive shape of the water-leaving reflectance. Salama and Shen (2009) proposed an analytical approach to consider and quantify this variability (Figure 19). Their method was validated with in situ measurements and successfully applied on data obtained from orbital ocean color and geostationary sensors. Deriving water-quality parameters from geostationary satellite in open coastal areas (Salama and Shen, 2009; Neukermans et al., 2008) is of unprecedented benefit. It will facilitate resolving the temporal dynamic of marine bio-geophysical parameters and overcome cloud covers.

2.14.9.3 Retrieval Algorithms Most developed algorithms for water-quality retrievals in inland and coastal waters are empirical in nature. This empiricism limits their application to a specific range of concentrations, area, and season. Kallio et al. (2001) studied different algorithms to estimate chlorophyll a in lakes. These algorithms were empirical and estimated one variable using band ratio of approximately 0.675 and 0.705 mm (Dekker et al., 1992; Gitelson et al., 1993). A generalized retrieval algorithm is, however, hindered by the large natural variability of inland waters. Significant efforts on improving the accuracy of air- and space-bornederived water-quality parameters are therefore required for inland and near-coastal waters. Many studies have used semi-

1.3

0.1 116°9′E

116°9′E

Aerosol reflectance 865 nm 0 (a)

29°31′N

29°31′N

116°9′E

29°31′N

29°31′N

116°9′E

0

b b(spm)(550) m−1

(b)

Figure 19 (a) Derived aerosol reflectance above the Poyang Lake. (b) Derived SPM concentration in the Poyang Lake. Notice that they are totally uncorrelated.

383

analytical models to derive water-quality parameters in lakes (Hoogenboom et al., 1998; Gons et al., 2002). Derived waterquality parameters from multivariable inversion methods are ambiguous and not unique (Sydor et al., 2004). Other promising methods are used when the inversion employs lookup tables (Van Der Woerd and Pasterkamp, 2008). Salama et al. (2009) showed that the inversion is very sensitive to the spectral shape parameters of SPM backscattering and absorption of dissolved organic matter. Including these two parameters has enhanced the retrieval in inland waters (Figure 20).

2.14.9.4 Uncertainty Estimates Reliable methods for uncertainty quantification of waterquality EO products are important for sensor and algorithm validation, assessment, and operational monitoring. High accuracy in both observations and algorithms may reduce considerable ranges of errors. EO-derived water-quality parameters have, however, an inherent stochastic component. This is due to the dynamic nature of water, intrinsic fluctuations, model approximations, correction schemes, and inversion methods. Quantitative measures of uncertainty support water quality and ocean-color product validation, especially with the introduction of the new AERONET-OC network (Zibordi, 2006). Due to stochasticity of the measurements, as well as model approximations and inversion ambiguity, the retrieved inherent optical properties (IOPs) are not the only possible set that caused the observed spectrum (Duarte et al., 2003; Sydor et al., 2004). Instead, many other sets of IOPs may be derived. Each of these sets has an unknown probability of being the derived ocean-color product. The probability distribution of the estimated IOPs provides, therefore, all the necessary information about the variability and uncertainties of derived water-quality parameters. Several efforts have been carried out to resolve the uncertainty of the derived IOPs. Duarte et al. (2003) analyzed the sensitivity of the observed RS reflectance due to variable concentrations of water constituents. Salama (2003) proposed a stochastic technique to quantify and separate the source of errors of IOPs derived from hyperspectral airborne measurements. Maritorena and Siegel (2005) employed a nonlinear regression technique for consistent merging of different ocean-color-derived products. Wang et al. (2005) performed a detailed study on the uncertainties of ocean-color model inversion related to fluctuations in each of the IOPs and their spectral shape. In general, these studies used the method of Bates and Watts (1988) to construct the confidence interval around the derived IOPs following different approaches, however. It is adequate as long as model inversion has a wellconditioned Jacobian matrix of the minimum cost function. Recently, Salama and Stein (2009) developed a generic method to quantify the uncertainties in the derived waterquality products based on their sources, namely, model approximations, measurement noise, and atmosphere correction. The method was evaluated and validated using oceancolor data sets (Figure 21). The method has promising applications for inland and coastal water. Moreover, it provides vital input to SPM assimilation models (Eleveld et al., 2008) and EO product merging (Pottier et al., 2006).

384

Observation of Hydrological Processes Using Remote Sensing

Chlorophyll a mg m−3

Poyang lake RMSE = 0.03 g m−3 2 R = 0.95 30

Wolderwijd and Veluemeer −3 RMSE = 0.32 mg m 2 R = 0.78

Wolderwijd and Veluemeer RMSE = 0.342 g m−3 R 2 = 0.95

−3

8

Poyang lake RMSE = 0.06 mg m−3 2 R = 0.94

Derived SPM g m

Derived chlorophyll a mg m−3

10

SPM g m−3

40

6

4

20

10 2

Poyang Lake 1:1 line Wolderwijd and Veluemeer

Poyang lake 1:1 line Wolderwijd and Veluemeer

0 0

2 4 6 8 Measured chlorophyll a mg m−3

0 0

10

10

20 30 Measured SPM g m−3

40

Figure 20 Results from remote-sensing inversion in inland waters. The spectral shape parameters were also derived from the inversion. (a) Derived chlorophyll a and (b) derived suspended particulate matter.

Total absorption

Scattering; SPM

10

15 r 2 = 0.88 2 = 0.005 r reg f = 0.93

1:1 line Regression

5

0

-5 -5

(a)

This work Derived error of a total(440), m−1

Derived error of b spm(550), m−1

15

0

5

10

15

Known RMSE of bspm(550), m−1

10

r 2 = 0.67 2 = 0.001 r reg f = 0.86

5

0

-5 -5

(b)

0

5

10

15

Known RMSE of a total(440), m−1

Figure 21 Estimated total error (dot symbols) on IOPs derived from SeaWiFS spectra of the Nomad data set. Nonlinear regression results are superimposed as plus symbols; r2 and r2reg are the correlation coefficients for dot and plus symbols respectively; f is the fraction of successful retrievals. (a) Errors of scattering coefficient; (b) error of total absorption coefficient. From Salama MS and Stein A (2009) Error decomposition and estimation of inherent optical properties. Applied Optics 48: 4947–4962.

2.14.9.5 Concluding Remarks and Future Perspective We have summarized the three requirements for reliable retrievals of water-quality parameters from RS data in inland and coastal waters. Although EO operational products are still under development, there are few issues that need extra attention in the future: 1. The red, NIR, and even shortwave IR bands, with sufficient signal-to-noise ratio, are necessary for RS of inland waters. They improve the accuracy of derived IOPs. 2. Improved parametrization of IOPs is needed for inland waters. The improvement should (a) account for different phytoplankton species and (b) deconvolve the overlapped absorptions at the blue.

3. Reliable methods to account for absorbing aerosol and adjacency effect in inland waters. 4. Studying the effects of climate change on water quality and a better understanding of the role of water quality of large inland lakes on the radiative energy budget on a subcatchment scale.

2.14.10 Water Use in Agro- and Ecosystems 2.14.10.1 Introduction A comprehensive review of reflective (and partly thermal) RS techniques applied to agro-hydrology and ecological systems was given by Dorigo et al. (2007). Understanding the

Observation of Hydrological Processes Using Remote Sensing

opportunistic nature of RS acquisitions, the traditional approach selected in most cases for the evaluation of RS-derived biophysical variables consists of the statistical comparison between field ancillary data and a corresponding RS data subset. The analysis ends by evaluating the strength of the correlation between these data sets. Actual water use, transpired by the crops and evaporated from the soil, is the main output of a vast number of soil– vegetation–atmosphere transfer (SVAT) algorithms and surface energy balance (SEB) models. In the SVAT and SEB sequence, a great number of submodels are required to retrieve land properties from reflective optical measurable from RS instruments. As such, crop-water requirements from RS (AET-RS) processes are heavily demanding in terms of data input and modeling. Along with the image process and analysis, a dedicated number of intermediate products are elaborated which are simultaneously essential for many other ecohydrological applications. We consider that a good overview of the use of water in agro- and ecosystem can be tackled by reviewing SVAT and SEB models. Efforts focused on the use of AET-RS and its pre- and postelaborated products are not only to improve water management in irrigated lands, but also associated with this, in irrigation planning, and irrigation monitoring, leading to performance indicators (Bos et al., 2005), water competition and water strategy at basin level (Bos et al., 2009), soil moisture retrievals (Wang and Qu, 2009), and several other categories of hydrological modeling benefiting from this approach. AET-RS estimation opens the essential spatial dimension to a diversity of agro-ecological models on the one hand, and welcomes continuous model output to cover the typical temporal gaps in RS imagery. Increasingly, data assimilation techniques are used to integrate RS information into continuous modeling with success (Loew, 2007). In this section, the topic of AET-RS products and the suggested link to FAO crop factors and irrigation (FAO, Food and Agriculture Organization; Allen et al., 1998) is selected as an example of a variety of opportunities that congregates much of the knowledge of RS in agro- and ecosystem.

2.14.10.2 Continuous Evaluation of Crop Water Use with Support from RS Despite the intensive research in the field of RS to evaluate energy fluxes on an instantaneous basis, setting an operational scheme for practical agronomical purposes proves more troublesome. The techniques to elaborate energy flux maps from RS are temporally intermittent, as ephemeral as the image opportunity. They depend on the availability of the image that is affected by spatial resolution (nadir and off-nadir views), revisiting time, and, mainly, cloudiness. High-resolution thermal sensors are less available today (2009) than it was in the past, probably after the failure in establishing operational sequences for use in the market. As such, low-resolution thermal sensors are the only available RS source with adequate temporal revisiting time. The spatial resolution is, at the same time, the main limitation of these sensors for applications at field scale. The elaboration of a single flux map requires the highest level of expertise and continuous upgrading. Updated preprocessing including sensor calibration, atmospheric correction,

385

narrow-to-broadband conversions, emissivity retrievals, angulargeometrical effects, and instantaneous-to-daily integration are some of the necessary steps to be performed. The processing time is much slower than the dynamics of the ET process, such that this might be the cause of the few operational efforts beyond the framework of international projects. The selected criteria for classification of the approaches to evaluate AET-RS regarding the FAO Kc (crop factor) link is based on the relative weight that the RS and modeling components have on the final product. In view of the extensive bibliography available, only a few main references are indicated here. Continuous monitoring of the crop-water requirements using RS images exclusively can be attempted in areas where both clear days and image coverage are frequent. Daily AET is estimated from SEB models (see Section 2.14.4) on the available cloud-free images. Auxiliary ground data are collected from meteorological stations on daily (or shorter) basis to complement the database required for the SEB. The net radiation ‘Rn’ is calculated on a daily basis from a combination of RS and meteorological stations (Hurtado and Sobrino, 2001) or from ground stations when images are not available (Allen et al., 1998; Irmak et al., 2003). For the days when images are available, surface albedo (Liang et al., 2003), surface temperature (sensor dependent) (Coll et al., 1994, 2005; Gillespie et al., 1998), emissivity (Valor and Caselles, 1996), and fractional vegetation coverage (Su, 2002) can be evaluated. Considering the dynamic behavior of these parameters in the diurnal cycle, the intention is to describe a continuous pixel-based temporal evolution of them from day to day, as good as possible. If the daily values of each parameter are known on a daily basis, a ‘potential image’ of these parameters is obtained. After that, the evaporative fraction ‘EFi’ can be evaluated (Su, 2002). This method also considers the evaporative fraction being conservative during the daily course (Brutsaert and Sugita, 1992), providing a method to scale instantaneous evaporative fraction to daily evaporative fraction. However, several authors have shown the variability of the evaporative fraction during the day (Chehbouni et al., 2008; Crago, 1996) to the point that the original hypothesis stated by the one-time AET-RS must be reviewed for the particular situation considered. The soil heat flux averaged over the day is usually assumed zero (Brutsaert, 2008), so the AET for any pixel at any day ‘i’ is calculated as: Rni * EFi. As AET and potential evapotranspiration (PET) are obtained on a daily basis, direct application to irrigation becomes possible. The approach is suited for irrigated lands in areas of data scarcity where clear skies are common as shown in a case study in Morocco (Jacobs et al., 2008). A second approach estimates AET replacing the Penman– Monteith method with the Priesley and Taylor (PT) method that has proved to be reasonable for wet irrigated areas (de Bruin, 1987; McNaughton and Spriggs, 1989). Under these circumstances, the evaluation of Kc * Ks ¼ AET/ET0 can be resolved from the balance of radiative fluxes at the surface only (Mekonnen and Bastiaanssen, 2000). If the area is under irrigation, then the water availability factor Ks ¼ 1, and a locally adjusted value of Kc can be evaluated directly from RS measurements.

386

Observation of Hydrological Processes Using Remote Sensing

This method of evaluating a Kc becomes also suitable for heterogeneous cropping patterns in the same land, typical of many agricultures in Asia and Africa. The fractional cover of each crop can be evaluated and the average Rn, G, and H used in the evaluation of a ‘composite Kc’ can be obtained from the homogeneous covers by using their fractions as weights. As AET is estimated from PT, the values could be verified by using an SEB method from RS, which allows monitoring. The reference ET0 in the above method is obtained with the PT method for grasslands. This approach is used in irrigated areas with data scarcity in order to improve the estimations of Kc and reduce the need of images. As Kc varies in terms of weeks, this approach is suitable to build adjusted Kc factors for irrigated crops after some seasons of analysis. After the Kc is evaluated, it can be applied for crop specifics without the requirement of images. The main deficiency of the purely RS-based models is the uncertainty of clear image acquisition in the right time window. As such, major attention is turned to dual approaches where a model does the calculations on crop-water requirements and RS images are used to feed the model. Difference in these methods is centered on how the images are used as model input. The first group of approaches using SVAT models such as SWAT (Arnold and Fohrer, 2005; Arnold et al., 1998), CRIWAR (Bos et al., 2009), SWATRE (Belmans, 1983), or SIMGRO (Querner and van Bakel, 1989) is applied independently or combined with ground data to make the actual evaluation of the crop-water requirements. As the models are based on water balance, soil moisture and AET are estimated at every time step and for each land-use location. Due to the balance agreement between soil moisture and evaporation, a good calibration of the soil moisture ensures a good estimate of the ET from the model, mainly in the case of irrigation (surplus). After the model is calibrated, AET is a product that can be used on daily basis. In well-managed irrigation schemes, the assessment of water efficiency and uniformity of water distribution is preferred over the strict use of the Kc factor for cropwater requirements estimates. These two items are considered essential in the evaluation of irrigation performances (Bos et al., 2005). In this type of appraisal model, the concept of Kc is replaced for the ‘relative ET’ defined as RE ¼ AET/PET, where PET can be obtained from standard meteorological measurements (Doorenbos and Pruitt, 1977). Irrigation efficiency in the command area is then evaluated as the combination of three measurable properties:

•

• •

Optimal water requirement ensures that no stress occurs. In general, when AET/PET X0.75, this condition prevails. The value of 0.75 is generally adopted, although it might be adjusted locally. Water efficiency is defined as the ratio of AET to irrigation water depth given to each unit of parcel or crop. Uniformity refers to the homogeneity of the distribution of water in space and time inside the command area. As AET and RE are obtained on a daily and pixel basis, the coefficient of variation of these variables is used to evaluate it. A threshold is set to evaluate uniformity.

As presented here, the approach does not use AET derived from imagery as part of the procedure. Then, images obtained on clear days are used to derive AET establishing a strong basis of comparison with the information produced by the model. AET estimated by the image is used to evaluate the three indicators mentioned earlier, allowing real-time monitoring and the detection of flaws in the model that then can be corrected. This approach was used very successfully by Roerink et al. (1997) in the irrigation fields in Mendoza, Argentina, with irrigation performance methods developed by Bos et al. (1994). For operational purposes, the most accepted approach is to allow a continuous SVAT model to perform the calculations of AET and crop-water requirement. The model is calibrated mainly against soil moisture and groundwater table, both key variables in the process that can also be easily monitored at point scale. No SEB-RS model is used in this case. In this sense, this approach is similar to the previous method. The difference occurs on the use of the imagery as input to the model. RS imagery is used only to evaluate crop patterns and land parameters or properties that are required for the model. From a sequence of high-resolution images, a selected vegetation index evolution in time is estimated at pixel level (usually the Normalized Difference Vegetation index (NDVI) is used in many SEB). From the individual points in time, smoothing, filtering, or mainly harmonic techniques are normally adopted to achieve a time-continuous evolution (Verhoef, 1996). The NDVI time series and up-to-date land-use maps suffice to evaluate basis Kc (Kcb) described in Allen et al. (1998) and the fractional cover (fc), both needed to input in the model (Bausch and Neale, 1987; Bausch, 1995; Gonzalez Piqueras, 2006; Heilman et al., 1982; Ray and Dadhwal, 2001; Valor and Caselles, 1996). The model ‘HidroMORE’ was the result of the application of Irrigation Advisory Services (IAS) in Mediterranean areas, and validation of the results are given by Rubio et al. (2003). This approach was successfully used in a demonstration European project called ‘Demeter’ (Jochum and Calera, 2006) and in worldwide international projects such as ‘Pleiades’. In general, AET derived from SEB algorithms using RS imagery needs strict field (on-site) validation and local finetuning. On wet conditions on typically well-covered irrigated fields, the actual evaporation approaches to the potential one and in this situation most of the SEB methods work well (Bastiaanssen et al., 1998; Norman et al., 1995b; Su, 2002). Kite and Droogers (2000) reviewed some models and RS retrieval under the same conditions and warned about the great variability in the results between FAO-24 and satellite methods. There is a general agreement in the circle of experts that some high degree of expertise is required for the application of AET-RS SEB models, a warning that must not be overlooked.

2.14.10.3 Drought Indices and Soil Moisture Monitoring The partition of available energy at the Earth surface is largely controlled by the available soil moisture. The relation between soil moisture and ET is very tight, to the point that irrigation supply (moisture) can be estimated directly from accumulated ET as well as from monitoring soil moisture changes. However, from the water management point of view these two processes

Observation of Hydrological Processes Using Remote Sensing

are very different in the sense that ET can be monitored by RS but cannot be controlled easily and the soil moisture can be controlled but it is not directly observable in the rooting depth. As we can only adjust the water allocated to crops, the soil moisture becomes the most relevant of all parameters in the water cycle for agro-systems. In watershed management, there are situations requiring continuous monitoring of soil moisture as during droughts, which is the most devastating form of agricultural deficiency. The evaluation of droughts is done through methods that account for the endurance of low soil moisture values through time on a pixel basis. RS in the visible and thermal wavelengths is unable to directly measure soil moisture on the ground. As such, indirect methods were designed to approach it. There are three approaches to monitor droughts purely from standard passive RS. The initial methods were based on the fact that during drought conditions a sudden change in soil moisture would be followed by a distinctive jump in the spectral reflectance of the observed pixel. In a soil, more humidity implies less reflectance. Bowers and Hanks (1965) and Bowers and Smith (1972) linearly related the soil moisture and the spectral bands where soil moisture is an energy absorber. The property that bodies oppose to temperature changes is called ‘thermal inertia’. It can be evaluated from multitemporal imagery (day and night) from the visible and thermal bands. Pratt and Ellyet (1979) presented a modification of the model to map soil moisture, and later Price (1985) used the energybalance concept to add certainty to the model. All these approaches were successful under controlled conditions in areas of sparse vegetation, as soil contrast needs to be only affected by moisture. More appropriate was the attempt to monitor soil moisture under vegetated conditions. Moisture depletion affects plant physiology and, in particular, the reflective properties of the vegetation. In the absence of vegetation water as incoming radiation needs dissipation, vegetation reacts by increasing both the reflectivity of the leaves (albedo) and the sensible heat (surface temperature). The accounting of the difference between the surface and air temperature in time leads to the design of the crop-water stress index (Idso et al., 1981, 1975; Jackson et al., 1981), which was suitable for evaluating stress for full cover situations. Moran et al. (1994) developed a water deficit index (WDI) using vegetation indices to account for partially vegetation covered areas, using a composite of the surface temperature for vegetation and land, an approach that was later on extended to SEB models. As air temperature was needed in this type of method and the availability of ground meteorological measurements was restrictive, the use of the variability of the canopy surface temperature allowed a pure RS approach, especially applicable to nonfully covered areas (Gonzalez Dugo et al., 2005). Vegetation indices and surface temperature were also used in the development of indices dedicated to drought monitoring (Carlson et al., 1991, 1994; Kogan, 1990; Sandholt et al., 2002). Indices are rarely absolute, as they might not directly compare to a similar degree of drought severity, which leads to the designs of regional indices that can be compared.

387

RS-derived SEB subproducts were earlier involved in the calculation of drought indices. Using air and surface temperature and its variation, albedo, and net radiation in a consistent framework allowed quantifying droughts spatially. It was only after AET-SEB product became operational, for example, SEBI (Menenti and Choudhury, 1993), S-SEBI (Roerink et al., 2000), SEBAL (Bastiaanssen, 1995), TSEB (Kustas and Norman, 1999; Kustas et al., 2004; Norman et al., 1995a), ALEXI (Anderson et al., 1997; Mecikalski et al., 1999), and Dis-ALEXI (Norman et al., 1995a), the derived energy fluxes, directly conditioned by soil moisture, could be used for the evaluation of drought indices. Su et al. (2003) used outputs of SEBS (Su, 2002) to propose a fully RS-derived Drought Severity Index (DSI) for the North China Plain for low-resolution imagery. The results showed that the relative evaporation (actual latent heat flux over potential latent heat flux) can be used to predict soil moisture within one standard deviation, but the effects of cloudcontaminated pixels highly condition the applicability of the approach. Soil moisture derived from MW RS techniques can also be used for drought monitoring; however, the applicability is limited to continental scale as the most available products have rather coarse spatial resolution (tens of kilometers) and hence not applicable for most agricultural applications. More details on MW-derived soil moisture can be found in Section 2.14.7. More recently, it has been shown that time series of soil moisture derived from high-resolution MW sensors (tens of meters), such as ASAR (Van der Velde and Su, 2009). However, at the time of writing, such data are restricted to selected experimental areas where access of data is guaranteed.

2.14.10.4 Algorithm Retrievals and Operability AET estimates from RS are now available from a number of empirical, physically based, and mixed-approach methods. The scientific community agrees that there is no one single approach that best suits all cases, but some methods are preferred over others according to specific land-cover patterns and wetness characteristics of the site. In order to make RS information operational, continuous modeling, testing, and validation are required. The Demeter and Pleiades projects are examples of European efforts in using the latest observation and communication technologies to narrow the gap between research and operational needs. The reader is referred to the cited literature for detailed information on the different retrieval algorithms; we will only briefly describe the practical issues used in the SEBS algorithm for practical applications.

2.14.10.5 SEBS Algorithm The SEBS algorithm (Su, 2002) is a single-source physical model for the evaluation of the energy-balance fluxes from RS imagery. A full description of the model can be found in Su (2002, 2005) and the corresponding open-source software SEBS4ILWIS from the International Institute for Geo-information Science and Earth Observation. SEBS is a column model, which means that information on adjacent pixels is not affecting the pixel where the calculations

388

Observation of Hydrological Processes Using Remote Sensing

are done. This means that, in the ideal case, the information required as input needs to be measured independently at the site. However, in practice, this is never the case and interpolation techniques are always required. The examples presented at the end of the section were evaluated at measuring points. In the following, the values between brackets are examples taken from one point during this analysis. In the case that SEBS applied to imagery, the input indicated with (*) should desirably be a map. SEBS for ILWIS Open Source software also provides a set of routine for bio-geophysical parameter extraction. It uses satellite EO data, in combination with meteorological information as inputs, to produce the evaporative fraction, net radiation, and soil heat flux parameters. The main steps using the MODIS data as an example are as follows (SEBS core is theoretically independent of the sensor used): 1. Reprojecting and converting MODIS level-1 B data with the ModisSwathTool software. 2. Importing images into ILWIS. 3. Preprocessing for SEBS:

• • • • • •

raw data to radiance/reflectance conversion; brightness temperature computation; SMAC for atmospheric correction; land surface albedo computation; land surface emissivity, NDVI, vegetation proportion, and emissivity difference computation; and land surface temperature computation.

4. SEBS core model for bio-geophysical parameter extraction. Inputs in SEBS. Meteorological information from meteorological stations or atmospheric model out fields is used at the time of the satellite pass:

• • • • • • • •

Reference height (zref): height from the ground where measurements of temperature, wind, pressure, and specific humidity are made [m]. Specific humidity: [kg kg1]. Wind speed (uref): [m s1]. Air temperature at reference height (Ta): [1C]. Air pressure at reference height: [Pa]. Air pressure at land surface: [Pa]. PBL height (hi): height of the planetary boundary layer (PBL) in [m] that can be estimated by radiosounding or using atmospheric model outputs. (default hi ¼ 1000 m). Incoming global solar radiation: [W m2] (636 W m2).

Input normally derived after image preprocessing:

• • •

Atmospherically corrected broadband albedo (a) [–] (0.18). Surface emissivity (e0) (0.98). Atmospherically corrected surface temperature [K] (296 K).

Input derived from land-cover properties. Land properties affect roughness, the proven most sensitive information in all SEBSVAT methods. A good estimation of aerodynamic roughness is the key for success:

•

The percentage of fractional vegetation cover (fc) – it controls the partition of energy fluxes between vegetation and bare soil.

• •

The land use that contains roughness classes (zom) and displacement zero heights (d0), all in (m), associated with vegetation height values. The leaf area index is included in the deduction of the aerodynamic roughness for heat transport (zoh).

Outputs of SEBS. After the successful completion of the SEBS operation in ILWIS, following raster maps are generated:

• • • • • • • • •

sebs_ evap: evaporative fraction [–] sebs_daily_evap: daily evaporation [mm d1] sebs_evap_relative: relative evaporation [–] sebs_G0: soil heat flux [W m2] sebs_H_dry: sensible heat flux at the dry limit [W m2] sebs_H_i: sensible heat flux [W m2] sebs_H_wet: sensible heat flux at the wet limit [W m2] sebs_Rn: net radiation [W m2] sebs_LE: latent heat flux [W m2]

More details on the operation of the software SEBS4ILWIS can also be obtained from the online help.

2.14.10.6 Evaluation Example A simple application example is presented for part of the Guaren˜as catchment in the Duero Basin, Spain, which is being monitored with the Network of Soil Moisture Measurement Stations of the University of Salamanca (REMEDHUS). Measurements started from June 1999 to the present. The network consists of a series of 23 soil moisture stations, three meteorological stations, and discharge gages. First, RS-derived soil moisture estimates were compared to ground-truth data. The RS soil moisture retrievals were obtained indirectly from the SEBS RS model. AET retrievals using SEBS (Su, 2002) were calculated using the software SEBS4ILWIS (ITC, 2008; Wang et al., 2008). The top soil moisture values of the 23 stations were recorded simultaneously for 13 clear-day MODIS images taken during 2007. All VIS and NIR bands were radiometrically and atmospherically corrected using the SMAC (Rahman and Dedieu, 1994) version implemented in the Integrated Land and Water Information System (ILWIS, a GIS system). Surface temperature was obtained using a split window technique (Sobrino and Raissouni, 2003) and albedo estimates following (Liang, 2001). Due to the very low resolution (B1000 m in the thermal band) of the imagery, the comparison between the ground measurement and the RS soil moisture derived at the pixel where each ground station was located was futile, because the soil moisture at the spot did not represent the one of the pixel. However, a more representative and realistic approach was to make use of the simultaneous information of all 23 stations and the corresponding RS-derived soil moisture at the pixels where the stations were located. The RS-derived soil moisture average was obtained for the network. To estimate RS soil moisture, it was considered that the relative soil moisture (the ratio of the actual soil moisture to the soil moisture at the limiting wet case) was equal to the relative ET that could be evaluated after SEBS (Su et al., 2003). The soil moisture at limiting wet case was evaluated after laboratory analysis. The results illustrate that there is a good

Observation of Hydrological Processes Using Remote Sensing

389

Field average soil moisture

Ground measured (%)

20

16

12

8 Y = 0.85 + 0.01 R 2 = 0.85

4

0 0

4

8

12

16

20

SEBS derived (%) Figure 22 SEBS-derived vs. ground-measured soil moisture for the REMEDHUS project. Data courtesy: The University of Salamanca.

Table 4

Comparison between FAO 56 and remote-sensing-derived Kc factors after SEBS analysis

Crop Stage

Image date

Daily ETact (mm d1)

Average daily ETact (mm d1)

Ten-day average ET0 (mm d1)

Kc average calculated ETact/ET0

Kc FAO guide lines

Initial Crop development

(14 Nov. 2007) 16 Dec. 2007) (8 Mar. 2007) (27 Apr. 2008) (1 May. 2008) (18 Jun. 2008)

1.09 0.99 2.17 4.47 3.65 4.98

1.09 0.99 2.17 4.06

1.57 0.83 2.18 4.23

0.70 1.19 1.00 1.19

0.70 0.7–1.15 0.7–1.15 1.15

4.98

5.75

0.87

1.15–0.25

Mid-season Late season

agreement between SEBS-estimated soil moisture values and ground-measured values on the field scale level with a strong correlation. Figure 22 shows the results of the comparison. The general dry condition for the soils in the region was captured by the model. This information suffices for modeling purposes of the surface water that was done with the HVB model (Bergstro¨m, 1995). As a second example, SEBS was used to calculate AET from the available imagery. Then, RS-derived Kc factor for four stages of wheat development was derived and compared with the tabulated values of the FAO guidelines. The single crop coefficient (Allen et al., 1998) is used to calculate crop ET. With the available imageries, the Kc was computed for four stages of wheat development, namely, the initial, the crop development, the mid-season, and the late season and then compared with the tabulated values of the FAO guidelines. In the study area, the sowing dates for winter wheat vary from October to November and the harvesting dates vary from June to July. The subset was selected in a way of having a clear wheat zone close to meteorological stations. The nearest two meteorological stations, Canizal and VA_02, were selected to calculate a 10-day average crop reference ET. The 10-day average ET0 was calculated based on the estimations of 5 consecutive days before and after the imagery date. Table 4 shows the results of the comparison. The RS values of Kc are in agreement with the values given in the FAO guidelines; however, the procedure allows local fitting of the

Kc. The procedure is universally applicable mainly in irrigated areas, as it is recommended for spatial irrigation performance studies (Bos et al., 2005).

Acknowledgment This work was partially funded by EUMETSAT Satellite Application Facility on Climate Monitoring.

References Ackerman SA, Strabala KI, Menzel WP, Frey RA, Moeller CC, and Gumley LE (1998) Discriminating clear sky from clouds with MODIS. Journal of Geophysical Research 103(D24): 32141--32157. Adam JC, Clark EA, Lettenmaier DP, and Wood EF (2006) Correction of global precipitation products for orographic effects. Journal of Climate 19: 15--38. Adler RF, Huffman GJ, Chang A, et al. (2003) The version-2 Global Precipitation Climatology Project (GPCP) monthly precipitation analysis (1979–present). Journal of Hydrometeorology 4: 1147--1167. Adler RF and Negri AJ (1988) A satellite IR technique to estimate tropical convective and stratiform rainfall. Journal of Applied Meteorology 27: 30--51. Albert A and Gege P (2006) Inversion of irradiance and remote sensing reflectance in shallow water between 400 and 800 nm for calculations of water and bottom properties. Applied Optics 45: 2331--2343. Albertson JD (1996) Large Eddy Simulation of Land–Atmosphere Interaction. PhD Thesis, University of California, Davis, 185pp.

390

Observation of Hydrological Processes Using Remote Sensing

Albertson JD, Kustas WP, and Scanlon TM (2001) Large-eddy simulation over heterogeneous terrain with remotely sensed land surface conditions. Water Resources Research 37: 1939--1953. Alemseged TH and Rientjes THM (2007) Spatio-temporal rainfall mapping from space: Setbacks and strengths. In: Proceedings of the 5th International Symposium on Spatial Data Quality SDQ 2007, Modelling Qualities in Space and Time, 9pp. Enschede, The Netherlands, 13–15 June. Enschede: ITC. Allen RG, Pereira LS, Raes D, and Smith M (1998) Crop Evapotranspiration. Guidelines for Computing Crop Water Requirements. In: Irrigation and Drainage Paper, FAO 56, 300pp. FAO, Rome. Alley WM, Healy RW, LaBaugh JW, and Reilly TE (2002) Flow and storage in groundwater systems. Science 296: 1985--1990. Alsdorf DE, Fu L-L, Mognard NM, et al. (2007) Measuring the global oceans and terrestrial fresh water from space. EOS, Transactions, American Geophysical Union 18(24): 253--257. Alsdorf DE and Lettenmaier DP (2003) Tracking fresh water from space. Science 301: 1485--1488. Alsdorf DE, Lettenmaier DP, Vo¨ro¨smarty C, and the NASA Surface Water Working Group (2003) The need for global, satellite-based observations of terrestrial surface waters. EOS, Transactions, American Geophysical Union 84(269): 275--276. Alsdorf DE, Melack JM, Dunne T, Mertes LAK, Hess LL, and Smith LC (2000) Interferometric radar measurements of water level changes on the Amazon floodplain. Nature 404: 174--177. Anagnostou EN, Negri AJ, and Adler RF (1999) A satellite infrared technique for diurnal rainfall variability studies. Journal of Geophysical Research 104: 31477--31488. Anderson MC, Norman JM, Diak GR, Kustas WP, and Mecikalski JR (1997) A twosource time-integrated model for estimating surface fluxes using thermal infrared remote sensing. Remote Sensing of Environment 60: 195--216. Anderson MC, Norman JM, Kustas WP, Li F, Prueger JH, and Mecikalski JR (2005) Effects of vegetation clumping on two-source model estimates of surface energy fluxes from an agricultural landscape during SMACEX. Journal of Hydrometeorology 6: 892--909. Anderson MG and McDonnell JJ (eds.) (2005) Encyclopedia of Hydrological Sciences, Vol. 2, Part 5, Remote Sensing, pp. 657--996. Chichester: Wiley. Anthes RA, Bernhardt PA, Chen Y, et al. (2008) The COSMIC/FORMOSAT-3 mission: Early results. Bulletin of the American Meteorological Society 89: 313--333. Arnold JG and Fohrer N (2005) SWAT2000: Current capabilities and research opportunities in applied watershed modeling. Hydrological Processes 19(3): 563--572. Arnold JG, Srinivasan RS, Muttiah RS, and Williams JR (1998) Large-area hydrologic modeling and assessment: Part I. Model development. Journal of the American Water Resources Association 34(1): 73--89. Aumann HH, Chahine MT, Gautier C, et al. (2003) AIRS/AMSU/HSB on the aqua mission: Design, science objectives, data products, and processing systems. IEEE Transactions on Geoscience and Remote Sensing 41: 253--264. Avissar R and Pielke RA (1989) A parameterization of heterogeneous land surfaces for atmospheric numerical models and its impact on regional meteorology. Monthly Weather Review 117: 2113--2136. Ba MB and Gruber A (2001) GOES Multispectral Rainfall Algorithm (GMSRA). Journal of Applied Meteorology 40: 1500--1514. Bamber JL and Rivera A (2007) A review of remote sensing methods for glacier mass balance determination. Global and Planetary Change 59: 138--148. Bartalis Z, Wagner W, Naeimi V, et al. (2007) Initial soil moisture retrievals from the METOP-A advanced scatterometer (ASCAT). Geophysical Research Letters 34: L20401 (1–5, doi:10.1029/2007GL031088). Bartholic JF, Namken LN, and Wiegand CL (1972) Aerial thermal scanner to determine temperatures of soil and of crop canopy differing in water stress. Agronomy Journal 64: 603--608. Bastiaanssen WGM (1995) Regionalization of Surface Flux Densities and Moisture Indicators in Composite Terrain – a Remote Sensing Approach under Clear Skies in Mediterranean Climates, 273pp. PhD Thesis, Wageningen Agricultural University, Wageningen. Bastiaanssen WGM, Menenti M, Feddes RA, and Holtslag AAM (1998a) A remote sensing surface energy balance algorithm for land (SEBAL): 1. Formulation. Journal of Hydrology 212–213: 198--212. Bastiaanssen WGM, Pelgrum H, Wang J, et al. (1998b) A remote sensing surface energy balance algorithm for land (SEBAL): 2. Validation. Journal of Hydrology 212–213: 213--229. Bates D and Watts D (1988) Nonlinear Regression Analysis and Its Applications. New York, NY: Wiley.

Batra N, Islam S, Venturini V, Bisht G, and Jiang L (2006) Estimation and comparison of evapotranspiration from MODIS and AVHRR sensors for clear sky days over the Southern Great Plains. Remote Sensing of Environment 103: 1--15. Baum BA, Arduini RF, Wielicki BA, Minnis P, and Tsay S-C (1994) Multilevel cloud retrieval using multispectral HIRS and AVHRR data: Nighttime oceanic analysis. Journal of Geophysical Research 99: 5499--5514. Bausch WC (1995) Remote sensing of crop coefficients for improving the irrigation scheduling of corn. Agricultural Water Management 27: 55--68. Bausch W and Neale CMU (1987) Crop coefficients derived from reflected canopy radiation: A concept. Transactions of the ASAE 30(3): 703--709. Becker F and Li Z-L (1995) Surface temperature and emissivity at various scales: Definition, measurement and related problems. Remote Sensing Reviews 12: 225--253. Becker MW (2006) Potential for satellite remote sensing of groundwater. Groundwater 44: 306--318. Belmans C (1983) Simulation model of the water balance of a cropped soil: SWATRE. Journal of Hydrology 63(3–4): 271--286. Bennartz R and Fischer J (2001) Retrieval of columnar water vapor over land from backscattered solar radiation using the Medium Resolution Imaging Spectrometer. Remote Sensing of Environment 78: 274--283. Bergstro¨m S (1995) The HBV model. In: Sing VP (ed.) Computer Models of Watershed Hydrology, pp. 443--476. Highlands Ranch, CO: Water Resources Publications. Berthier E and Toutin T (2008) SPOT5-HRS digital elevation models and the monitoring of glacier elevation changes in North-West Canada and South-East Alaska. Remote Sensing of Environment 112(5): 14--28. Bertoldi G, Albertson JD, Kustas WP, Li F, and Anderson MC (2007) On the opposing roles of air temperature and wind speed variability in flux estimation from remotely sensed land surface states. Water Resources Research 43: W10433 (doi:10.1029/ 2007WR005911). Beven K (2001) How far can we go in distributed hydrological modeling. Hydrology and Earth System Sciences 5: 1--12. Birkett CM (1995) The contribution of TOPEX/POSEIDON to the global monitoring of climatically sensitive lakes. Journal of Geophysical Research 100: 25179--25204. Birkett CM (1998) Contribution of the TOPEX NASA radar altimeter to the global monitoring of large rivers and wetlands. Water Resources Research 34: 1223--1239. Birkett CM, Mertes LAK, Dunne T, Costa M, and Jasinski J (2004) Altimetric remote sensing of the Amazon: Application of satellite radar altimetry. Journal of Geophysical Research 107(D20): 8059 (doi:10.1029/2001JD000609). Boegh E, Soegaard H, Hanan N, Kabat P, and Lesch L (1999) A remote sensing study of the NDVI-Ts relationship and the transpiration from sparse vegetation in the Sahel based on high-resolution satellite data. Remote Sensing of Environment 69: 224--240. Boegh E, Soegaard H, and Thomsen A (2002) Evaluating evapotranspiration rates and surface conditions using Landsat TM to estimate atmospheric resistance and surface resistance. Remote Sensing of Environment 79: 329--343. Boegh E, Thorsen M, Butts MB, et al. (2004) Incorporating remote sensing data in physically based ditributed agro-hydrological modeling. Journal of Hydrology 287: 279--299. Bolle H-J (ed.) (2003). Mediterranean Climate – Variability and Trends, 372pp. Berlin, Springer. Bos MG, Burton MA, and Molden DJ (2005) Irrigation and Drainage Performance Assessment. Practical Guidelines, 158pp. Trowbridge: Cromwell Press. Bos MG, Kselik RAL, Allen RG, and Molden DJ (2009) Water Requirements for Irrigation and the Environment. Dordrecht: Springer. Bos MG, Murray-Rust DH, Merrey D, Johnson HG, and Snellen WB (1994) Methodologies for assessing performance of irrigation and drainage management. Irrigation and Drainage System 7: 231--261. Bowers SA and Hanks RJ (1965) Reflection of radiant energy from soil. Soil Science 100(2): 130--138. Bowers SA and Smith SJ (1972) Spectrophotometric determination of soil water content. Soil Science Society of America Proceedings 36: 978--980. Bringi VN and Chandrasekar V (2001) Polarimetric Doppler Weather Radar, 636pp. New York, NY: Cambridge University Press. Brogniez H, Le´mond J, Roca R, and Picon L (2007) Variability of free tropospheric humidity from Meteosat over the tropics: 1983–2005. In: EUM P.50 – Joint 2007 EUMETSAT and 15th AMS Conference, 6pp. http://www.eumetsat.int/Home/Main/ AboutEUMETSAT/Publications/ConferenceandWorkshopProceedings/2007/ groups/cps/documents/document/pdf_conf_p50_s7_02_brogniez_v.pdf (accessed September 2010). Brogniez H, Roca R, and Picon L (2005) Evaluation of the distribution of subtropical free tropospheric humidity in AMIP-2 simulations using METEOSAT water vapor channel data. Geophysical Research Letters 32: L19708 (doi:10.1029/ 2005GL024341).

Observation of Hydrological Processes Using Remote Sensing

Brown L, Thorne R, and Woo M-K (2008) Using satellite imagery to validate snow distribution simulated by a hydrological model in large northern basins. Hydrological Processes 22(15): 2777--2787. Brutsaert W (1982) Evaporation into the Atmosphere. 299pp. Dordrecht: Reidel. Brutsaert W (1992) Stability correction functions for the mean wind speed and temperature in the unstable surface layer. Geophysical Research Letters 19: 469--472. Brutsaert W (2008) Hydrology. An Introduction. New York, NY: Cambridge University Press. Brutsaert W and Sugita M (1992) Application of self-preservation in the diurnal evolution of the surface energy budget to determine daily evaporation. Journal of Geophysical Research 97(18): 377--382. Buehler S and John V (2005) A simple method to relate microwave radiances to upper tropospheric humidity. Journal of Geophysical Research 110: D02110 (doi:10.1029/2004JD005111). Buehler SA, et al. (2008) An upper tropospheric humidity data set from operational satellite microwave data. Journal of Geophysical Research 113: D14110 (doi:10.1029/2007JD009314). Burrows JP, et al. (1999) The global ozone monitoring experiment (GOME): Mission concept and first scientific results. Journal of the Atmospheric Sciences 56: 151--175. Carbone RE and Tuttle J (2008) Rainfall occurrence in the U.S. warm season: The diurnal cycle. Journal of Climate 21: 4132--4146. Carder K, Cattrall C, and Chen F (1999) MODIS Clear Water Epsilons Algorithm Theoretical Basis Document ATBD, 72pp. http://modis.gsfc.nasa.gov/data/atbd/ atbd_mod21.pdf (accessed September 2010). Carlson TN, Capehart WJ, and Gillies RR (1995) A new look at the simplified method for remote sensing of daily evapotranspiration. Remote Sensing of Environment 54: 161--167. Carlson TN, Gillies RR, and Perry EM (1994) A method to make use of thermal infrared temperature and NDVI measurements to infer surface soil water content and fractional vegetation cover. Remote Sensing Reviews 9: 161--173. Carlson TN, Perry EM, and Schmugge TJ (1991) Remote sensing estimation of soil moisture availability and fractional vegetation cover for agricultural fields. Agricultural and Forest Meteorology 52: 45--69. Cazenave A, Milly PCD, Douville H, Benveniste J, Kosuth P, and Lettenmaier DP (2004) Space techniques used to measure change in terrestrial waters. EOS, Transactions, American Geophysical Union 85(6): 58--59. Ceballos A, Scipal K, Wagner W, and Martinez-Fernandez J (2005) Validation of ERS scatterometer-derived soil moisture data in the central part of the Duero Basin, Spain. Hydrological Processes 19(8): 1549--1566. CEOS (2009) The Earth Observation Handbook. http://www.eohandbook.com (accessed September 2010). Chehbouni A, Hoedjes JCB, Rodriguez J-C, et al. (2008) Using remotely sensed data to estimate area-averaged daily surface fluxes over a semi-arid mixed agricultural land. Agricultural and Forest Meteorology 148: 330--342. Chehbouni A, Seen DL, Njoku EG, Lhomme JP, Monteny B, and Kerr YH (1997) Estimation of sensible heat flux over sparsely vegetated surfaces. Journal of Hydrology 188–189: 855--868. Chehbouni A, Seen DL, Njoku EG, and Monteney BM (1996) Examination of the difference between radiative and aerodynamic surface temperatures over sparsely vegetated surfaces. Remote Sensing of Environment 58: 177--186. Chen JL, Tapley BD, and Wilson CR (2006) Alaskan mountain glacial melting observed by satellite gravimetry. Earth and Planetary Science Letters 248: 368--378. Chomko R, Gordon H, Maritorena S, and Siegel D (2003) Simultaneous retrieval of oceanic and atmospheric parameters for ocean color imagery by optimization: a validation. Remote Sensing of Environment 84: 208--220. Choudhury BJ, Schmugge TJ, Chang A, and Newton RW (1979) Effect of surface roughness on the microwave emission from soils. Journal of Geophysical Research 84: 5699--5706. Clark MP, Slater AG, Barrett AP, et al. (2006) Assimilation of snow covered area information into hydrologic and land-surface models. Advances in Water Resources 29: 1209--1221. Coakley JA Jr. and Bretherton FP (1982) Cloud cover from high-resolution scanner data: Detecting and allowing for partially filled fields of view. Journal of Geophysical Research 87: 4917--4932. Coll C, Caselles V, Sobrino JA, and Valor E (1994) On the atmospheric dependence of the split-window equation for land surface temperature. International Journal of Remote Sensing 15(1): 105--122. Coll C, et al. (2005) Ground measurements for the validation of land surface temperatures derived from AATSR and MODIS data. Remote Sensing of Environment 97(3): 288--300.

391

Collier CG (1983) A weather radar correction procedure for real-time estimation of surface rainfall. Quarterly Journal of the Royal Meteorological Society 109: 589--608. Crago R (1996) Conservation and variability of the evaporative fraction during the daytime. Journal of Hydrology 180(1–5): 173--194. Crowley JW, Mitrovica JX, Bailey RC, Tamisiea ME, and Davis JL (2006) Land water storage within the Congo Basin inferred from GRACE satellite gravity data. Geophysical Research Letters 33: L19402. Curry JA, Hobbs PV, King MD, et al. (2000) FIRE Arctic clouds experiment. Bulletin of the American Meteorological Society 81: 5--29. Curtis S, Salahuddin A, Adler RF, Huffman GJ, Gu G, and Hong Y (2007) Precipitation extremes estimated by GPCP and TRMM: ENSO relationships. Journal of Hydrometeorology 8: 678--689. Davolio S and Buzzi A (2003) A nudging scheme for the assimilation of precipitation data into a mesoscale model. Weather Forecasting 19: 855--871. de Bruin HAR (1987) A model for the Priestley–Taylor parameter. Journal of Climate and Applied Meteorology 12: 572--578. de Bruin HAR, van den Hurk BJJM, and Kohsiek W (1995) The scintillation method tested over a dry vineyard area. Boundary-Layer Meteorology 76: 25--40. De Jeu R (2003) Retrieval of Land Surface Parameters Using Passive Microwave Remote Sensing, 122pp. PhD Dissertation, VU University Amsterdam, Amsterdam. De Jeu R, et al. (2008) Global soil moisture patterns observed by space borne microwave radiometers and scatterometers. Surveys in Geophysics 28: 399--420. De Lathauwer E, Timmermans WJ, Satalino G, Mattia F, Loew A, and Pauwels VRN (2009) Assessment of the potential use of remotely sensed soil moisture values for spatially distributed latent heat flux estimation. JSTARS-IEEE (submitted). De Wit A and van Diepen C (2007) Crop model data assimilation with the Ensemble Kalman Filter for improving regional crop yield forecasts. Agricultural and Forest Meteorology 146(1–2): 38--56. Dekker A, Malthus T, Wijnen M, and Seyhan E (1992) Remote sensing as a tool for assessing water quality in Loosdrecht lakes. Hydrobiologia 233: 137--159. Delanoe¨ J and Hogan RJ (2008) A variational scheme for retrieving ice cloud properties from combined radar, lidar, and infrared radiometer. Journal of Geophysical Research 113: D07204 (doi:10.1029/2007JD009000). Deneke HM, Roebeling RA, and Feijt A (2008) Estimation surface solar irradiance from METEOSAT SEVIRI-derived cloud properties. Remote Sensing of Environment 112: 3131--3141. Derksen C, Walker A, and Goodison B (2003) A comparison of 18 winter seasons of in situ and passive microwave derived snow water equivalent estimates in Western Canada. Remote Sensing of Environment 88(3): 271--282. Dickey JO, Bentley CR, Bilham R, et al. (1997) Satellite gravity and the geosphere. National Research Council Report, 112pp. Washington, DC: National Academies Press. DiGiacomo P, Hook S, Neumann A, et al. (2007) GEO inland and nearshore coastal water quality remote sensing. In: Bauer M, Dekker A, DiGiacomo P, et al. (eds.), GEO Group on Earth Observation, 31pp. http://www.earthobservations.org/ meetings/20070327_29_water_quality_workshop_report.pdf (accessed September 2010). Dobson MC, Ulaby FT, Hallikainen MT, and Elrayes MA (1985) Microwave dielectric behavior of wet soil. 2. Dielectric mixing models. IEEE Transactions on Geoscience and Remote Sensing 23(1): 35--46. Doerffer R and Schiller H (2007) The MERIS case 2 water algorithm. International Journal of Remote Sensing 28: 517--535. Doorenbos J and Pruitt WO (1977). Guidelines for Predicting Crop Water Requirements, Irrigation and Drainage Paper, FAO 33, 193pp. FAO, Rome. Dorigo WA, et al. (2007) A review on reflective remote sensing and data assimilation techniques for enhanced agroecosystem modeling. International Journal of Applied Earth Observation and Geoinformation 9: 165--193. Dozier J and Painter TH (2004) Multispectral and hyperspectral remote sensing of alpine snow properties. Annual Review of Earth and Planetary Science 32: 465--494. Draper C, Walker J, Steinle P, De Jeu R, and Holmes T (2009) Evaluation of AMSR-E derived soil moisture over Australia. Remote Sensing of Environment 113: 703--710. Drusch M, Holmes TRH, Rosnay PD, and Balsamo G (2009) Comparing ERA-40 based L-band brightness temperatures with Skylab observations: A preliminary calibration/validation study in preparation for the SMOS mission. Journal of Hydrometeorology 10(1): 213--226. Duarte J, Ve´lez-Reyes M, Tarantola S, Gilbes F, and Armstrong R (2003) A probabilistic sensitivity analysis of water-leaving radiance to water constituents in coastal shallow waters. In: Frouin RJ, Gilbert GD, and Pan D (eds.) Ocean Remote Sensing and Imaging, Proceedings of SPIE, vol. 5155, pp. 162–173.

392

Observation of Hydrological Processes Using Remote Sensing

Dubois PC, van Zyl J, and Engman ET (1995) Measuring soil moisture with imaging radars. IEEE Transactions on Geoscience and Remote Sensing 33: 915--926. Eagleman JR and Ulaby FT (1975) Remote sensing of soil moisture by SKYLAB radiometer and scatterometer sensors. Journal of the Astronautical Science 23: 147--159. Ebert EE, Janowiak J, and Kidd C (2007) Comparison of near-real-time precipitation estimates from satellite observations and numerical models. Bulletin of the American Meteorological Society 88: 47--64. Eleveld M, van der Woerd H, El Serafy G, Blaas M, van Kessel T, and de Boer G (2008) Assimilation of remotely sensed observation in a sidement transport model. In: Ocean Optics. Barga, Italy. Halifax: Lewis Conference Services International. http:// oceanopticsconference.org/abstracts (accessed September 2010). Engman E and Chaunhan N (1995) Satus of microwave soil moisture measurements with remote sensing. Remote Sensing of Environment 51: 189--198. Entekhabi D, Njoku E, O’Neill P, et al. (2010) The soil moisture active and passive (SMAP) mission. Proceedings of the IEEE 98: 704--716 (doi:10.1109/ JPROC.2010.2043918). ESA (2006) The Changing Earth, ESA SP-1304, 85pp. http://esamultimedia.esa.int/ docs/SP-1304.pdf (accessed September 2010). Eyre R (1989) Inversion of cloudy satellite sounding radiances by nonlinear optimal estimation. I: Theory and simulation for TOVS. Quarterly Journal of the Royal Meteorological Society 115: 1001--1026. Ezzahar J, Chehbouni A, Hoedjes JCB, and Chehbouni A (2007) On the application of scintillometry over heterogeneous grids. Journal of Hydrology 334: 493--501. Ferraro RR, Weng F, Grody NC, et al. (2005) NOAA operational hydrological products derived from the Advanced Microwave Sounding Unit. IEEE Transactions on Geoscience and Remote Sensing 43: 1036--1049. Fetzer EJ, Lambrigtsen BH, Eldering A, Aumann HH, and Chahine MT (2006) Biases in total precipitable water vapor climatologies from Atmospheric Infrared Sounder and Advanced Microwave Scanning Radiometer. Journal of Geophysical Research 111: D09S16 (doi:10.1029/2005JD006598). Fetzer EJ, et al. (2008) Comparison of upper tropospheric water vapor observations from the Microwave Limb Sounder and Atmospheric Infrared Sounder. Journal of Geophysical Research 113: D22110 (doi:10.1029/2008JD010000). Floricioiu D, Eineder M, Rott H, and Nagler T (2008) Velocities of major outlet glaciers of the Patagonia Icefield observed by TerraSAR-X. In: Proceedings of the IEEE International Geoscience and Remote Sensing Sympsium 2008, paper no. 2241, pp. 31–34 (ISBN 978-1-4244-2808). Foken T (2008) The energy balance closure problem: an overview. Ecological Application 18: 1351--1367. Foster JL and Chang ATC (1993) Snow cover. In: Gurney RJ, et al. (eds.) Atlas of Satellite Observations Related to Global Change, pp. 361--370. Cambridge: Cambridge University Press. Foster JL, Sun C, Walker JP, et al. (2005) Quantifying the uncertainty in passive microwave snow water equivalent observations. Remote Sensing of Environment 94: 187--203. Francois C, Ottle C, and Prevot L (1997) Analytical parameterization of canopy directional emissivity and directional radiance in the thermal infrared. Application on the retrieval of soil and foliage temepratures using two directional measurements. International Journal of Remote Sensing 18: 2587--2621. Frappart F, DoMinh K, L’Hermitte J, et al. (2006) Water volume change in the lower Mekong from satellite altimetry and imagery data. Geophysical Journal International 167: 570--584. French A, Schmugge TJ, and Kustas WP (2002) Estimating evapotranspiration over El Reno, Oklahoma with ASTER imagery. Agronomie 22: 105--106. French AN, Jacob F, Anderson MC, et al. (2005) Surface energy fluxes with the Advanced Spaceborne Thermal Emission and Reflection radiometer (ASTER) at the Iowa 2000 SMACEX site (USA). Remote Sensing of Environment 99: 55--65. Fu L-L, Alsdorf DE, Rodriguez E, et al. (2009) The SWOT (surface water and ocean topography) mission: Spaceborne radar interferometry for oceanographic and hydrological applications. soumis a` OCEANOBS’09. Gentine P, Entekhabi D, Chehbouni A, Boulet G, and Duchemin B (2007) Analysis of evaporative fraction diurnal behaviour. Agricultural and Forest Meteorology 143: 13--29. GEO (2005) GEOSS. 10-Year Implementation Plan. Reference Document, GEO 1000R, GEO 1000R/ESA SP-1284, February 2005, Noordwijk, The Netherlands, 73pp. GEO (2007) The Full Picture. http://www.earthobservations.org/documents/ the_full_picture.pdf. Gillies RR, Carlson TN, Cui J, Kustas WP, and Humes KS (1997) A verification of the triangle method for obtaining surface soil water content and energy fluxes from remote measurements of Normalized Difference Vegetation Index (NDVI) and

surface radiant temperature. International Journal of Remote Sensing 18: 3145--3166. Gillespie AR, et al. (1998) A temperature and emissivity separation algorithm for advanced spaceborne thermal emission and reflection radiometer ASTER images. IEEE Transactions on Geoscience and Remote Sensing 21: 2127--2132. Gitelson A, Garbuzov G, Szilagyi F, Mittenzwey K, Karnieli A, and Kaiser A (1993) Quantitative remote sensing methods for real-time monitoring of inland waters quality. International Journal of the Remote Sensing 14: 1269--1295. Gons H, Rijkeboer M, and Ruddick K (2002) A chlorophyll-retrieval algorithm for satellite imagery (Medium Resolution Imaging Spectrometer) of inland and coastal waters. Journal of Plankton Research 24: 947--951. Gonzalez Dugo MP, Moran MS, Mateos L, and Bryant R (2005) Canopy temperature variability as an indicator of crop water stress severity. Irrigation Science 24: 233--240. Gonzalez Piqueras J (2006) Evapotranspiracion de la cubierta vegetal mediante la determinacion del coeficiente de cultivo por teledteccion. PhD Thesis, Universidad de Valencia, Valencia. Gordon H, Du T, and Zhang T (1997) Remote sensing of ocean color and aerosol properties: Resolving the issue of aerosol absorption. Applied Optics 36: 8670--8684. Gordon H and Morel A (1983) Remote Assessment of Ocean Color for Interpretation of Satellite Visible Imagery: A Review. New York, NY: Springer. Gordon H and Wang M (1994) Retrieval of water-leaving radiance and aerosol optical thickness over the ocean with SeaWiFS: A preliminary algorithm. Applied Optics 33: 443--452. Gould R, Arnone R, and Martinolic P (1999) Spectral dependence of scattering coefficients in case I and case II waters. Applied Optics 38: 2377--2383. Goutorbe JP, Dolman AJ, Gash JHC, et al. (eds.) (1997) HAPEX-Sahel. Journal of Hydrology 188–189: 1–1079. Gowda PH, Chavez JL, Colaizzi PD, Evett SR, Howell TA, and Tolk JA (2007) Remote sensing based energy balance algorithms for mapping ET: Current status and future challenges. Transactions of the ASABE 50: 1639–1168. Green AE, Astill MS, McAneney KJ, and Nieveen JP (2001) Path-averaged surface fluxes determined from infrared and microwave scintillometers. Agricultural and Forest Meteorology 109: 233--247. Greenwald TJ, Stephens GL, Vonder Haar TH, and Jackson DL (1993) A physical retrieval of cloud liquid water over the global oceans using SSM/I observations. Journal of Geophysical Research 98: 18471--18488. Gupta SK, Ritchey AC, Wilber AC, Whitlock CH, Gibson GG, and Stackhouse PW (1999) A climatology of surface radiation budget derived from satellite data. Journal of Climate 12: 2691--2710. Hall DK, Riggs GA, Salomonson VV, DiGirolamo NE, and Bayr KJ (2002) MODIS snow-cover products. Remote Sensing of Environment 83: 181--194. Hallikainen M, Ulaby FT, and Deventer T (1987) Extinction behavior of dry snow in the 18- to 90-GHz range. IEEE Transactions on Geoscience and Remote Sensing 25: 737--745. Han Q, Rossow W, Welch R, White A, and Chou J (1995) Validation of satellite retrievals of cloud microphysics and liquid water path using observations from FIRE. Journal of the Atmospheric Sciences 52: 4183--4195. Han SC and Ditmar P (2008) Localized spectral analysis of global satellite gravity fields for recovering time-variable mass redistributions. Journal of Geodesy 82: 423--430. Han SC, Kim H, Yeo I-Y, et al. (2009) Dynamics of surface water storage in the Amazon inferred from measurements of inter-satellite distance change. Geophysical Research Letters 36: L09403. Hanssen RF (2001) Radar Interferometry. 308 pp. Dordrecht: Kluwer Academic. Heilman JL, Heilman WE, and Moore DG (1982) Evaluating the crop coefficient using spectral reflectance. Agronomy Journal 74: 967--971. Hillel D (1998) Environmental Soil Physics. Boston: Academic Press. Hipp J (1974) Soil electromagnetic parameters as functions of frequency, soil density, and soil moisture. Proceedings of the IEEE 62: 98--103. Hitschfeld W and Bordan J (1954) Errors inherent in the radar measurement of rainfall at attenuating wavelengths. Journal of Meteorology 11: 58--67. Holleman I (2007) Bias adjustment and long-term verification of radar-based precipitation estimates. Meteorological Applications 14: 195--203. Holleman I, Delobbe L, and Zgonc A. (2008). Update on the European weather radar network (OPERA). In: Proceedings of ERAD, 2008, pp. 1–5. http://www.wmo.int/ pages/prog/www/IMOP/publications/IOM-96_TECO-2008/1(11)_Holleman_The_ Netherlands.pdf (accessed September 2010). Hollinger SE and Daughtry CST (1999) Southern Great Plains 1997 hydrological experiment: Vegetation sampling and data documentation. Technical Report, Atmosphere Environment, Section III. State Water Survey, Champaign, 1999.

Observation of Hydrological Processes Using Remote Sensing

Hong Y and Adler RF (2008) Predicting global landslide spatiotemporal distribution: Integrating landslide susceptibility zoning techniques and real-time satellite rainfall estimates. International Journal of Sediment Research 23: 249--257. Hoogenboom H, Dekker A, and De Haan J (1998) Retrieval of chlorophyll a and suspended matter in inland waters from CASI data by matrix inversion. Canadian Journal of Remote Sensing 24: 144--152. Hu C, Carder K, and Muller-Karger F (2000) Atmospheric correction of SeaWiFS imagery over turbid coastal waters: A practical method. Remote Sensing of Environment 74: 195--206. Huang H-L, et al. (2004) International MODIS and AIRS processing package (IMAPP): A direct broadcast software package for the NASA Earth Observing System. Bulletin of the American Meteorological Society 85: 159--161. Hurtado R and Sobrino JA (2001) Daily net radiation estimated from air temperature and NOAA–AVHRR data: A case study for the Iberian Peninsula. International Journal of Remote Sensing 22(8): 1521--1533. Idso SB, Kackson RD, and Pinter PJ (1981) Normalizing the stress degree of water for environmental variability. Agricultural and Forest Meteorology 24: 45--55. Idso SB, Schmugge TJ, and Jackson RD (1975) The utility of surface temperature measurements for remote sensing of soil water studies. Journal of Geophysical Research 80: 3044--3049. Idso SB, Schmugge TJ, Jackson RD, and Reginato RJ (1975) The utility of surface temperature measurements for the remote sensing of the soil water status. Journal of Geophysical Research 80: 3044--3049. IPCC (2008) Technical Paper on Climate Change and Water. IPCC, April 2008, 240pp. Geneva, Switzerland: World Meteorological Organization. Irmak S, et al. (2003) Predicting daily net radiation using minimum climatological data. Journal of Irrigation and Drainage Engineering 129(4): 256--269. ITC (2008) Integrated Land and Water Information System 3.6. Open Source. 52 North. ITC. Jackson D and Bates J (2001) Upper tropospheric humidity algorithm assessment. Journal of Geophysical Research 106: 32259--32270. Jackson RD, Idso SB, and Reginato RJ (1981) Canopy temperature as a crop water stress indicator. Water Resources Research 17: 1133--1138. Jackson RD, Reginato RJ, and Idso SB (1977) Wheat canopy temperature: A practical tool for evaluating water requirements. Water Resources Research 13: 651--656. Jackson TJ, et al. (1999a) Soil moisture mapping at regional scales using microwave radiometry: The Southern Great Plains hydrology experiment. IEEE Transactions on Geoscience and Remote Sensing 37(8): 2136--2149. Jackson TJ, Le Vine DM, Hsu AY, et al. (1999b) Soil moisture mapping at Regional Scales using microwave radiometry: The Southern Great Plains hydrology experiment. IEEE Transactions on Geoscience and Remote Sensing 37(5): 2136--2151. Jacobowitz H, et al. (2003) The advanced very high resolution radiometer pathfinder atmosphere (PATMOS) climate dataset: A resource for climate research. Bulletin of the American Meteorological Society 84(6): 785--793. Jacobs C, Roerink G, and Hammani A (2008) Crop water stress detection from remote sensing using the SSEBI-2 algorithm: A case study in Morocco. Unpublished lecture, Alterra. Jia L, Su Z, van den Hurk B, et al. (2003) Estimation of sensible heat flux using the Surface Energy Balance System (SEBS) and ATSR measurements. Physics and Chemistry of the Earth 28: 75--88. Jiang L and Islam S (2001) Estimation of surface evaporation map over southern Great Plains using remote sensing data. Water Resources Research 37: 329--340. Jochum MAO and Calera A (2006) Operational space-assisted irrigation advisory services: Overview of and lessons learned from the project DEMETER, AIP conference. Earth Observation for Vegetation Monitoring and Water Management 852: 3--13. John VO and Soden BJ (2007) Temperature and humidity biases in global climate models and their impact on climate feedbacks. Geophysical Research Letters 34: L18704 (doi:10.1029/2007GL030429). Jolivet D and Feijt A (2005) Quantification of the accuracy of LWP fields derived from NOAA-16 advanced very high resolution radiometer over three ground stations using microwave radiometers. Journal of Geophysical Research 110: D11204 (doi:10.1029/2004JD005205). Joss J and Waldvogel A (1990) Precipitation measurement and hydrology. In: Atlas D (ed.) Radar in Meteorology, pp. 577--606. Boston, MA: American Meteorological Society. Kabat P, Prince SD, and Prihodko L (eds.) (1997) HAPEXnSahel West Central Supersite: Methods, Measurements and Selected Results, Report 130. Wageningen: DLO Winand Staring Center. Kallio K, Kutser T, Hannonen T, et al. (2001) Retrieval of water quality from airborne imaging spectrometry of various lake types in different seasons. Science of the Total Environment 208: 59--77.

393

Kalma JD and Calder IR (1994) Land surface processes in large scale hydrology. Operational Hydrology Report No. 40, 60pp. Geneva, Switzerland: World Meteorological Organization. Kalma JD, McVicar TR, and McCabe MF (2008) Estimating land surface evaporation: A review of methods using remotely sensed surface temperature data. Surveys in Geophysics 29: 421--469. Kargel JS, et al. (2005) Multispectral imaging contributions to global land ice measurements from space. Remote Sensing of Environment 99: 187--219. Karlsson K (2003) A 10 year cloud climatology over Scandinavia derived from NOAA advanced very high resolution radiometer imagery. International Journal of Climatology 23: 1023--1044. Kern M, et al. (2008) CoReH2O: Candidate earth explorer core missions. Reports for Assessment, ESA SP-1313(3), 104pp. (ISSN 0379-6566). Noordwijk: Mission Science Division, ESA-ESTEC. Kerr YH, et al. (2001) Soil moisture retrieval from space: The Soil Moisture and Ocean Salinity (SMOS) mission. IEEE Transactions on Geoscience and Remote Sensing 39: 1729--1735. Key J, et al. (2007) IGOS Cryosphere Theme Report. WMO/TD-No. 1405, 100pp., August 2007. http://cryos.ssec.wisc.edu/docs/cryos_theme_report.pdf (accessed September 2010). Kidd C, Levizzani V, Turk FJ, and Ferraro RR (2009) Satellite precipitation measurements for water resource monitoring, Journal of the American Water Resources Association 45: 567--579. Kim SW, Hong SH, and Won JS (2005) An application of L-band synthetic aperture radar to tide height measurement. IEEE Transactions on Geoscience and Remote Sensing 4: 2708--2710. Kirdiashev KP, Chukhlantsev AA, and Shutko AM (1979) Microwave radiation of the earth’s surface in the presence of vegetation cover. Radiotekhnika i Elektronika 24(8): 256--264. Kite GW and Droogers P (2000) Comparing evapotranspiration estimates from satellites, hydrological models and field data. Journal of Hydrology 229: 3--18. Kogan FN (1990) Remote sensing of weather impacts on vegetation in nonhomogeneous areas. International Journal of Remote Sensing 11: 1405--1419. Koistinen J (1991) Operational correction of radar rainfall errors due to the vertical profile of reflectivity. In: 25th Radar Meteorological Conference, pp. 91–96. Paris, France. Reading: American Meteorological Society. Kolberg S, Rue H, and Gottschalk L (2006) A Bayesian spatial assimilation scheme for snow coverage observations in a gridded snow model. Hydrology and Earth System Sciences 10: 369--381. Kouraev AV, Zakharova EA, Samain O, Mognard NM, and Cazenave A (2004) Ob’ river discharge from TOPEX/Poseidon satellite altimetry data. Remote Sensing of Environment 93: 238--245. Kuchinke C, Gordon H, and Franz B (2009a) Spectral optimization for constituent retrieval in case 2 waters I: Implementation and performance. Remote Sensing of Environment 113: 571--587. Kuchinke C, Gordon H, Harding L, Jr, and Voss K (2009b) Spectral optimization for constituent retrieval in case 2 waters II: Validation study in the Chesapeake Bay. Remote Sensing of Environment 113: 610--621. Kuligowski R and Barros A (2001) Combined IR-microwave satellite retrieval of temperature and dewpoint profiles using artificial neural networks. Journal of Applied Meteorology 40: 2051--2067. Kummerow CD, Hong Y, Olson WS, et al. (2001) The evolution of the Goddard Profiling Algorithm (GPROF) for rainfall estimation from passive microwave sensors. Journal of Applied Meteorology 40: 1801--1820. Kustas WP (1990) Estimates of evapotranspiration within a one- and two-layer model of heat transfer over partial vegetation cover. Journal of Applied Meteorology 29: 704--715. Kustas WP, Diak GR, Norman JM. (2001). Time difference methods for monitoring regional scale heat fluxes with remote sensing. In: Lakshmi V, Albertson J, and Schaake J. (eds.), Observations and Modeling of the Land Surface Hydrological Processes, American Geophysical Union Water Science and Application Series, vol. 3, pp. 15–29. Washington, DC: American Geophysical Union. Kustas WP and Goodrich DC (1994) Monsoon’90 multidisciplinary experiment, preface. Water Resources Research 30: 1211--1225. Kustas WP, Humes KS, Norman JM, and Moran MS (1996) Single- and dual-source modeling of surface energy fluxes with radiometric surface temperature. Journal of Applied Meteorology 35: 110--121. Kustas WP and Norman JM (1996) Use of remote sensing for evapotranspiration monitoring over land surfaces. Hydrological Sciences 41: 495--516. Kustas WP and Norman JM (1997) A two-source approach for estimating turbulent fluxes using multiple angle thermal infrared observations. Water Resources Research 33: 1495--1508.

394

Observation of Hydrological Processes Using Remote Sensing

Kustas WP and Norman JM (1999) Evaluation of soil and vegetation heat flux predictions using a simple two-source model with radiometric temperatures for partial canopy cover. Agricultural and Forest Meteorology 94: 13--29. Kustas WP and Norman JM (2000) A two-source energy balance approach using directional radiometric temperature observations for sparse canopy covered surfaces. Agronomy Journal 92: 847--854. Kustas WP, Norman JM, Schmugge TJ, and Anderson MC (2004) Mapping surface energy fluxes with radiometric temperature. In: Quattrochi DA and Luvall JC (eds.) Thermal Remote Sensing in Land Surface Processes, pp. 205--253. Boca Raton, FL: CRC Press. Lagouarde J-P and McAneney KJ (1992) Daily sensible heat flux estimation from a single measurement of surface temperature and maximum air temperature. Boundary-Layer Meteorology 59: 341--362. Laing AG, Carbone RE, Levizzani V, and Tuttle J (2008) The propagation and diurnal cycles of deep convection in northern tropical Africa. Quarterly Journal of the Royal Meteorological Society 134: 93--109. Laviola S and Levizzani V (2008) Rain retrieval using 183 GHz absorption lines. In: IEEE Proceedings of the MicroRad 2008, 10th Specialist Meeting of Microwave Radiometry and Remote Sensing of the Environment, pp. 1–4 (doi: 10.1109/ MICRAD.2008.4579505). Firenze, 11–14 March. Leese J, Jackson TJ, Pitman A, and Dirmeyer P (2000) Meeting summary, GEWEX/ BAHC International Workshop on soil moisture monitoring, analysis, and prediction for hydrometeorological and hydroclimatological applications. Bulletin of the American Meteorological Society 39: 1248--1268. LeFavour G and Alsdorf D (2005) Water slope and discharge in the Amazon River estimated using the shuttle radar topography mission digital elevation model. Geophysical Research Letters 32: L17404 (doi:10.1029/2005GL023836). Lemke P, Ren J, et al. (2007) Observations: Changes in snow, ice and frozen ground. In: Climate Change 2007: The Physical Science Basis, Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, pp. 337–383. Cambridge and New York, NY: Cambridge University Press. Levizzani V, Bauer P, and Turk FJ (eds.) (2007) Measuring Precipitation from Space. Dordrecht: Springer., 722pp. Lhomme JP, Monteney B, and Amadou M (1994) Estimating sensible heat flux from radiometric temperature over sparse millet. Agricultural and Forest Meteorology 68: 77--91. Li F, Kustas WP, Prueger JH, Neale CMU, and Jackson TJ (2005) Utility of remote sensing based two-source energy balance model under low and high vegetation cover conditions. Journal of Hydrometeorology 6: 878--891. Li J, Wolf WW, Menzel WP, Zhang WJ, Huang HL, and Achtor TH (2000) Global soundings of the atmosphere from ATOVS measurements: The algorithm and validation. Journal of Applied Meteorology 39: 1248--1268. Liang S (2001) Narrowband to broadband conversions of land surface albedo I: Algorithms. Remote Sensing of Environment 76(2): 213--238. Liang S (2004) Quantitative Remote Sensing of Land Surface. Hoboken, NJ: Wiley. Liang S, et al. (2003) Narrowband to broadband conversions of land surface albedo: II. Validation. Remote Sensing of Environment 84(1): 25--41. Liebe HJ (2004) MPM – an atmospheric millimeter-wave propagation model. International Journal of Infrared and Millimeter Waves 10(8): 631--650. Liu Y, De Jeu RAM, van Dijk AIJM, and Owe M (2007) TRMM-TMI satellite observed soil moisture and vegetation density (1998–2005) show strong connection with El Nino in eastern Australia. Geophysical Research Letters 34 doi:10.1029/ 2007GL030311. Loew A (2007) Remote sensing data assimilation. Special Issue. Remote Sensing of Environment 112(4): 1257. Loew A, Holmes T, and De Jeu R (2009) The European heat wave 2003: Early indicators from multisensoral microwave sensing. Journal of Geophysical Research 114: D05103 (doi:10.1029/2008JD010533). Loiselet M, Stricker N, Menard Y, and Luntama JP (2000) GRAS – Metop’s GPS-based atmospheric sounder. ESA Bulletin 102: 38--44. Longe´pe´ N, Allain S, Ferro-Famil L, Pottier E, and Durand Y (2009) Snowpack characterization in mountainous regions using C-band SAR data and a meteorological model. IEEE Transactions on Geoscience and Remote Sensing 47(2): 406--418. Lu Z, Crane M, Kwoun O, Wells C, Swarzenski C, and Rykhus R (2005) C-band radar observes water-level change in coastal Louisiana swamp forests. EOS, Transactions, American Geophysical Union 86: 141--144. Luojus KP, Pulliainen JT, Metsa¨ma¨ki SJ, and Hallikainen MT (2007) Snow-covered area estimation using satellite radar wide-swath images. IEEE Transactions on Geoscience and Remote Sensing 45(4): 978--989. Luthcke SB, Zwally HJ, Abdalati W, et al. (2006) Recent Greenland ice mass loss by drainage system from satellite gravity observations. Science 314: 1286--1289.

Magnu´sson E, Rott H, Bjo¨rnsson H, and Pa´lsson F (2007) The impact of jo¨kulhlaups on basal sliding observed by SAR interferometry on Vatnajo¨kull. Journal of Glaciology 35(181): 232--240. Maheu C, Cazenave A, and Mechoso C (2003) Water level fluctuations in the Plata Basin (South America) from Topex/Poseidon Satellite Altimetry. Geophysical Research Letters 30: 1143--1146. Mann J and Lenschow DH (1994) Errors in airborne flux measurements. Journal of Geophysical Research 99: 14519--14526. Maritorena S and Siegel D (2005) Consistent merging of satellite ocean color data sets using a bio-optical model. Remote Sensing of Environment 94: 429--440. Marshall J and Palmer W (1948) The distribution of raindrops with size. Journal of Meteorology 5: 165--166. Massman WJ (1999) A model study of kB-1 H for vegetated surfaces using ‘localized near-field’ Lagrangian theory. Journal of Hydrology 223: 27--43. Ma¨tzler C (1987) Applications of the interaction of of microwaves with the natural snow cover. Remote Sensing Reviews 2: 259--387. McCabe MF and Wood EF (2006) Scale influences on the remote estimation of evapotranspiration using multiple satellite sensors. Remote Sensing of Environment 105: 271--285. McNaughton KG and Spriggs TW (1989) An Evaluation of the Priestley and Taylor Equation and the Complementary Relationship Using Results from a MixedLayer Model of the Convective Boundary Layer, pp 89–104. Wallingford: IAHS Red Book. McVicar TR and Jupp DLB (2002) Using covariates to spatially interpolate moisture availability in the Murray–Darling Basin: A novel use of remotely sensed data. Remote Sensing of Environment 79: 199--212. Mecikalski JR, Diak GR, Anderson MC, and Norman JM (1999) Estimating fluxes on continental scales using remotely-sensed data in an atmospheric–land exchange model. Journal of Applied Meteorology 38: 1352--1369. Meesters AGCA, De Jeu RAM, and Owe M (2005) Analytical derivation of the vegetation optical depth from the microwave polarization difference index. IEEE Geoscience and Remote Sensing Letters 2(2): 121--123. Mekonnen G and Bastiaanssen WGM (2000) A new simple method to determine crop coefficients for water allocation planning from satellites: Results from Kenya. Irrigation and Drainage Systems 14: 237--256. Menenti M and Choudhury BJ (1993) Parametrization of land surface evapotranspiration using a location-dependent potential evapotranspiration and surface temperature range. In: Bolle HJ, Feddes RA, and Kalma JD (eds.) Exchange Processes at the Land Surface for a Range of Space and Time Scales, Proceedings of the Yokohama Symposium, July 1993, pp. 561–568. Yokohama: IAHS. Menenti M, Jia L, and Bastiaanssen WGM (2004) Energy and water flow through the soil–vegetation–atmosphere system: The fiction of measurements and the reality of models. In: Feddes RA, de Rooij GH, and van Dam JC (eds.) Frontis Workshop on Un-Saturated Zone Modeling: Progress, Challenges and Applications, pp. 211– 229. Wageningen, The Netherlands, 3–5 October 2004. Mercier F, Cazenave A, and Maheu C (2002) Interannual lake level fluctuations in Africa from Topex–Poseidon: Connections with ocean–atmosphere interactions over the Indian Ocean. Global and Planetary Change 32: 141--163. Merlin O and Chehbouni A (2004) Different approaches in estimating heat flux using dual angle observations of radiative surface temperature. International Journal of Remote Sensing 25: 275--289. Michelson DB, Andersson T, Koistinen J, et al. (2000) BALTEX Radar Data Centre Products and their methodologies. Report RMK No. 90, Swedish Meteorological and Hydrological Institute. Mieruch S, Noel S, Bovensmann H, and Burrows JP (2008) Analysis of global water vapor trends from satellite measurements in the visible spectral range. Atmospheric Chemistry and Physics 8: 491--504. Milz M, Buehler SA, and John VO (2009) Comparison of AIRS and AMSU-B monthly mean estimates of upper tropospheric humidity. Geophysical Research Letters 36: L10804 (doi:10.1029/2008GL037068). Minnis P, et al. (2003) A global cloud database from VIRS and MODIS for CERES. Proceedings of the SPIE 4891: 115--126. Mironov VL, Dobson C, Kaupp VH, Komarov VA, and Kleshchenko VN (2004) Generalized refractive mixing dielectric model for moist soils. IEEE Transactions on Geoscience and Remote Sensing 42: 773--785. Mladenova I, Lakshmi V, Walker J, Long D, and De Jeu R (2009) Observations for disaggregation of radiometer derived soil moisture estimates over the NAFE’06 study area. IEEE Geoscience and Remote Sensing Letters 6(4): 640--643 (doi:10.1109/LGRS.2009.2021492). Mognard NM and Alsdorf DE (2006) Why a hydrology mission needs two-dimensional acquisitions of water surface elevations? In: Proceedings of the 15 Years of Progress in Radar Altimetry Symposium Venice, ESA Special Publication SP-614. Italy, 13–18 March 2006.

Observation of Hydrological Processes Using Remote Sensing

Mognard NM, Alsdorf DE, Cazenave A, and Rodriguez E (2007) The water and terrestrial elevation recovery hydrosphere mapper (WATER-HM): A dedicated surface water mission. In: Proceedings of the 2nd Space for Hydrology Workshop, ESA Workshop Proceedings Publication, WPP-280. 11–14 November 2007. Geneva: WMO. Molotoch NP (2009) Reconstructing snow water equivalent in the Rio Grande headwaters using remotely sensed snow cover data and a spatially distributed snowmelt model. Hydrological Processes 23: 1076--1089. Moran MS, Clarke T, Inoue Y, and Vidal A (1994) Estimating crop water deficit using the relation between surface-air temperature and spectral vegetation index. Remote Sensing of Environment 49: 246--263. Moran MS and Jackson RD (1991) Assessing the spatial distribution of evapotrnaspiration using remotely sensed inputs. Journal of Environmental Quality 20: 725--737. Mueller RW, Matsoukas C, Gratzki A, Behr HD, and Hollmann R (2009) The CM-SAF operational scheme for the satellite based retrieval of solar surface irradiance F a LUT based eigenvector hybrid approach. Remote Sensing of Environment 113: 1012--1024. Muller-Krager F, Hu C, Andre´foue¨t S, Varela R, and Thunell R (2005) The color of the coastal ocean and applications in the solution of research and management problems. In: Miller R, Del Castillo C, and McKee B (eds.) Remote Sensing of Coastal Aquatic Environments: Technologies, Techniques and Applications, pp. 101--127. Dordrecht: Springer. Nagler T and Rott H (2000) Retrieval of wet snow by means of multitemporal SAR data. IEEE Transactions on Geoscience and Remote Sensing 38(2): 754--765. Nagler T and Rott H (2005) Snow classification algorithm for Envisat ASAR. In: Proceedings of the Envisat & ERS Symposium, ESA SP-572. Salzburg, Austria, 6–10 September 2004. http://earth.esa.int/workshops/salzburg04/papers_posters/ 4C4_nagler_169.pdf (accessed September 2010). Nagler T, Rott H, Malcher P, and Mu¨ller F (2008) Assimilation of meteorological and remote sensing data for snowmelt runoff forecasting. Remote Sensing of Environment 112: 1408--1420. Nakajima T and King MD (1990) Determination of the optical thickness and effective particle radius of clouds from reflected solar radiation measurements. Part I. Theory. Journal of the Atmospheric Science 47: 1878--1893. Nauss T and Kokhanovsky AA (2007) Assignment of rainfall confidence values using multispectral satellite data at mid-latitudes: First results. Advanced Geosciences 10: 99--102. Nemani RR and Running SW (1989) Estimation of regional surface resistance to evapotranspiration from NDVI and thermal-IR AVHRR data. Journal of Applied Meteorology 28: 276--284. Neukermans G, Nechad B, and Ruddick K (2008) Optical remote sensing of coastal waters from geostationary platforms: A feasibility study – Mapping Total Suspended Matter with SEVIRI. In: Ocean Optics. Barga, Italy, 6–10 October 2008. Halifax: Lewis Conference Services International. Nieuwenhuis GJA, Smidt EH, and Thunnissen HAM (1985) Estimation of regional evapotranspiration of arable crops from thermal infrared images. International Journal of Remote Sensing 6: 1319--1334. Niu GY, Seo KW, Yang ZL, et al. (2007) Retrieving snow mass from GRACE terrestrial water storage change with a land surface model. Geophysical Research Letters 34: L15704. Njoku E, Ashcroft P, Chan T, and Li L (2005) Global survey and statistics of radiofrequency interference in AMSR-E land observations. IEEE Transactions on Geoscience and Remote Sensing 43(5): 938--947. Njoku EG and Chan SK (2006) Vegetation and surface roughness effects on AMSR-E land observations. Remote Sensing of Environment 100(2): 190--199. Njoku EG, Jackson TJ, Lakshmi V, Chan TK, and Nghiem SV (2003) Soil moisture retrieval from AMSR-E. IEEE Transactions on Geoscience and Remote Sensing 41(2): 215--229. Njoku EG and Kong J (1977) Theory for passive microwave remote sensing of near-surface soil moisture. Journal of Geophysical Research 82(20): 3108-3118. Noel S, Buchwitz M, Bovensmann H, Hoogen R, and Burrows J P (1999) Atmospheric water vapor amounts retrieved from GOME satellite data. Geophysical Research Letters 26: 1841--1844. Norman J, Kustas WP, and Humes KS (1995) A two-source approach for estimating soil and vegetation energy fluxes in observations of directional surface temperature. Agricultural and Forest Meteorology 77: 263--293. NRC (National Research Council) (2007) Earth Science and Applications from Space, Earth Science and Applications from Space: National Imperatives for the Next Decade and Beyond, 456pp. Washington, DC: National Academies Press. Overgaard J, Rosjberg D, and Butts MB (2006) Land-surface modelling in hydrological perspective – a review. Biogeosciences 3: 229--241.

395

Owe M, De Jeu R, and Holmes T (2008) Multisensor historical climatology of satellitederived global land surface moisture. Journal of Geophysical Research 113: F01002 (doi:10.1029/2007JF000769). Parajka J, et al. (2006) Assimilating scatterometer soil moisture data into conceptual hydrologic models at the regional scale. Hydrology and Earth System Sciences 10(3): 353--368. Pathe C, Wagner W, Sabel D, Doubkova M, and Basara J (2009) Using ENVISAT ASAR Global Mode data for surface soil moisture retrieval over Oklahoma, USA. IEEE Transactions on Geoscience and Remote Sensing 47(2): 468--480. Paul F, Ka¨a¨b A, Maisch M, Kellenberger T, and Haeberli W (2002) The new remote sensing-derived Swiss glacier inventory: I. Methods. Annals of Glaciology 34: 355--361. Pellarin T, Wigneron JP, Calvet JC, and Waldteufel P (2003) Global soil moisture retrieval from a synthetic L-band brightness temperature data set. Journal of Geophysical Research 108: 9–1. Pinker R, Frouin R, and Li Z (1995) A review of satellite methods to derive surface shortwave irradiance. Remote Sensing of Environment 51: 108--124. Platnick S (2001) A superposition technique for deriving mean photon scattering statistics in plane-parallel cloudy atmospheres. Journal of Quantitative Spectroscopy and Radiative Transfer 68: 57--73. Platnick S, et al. (2003) The MODIS cloud products: algorithms and examples from Terra. IEEE Transactions on Geoscience and Remote Sensing 41: 459--473. Pottier C, Garc¸on V, Larnicol G, Sudre J, Schaeffer P, and Le Traon P-Y (2006) Merging SeaWiFS and MODIS/Aqua ocean color data in North and Equatorial Atlantic using weighted averaging and objective analysis. IEEE Transactions on Geoscience and Remote Sensing 44: 3436--3451. Pougatchev P, et al. (2009) IASI temperature and water vapor retrievals – error assessment and validation. Atmospheric Chemistry and Physics Dicussions 9: 7971--7989. Pratt DA and Ellyet CD (1979) The thermal inertia approach to mapping of soil moisture and geology. Remote Sensing of Environment 8: 147--158. Price JC (1982) Estimation of regional scale evapotranspiration through analysis of satellite thermal-infrared data. IEEE Transactions on Geoscience and Remote Sensing GE-20: 286--292. Price JC (1985) On the analysis of thermal infrared imagery, the limited utility of apparent thermal inertia. Remote Sensing of Environment 18: 59--73. Price JC (1990) Using spatial context in satellite data to infer regional scale evapotranspiration. IEEE Transactions on Geoscience and Remote Sensing 28: 940--948. Probert-Jones JR (1962) The radar equation in meteorology. Quarterly Journal of the Royal Meteorological Society 88: 486--495. Pu Z, Xu L, and Salomonson VV (2007) MODIS/Terra observed seasonal variations of snow cover over the Tibetan Plateau. Geophysical Research Letters 34(6): L06706. Pulliainen J (2006) Mapping of snow water equivalent and snow depth in boreal and sub-arctic zones by assimilating space-borne microwave radiometer data and ground-based observations. Remote Sensing of Environment 101: 257--269. Pulliainen JT, Hallikainen MT, and Grandell J (1999) HUT snow emission model and its applicability to snow water equivalent retrieval. IEEE Transactions on Geoscience and Remote Sensing 37(8): 1378--1390. Pulliainen JT, Manninen T, and Hallikainen M (1998) Application of ERS-1 Wind Scatterometer data to soil frost and soil moisture monitoring in boreal forest zone. IEEE Transactions on Geoscience and Remote Sensing 36(3): 849--863. Quattrochi DA and Luvall JC (1999) Thermal infrared remote sensing for analysis of landscape ecological processes: Methods and applications. Landscape Ecology 14: 577--598. Querner EP and van Bakel PJT (1989) Description of the Regional Groundwater Flow Model SIMGRO. Wageningen: DLO-Staring Centrum. Rahman H and Dedieu G (1994) SMAC: A simplified method for the atmospheric correction of satellite measurements in the solar spectrum. International Journal of Remote Sensing 15(1): 123--143. Randel DL, Vonder TH, Haar MA, et al. (1996) A new global water vapor dataset. Bulletin of the American Meteorological Society 77: 1233--1246. Rast M, Bezy JL, and Bruzzi S (1999) The ESA Medium Resolution Imaging Spectrometer MERIS – a review of the instrument and its mission. International Journal of Remote Sensing 20: 1681--1702. Raup BH, et al. (2007) Remote sensing and GIS technology in the Global Land Ice. Measurements from Space (GLIMS) Project. Computers and Geosciences 33: 104--125. Ray SS and Dadhwal VK (2001) Estimation of crop evapotranspiration of irrigation command area using remote sensing and GIS. Agricultural Water Management 49(3): 239--249. Rees WG (2001) Physical Principles of Remote Sensing. Cambridge: Cambridge University Press.

396

Observation of Hydrological Processes Using Remote Sensing

Rinehart RE (2004) Radar for Meteorologists, 4th edn., 482pp. Columbia, MO: Rinehart Publications Riou C, Itier B, and Seguin B (1988) The influence of surface roughness on the simplified relationship between daily evaporation and surface temperature. International Journal of Remote Sensing 9: 1529--1533. Robinson DA, Dewey KF, and Heim RR (1993) Global snow cover monitoring: an update. Bulletin of the American Meteorological Society 74: 1689--1696. Rodell M, Chao BF, Au AY, Kimball J, and McDonald K (2003) Global biomass variation and its geodynamic effects, 1982–1998. Earth Interactions 9(2): 1--19. Rodell M, Chen J, Kato H, Famiglietti JS, Nigro J, and Wilson CR (2007) Estimating ground water storage changes in the Mississippi River Basin (USA) using GRACE. Hydrogeology Journal 15: doi:10.1007/s10040-006-0103-7. Rodell M and Famiglietti JS (1999) Detectability of variations in continental water storage from satellite observations of the time dependent gravity field. Water Resources Research 35: 2705--2723. Rodell M and Famiglietti JS (2001) An analysis of terrestrial water storage variations in Illinois with implications for the Gravity Recovery and Climate Experiment (GRACE). Water Resources Research 37: 1327--1340. Rodell M, Famiglietti JS, Chen J, et al. (2004a) Basin scale estimates of evapotranspiration using GRACE and other observations. Geophysical Research Letters 31: L20504 (doi:10.1029/2004GL020873). Rodell M and Houser PR (2004) Updating a land surface model with MODIS-derived snow cover. Journal of Hydrometeorology 5: 1064--1075. Rodell M, Houser PR, Jambor U, et al. (2004b) The Global Land Data Assimilation System. Bulletin of the American Meteorological Society 85(3): 381--394. Rodell M, Velicogna I, and Famiglietti JS (2009) Satellite-based estimates of groundwater depletion in India. Nature, doi:10.1038/nature08238. Rodriguez E, Morris C, Belz J, et al. (2005) An assessment of the SRTM topographic products. Technical Report JPL D-31639, Jet Propulsion Laboratory, Pasadena, CA, 143pp. Roebeling RA, Deneke HM, and Feijt AJ (2008) Validation of cloud liquid water path retrievals from SEVIRI using one year of CloudNET observations. Journal of Applied Meteorology and Climatology 47: 206--222. Roebeling RA, Feijt AJ, and Stammes P (2006) Cloud property retrievals for climate monitoring: Implications of differences between SEVIRI on METEOSAT-8 and AVHRR on NOAA-17. Journal of Geophysical Research 111: D20210 (doi:10.1029/ 2005JD006990). Roebeling RA and Holleman I (2009) Development and validation of rain rate retrievals from SEVIRI using weather radar observations. Journal of Geophysical Research (submitted). Roerink GJ, Bastiaanssen WGM, Chambouleyron J, and Menenti M (1997) Relating crop water consumption to irrigation water suply by remote sensing. Water Resources Management 11: 445--465. Roerink GJ, Su Z, and Menenti M (2000) S-SEBI: A simple remote sensing algorithm to estimate the surface energy balance. Physics and Chemistry of the Earth (B) 25: 147--157. Rosenfeld D and Gutman G (1994) Retrieving microphysical properties near the tops of potential rain clouds by multispectral analysis of AVHRR data. Atmospheric Research 34: 259--283. Rosenfeld D and Lensky IM (1998) Satellite-based insights into precipitation formation processes in continental and maritime clouds. Bulletin of the American Meteorological Society 79: 2457--2476. Rossa A, Bruen M, Macpherson B, Holleman I, Michelson D, and Michaelides S (eds.) (2005) COST Action 717: Use of Radar Observations in Hydrology and NWP Models, EUR 21954, 292pp. Brussels: European Union. Rossow WB and Garder L (1993) Cloud detection using satellite measurements of infrared and visible radiances for ISCCP. Journal of Climate 6: 2341--2369. Rossow WB and Schiffer RA (1999) Advances in understanding clouds from ISCCP. Bulletin of the American Meteorological Society 80: 2261--2287. Rott H (1997) Capabilities of microwave sensors for monitoring areal extent and physical properties of the snowpack. In: Sooroshian S, Gupta HV, and Rodda JC (eds.) Land Surface Processes in Hydrology, NATO ASI Series I: Global Environmental Change, vol. 46, pp. 135–167. Berlin: Springer. Rott H, Nagler T, Malcher P, and Mu¨ller F (2008) A satellite-based information system for glacier monitoring and modelling. In: Gomarasca MA (ed.) GeoInformation in Europe, Proceedings of the 27th EARSeL Symposium, pp. 395–402 (ISBN 9789059660618). Rotterdam: Millpress. Rowlands DD, Luthcke SB, Klosko SM, et al. (2005) Resolving mass flux at high spatial and temporal resolution using GRACE intersatellite measurements. Geophysical Research Letters 32: L04310 (doi:10.1029/2004GL021908). Rubio E, Mejuto M, Calera A, Vela A, Castan˜o S, and Moratalla A (2003) Validation of an operational modelling of evapotranspiration and direct recharge. In: Proceedings of the 10th International Symposium. Barcelona: SPIE.

Ruddick K, Ovidio F, and Rijkeboer M (2000) Atmospheric correction of SeaWiFS imagery for turbid coastal and inland waters. Applied Optics 39: 897--912. Rudiger C, et al. (2009) An Intercomparison of ERS-Scat, AMSR-E soil moisture observations with model simulations over France. Journal of Hydrometeorology 10: 431--447. Sahoo AK, et al. (2008) Evaluation of AMSR-E soil moisture results using the in-situ data over the Little River Experimental Watershed, Georgia. Remote Sensing of Environment 112(6): 3142--3152. Salama MS and Shen F (2009) Simultaneous atmospheric correction and quantification of suspended particulate matters from orbital and geostationary earth observation sensors. Estuarine, Coastal and Shelf Science (in press). Salama MS and Stein A (2009) Error decomposition and estimation of inherent optical properties. Applied Optics 48: 4947--4962. Salama MS, Su Z, Mannaerts CM, and Verhoef W (2009) Deriving inherent optical properties and associated uncertainties for the Dutch inland waters during the eagle campaign. Hydrology and Earth System Sciences 13: 1113--1121. Salama S (2003) Optical remote sensing for the estimation of marine bio-geophysical quantities. In: Hydraulics, p. 175. Leuven: Katholieke Universiteit Leuven. Salama S, Monbaliu J, and Coppin P (2004) Atmospheric correction of Advanced Very High Resolution Radiometer imagery. International Journal of Remote Sensings 25: 1349--1355. Salminen M, Pulliainen J, Metsa¨ma¨ki S, Kontu A, and Suokanerva H (2009) The behaviour of snow and snow-free surface reflectance in boreal forests: Implications to the performance of snow covered area monitoring. Remote Sensing of Environment 113: 907--918. Sanchez JM, Kustas WP, Caselles V, and Anderson MC (2008) Modelling surface energy fluxes over maize using a two-source patch model and radiometric soil and canopy temperature observations. Remote Sensing of Environment 112: 1130--1143. Sandholt I, Rasmussen K, and Andersen J (2002) A simple interpretation of the surface temperature/vegetation index space for assessment of surface moisture status. Remote Sensing of Environment 79: 213--224. Santer BD, et al. (2007) Identification of human-induced changes in atmospheric moisture content. Proceedings of the National Academy of Sciences 104(39): 15248--15253. Saunders RW and Kriebel KT (1988) An improved method for detecting clear sky and cloudy radiances from AVHRR radiances. International Journal of Remote Sensings 9: 123--150. Schanda E (1986) Physical Fundamentals of Remote Sensing, 187pp. Berlin and Heidelberg: Springer. Schmetz J, Pili P, Tjemkes S, et al. (2002) An introduction to Meteosat Second Generation (MSG). Bulletin of the American Meteorological Society 83: 977--992. Schmetz J and Turpeinen O (1988) Estimation of the upper tropospheric relative humidity field from METEOSAT water vapor image data. Journal of Applied Meteorology 27: 889--899. Schmugge T, Gloersen P, Wilheit T, and Geiger F (1974) Remote-sensing of soilmoisture with microwave radiometers. Journal of Geophysical Research 79(2): 317--323. Schuepp PH, MacPherson JI, and Desjardins RL (1992) Adjustment of footprint correction for airborne flux mapping over the FIFE site. Journal of Geophysical Research 97: 18455--18466. Schulz J, et al. (2009) The EUMETSAT satellite application facility on climate monitoring (CM-SAF): Mission concept and first results. Atmospheric Chemistry and Physics 9: 1687--1709. Scipal K, Drusch M, and Wagner W (2008) Assimilation of a ERS scatterometer derived soil moisture index in the ECMWF numerical weather prediction system. Advances in Water Resources 31(8): 1101--1112. Scott N, et al. (1999) Characteristics of the TOVS Pathfinder Path-B dataset. Bulletin of the American Meteorological Society 80: 2679--2701. Seguin B and Itier B (1983) Using midday surface temperature to estimate daily evporation from satellite thermal IR data. International Journal of Remote Sensings 4: 371--383. Sellers PJ, Hall FG, Asrar G, Strebel DE, and Murphy RE (1992) An overview of the First International Satellite Land Surface Climatology Project (ISLSCP) Field Experiment (FIFE). Journal of Geophysical Research 97(D17): 18345--18371. Shamir E and Georgakakos KP (2006) Distributed snow accumulation and ablation modeling in the American River basin. Advances in Water Resources 29: 558--570. Shen F, Salama MS, Zhou Y, Li J, Su Z, and Kuang D (2010) Remote-sensing reflectance characteristics of highly turbid estuarine waters – a comparative experiment of the Yangtze River and the Yellow River. International Journal of Remote Sensing 31(10): 2639--2654.

Observation of Hydrological Processes Using Remote Sensing

Shiklomanov IA (1993) World fresh water resources. In: Gleick PH (ed.) Water in Crisis. New York, NY: Oxford University Press. Shuttleworth JW and Wallace JS (1985) Evaporation from sparse crops – an energy combination theory. Quarterly Journal of the Royal Meteorological Society 111: 839--855. Shuttleworth WJ (1988) Macrohydrology – the new challenge for process hydrology. Journal of Hydrology 100: 31--56. Shuttleworth WJ, Gurney RJ, Hsu AY, and Ormsby JP (1989) FIFE: The Variation in Energy Partition at Surface Flux Sites, vol. 186, pp. 67–74. Wallingford: IAHS Publication. Siegel D, Wang M, Maritorena S, and Robinson W (2000) Atmospheric correction of satellite ocean color imagery: The black pixel assumption. Applied Optics 39: 3582--3591. Simeoni D, Singer C, and Chalon G (1997) Infrared atmospheric sounding interferometer. Acta Astronautica 40: 113--118. Sirguey P, Mathieu R, and Arnaud Y (2009) Subpixel monitoring of the seasonal snow cover with MODIS at 250 m spatial resolution in the Southern Alps of New Zealand: Methodology and accuracy assessment. Remote Sensing of Environment 113: 160--181. Slater AG and Clark MP (2006) Snow data assimilation via an ensemble Kalman filter. Journal of Hydrometeorology 7: 478--493. Smith LC (1997) Satellite remote sensing of river inundation area, stage, and discharge: A review. Hydrological Processes 11: 1427--1439. Sobrino JA, Jime´nez-Mun˜oz JC, So`ria G, et al. (2008) Thermal remote sensing in the framework of the SEN2FLEX project: Field measurements, airborne data and applications. International Journal of Remote Sensing 29(17): 4961--4991. Sobrino JA, Jimenez-Munoz JC, Zarco-Tejada PJ, et al. (2009) Thermal remote sensing from Airborne Hyperspectral Scanner data in the framework of the SPARC and SEN2FLEX projects: An overview. Hydrology and Earth System Sciences 13: 2031--2037. Sobrino JA and Raissouni N (2003) Surface temperature and water vapour retrieval from MODIS data. International Journal of Remote Sensings 24(24): 5162--5182. Soden BJ and Lanzante JR (1996) An assessment of satellite and radiosonde climatologies of upper-tropospheric water vapor. Journal of Climate 9: 1235--1250. Soer GJR (1980) Estimation of regional evapotranspiration and soil moisture conditions using remotely sensed crop surface temperatures. Remote Sensing of Environment 9: 27--45. Sohn BJ and Smith EA (2003) Explaining sources of discrepancy in SSM/I water vapor algorithms. Journal of Climate 16: 3229--3255. Solberg R, Huseby RB, Koren H, and Malnes E (2008) Time-series fusion of optical and SAR data for snow cover area mapping. In: Proceedings of the 5th EARSeL LIS-SIG Workshop: Remote Sensing of Land Ice and Snow. Bern, 11–13 February 2008. http://www.conferences.earsel.org/abstract/show/395 (accessed September 2010). Stamnes K, Li W, Yan B, et al. (2003) Accurate and self-consistent ocean color algorithm: Simultaneous retrieval of aerosol optical properties and chlorophyll concentrations. Applied Optics 42: 939--951. Stokstad E (1999) Scarcity of rain, stream gages threatens forecasts. Science 285: 1199. Stone LR and Horton ML (1974) Estimating evapotrnapsiration using canopy temperatures: Field evaluation. Agronomy Journal 66: 450--454. Strassberg G, Scanlon BR, and Rodell M (2007) Comparison of seasonal terrestrial water storage variations from GRACE with groundwater-level measurements from the High Plains Aquifer (USA). Geophysical Research Letters 34: L14402 (doi:10.1029/2007GL030139). Strozzi T, Luckman A, Murray T, Wegmu¨ller U, and Werner CL (2002) Glacier motion estimation using SAR offset-tracking procedures. IEEE Transactions on Geoscience and Remote Sensing 40(119): 2384--2391. Stuefer M, Rott H, and Skvarca P (2007) Glaciar Perito Moreno, Patagonia: Climate sensitivities and glacier characteristics preceding the 2003/2004 and 2005/2006 damming events. Journal of Glaciology 53(180): 1--13. Su Z (2002) The Surface Energy Balance System (SEBS) for estimation of turbulent heat fluxes. Hydrology and Earth System Sciences 6: 85--99. Su Z (2005) Estimation of the surface energy balance. In: Anderson MG and McDonnell JJ (eds.) Encyclopedia of Hydrological Sciences, pp. 731--752. Chichester: Wiley. Su Z and Pelgrum H (1999) and Menenti MAggregation effects of surface heterogeneity in land surface processes. Hydrology and Earth System Sciences 3: 549--563. Su Z, Schmugge T, Kustas WP, and Massman WJ (2001) An evaluation of two models for estimation of the roughness height for heat transfer between the land surface and the atmosphere. Journal of Applied Meteorology 40: 1933--1951.

397

Su Z, Timmermans WJ, Gieske A, et al. (2008) Quantification of land–atmosphere exchanges of water, energy and carbon dioxide in space and time over the heterogeneous Barrax site. International Journal of Remote Sensing 29(17): 5215--5235. Su Z, Timmermans WJ, Tol Cvd, et al. (2009) EAGLE 2006 – multi-purpose, multiangle and multi-sensor in-situ, airborne and space borne campaigns over grassland and forest. Hydrology and Earth System Sciences 13: 833--845. Su Z, et al. (2003) Assessing relative soil moisture with remote sensing data: Theory and experimental validation. Physics and Chemistry of the Earth 28(1–3): 89--101. Surussavadee C and Staelin DH (2008) Global millimeter-wave precipitation retrievals trained with a cloud-resolving numerical weather prediction model, part I: Retrieval design. IEEE Transactions on Geoscience and Remote Sensing 46: 99--108. Swenson S, Famiglietti J, Basara J, and Wahr J (2008) Estimating profile soil moisture and groundwater variations using GRACE and Oklahoma Mesonet soil moisture data. Water Resources Research 44: W01413. Swenson S, Yeh PJF, Wahr J, and Famiglietti J (2006) A comparison of terrestrial water storage variations from GRACE with in situ measurements from Illinois. Geophysical Research Letters 33: L08607 (doi:10.1029/2005GL025489). Sydor M, Gould R, Arnone R, Haltrin V, and Goode W (2004) Uniqueness in remote sensing of the inherent optical properties of ocean water. Applied Optics 43: 2156--2162. Syed TH, Famiglietti JS, Chen J, et al. (2005) Total basin discharge for the Amazon and Mississippi River basins from GRACE and a land–atmosphere water balance. Geophysical Research Letters 32: L24404 (doi:10.1029/2005GL024851). Tapley BD, Bettadpur S, Ries JC, Thompson PF, and Watkins MM (2004) GRACE measurements of mass variability in the Earth system. Science 305: 503--505. Timmermans J, Verhoef W, Tol Cvd, and Su Z (2009a) Retrieval of canopy component temperatures through Bayesian inversion of directional thermal measurements. Hydrology and Earth System Sciences Discussions 6: 3007--3040. Timmermans WJ, Bertoldi G, Albertson JD, Olioso A, Su Z, and Gieske ASM (2008) Accounting for atmospheric boundary layer variability on flux estimation from RS observations. International Journal of Remote Sensing 29: 5275--5290. Timmermans WJ, Jime´nez-Mun˜oz JC, Hidalgo V, et al. (2009b) Surface energy fluxes from high resolution data during AgriSAR-2006: A model inter-comparison. Journal of Selected Topics in Applied Earth Observations and Remote Sensing – IEEE (in review). Timmermans WJ, Kustas WP, Anderson MC, and French AN (2007) An intercomparison of the surface energy balance algorithm for land (SEBAL) and the two-source energy balance (TSEB) modeling schemes. Remote Sensing of Environment 108: 369--384. Timmermans WJ, Su Z, and Olioso A (2009c) Footprint issues in scintillometry over heterogeneous landscapes. Hydrology and Earth System Sciences Discussions 6: 2099--2127. Todd MC, Barrett EC, Beaumont MJ, and Green JL (1995) Satellite identification of rain days over the upper Nile river basin using an optimum infrared rain/no-rain threshold temperature model. Journal of Applied Meteorology 34: 2600--2611. Tol Cvd, Tol Svd, Verhoef A, et al. (2009) A Bayesian approach to estimate sensible and latent heat over vegetation. Hydrology and Earth System Sciences 13: 749--758. Trenberth KE, Fasullo JT, and Kiehl J (2009) Earth’s global energy balance. Bulletin of the American Meteorological Society 311–323: doi:10.1175/2008BAMS2634.1. Trenberth KE, Fasullo J, and Smith L (2005) Trends and variability in column-integrated atmospheric water vapor. Climate Dynamics 24: 741--758 (doi:10.1007/s00382-0050017-4). Trenberth KE, Smith L, Qian T, Dai A, and Fasullo J (2007) Estimates of the global water budget and its annual cycle using observational and model data. Journal of Hydrometeorology 8: 758--769. Twomey S (1977) Atmospheric Aerosols, 302pp. Elsevier. Ulaby FT (1976) Passive microwave remote-sensing of Earth’s surface. IEEE Transactions on Antennas and Propagation 24(1): 112--115. Ulaby FT, Moore RK, and Fung AK (eds.) (1981) Microwave Remote Sensing: Active and Passive, Vol. I: Microwave Remote Sensing Fundamentals and Radiometry, 456pp. Reading, MA: Addison-Wesley. Advanced Book Program. Valor E and Caselles V (1996) Mapping land surface emissivity from NDVI: Application to European, African and South American areas. Remote Sensing of Environment 57(3): 167--184. Van der Velde R and Su Z (2009) Dynamics in land surface conditions on the Tibetan Plateau observed by advanced synthetic aperture radar, ASAR. Hydrological Sciences Journal 54(6): 1079--1093. Van Der Woerd H and Pasterkamp R (2008) HYDROPT: A fast and flexible method to retrieve chlorophyll-a from multispectral satellite observations of optically complex coastal waters. Remote Sensing of Environment 112: 1795--1807.

398

Observation of Hydrological Processes Using Remote Sensing

van Meijgaard E and Crewell S (2005) Comparison of model predicted liquid water path with ground-based measurements during CLIWA-NET. Atmospheric Research, Special Issue: CLIWA-NET: Observation and Modelling of Liquid Water Clouds 75(3): 201--226 (doi:10.1016/j.atmosres.2004.12.006). Velde R, Su Z, Ek M, Rodell M, and Ma Y (2009) Influence of thermodynamic soil and vegetation parameterizations on the simulation of soil temperature states and surface fluxes by the Noah LSm over a Tibetan Plateau site. Hydrology and Earth System Sciences 13: 759--777. Velicogna I and Wahr J (2006) Measurements of time-variable gravity show mass loss in Antarctica. Science 311: 1754--1756. Venturini V, Bisht G, Islam S, and Jiang L (2004) Comparison of evaporative fractions estimated from AVHRR and MODIS sensors over South Florida. Remote Sensing of Environment 93: 77--86. Verhoef W (1996) Application of Harmonic Analysis of NDVI Time Series, Report No. 108. Wageningen, NL: DLO Winnand Staring Centre. Verhoest NEC, Lievens H, Wagner W, A´lvarez-Mozos J, Moran MS, and Mattia F (2008) On the soil roughness parameterization problem in soil moisture retrieval of bare surfaces from synthetic aperture radar. Sensors 8: 4213--4248. Verstraeten W, Veroustraete F, and Feyen J (2008) Assessment of evapotranspiration and soil moisture content across different scales of observation. Sensors 8: 70--117. Vignal B and Krajewski W (2001) Large-sample evaluation of two methods to correct range-dependent error for WSR-88 D rainfall estimates. Journal of Hydrometeorology 2: 490--504. Vikhamar D and Solberg R (2003) Snow-cover mapping in forests by constrained linear spectral unmixing of MODIS data. Remote Sensing of Environment 88(3): 309  323. Voisin N, Wood AW, and Lettenmaier DP (2008) Evaluation of precipitation products for global hydrological prediction. Journal of Hydrometeorology 9: 388--407. Vo¨ro¨smarty CJ, Askew A, Barry R, et al. (2001) Global water data: A newly endangered species. EOS, Transactions, American Geophysical Union 82(5): 54--58. Wagner W, et al. (2003) Evaluation of the agreement between the first global remotely sensed soil moisture data with model and precipitation data. Journal of Geophysical Research – Atmospheres 108(D19): 4611. Wagner W, et al. (2007a) Operational readiness of microwave remote sensing of soil moisture for hydrologic applications. Nordic Hydrology 38(1): 1--20. Wagner W, Lemoine G, and Rott H (1999) A method for estimating soil moisture from ERS scatterometer and soil data. Remote Sensing of Environment 70: 191--207. Wagner W, Naeimi V, Scipal K, De Jeu R, and Martinez-Fernandez J (2007b) Soil moisture from operational meteorological satellites. Hydrogeology Journal 15(1): 121--131. Wagner W, Verhoest NEC, Ludwig R, and Tedesco M (2009) Editorial, Remote sensing in hydrological sciences. Hydrology and Earth System Sciences 13: 813--817. Wahr J, Molenaar M, and Bryan F (1998) Time-variability of the Earth’s gravity field: Hydrological and oceanic effects and their possible detection using GRACE. Journal of Geophysical Research 103(30): 205--230. Walker JP, Houser PR, and Willgoose GR (2004) Active microwave remote sensing for soil moisture measurement: A field evaluation using ERS-2. Hydrological Processes 18: 1975--1997. Wang JR and Choudhury BJ (1981) Remote sensing of soil moisture content over bare field at 1. GHz frequency. Journal of Geophysical Research 4(86): 5277--5287. Wang JR and Schmugge TJ (1980) An empirical model for the complex dielectric permittivity of soils as a function of water content. IEEE Transactions on Geoscience and Remote Sensing 18: 288--295. Wang L, Parodi GN, and Su Z (2008) SEBS module BEAMS: A practical tool for surface energy balance estimates from remote sensing data. In: Agency ES (ed.) Proceedings of the 2nd MERIS/(A)ATSR Workshop, ESA SP-666, November 2008, Tool and Services, 8pp. Frascati, Italy, 22–26 September 2008. Noordwijk: ESA Communication Production Office ESTEC. Wang L and Qu JJ (2009) Satellite remote sensing applications for surface soil moisture monitoring: A review. Frontiers of Earth Science in China 3(2): 237--247. Wang P, Boss E, and Roesler C (2005) Uncertainties of inherent optical properties obtained from semianalytical inversions of ocean color. Applied Optics 44: 4074--4084. Wen J and Su Z (2003) A time series based method for estimating relative soil moisture with ERS wind scatterometer data. Geophysical Research Letters 30(7): 1397 (doi:10.1029/ 2002GL016557). Weng F, Zhao L, Ferraro RR, Poe G, Li X, and Grody NC (2003) Advanced microwave sounding unit cloud and precipitation algorithms. Radio Science 38: 8068 (doi:10.1029/2002RS002679). Wentz FJ (1997) A well-calibrated ocean algorithm for special sensor microwave/ imager. Journal of Geophysical Research 102: 8703--8718.

Weydahl DJ (2001) Analysis of ERS tandem coherence from glaciers, valleys and fjord ice in Svalbard. IEEE Transactions on Geoscience and Remote Sensing 39: 2029--2039. Wielicki BA, Barkstrom BR, Harrison EF, Lee RB, Smith GL, and Cooper JE (1996) Clouds and the Earth’s radiant energy system (CERES): An earth observing system experiment. Bulletin of the American Meteorological Society 77: 853--868. Wigneron JP, Laguerre L, and Kerr Y (2001) A simple parmeterization of the L-band microwave emission from rough agricultural soils. IEEE Transactions on Geoscience and Remote Sensing 39: 1697--1707. Wilson JW (1970) Integration of radar and raingage data for improved rainfall measurement. Journal of Applied Meteorology 9: 489--497. Wood EF (1998) Hydrologic measurements and observations: An assessment of needs. In: Hydrologic Sciences: Taking Stock and Looking Ahead, pp. 67--86. Washington, DC: National Academies Press. Woodhouse IH and Hoekman DH (2000) A model-based determination of soil moisture trends in Spain with the ERS-scatterometer. IEEE Transactions on Geoscience and Remote Sensing 38(4): 1783--1793. Yeh PJ-F, Swenson SC, Famiglietti JS, and Rodell M (2006) Remote sensing of groundwater storage changes in Illinois using the Gravity Recovery and Climate Experiment (GRACE). Water Resources Research 42: doi:10.1029/2006WR005374. Zaitchik BF, Rodell M, and Reichle RH (2008) Assimilation of GRACE terrestrial water storage data into a land surface model: Results for the Mississippi River Basin. Journal of Hydrometeorology 9(3): 535--548 (doi:10.1175/2007JHM951.1). Zhan X, Kustas WP, and Humes KS (1996) An intercomparison study on models of sensible heat flux over partial canopy surfaces with remotely sensed surface temperature. Remote Sensing of Environment 58: 242--256. Zhao D, Su ZB, and Zhao M (2006) Soil moisture retrieval from satellite images and its application to heavy rainfall simulation in eastern China. Advances in Atmospheric Sciences 23(2): 299--316. Zhao F and Nakajima T (1997) Simultaneous determination of water-leaving reflectance and aerosol optical thickness from Coastal Zone Color Scanner measurements. Applied Optics 36: 6949--6956. Zibordi G (2006) A network for standardized ocean color validation measurements. EOS, Transactions, American Geophysical Union 87: 293--297. Zribi M, Baghdadi N, Holah N, and Fafin O (2005) New methodology for soil moisture estimation and its application to ENVISAT-ASAR multi-incidence data inversion. Remote Sensing of Environment 96: 485--496.

Relevant Websites http://grace.sgt-inc.com Access Program; Mass Anomalies. http://www.cloud-net.org Cloudnet. http://www.demeter-ec.net DEMETER. http://earth.esa.int ESA Earthnet. http://www.eumetnet.eu.org EUMETNET, the Network of European Meteorological Services. http://www.knmi.nl EUMETNET, the Network of European Meteorological Services; Opera. http://ec.europa.eu European Commission; GMES. http://geoid.colorado.edu Fedora Core Test Page; Grace. http://www-app2.gfz-potsdam.de GFZ Potsdam, Department 1: The Grace Mission. http://www.glims.org GLIMS: Global Land Ice Measurements from Space. http://www.gewex.org Global Energy and Water Cycle Experiment: GEWEX. http://www.isac.cnr.it ISAC; CGMS, IPWG; IPWG-5, Hamburg, Germany. http://www.itc.nl ITC, Faculty of Geo-Information Science and Earth Observation. http://www.legos.obs-mip.fr LEGOS; Hydrology from Space. http://grace.jpl.nasa.gov NASA; Grace Tellus.

Observation of Hydrological Processes Using Remote Sensing

http://daac.ornl.gov ORNL, DAAC; The First ISLSCP Field Experiment (FIFE). http://www.pleiades.es PLEIADES. http://postel.mediasfrance.org Postel, December 2008: Globcover Validation Report and New Regional Land Cover Products Available. http://modis-snow-ice.gsfc.nasa.gov The Modis Snow/Ice Global Mapping Project.

http://www.ars.usda.gov United States Department of Agriculture; Agricultural Research Service; Monsoon’90; Soil Moisture Experiments. http://www.csr.utexas.edu University of Texas at Austin, Center for Space Research. http://free.vgt.vito.be Vegetation, Free Vegetation Products. http://www.wacmos.org WACMOS.

399

2.15 Hydrogeophysics SS Hubbard, Lawrence Berkeley National Laboratory, Berkeley, CA, USA N Linde, University of Lausanne, Lausanne, Switzerland & 2011 Elsevier B.V. All rights reserved.

2.15.1 Introduction to Hydrogeophysics 2.15.2 Geophysical Methods 2.15.2.1 Electrical Resistivity Methods 2.15.2.2 IP Methods 2.15.2.3 SP Methods 2.15.2.4 Controlled-Source Inductive EM Methods 2.15.2.5 GPR Methods 2.15.2.6 Seismic Methods 2.15.2.7 Surface Nuclear Magnetic Resonance 2.15.2.8 Gravity 2.15.2.9 Magnetics 2.15.2.10 Well Logging 2.15.3 Petrophysical Models 2.15.3.1 Electrical Conductivity 2.15.3.1.1 Archie’s law 2.15.3.1.2 Waxman–Smits law 2.15.3.1.3 The Johnson, Koplik, and Schwartz model 2.15.3.1.4 Self-similar models 2.15.3.2 Dielectric Permittivity 2.15.3.2.1 Volume averaging 2.15.3.2.2 Topp’s equations 2.15.3.3 Complex Conductivity 2.15.3.3.1 Cole–Cole model 2.15.4 Parameter Estimation/Integration Methods 2.15.4.1 Key Components, Constraints, Metrics, and Steps in Parameter Estimation 2.15.4.1.1 Model space and initial model 2.15.4.1.2 Objective function (systems of equations) 2.15.4.1.3 Inversion step or model proposal 2.15.4.1.4 Geophysical model or model population 2.15.4.2 Example Parameter Estimation Approaches 2.15.4.2.1 Direct mapping approaches 2.15.4.2.2 Integration approaches (geostatistical, Bayesian) 2.15.4.2.3 Joint inversion or fully coupled hydrogeophysical inversion 2.15.5 Case Studies 2.15.5.1 Subsurface Architecture Delineation 2.15.5.1.1 3D resistivity mapping of a Galapagos volcano aquifer 2.15.5.1.2 High-resolution GPR imaging of alluvial deposits 2.15.5.1.3 Subsurface flow architecture delineation using seismic methods 2.15.5.1.4 Fracture zonation characterization using azimuthal electrical methods 2.15.5.2 Delineation of Anomalous Fluid Bodies 2.15.5.2.1 Electrical resistivity to delineate high-ionic-strength plume boundaries 2.15.5.2.2 SP imaging of redox potentials associated with contaminated plumes 2.15.5.3 Hydrological Process Monitoring 2.15.5.3.1 Soil moisture monitoring 2.15.5.3.2 Saline tracer monitoring in fractured rock using time-lapse GPR methods 2.15.5.3.3 Seasonal changes in regional saltwater dynamics using time-lapse EM methods 2.15.5.4 Hydrogeological Parameter or Zonation Estimation for Improving Flow Predictions 2.15.5.4.1 Hydraulic conductivity and zonation estimation using GPR and seismic methods 2.15.5.4.2 Joint modeling to estimate temporal changes in moisture content using GPR 2.15.6 Summary and Outlook Acknowledgments References

401 403 403 404 404 405 405 406 406 407 407 407 407 408 408 408 408 409 409 409 409 410 410 410 410 410 410 413 413 413 413 414 414 414 414 415 415 416 417 417 417 419 420 420 422 423 424 424 425 427 430 430

401

402

Hydrogeophysics

2.15.1 Introduction to Hydrogeophysics

(e.g., Gelhar, 1993). For example, the distributions of microfractures and geological formations both influence the hydraulic conductivity and thus subsurface flow, albeit over dramatically different spatial scales. Similarly, different hydrological processes may exert varying degrees of control on subsurface flow and transport as a function of the scale: the overall system response of the particular problem may be dominated by seasonal precipitation patterns or by surface– groundwater interactions at the catchment scale; by the influence of groundwater pumping wells, gradients, and heterogeneity-induced mixing at the local scale; and by microbe–mineral interactions and diffusion at the grain scale (Figure 1). The level of subsurface characterization required for a particular problem depends therefore on many factors,

The shallow subsurface of the Earth is an extremely important geological zone, one that yields our water resources, supports our agriculture and ecosystems, influences our climate, and serves as the repository for our contaminants. The need to develop sustainable water resources for increasing population, agriculture, and energy needs and the threat of climate and land-use change on ecosystems contribute to an urgency associated with improving our understanding of flow and transport processes in the shallow subsurface. Developing a predictive understanding of subsurface flow and transport is complicated by the disparity of scales across which controlling hydrological properties and processes span

Evapotranspiration Precipitation

Coupling of hydrological system with ecosystem and climate

O2

Coupling of groundwater, vadose zone, and surface waters

O2

River stage fluctuation

Preferential transport and mixing induced by heterogeneity

U(VI) Qtz

Interaction between minerals, pore water, and microbes

Qtz Biofilm Fe(OH)3

Qtz

Figure 1 Subsurface flow and transport is impacted by coupled processes and properties that preferentially exert influence over a wide range of spatial scales, rendering characterization based on borehole data only challenging. Modified from US DOE (2010) Complex Systems Science for Subsurface Fate and Transport, Report from the August 2009 Workshop, DOE/SC-0123, U.S. Department of Energy Office of Science (www.science.doe.gov/ober/BER_workshops.html).

Hydrogeophysics

including: the level of heterogeneity relative to the characterization objective, the spatial and temporal scales of interest, and regulatory or risk drivers. In some cases, reconnaissance efforts that delineate major characteristics of the study site may be sufficient, while other investigations may require a much more intensive effort. Conventional techniques for characterizing or monitoring the hydrogeological properties that control flow and transport typically rely on borehole access to the subsurface. For example, established hydrological characterization methods (such as pumping, slug, and flowmeter tests) are commonly used to measure hydraulic conductivity in the vicinity of the wellbore (e.g., Freeze and Cherry, 1979; Butler, 2005; Molz et al., 1994), and wellbore fluid samples are often used for water-quality assessment (e.g., Chapelle, 2001). Unfortunately, data obtained using borehole methods may not capture sufficient information away from the wellbore to describe the key controls on subsurface flow. The inability to characterize controlling properties at a high-enough spatial resolution and over a large-enough volume for understanding and predicting flow and transport processes using borehole methods often hinders our ability to predict and optimally manage associated resources. The field of hydrogeophysics has developed in recent years to explore the potential that geophysical methods have for characterization of subsurface properties and processes relevant for hydrological investigations. Because geophysical data can be collected from many different platforms (such as from satellites and aircrafts, at the ground surface of the Earth, and within and between wellbores), integration of geophysical data with direct hydrogeological or geochemical measurements can provide characterization information over a variety of spatial scales and resolutions. The main advantage of using geophysical data over conventional measurements is that geophysical methods can provide spatially extensive information about the subsurface in a minimally invasive manner at a comparatively high resolution. The greatest disadvantage is that the geophysical methods only provide indirect proxy information about subsurface hydrological properties or processes relevant to subsurface flow and transport. Hydrogeophysical investigations strive to provide information that can be used to (1) develop insights about complex hydrological processes, (2) serve as input data to construct flow and transport models, and (3) guide the management of subsurface water resources and contaminants. The field of hydrogeophysics builds on many decades of experience associated with the mining and petroleum industries, which have relied heavily on geophysical methods to guide the exploration of ore and hydrocarbons, respectively. Because geophysics has been used as a tool in these industries for so long, there is a relatively good understanding about methods and optimal data acquisition approaches for given problems, as well as about petrophysical relationships associated with the consolidated, high-pressure, and hightemperature subsurface environments common to those industries. However, such subsurface conditions are quite different from the shallow, low-temperature, low-pressure, and weakly consolidated environments that typify most hydrogeological investigation sites. The parameter values and the functional form of the petrophysical relationships that link

403

geophysical properties to subsurface parameters, as well as the geophysical response itself, can vary dramatically between different types of environments. In the last decade, many advances have been made that facilitate the use of geophysical data for shallow subsurface hydrogeological characterization. These advances include those associated with instrument development, interpretation procedures, petrophysical relationships relevant to near subsurface environments, integration or joint inversion approaches for combining multiple data sets, and coupled hydrological and geophysical modeling. Simultaneously, the number of publications related to hydrogeophysics has dramatically increased and are now common contributions to hydrological and geophysical journals such as Water Resources Research, Journal of Hydrology, Vadose Zone Journal, and Geophysics. Most hydrological and Earth science professional meetings (such as American Geophysical Union, Geological Society of America, and European Geosciences Union) now commonly host one or more hydrogeophysical special sessions at their annual meetings. These meetings have created an active environment where geophysicists and hydrologists can interact to learn about each other’s methods and challenges. Many groundbreaking hydrogeophysical studies have now been published by researchers with a formal hydrological training and it is becoming more common for geophysicists to strive to gain hydrological insights in addition to focusing primarily on advancing geophysical instrumentation and methodology. Some of the fairly recent hydrogeophysical advances are summarized in two edited books Hydrogeophysics (Rubin and Hubbard, 2005) and Applied Hydrogeophysics (Vereecken et al., 2006), as well as by numerous individual publications. Generally, hydrogeophysical characterization and monitoring objectives can often be categorized into the following three categories: 1. hydrological mapping of subsurface architecture or features (such as interfaces between key geological units, water table, or contaminant plume boundaries); 2. estimating subsurface properties or state variables that influence flow and transport (such as hydraulic conductivity or soil moisture); and 3. monitoring subsurface processes associated with natural or engineered in situ perturbations (such as infiltration through the vadose zone and tracer migration). There are several components that are common to most hydrogeophysical studies. First and foremost, it is critical to collect high-quality geophysical data sets using the geophysical method or methods that are most likely to provide data that can help to resolve the hydrogeological characterization or monitoring objective and that work well in the given environment. Although the corresponding geophysical properties (such as electrical conductivity/resistivity from electrical and electromagnetic (EM) methods or dielectric constant from ground penetrating radar (GPR) methods) can be used to infer hydrogeological properties or structures, petrophysical relationships must be developed and invoked at some stage to link the geophysical properties or data with the property or variable of interest (such as hydraulic conductivity or water

404

Hydrogeophysics

content). Integration or joint inversion methodologies are used to systematically integrate or fuse disparate data sets (geophysical and hydrogeological) to obtain a meaningful interpretation that honors all data and physical laws. The ultimate step is the use of the integrated hydrogeophysical property or state model to interpret complex subsurface system processes or to guide the optimal management of subsurface water resources and contaminants. The key objectives of this chapter are to familiarize hydrogeologists and water-resource professionals with the state of the art as well as the existing challenges associated with hydrogeophysics. We provide a review of the key components of many hydrogeophysical studies as well as example case studies that are relevant to understanding of hydrological behavior at the field scale. The remainder of this chapter is organized as follows. A brief description of some of the key geophysical methods that are used in hydrogeophysics is provided in Section 2.15.2. Descriptions of theoretical and empirical petrophysical relationships that can be used to link the geophysical attributes to the hydrogeological property of interest are discussed in Section 2.15.3. Section 2.15.4 reviews parameter estimation and integration methods that are used to combine disparate data sets for a consistent interpretation of critical flow and transport properties. Finally, in Section 2.15.5 we present various case studies that illustrate the use of geophysical data sets, petrophysics, and estimation methods to investigate near subsurface systems, with a particular emphasis on case studies that are conducted over field scales relevant to water resources and contaminant remediation.

2.15.2 Geophysical Methods The purpose of this section is to introduce some of the geophysical techniques that are most commonly used for hydrogeological studies, including electrical resistivity tomography (ERT), induced polarization (IP), electromagnetic induction (EMI), self-potential (SP), GPR, seismic, surface nuclear magnetic resonance (SNMR), gravity, magnetics, and wellbore logging techniques. For each method, we provide a brief description of the underlying physical principles and instrumentation, common acquisition strategies, and general data reduction and interpretation methods. We restrict our discussions to practical use and limitations of common geophysical methods; geophysical theory (e.g., Telford et al., 1990) is beyond the scope of this discussion. For detailed information, references are given for each geophysical method. This discussion of classical geophysical methods is envisioned to compliment existing literature on what are typically considered to be hydrological sensors or measurement approaches, even though they rely on geophysical mechanisms; examples include soil moisture probes (time domain reflectometer and capacitance probes), EM wellbore flowmeters, and various remote-sensing sensors deployed from airborne or space-borne platforms. Reviews of these methods are provided by Vereecken et al. (2008) and Butler (2005). The discussion of geophysical methods provided here is augmented by Section 2.15.3, where several petrophysical relationships are provided that may permit the transfer of

geophysical measurements and models into estimates of hydrological parameters.

2.15.2.1 Electrical Resistivity Methods For groundwater studies, electrical resistivity methods have perhaps been more frequently used than any other geophysical method. Resistivity is a measure of the ability to resist electrical current flow through materials; it is the inverse of electrical conductivity and is an intrinsic property of the material. In electrical resistivity methods, a typically lowfrequency (o1 Hz) current is injected into the ground between two current electrodes, while one or more pairs of potential electrodes are used to measure electrical potential differences. At the low frequencies measured, energy loss via ionic and electronic conduction dominates. Ionic conduction results from the electrolyte filling the interconnected pore space (Archie, 1942) as well as from surface conduction via the formation of an electrical double layer at the grain-fluid interface (e.g., Revil and Glover, 1997, 1998). Electronic conduction resulting from the formation of continuous conductive pathways by metallic minerals is typically not important for most environmental applications. The current distribution can be visualized by equipotential surfaces, with current flow lines running perpendicular to these surfaces. The fraction of total current flow that penetrates to a particular depth is a function of the current electrode spacing and location, the electrical resistivity distribution of the subsurface materials, and the topography. Most resistivity surveys utilize a four-electrode measurement approach. To obtain a value for subsurface resistivity, two potential electrodes are placed at some distance from the current electrodes, and the difference in electrical potential or voltage is measured. This measurement, together with the injected current and the geometric factor which is a function of the particular electrode configuration and spacing, can be used to calculate resistivity for uniform subsurface conditions following Ohm’s law. Common electrode configurations include the Wenner, the Schlumberger, and the dipole–dipole arrays. In real heterogeneous (nonuniform) subsurface environments, the more general term ‘apparent resistivity’ is used, which refers to the resistivity of an equivalent uniform media. There are several modes of acquiring electrical data. Profiling is undertaken by moving the entire array laterally along the ground surface by a fixed distance after each reading to obtain apparent resistivity measurements over a relatively constant depth as a function of distance. As profiles give lateral variations in electrical conductivity but not information about vertical distribution, the interpretation of profile data is generally qualitative, and the primary value of the data is to delineate sharp lateral contrasts associated with vertical/near vertical contacts. Vertical electrical sounding (VES) curves give information about the vertical variations in electrical conductivity at a single ground surface location assuming an idealized one-dimensional (1 D) resistivity structure. For example, soundings with the Wenner array are obtained by expanding the array along a straight line so that the spacing between the individual electrodes remains equal for each measurement, but increases after each measurement.

Hydrogeophysics

The depth of investigation for a given measurement is a function of the electrode spacing as well as the subsurface resistivity contrasts; as the electrode spacing is increased, the data are increasingly sensitive to deeper structures. Modern multichannel geoelectrical equipment now includes multiplexing capabilities and automatic and autonomous computer acquisition, which greatly facilitate data acquisition within acceptable timeframes. Such surface imaging, now commonly called electrical resistivity tomography or ERT, allows the electrodes (tens to hundreds) to be used alternatively as both current and potential electrodes to obtain 2D or 3D electrical resistivity models (e.g., Gu¨nther et al., 2006). In fact, when performing ERT, it is limiting to restrict the measurement sequence to a given configuration type, since optimal data sets often consist of a combination of traditional and nontraditional configuration types (Stummer et al., 2004; Wilkinson et al., 2006). With the development of advanced and automated acquisition systems, robust inversion routines, and the capability of recording tens of thousands of measurements per hour, ERT has proved to be useful for dynamic process monitoring using electrodes placed at the ground surface or in wellbores. A review of surface and crosshole ERT methods for hydrogeological applications is given by Binley and Kemna (2005), and discussion of petrophysical relationships that link the electrical properties with hydrological properties of interest is described in Section 2.15.3.1.

2.15.2.2 IP Methods IP methods measure both the resistive and capacitive properties of subsurface materials. IP measurements can be acquired using the same four-electrode geometry that is conventionally used for electrical resistivity surveys, although IP surveys typically employ nonpolarizing electrodes. Surveys can be conducted in the time domain as well as in the frequency domain. In the time domain, the current is injected and the decay of the voltage over time is measured. Frequency-domain methods measure the impedance magnitude and phase shift of the voltage relative to an injected alternating current. Spectral induced polarization (SIP) methods measure the polarization relaxation over many frequencies (typically over the range of 0.1–1000 Hz). The voltage decay (in the time domain) and spectral response (in the frequency domain) are caused by polarization of ions in the electrical double layer at the mineral–fluid interface, by accumulation of electrical charges at pore space constrictions (e.g., pore throats), and by conduction in the pore fluid and along the fluid-grain boundaries. More information about IP methods is provided by Binley and Kemna (2005) and Leroy and Revil (2009). The linkage between IP attributes, granulometric properties, and interfacial phenomena suggests that it also holds significant potential for exploring hydrogeological properties (e.g., Slater and Lesmes, 2002) as well as complex biogeochemical processes associated with contaminant remediation (e.g., Williams et al., 2005, 2009; Slater et al., 2007). Section 2.15.3.3 provides discussion of petrophysical models associated with SIP data sets.

405

2.15.2.3 SP Methods SP is a passive method where naturally occurring electric fields (voltage gradients) are measured at the ground surface or in wellbores using nonpolarizable electrodes and a high-impedance voltmeter. Electrical potentials measured with the SP method obey a Poisson’s equation with a source term given by the divergence of an electrical source current density (e.g., Minsley et al., 2007). The source current density has several possible contributors, including those associated with ground water flow, redox phenomena, and ionic diffusion. The electrokinetic contribution associated with the flow of ground water in a porous medium (or more precisely, with the drag of charges contained in the diffuse layer that surrounds mineral surfaces) has been recognized for many decades and has been used to qualitatively interpret SP signals in terms of seepage beneath dams or to map groundwater flow (e.g., Poldini, 1938). However, only more recently have such data sets been used to quantify hydrological properties by coupling equations that represent volumetric fluid flux and volume current density, which are linked by a coupling coefficient (e.g., Sill, 1983; Revil et al., 2003). The underlying physics of the redox and ionic diffusion contributions are now better understood and current research is advancing our ability to use SP for quantitative hydrogeochemical characterization (such as for characterizing field-scale redox gradients; refer to case study provided in Section 2.15.5.2). The SP method is the only geophysical method that is directly sensitive to hydrological fluxes (e.g., Sill, 1983). Even if several alternative formulations exist to describe electrokinetic phenomena, we consider here the case where the SP sources are expressed in terms of Qv, where Qv is expressed as excess charge in the diffuse double layer per saturated pore volume. The relative movement of an electrolyte with respect to mineral grains with a charged surface area results in socalled streaming currents (e.g., Sill, 1983). These currents are intimately linked to the Darcy velocity U and an excess charge Qv along the hydrological flow paths. A practical formulation of the streaming currents that corresponds to this parametrization is (e.g., Revil and Linde, 2006)

Js ¼ Qv U

ð1Þ

This equation is only strictly valid when the size of the double layer is comparable to the size of the pores (see Revil and Linde (2006) for a description of chemico-electromechanical coupling under such conditions) and when internal permeability variations within the averaged volume are small. Equation (1) can be used in heterogeneous media or in coarse sediments when we replace Qv with an effective Qeff v that is scaled with the relative contributions to permeability of all flow paths in the media (Linde, 2009). It is straightforward to deduce Qeff v of aquifer materials in the laboratory using the relationship (Revil and Leroy, 2004)

Csat ¼

Qeff v k mw s

ð2Þ

where k is the permeability and mw is the dynamic water viscosity. The voltage coupling Csat can at moderate to high

406

Hydrogeophysics

permeabilities be obtained using a simple experimental setup, for example, using the type of column experiment presented by Suski et al. (2006). The dependence of Qeff v;sat with water content depends on the geological media considered, but it is to a first order inversely related to water content (Linde et al., 2007; Linde, 2009). The source current that is responsible for observed electrical potential signals associated with these processes is given by the divergence of Equation (1) (e.g., Linde et al., 2007).

2.15.2.4 Controlled-Source Inductive EM Methods Controlled-source inductive EM methods use a transmitter to pass a time- or frequency-varying current through a coil or dipole placed on the Earth’s surface, in boreholes, mounted on an aircraft or towed behind a ship. Governed by Maxwell’s equations and typically operating in the 1–15 kHz range, this alternating current produces a time-varying primary magnetic field, which in turn interacts with the conductive subsurface to induce time-varying eddy currents. These eddy currents give rise to a secondary EM field. Attributes of this secondary magnetic field, such as amplitude, orientation, and/ or phase shift, can be measured by a receiver coil. By comparing these attributes with those of the primary field, information about the presence of subsurface electrical conductors or the subsurface electrical conductivity distribution can be inferred. Because a conductive subsurface environment or target is required to set up the secondary field measured with inductive EM methods, EM methods are best suited for use when attempting to detect the presence of highconductivity subsurface targets, such as saltwater saturated sediments or clay layers. However, because coils do not require contact with the ground, EM methods are often more successful on electrically resistive or paved ground than the classical DC resistivity method, which requires electrode contact. As with ERT and SIP data, EM induction data can often be collected in profile or sounding mode. The mode of acquisition and the resolution and depth penetration of the data are dictated by the electrical conductivity distribution of the subsurface and the coil spacing and source configuration. For frequency domain systems, high transmitter frequencies permit high-resolution investigation of subsurface conductors at shallow depths, while lower transmitter frequencies permit deeper observations but at a loss in resolution. Time domain systems measure the secondary magnetic field as a function of time, and early-time measurements yield information about the near surface, while later-time measurements are increasingly influenced by the electrical properties at larger depths. The depth of penetration and resolution are also governed by coil configuration; the measurements from larger coil separations are influenced by electrical properties at greater depths, while smaller coil spacings sample from the near surface. A review and discussion of the use of controlled-source EM methods for hydrogeological investigations is given by Everett and Meju (2005). It should be noted that it is also possible to use civilian and military radio transmitters, operating in the 10–250 kHz frequency range, as the source signal. These are the signals used in the popular

very low frequency (VLF) (e.g., Pedersen et al., 1994) and radio magnetotelluric (RMT) (e.g., Linde and Pedersen, 2004) techniques.

2.15.2.5 GPR Methods GPR methods use EM energy at frequencies of B10 MHz to 1 GHz to probe the subsurface. At these frequencies, the separation (polarization) of opposite electric charges within a material that has been subjected to an external electric field dominates the electrical response. GPR systems consist of an impulse generator which repeatedly sends a particular voltage and frequency source to a transmitting antenna. When the source antenna is placed on or above the ground surface, waves are radiated downward into the soil. In general, GPR performs better in unsaturated coarse or moderately coarse textured soils; GPR signal strength is strongly attenuated in electrically conductive environments (such as systems dominated by the presence of clays or high ionic strength pore fluids). Together, the electrical properties of the host material and the frequency of the GPR signal primarily control the resolution and the depth of penetration of the signal. Increasing the frequency increases the resolution but decreases the depth of penetration. GPR data sets can be collected in the time or in the frequency domain. Time-domain systems are most commonly used in near-surface investigations. Generally, one chooses a radar center frequency that yields both sufficient penetration and resolution; for field applications this is often between 50 and 250 MHz. However, significant advances have been made in the development of frequency domain systems. Lambot et al. (2004a) describe a stepped-frequency continuous-wave radar deployed using an off-ground horn antenna over the frequency range of 0.8–3.4 GHz. The wide bandwidth and off-ground configuration permits more accurate modeling of the radar signal, thus potentially leading to improved estimates of subsurface parameters (Lambot et al., 2004b, 2006). The most common ground surface GPR acquisition mode is surface common-offset reflection, in which one (stacked) trace is collected from a transmitter–receiver antenna pair pulled along the ground surface. With this acquisition mode, GPR antennas can be pulled along or above the ground surface at walking speed. When the EM waves in the ground reach a contrast in dielectric constants, part of the energy is reflected and part is transmitted deeper into the ground. The reflected energy is displayed as 2D profiles that indicate the travel time and amplitude of the reflected arrivals; such profiles can be displayed in real time during data collection and can be stored digitally for subsequent data processing. An example of the use of GPR profiles for interpreting subsurface stratigraphy is provided in Section 2.15.5.1. The velocity of the GPR signal can be obtained by measuring the travel time of the signal over a known distance between the transmitter and the receiver. The propagation phase velocity (V) and signal attenuation are controlled by the dielectric constant (k) and the electrical conductivity of the subsurface material through which the wave travels. At the high-frequency range used in GPR, the velocity in a low electrical conductivity material can be related to the dielectric

Hydrogeophysics

constant, also known as the dielectric permittivity, as (Davis and Annan, 1989)

407

where c is the propagation velocity of EM waves in free space 8 (3 10 m s1). Approaches that facilitate EM velocity analysis include surface common-midpoint (CMP), crosshole tomography acquisition, as well as analysis of the groundwave arrival recorded using common-offset geometries. Full-waveform inversion approaches have recently been developed (e.g., Ernst et al., 2007; Sassen and Everett, 2009) that offer potential for improved subsurface property characterization over methods based on travel times alone. Discussion of petrophysical relationships that link dielectric permittivity with hydrological properties of interest is described in Section 2.15.3.2. A review of GPR methods applied to hydrogeological applications is given by Annan (2005).

distribution using many first arrival travel times corresponding to refracted energy for many combinations of transmitter and receiver locations. Refraction techniques are most appropriate when there are only a few shallow (o50 m) targets of interest, or where one is interested in identifying gross lateral velocity variations or changes in interface dip. Seismic refraction methods yield much lower resolution than seismic reflection and crosshole methods. However, because refraction methods are inexpensive and acquisition may be more successful in unsaturated and unconsolidated environments, they are often chosen over reflection methods for applications such as determining the depth to the water table and to the top of bedrock, the gross velocity structure, or for locating significant faults. With crosshole seismic tomographic data, the multiple sampling of the inter-wellbore area via raypaths that emanate from instruments lowered down boreholes permits very detailed estimation of the velocity structure that can be used to estimate hydrogeological properties. A review of shallow seismic acquisition and processing techniques is given by Steeples (2005).

2.15.2.6 Seismic Methods

2.15.2.7 Surface Nuclear Magnetic Resonance

Seismic methods common to hydrological investigations use high-frequency (B100–5000 Hz) pulses of acoustic energy to probe the subsurface. These pulses are generally artificially produced (using weight drop, hammers, explosives, piezoelectric transducers, etc.) and propagate outward as a series of wavefronts. The passage of the wavefront creates a motion that can be detected by a sensitive geophone or hydrophone. According to the theory of elasticity upon which seismic wave propagation is based, several different waves are produced by a disturbance; these waves travel with different propagation velocities that are governed by the elastic constants and density of the material. The P-wave energy is transmitted by a back-and-forth particle movement in the direction of the propagating wave. Transverse waves, also called S (secondary or shear)-waves, have lower velocities than the P-wave and thus arrive later in the recording. P-wave arrivals are the easiest to detect and most commonly used arrival; we focus here exclusively on information available from P-waves. The principles of seismic reflection, refraction, and tomographic methods are briefly described below. The surface reflection technique is based on the return of reflected P-waves from boundaries where velocity and density (or seismic impedance) contrasts exist. Processing of seismic reflection data generally produces a wiggle-trace profile that resembles a geologic cross section. However, due to the lack of well-defined velocity contrasts and strong signal interference in shallow unconsolidated and unsaturated materials, seismic reflection approaches to image near subsurface architecture can be challenging. With refraction methods, the incident ray is refracted along the target boundary before returning to the surface. The refracted energy arrival times are displayed as a function of distance from the source, and interpretation of this energy can be accomplished by using simple software or forward modeling techniques. As with GPR methods, the arrival times and distances can be used to obtain velocity information directly. More advanced applications include multi-dimensional inversion for the subsurface velocity

SNMR is a geophysical method that takes advantage of the NMR response of hydrogen protons, which are components of water molecules, to estimate water content. This method involves the use of a transmitting and a receiving loop to induce and record responses to an EM excitation induced at the resonance frequency of protons (the Larmor frequency). Under equilibrium conditions, the protons of water molecule hydrogen atoms have a magnetic moment that is aligned with the Earth’s local magnetic field. Upon excitation, the axis of the precession is modified. When the external field is removed, relaxation occurs as a function of the spatial distribution, amount, and mobility of water; this relaxation manifests itself as an EM signal that decays over time. Through use and analysis of different excitation intensities, initial amplitudes, and decay time, approaches have been proposed to estimate the density distribution of hydrogen atoms as well as associated pore and grain size and water content. Although SNMR holds significant potential for directly investigating subsurface hydrological properties, it is still in an early stage of development and its resolving power is rather limited. As described by Yaramanci et al. (2005), advances are needed to overcome induction effects and inversion errors associated with multi-dimensional heterogeneities and regularization. A further problem with this method is that it is very sensitive to cultural EM noise and that the measured signals are often weak. Hertrich (2008) provides a review of SNMR for groundwater applications, and describes recent algorithm and method development.

kE

 c 2 V

ð3Þ

2.15.2.8 Gravity Measurements of changes in gravitational acceleration can be used to obtain information about subsurface density variations that can in turn be related to variations in lithology or moisture content. The common measuring device for this potential field method is a gravimeter, an instrument which is portable and easy to use. An extremely sensitive spring balance

408

Hydrogeophysics

inside the gravimeter measures differences in the weight of a small internal object from location to location; the weight differences are attributed to changes in the acceleration of gravity due to lateral variations in subsurface density. Measurements can be collected at a regional or local scale depending on the station spacing, which is usually less than half of the depth of interest. The theoretical response to the gravitational field due to factors such as the datum, latitude, terrain, drift, and regional gradient is typically compensated for prior to interpretation of the remaining gravity anomaly. Qualitative interpretation usually consists of constraining a profile or contoured anomaly map with other known geologic information to delineate, for example, the boundary of a sedimentary basin that overlies denser bedrock. A general review of the gravity technique and applications to environmental studies is given by Hinze (1990). More recently, microgravity studies have recently been performed in an attempt to quantify changes in water storage associated with hydrological processes (e.g., Krause et al., 2009) and to characterize cavities in karstic terrains (Styles et al., 2005).

2.15.2.9 Magnetics Magnetic methods obtain information related to the direction, gradient, or intensity of the Earth’s magnetic field. The intensity of the magnetic field at the Earth’s surface is a function of the location of the observation point in the primary earth magnetic field as well as from contributions from local or regional variations of magnetic material such as magnetite, the most common magnetic mineral. After correcting for the effects of the Earth’s natural magnetic field, magnetic data can be presented as total intensity, relative intensity, and vertical or horizontal gradient anomaly profiles or contour maps. Interpretation of magnetic surveys generally involves forward modeling or mapping of the anomalies correlating them with other known geologic information. As magnetic signatures depend to a large extent on magnetic mineral content, which is low in most sediments that comprise aquifers, magnetics is not commonly employed for hydrological investigations, but it can be a very powerful technique to locate lateral boundaries of landfills. Exceptions include mapping subsurface structures (basement topography, faults, and paleochannels), provided that a sufficient magnetic signature or contrast exists. A review of magnetic methods as applied to environmental problems is given by Hinze (1990).

volume of investigation of the borehole measurement is related to the log type, source–detector spacing, the borehole design, and the subsurface material. The well log measurements can be compared with each other and with direct measurements (such as from core samples) to develop sitespecific petrophysical relationships. Log data are also useful to tie hydrological and geological data collected at the wellbore location with geophysical signatures of property variations collected using surface or crosshole geophysical data. References for borehole geophysics applied to hydrogeologic investigations are given by Keys (1989) and Kobr et al. (2005).

2.15.3 Petrophysical Models To be useful in hydrology, geophysical data and hence the corresponding geophysical properties need to be sensitive to hydrological primary (e.g., total and effective porosity, and permeability) or state variables (e.g., salinity, water content, and pressure gradients). In this section, we introduce different petrophysical models that link hydrological and geophysical properties. We focus on models related to electrical properties, since they dominate hydrogeophysical applications through methods such as ERT, SIP, EM, and GPR (see Section 2.15.2). Pertinent models related to gravity, seismics, and borehole geophysical data are not considered here for brevity, but can be found in references such as Mavko et al. (1998), Scho¨n (1996), Gue´guen and Palciauskas (1994), and Carcione et al. (2007). A wealth of models for electrical properties in porous media has been proposed and only main results are summarized below; the reader is referred to Lesmes and Friedman (2005), Keller (1987), and Slater (2007) for more information. The petrophysical models discussed below were chosen because they are fairly general, and also because most of them share a similar parametrization. Purely mathematical models are useful to define bounds on properties, such as the classical Hashin–Shtrikman bounds (Hashin and Shtrikman, 1962). More common in hydrogeophysical studies is the use of semi-empirical models that partly incorporate geometrical and physical properties of the components that comprise the porous media. Examples of such models are Archie’s law (Archie, 1942) or the complex refractive index model (Birchak et al., 1974). In many cases, purely empirical relationships are obtained by fitting polynomial functions (e.g., Topp et al., 1980). Below, we briefly review petrophysical models associated with electrical conductivity, dielectric permittivity, complex conductivity, and electrokinetics.

2.15.2.10 Well Logging

2.15.3.1 Electrical Conductivity

Well logging refers to the process of recording and analyzing measurements collected discretely or continually within wellbores. Borehole measurements are made by lowering a probe into the borehole on the end of an electric cable. The probe, generally 2.5–10.0 cm in diameter and 0.5–10.0 m in length, typically encloses sources, sensors, and the electronics necessary for transmitting and recording signals. A variety of different types of wellbore probes are available; perhaps the most common for hydrological studies include: SP, electrical, EM, gamma–gamma, natural gamma, acoustic, temperature, flowmeter, neutron–neutron, televiewer, and caliper logs. The

The conductive and capacitive properties of an isotropic and homogeneous media can be represented by a complex conductivity (s*), a complex resistivity (r*), or a complex permittivity (e*):

s * ðoÞ ¼

1 ¼ ioe*ðoÞ r*ðoÞ

ð4Þ

pffiffiffiffiffiffiffi where o is the angular frequency and i ¼ 1. It is common practice to refer to the real-valued component of s * ðoÞ ¼ s0 ðoÞ þ is00 ðoÞ at low frequencies (say 0–250 kHz) as s and

Hydrogeophysics

the real–valued relative permittivity at high frequencies (10– 1000 MHz) as k ¼ e0 /e0, where e0 is the effective permittivity of the media and e0 is the permittivity of vacuum. It is important to note that in these frequency ranges both properties, s and k, have a weak frequency dependency (e.g., see Figure 4.1 in Lesmes and Friedman (2005)) that needs to be taken into account for quantitative comparisons. Low-frequency polarization s00 ðoÞ is discussed in Section 2.15.3.3.

409

2.15.3.1.1 Archie’s law

dominates over the surface conductivity term (e.g., Waxman and Smits, 1968; Johnson et al., 1986; Sen et al., 1988). One problem with Waxman and Smits’ model is that it uses an average Qv determined by titration while only the excess charge located along the conducting paths in the pore space will contribute to electrical flow. Another problem arises when surface conductivity becomes more important, since the electrical conduction paths change and can no longer be expressed by F (see Equation (6)) only (Johnson et al., 1986; Revil et al., 1998).

The aggregated empirical Archie’s first and second law (Archie, 1942), expressed here in terms of electrical conductivity, is probably the most commonly used model to interpret electrical conductivity in hydrological studies:

A fundamental length-scale parameter L was introduced by Johnson et al. (1986) as

s ¼ sw Snw fm ¼ sw Snw F 1

ð5Þ

where s is the bulk electrical conductivity of the media, sw is the electrical conductivity of the pore fluid, Sw is the water saturation, n is the water saturation exponent, f is the porosity, and m is the cementation exponent. The electrical formation factor F is defined in the absence of surface conductivity ss as (e.g., Revil et al., 1998)

  1 s ¼ fm  lim F ss -0 sw

ð6Þ

The attraction of Archie’s law in hydrological applications is obvious since it includes key properties, namely the electrical conductivity of the pore fluid related to salinity and the inverse of the electrical formation factor, which can be thought of as an effective interconnected porosity (Revil and Cathles, 1999). Archie’s law not only explains a lot of experimental data, but also is physically justified when surface conduction is negligible (Sen et al., 1981). In the vadose zone, the water saturation exponent may display significant hysteresis (Knight, 1991). Archie’s law is only valid for a continuous water phase, which might break down in dry areas where evaporation is significant (e.g., Shokri et al., 2009). Another more serious problem with this model is that surface conduction, which plays a role when significant clay and silt fractions are present in the media, is ignored.

2.15.3.1.2 Waxman–Smits law A number of models have been proposed to incorporate surface conduction. One of the most commonly used models that includes surface conduction in saturated media is the model of Waxman and Smits (1968):

1 s ¼ ðsw þ BQv Þ F

ð7Þ

where B is the equivalent conductance per ion and Qv is the density of counter ions per unit pore volume. Electrical conduction is here modeled as being composed of an electrical path in the pore volume and another parallel path at the mineral–water interface. This equation has been extensively used in the oil industry and it provides normally a good fit to experimental data when the electrolytic conductivity term

2.15.3.1.3 The Johnson, Koplik, and Schwartz model

Z 2 ¼Z L

jrc0 ðrÞj2 dS ð8Þ jrc0 ðrÞj2 dVp

where rc0 ðrÞ is the electrical potential gradient at position r in the absence of surface conductivity from a current source imposed from the sides and where the integration is performed over the mineral–water interface (S) and the pore volume (Vp), respectively. It follows that 2/L is an effective surface-to-pore-volume ratio weighted by the local strength of the electric field. This weighting eliminates contributions from dead-end pores (Johnson et al., 1986). Johnson et al. (1986) use a perturbation technique to derive the following equation:

s¼

  1 2Ss sw þ þ OðS2s Þ F L

ð9Þ

where the specific surface conductivity is given by (e.g., Schwartz et al., 1989)

Ss ¼

ZN

½sðeÞ  sw de

ð10Þ

0

where e measures the distance along a normal directed into the pore space from the grain boundary. The contributions to Ss become insignificant for values much larger than the Debye screening length that is at most some 100 A˚. The Ss is fairly well-known and is much less variable than L (Leroy and Revil, 2004). Neglecting second-order terms in Equation (9), OðS2s Þ, is only valid in the vicinity of the high-salinity limit. Schwartz et al. (1989) extended the theory of Johnson et al. (1986) to the low-salinity limit in which the electrical flow paths are determined by regions with significant surface conduction. They showed that Pade´ approximants (a ratio of two polynomials) are effective to interpolate between the high- and low-salinity limits. Johnson et al. (1986) also show that L can be used to predict permeability k with a high predictive power using the relation (see also Bernabe´ and Revil, 1995)

kE

L2 4F

ð11Þ

410

Hydrogeophysics

2.15.3.1.4 Self-similar models Another approach to model electrical conductivity is based on self-similar models with electrolytic conduction only (Sen et al., 1981) or with surface conductivity included (Bussian, 1983). Revil et al. (1998) extended the model of Bussian (1983) to explicitly model the different conduction paths taken by anions and cations. Tortuosity affecting the migration of the anions is given by Ff, but the dominant conduction paths for the cations shift toward the conduction paths defined by the distribution of Qv at the mineral–water interfaces as the salinity decreases. The ubiquitous presence of surface conductivity in geological porous media makes models of electrical conductivity alone uncertain tools in hydrological studies, since a moderately high electrical conductivity can be explained by either a fairly high sw with a well-connected pore space (i.e., low F) without any clay particles, or a low sw and a poorly connected pore space (i.e., high F) with a moderate clay fraction. The hydrological behaviors of these two types of media are fundamentally different and electrical conductivity data alone may not offer even qualitative information about the dominant hydrological properties (e.g., Purvance and Andricevic, 2000). To make quantitative predictions, it is therefore often necessary to have access to other types of geophysical (such as IP) or geological data or to perform time-lapse experiments, where temporal variations in the geophysical data are recorded (e.g., Binley et al., 2002).

2.15.3.2.1 Volume averaging Variations of electrical polarization at the frequencies used in ground-penetrating radar (GPR; 10–1000 MHz) are mainly determined by water content and less by mineralogy, even if polarizations of mineral grains need to be considered. Due to the need for complimentary data in many hydrogeophysical applications, it is common to use estimates of both electrical conductivity and the relative permittivity (e.g., Binley et al., 2002; Linde et al., 2006a). When explaining relative permittivity data, it can therefore be useful to use a relative permittivity model that shares a similar parametrization of the pore geometry as the one used to explain electrical conductivity. Such an approach was presented by Pride (1994) who used a volume-averaging approach to derive the following equation for relative permittivity



1 ðkw  ks Þ þ ks F

 ð12Þ

where kw is the relative permittivity of water (kw E 80) and ks is the relative permittivity of the solid (ks ¼ 3–8). This equation was extended by Linde et al. (2006a) to incorporate partial saturations as

k¼

ka ¼

n X

fi kai

ð14Þ

i¼1

where the subscript i indicates the contribution of each phase (e.g., rock matrix, water, and air). Equation (14) with a ¼ 0.5 is referred to as the complex refractive index model (Birchak et al., 1974)

pffiffiffi pffiffiffiffiffiffi pffiffiffiffiffi pffiffiffiffi k ¼ y kw þ ðf  yÞ ka þ ð1  fÞ ks

ð15Þ

where y is water content. Brovelli and Cassiani (2008) showed that this commonly used model is only valid when the cementation exponent m B 2 and when the dielectric contrast between phases are large. This means that Equation (15) is based on an implicit assumption about the connectedness of the pore space that in reality varies (e.g., m is typically B1.5 in unconsolidated aquifer materials; Lesmes and Friedman, 2005). Recently, Brovelli and Cassiani (2010) showed convincingly that an appropriately weighted combination of the lower and upper Hashin–Shtrikman bounds using the cementation factor could predict permittivity measurements very well.

2.15.3.2.2 Topp’s equations

2.15.3.2 Dielectric Permittivity

k¼

One of the most common petrophysical models used to estimate water content from relative permittivity data is the so-called Lichteneker–Rother model (e.g., Gue´guen and Palciauskas, 1994):

1 n S kw þ ð1  Snw Þka þ ðF  1Þks F w

ð13Þ

where ka is the relative permittivity of air (ka ¼ 1) (see Linde et al. (2006a) for a corresponding model for electrical conductivity with surface conductivity included).

A set of models that are purely empirical but have high predictive power in soils are the so-called Topp’s equations that were derived at high frequencies for different soil types. The general Topp equation (Topp et al., 1980) when the soil type is unknown is

k ¼ 3:03 þ 9:3y þ 146y 2  76:7y 3

ð16Þ

2.15.3.3 Complex Conductivity 2.15.3.3.1 Cole–Cole model We now focus on the frequency behavior of the imaginary component of the complex electrical conductivity (Equation (4)) s00 ðoÞ at low frequencies. Recent experiments suggest that the electrochemical polarization of a grain is dominated by the mineral/water interface of the Stern layer and Maxwell– Wagner effects associated with accumulation of electrical charges at pore throats (Leroy et al., 2008). SIP data (also referred to as complex conductivity; Kemna, et al., 2000) have been identified as the most promising method to develop robust inferences of polarization processes (Ghorbani et al., 2007) and potentially permeability in hydrological studies (Slater and Lesmes, 2002; Binley et al., 2005). The most common petrophysical model used in SIP is the phenomenological Cole–Cole model (Cole and Cole, 1941) or combinations of several Cole–Cole models. The Cole–Cole model can be expressed as

  s *ðoÞ ¼ s0 1 þ m

ðiotÞ c 1þðiotÞ c ð1  mÞ

 ð17Þ

Hydrogeophysics

where s0 is the conductivity at the DC limit, t is the mean relaxation time, c is an exponent that typically takes values in the range of 0.1–0.6, and m is the chargeability (m ¼ 1 – s0/ sN, where sN is the electrical conductivity at high frequency). Parameters of this model might be sensitive to specific surface area (Bo¨rner and Scho¨n, 1991; Slater et al., 2006), dominant pore-throat sizes (Scott and Barker, 2003), or effective grain sizes (Slater and Lesmes, 2002). Laboratory measurements on sandstone suggest a strong correlation between the relaxation time and the permeability (r2 ¼ 0.78) (Binley et al., 2005) and promising results have been reported from field applications (Ho¨rdt et al., 2007). It is likely that new physical models based on a more physical parametrization of the pore space that is consistent with the ones developed for other electrical properties (e.g., Leroy et al., 2008; Leroy and Revil, 2009) will help to gain a better understanding of the low-frequency polarization response and its sensitivity to hydrological parameters. In particular, it is important to develop a theory that holds at any frequency and that takes the characteristics of the electrical double layer and the surface chemistry into account.

2.15.4 Parameter Estimation/Integration Methods This section addresses how geophysical data and models can be used together with hydrological data and models to improve the imaging of hydrological properties or monitoring of hydrological processes. The approaches that have been presented in the literature differ mainly in how they represent the model parameter space; what importance and representation is given to a priori information; at what stage different data types are coupled; how uncertainties in the observations, the forward models, and the petrophysical models are treated. Figure 2 provides a schematic view of how geophysical and hydrological data and models can be integrated at different stages in the inversion process. The figure can also be seen as a general representation of how joint inversion can be carried out for the case of two data types. For an in-depth treatment of inversion theory, the reader is referred to Menke (1984), Parker (1994), McLaughlin and Townley (1996), and Tarantola (2005). Study objectives and the available budget will determine many of the choices made throughout the inversion process. These aspects are not incorporated in Figure 2, since it mainly serves to illustrate where interactions between geophysical and hydrological components of the inversion process might take place. In a given hydrogeophysical inversion method only a fraction of the links between the geophysical and hydrological compartments in Figure 2 is likely to be used.

2.15.4.1 Key Components, Constraints, Metrics, and Steps in Parameter Estimation 2.15.4.1.1 Model space and initial model A key choice in any inversion is to decide on the model parametrization used to represent the subsurface, the permissible ranges of model parameters, and the initial model (see box ‘Model space and initial model’). These choices will mainly be based upon prior knowledge (see box ‘Prior knowledge’). Prior knowledge is information about

411

characteristics of the system that we have obtained from other sources of information than the actual geophysical or hydrogeological data that we try to invert. Prior knowledge might, in this case, be related to information about the geological setting and previous exploratory or detailed studies. The link from box ‘Geophysical (or hydrological) system property data’ to box ‘Prior knowledge’ indicates also estimates of system properties that have been made outside the parameter estimation procedure (e.g., sonic log data transformed to P-wave velocities, neutron–neutron data transformed into porosity, and EM flowmeter data translated into relative variations in permeability) and we assume that these properties are known with an associated uncertainty.

2.15.4.1.2 Objective function (systems of equations) The number of independent parameters that can be inferred from hydrological and geophysical data depend on the data type, the experimental design, the number of data available, the data quality, and the forward model used. There exists however an upper limit of how many parameters one can independently estimate from a given data set. For this reason, we must find ways to constrain model space in order to obtain meaningful results, in addition to simply decrease computing time and memory use. In practice, it is necessary to explicitly constrain the parameter space by solving either an overdetermined problem with few model parameters or an underdetermined problem where a unique solution defined as the model that fits the data with the least model structure as defined by the regularization constraints used to stabilize the inverse solution. The zonation approach to model parametrization is to assume that the subsurface can be represented by a number of zones with similar physical properties, where the boundaries are either assumed to be known or updated during the inversion process. Possible applications where a zonation approach could be justified are the delineation of sand from interbedded clay layers or sediments from the underlying bedrock. The advantage of the zonation approach is that the number of model parameters can be kept relatively small and smoothness constraints across boundaries in the inversion may thus be avoided. The geostatistical approach is based on the assumption that the parameter field can be explained by a known or estimated spatial random variable with a certain correlation structure and deterministic trend. This parametrization is probably preferable when the parameters of interest vary in more or less random fashion and there is no clearly defined structure (see further discussion in Mclaughlin and Townley (1996)). Geophysical inversion is typically performed using a very fine model discretization where the aim of the inversion is to fit the data to a certain error level while minimizing deviations from an assumed prior model or spatial variability between neighboring cells. Regardless of the parametrization used, it is clear that prior knowledge should affect the objective function as indicated in Figure 2 (see arrow to box ‘Objective function (systems of equations)’). After defining the model parametrization, it is necessary to define a metric that defines what constitutes a good model and an algorithm that can be used to find such models. There are two main groups of inversion strategies: (1) deterministic

Geophysics

No

Hydrology

Model space and initial model

Prior knowledge

Prior knowledge

Model space and initial model

Inversion step or model proposal

Objective function (Systems of equations)

Objective function (Systems of equations)

Inversion step or model proposal

Forward modeling (Jacobian matrix)

Forward modeling (Jacobian matrix)

Geophysical state data

Hydrological state data

Geophysical system property data

Hydrological system property data

Proposed model acceptable or improved?

Yes

No

Yes Save model

Save model

Data misfit > target or stochastic

Proposed model acceptable or improved?

Petrophysical model

Data misfit > target or stochastic End inversion?

End inversion? Data and model integration Data misfit < target or sufficient number of realizations Geophysical model or model population

Final model or population of models

Data misfit < target or sufficient number of realizations Hydrological model or model population

Figure 2 This flowchart illustrates possible ways that geophysics and hydrology can be integrated in hydrogeophysical studies. Recent hydrogeophysical research indicates that it is important to tightly couple hydrology and geophysics throughout the inversion and modeling process; see possible connections from blue hydrological boxes to green geophysics boxes, and vice versa. Please refer to text for details.

Hydrogeophysics

inversion where one unique model is sought that describes the subsurface in some average sense (Menke, 1984; Parker, 1994) and (2) stochastic inversion where a probabilistic description of the model space is used and where a large population of possible models are sampled without specifying which is the best model, only how likely they are to explain the available data and any prior knowledge (Tarantola, 2005). Regardless of the inversion approach, the definition of the objective function will, to a large degree, determine the type of models that will be obtained for a given data set. Objective functions, at least in deterministic inversions, often include two different terms: (1) one data misfit term that characterizes how well a model explains the observed data and (2) one model misfit term that defines how a model corresponds with prior knowledge or any assumptions about how the model is likely to vary spatially. The most common approach is to quantify these two terms by using a least-squares formulation, where a weighted sum of the two squared misfit terms is penalized simultaneously, which typically works well when system properties are expected to vary smoothly and when data errors have an approximately Gaussian distribution. In this case, the data misfit is expressed as

w2d ¼ ðd  F½mÞT C1 d ðd  F½mÞ

ð18Þ

where d is an N  1 data vector (e.g., electrical resistances or observed drawdown at a pumping well); F[m] is a forward model operator response for a given model vector m of size M  1; superscript T indicates transposition; C1 d is the inverse of the data covariance matrix. It is commonly assumed that data errors are uncorrelated, rendering C1 d a diagonal matrix that contains the inverses of the estimated variances of the data errors; thus, more reliable data carry larger weight when evaluating the data fit. The corresponding model norm is

w2m ¼ ðm  m0 ÞT C1 m ðm  m0 Þ

ð19Þ

where m0 is a reference model of size M  1; C1 m is the inverse of the model covariance matrix, which characterizes the expected variability and correlation of model parameters (Maurer et al., 1998; Linde et al., 2006a). It should be noted that it is often common to neglect the term m0 and replace C1 m with a regularization term that approximates the square of a first or second derivative of the model. The objective function for a classical geophysical deterministic inversion is in the general least-squares case:

Wl ðmÞ ¼ ðm  m0 ÞT C1 m ðm  m0 Þ  1 ðd  F½mÞ T C1 þl d ðd  F½mÞ

ð20Þ

where l1 acts as a trade-off parameter between the smooth well-conditioned problem defined by a heavy penalty on deviations from the predefined model behavior (i.e., l is large) and the ill-conditioned problem defined by the data misfit term (i.e., l is small). In order to obtain models that display sharper contrasts between geological units or when data noise has a nonGaussian distribution, it is possible to use a method called iteratively reweighted least-squares based on the Ekblom

413

lp-norm (Farquharson, 2008), thereby approaching a formulation of the inverse problem where only absolute differences in misfit are penalized, while maintaining the numerical advantages of least-squares formulations. A number of alternative data and model norms have been proposed to obtain models that provide closer representations of the expected model behavior; these methods are all based on iterative reweighting of the model misfit terms (Zhdanov, 2009; Ajo-Franklin et al., 2007; Minsley et al., 2007). In order to evaluate the performance of a proposed model for a given objective function, it is necessary to have access to a forward model (see box ‘Forward modeling (Jacobian matrix)’), which is the model that numerically solves the governing partial differential equation for a given model, boundary conditions, and excitation (e.g., current injection, detonation of explosives, or water injection in a wellbore). The accuracy of the forward model is of key importance in any inversion scheme. In deterministic inversions, it is also important to have access to the Jacobian or sensistivity matrix that defines how sensitive the modeled data are to a given small perturbation of each model parameter. The objective function offers many opportunities to couple different data types to perform joint inversion by simply augmenting d and m with new data and model types, respectively. In order to perform joint inversion, it is necessary to define some sort of constraint such that the different data types and models interact in a meaningful way. These constraints can be of many types, such as structural constraints that penalize dissimilarity between two types of models (see arrows from box ‘Save model’). One possible approach to structural joint inversion is to assume that the gradients in two models should be parallel or anti-parallel, thereby providing models that are structurally similar (e.g., Gallardo and Meju, 2003, 2004; Linde et al., 2006a, 2008). When performing joint inversion, only one objective function is used. To decrease the number of model unknowns when solving the corresponding system of equations, it is also possible to use an iterative sequential approach where two different objective functions are used as indicated in Figure 2. Another approach is to use the final model from one method to define spatial statistics that can be used to constrain the other model (Saunders et al., 2005), or the models can be constrained by using system property data from another method (Dafflon et al., 2009) or key interfaces can be incorporated, such as the depth to bedrock determined by seismic refraction in hydrogeological modeling of hillslope processes. Such information enters the objective function through the box ‘Prior knowledge’. Another approach when an accurate petrophysical relation is known to exist is to couple different model or data types by directly assuming that a given petrophysical model (known or with a given functional form with unknown parameter values) exists (see box ‘Petrophysical models’). In this way, a geophysical model can be defined by a number of hydrological properties and state variables, and the geophysical data can thereby be directly incorporated into the hydrological inversion without the need to construct a geophysical model. In this case, only a geophysical forward model and a petrophysical model is needed to interpret the geophysical data within a hydrological inverse framework. These types of inversion methods are often referred to as fully coupled

414

Hydrogeophysics

hydrogeophysical inversion (Kowalsky et al., 2005; Pollock and Cirpka, 2010). One problem with such an approach is that not only the parameter values used in petrophysical models might change within the study area, but also their functional form if they are too simplified.

2.15.4.1.3 Inversion step or model proposal The next step corresponds to the box ‘Inversion step or model proposal’. For the deterministic case, the system of equations are solved for a given trade-off l of the different data and model misfit terms of the objective function. A large amount of numerical methods are available to solve this problem and a review of the most common methods is outside the scope of this chapter, but a good starting point is Golub and van Loan (1996). In a typical deterministic inversion, the inversion process continues until w2d E w2 , where w2 is a predefined * * target data misfit. A new inversion step using the model obtained in the previous model is carried out if w2d 4 w2 , where * typically also the value of l is decreased. If w2d o w2 , it is cus* tomary to repeat the previous inversion step with a larger l until w2d E w2 (see boxes ‘End inversion?’). In cases where no * convergence is obtained, one needs to change the inversion settings. In stochastic inversions, a proposed model is evaluated based on prior knowledge and the so-called likelihood function, which is closely related to the data misfit term. Bayesian theory offers a consistent and general framework to sequentially condition models to different data types. The posterior distribution of the model parameters m given data d is given by Bayes’ theorem

pðmjdÞ ¼ CpðdjmÞpðmÞ

ð21Þ

where C is a normalizing coefficient, pðdjmÞ is the likelihood function, and pðmÞ is the prior distribution of the model parameters (permissible range and the distribution within the range for each parameter). The likelihood functions provide information about how likely it is that a given model realization is responsible for the observed data. A main attraction of Bayesian sampling methods is that virtually any formulation of the likelihood function and the prior model can be used and it can differ between data and model types if performing joint inversion. The functional form of the petrophysical relationship can also be chosen in a flexible manner. The aim of Bayesian methods is generally to explore pðdjmÞ and this is often done by using Monte Carlo Markov Chain (MCMC) methods (e.g., Hastings, 1970; Mosegaard and Tarantola, 1995; Chen et al., 2006; Vrugt et al., 2009).

2.15.4.1.4 Geophysical model or model population Assessment of the uncertainty in the final inversion images obtained from deterministic inversion is often limited to classical linear uncertainty estimates based on the posterior model covariance matrix and resolution measures based on the resolution matrix. These estimates bear a strong imprint of the regularization used to create a stable solution (Alumbaugh and Newman, 2000). The estimated uncertainty of individual model parameters is therefore often vastly underestimated.

Different approaches have been proposed in the literature to address the variance and resolution properties of deterministic inversion models. One popular approach is simply to perform several inversions where the regularization operators or the initial model vary. This approach provides a qualitative assessment of parameters that are well resolved by the geophysical data (Oldenburg and Li, 1999). Another approach is to perform a most-squares inversion (Jackson, 1976), where the bounds within which a model parameter can vary are sought for a given small increase in data misfit. Kalscheuer and Pedersen (2007) present a nonlinear variance and resolution analysis that investigate for a given variance of a model parameter, the resulting resolution properties of this estimate. The advantage compared with classical resolution analysis (e.g., Alumbaugh and Newman, 2000; Friedel, 2003) is that resolution properties are calculated for the same model variance and that regularization operators do not influence resolution estimates. Nonlinearity is partly handled by introducing nonlinear semi-axis that takes nonlinearity in the model eigenvectors into account. Even if these methods provide a qualitative assessment of model resolution and parameter uncertainty, they provide limited insight with respect to the probability distribution of the underlying model parameters and their multi-dimensional cross-correlations.

2.15.4.2 Example Parameter Estimation Approaches 2.15.4.2.1 Direct mapping approaches The simplest application of geophysical data in quantitative hydrology is direct mapping (Linde et al., 2006b). In its simplest case, all boxes and arrows related to hydrology in Figure 2 are removed and the inversion is performed using a standard geophysical inversion method. It is assumed that a known petrophysical model exists and that it can be used to map the final geophysical model into a hydrological model. Such transformations can be useful, but it is important to understand that geophysical models are only smoothed descriptions of the real property distribution and that the estimates might be biased. Day-Lewis and Lane (2004) and Day-Lewis et al. (2005) have developed a framework to describe how resolution in geophysical images degrade as a function of experimental design and data errors for linear and linearized nonlinear problems. They also show how it is possible to establish apparent petrophysical models from a known intrinsic petrophysical model that take this smoothing into account and thereby transform the geophysical model into a more realistic hydrological model. Direct mapping approaches can be made more effective when defining the model space and initial model, as well as the objective function, using prior knowledge related to the hydrology.

2.15.4.2.2 Integration approaches (geostatistical, Bayesian) A more advanced approach is to combine site-specific hydrological system property data with geophysical models. We refer to this group of models as integration approaches and they are often based on concepts from geostatistics (Linde et al., 2006b). In this case, the geophysical inversion is performed in the same way as for direct mapping, but the petrophysical model and the model integration differ. One example of this

Hydrogeophysics

approach would be to update a model of hydraulic conductivity observed at observation wells with geophysical models that are partly sensitive to hydraulic conductivity (e.g., Chen et al., 2001). Such models can incorporate some of the uncertainty in the geophysical and petrophysical relationships in the resulting hydrological models, but they are bound to use either petrophysical models with parameters determined from laboratory measurements (which are often unsuitable in this context; Moysey et al., 2005) or empirical field-specific relationships (which may be invalid away from calibration points; Linde et al., 2006c). Direct mapping and integration approaches are useful routine tools, but they share several main limitations: (1) laboratory-based or theoretical petrophysical models often cannot be used directly, (2) the estimation of site-specific parameter values of the petrophysical models are not included within the inversion process, (3) there is no informationsharing between different data types during the inversion, (4) resulting uncertainty estimates are qualitative at best, and (5) they often provide physically impossible models (e.g., mass is not conserved when performing tracer tests, e.g., Singha and Gorelick, 2005).

2.15.4.2.3 Joint inversion or fully coupled hydrogeophysical inversion The hydrogeophysical research community has in the recent years developed approaches that do not suffer from some of the limitations of direct mapping and integration methods by using both hydrological and geophysical state data during the inversion process and by coupling the hydrological and geophysical models during the inversion. We refer to such approaches as joint inversion (Linde et al., 2006b) or, alternatively, as fully coupled hydrogeophysical inversion. These approaches often include one or more of the following: (1) hydrological flow and transport modeling form together with geophysical forward modeling an integral part of the parameter estimation process; (2) petrophysical relationships are inferred during the inversion process; (3) nonuniqueness is explicitly recognized and a number of equally possible models are evaluated. This type of approach has at least four main advantages: (1) mass conservation can be assured in time-lapse studies and physically impossible flow fields are avoided when incorporating flow and transport simulations within the inversion framework; (2) data sharing during the inversion makes it often possible to obtain more realistic models with a higher resolution; (3) unknown parameters of petrophysical models can be estimated during the inversion process; (4) and physically implausible model conceptualization might make it impossible to fit the data to a realistic error level. This last point is important, since it makes joint inversion well suited to distinguish not only between possible realistic models and inconsistent parameter distributions, but also between competing conceptual models. Joint inversion comes at a price since it is necessary to develop new inversion codes that are suitable to the available data and model objectives; recent hydrogeophysical joint or fully coupled inversion methodologies include Kowalsky et al. (2005, 2006); Chen et al. (2006, 2010), Linde et al. (2006a, 2008),Lambot et al. (2009), Pollock and Cirpka (2009), and Huisman et al. (2010). Such developments

415

can in practice be greatly facilitated by incorporating freely available forward codes or by using commercial multiphysics modeling packages. Despite this, the amount of work involved is typically more significant compared with direct mapping and data integration approaches.

2.15.5 Case Studies Several case studies are presented to illustrate the use of geophysical methods for delineating subsurface architecture (Section 2.15.5.1), delineating anomalous subsurface fluid bodies (Section 2.15.5.2), monitoring hydrological processes (Section 2.15.5.3), and estimating hydrological properties (Section 2.15.5.4). The examples are based primarily on published hydrogeophysical studies that were conducted to gain insights about field-scale system behavior, improve flow and transport predictions, or to provide input to water resources or contaminant remediation management decisions. Examples were chosen to illustrate the utility for a variety of different characterization objectives, geophysical methods, and hydrogeophysical estimation approaches. Each example provides a brief background of the study as well as references for readers interested in more information.

2.15.5.1 Subsurface Architecture Delineation Because geophysical properties are often sensitive to contrasts in physical and geochemical properties, geophysical methods can be useful for mapping subsurface architecture, defined here as a distribution of hydrogeologically distinct units. Using geophysical methods for subsurface mapping is perhaps the most well-developed application in hydrogeophysics, and it is often commonly performed using surface-based geophysical techniques. Examples of common mapping objectives in hydrogeological applications include the mapping of stratigraphy or the depth to bedrock or the water table. The ability to distinguish hydrogeologically meaningful boundaries using geophysical data depends on their sensitivity to subsurface physical properties, contrasts in these properties, and the resolution of the geophysical method at the characterization target depth. In this section, we describe the use of geophysical data for mapping subsurface architecture or features by presenting several case studies that differ in their choice of geophysical method, the scales involved, characterization objective, and interpretation or integration approach. These examples include the use of airborne EM data to map lithofacies relevant for water resources at an island; high-resolution GPR imaging of braided river deposits; seismic methods to estimate subsurface architecture in contaminated environments; and fracture characterization using azimuthal SP measurements.

2.15.5.1.1 3D resistivity mapping of a Galapagos volcano aquifer A fundamental limitation of most geophysical methods used in hydrogeophysics is that they have a limited ability to cover areas larger than B1 km2 within a reasonable time and budget at a high resolution. If funding permits, airborne geophysics can be very useful for gaining an overall view of the geological

416

Hydrogeophysics

structure at the watershed scale. Most airborne data that are collected by geological surveys around the world provide only a limited depth resolution (e.g., Pedersen et al., 1994), whereas systems developed by the mining industry are designed for deeper targets and for more pronounced anomalies such that data quality demands are lower than in hydrogeological applications. An exception is the helicopter-borne SkyTEM system that was purposefully developed for mapping of geological structures in the near surface for groundwater and environmental applications (Sørensen and Auken, 2004). This is a transient electromagnetic (TEM) system that uses a strong current flowing in the transmitter coil to induce weak secondary subsurface currents whose resulting magnetic fields are subsequently measured with a receiver coil. SkyTEM measurements provide similar data quality and resolution as ground-based TEM systems, but with the advantage that measurements are carried out at speeds exceeding 15 km h1 corresponding to a typical station spacing of 35–45 m. This system operates normally at altitudes of 15–20 m with the helicopter located at an altitude of 50 m. The system is a standalone system that can be attached to the cargo hook of any helicopter. It uses a four-turn 12.5 12.5 m2 transmitter loop with a low moment using one turn only and a high moment using all four turns. The receiver coil (0.5  0.5 m2) is located 1.5 m above a corner of the quadratic and rigidly fixed transmitter loop. D’Ozouville et al. (2008) illustrate the tremendous amount of information that airborne EM can provide in hydrological

studies in remote areas where only limited prior geological and geophysical work have previously been carried out. This study was motivated by the need to better understand the groundwater resources on the volcanic island of Santa Cruz in the Galapagos island, which is experiencing challenges in meeting the water demands of the island’s population and its many visitors. In order to provide an overall view of potential groundwater resources at the island, a SkyTEM survey of 900 km covering 190 km2 was carried out to obtain a detailed view of the island’s internal 3D electrical resistivity structure. Figure 3 shows the resulting electrical resistivity models from two profiles that cross the island. 3D inversion of TEM data is computationally infeasible for large data sets, and the inversions were performed using 1 D forward modeling. The strong lateral continuity of these models is the result of lateral model constraints that are imposed during the inversion (Viezzoli et al., 2008). Four hydrogeological units were interpreted and they are indicated as I–IV in Figure 3. Unit I represents unsaturated fractured basalts with resistivities above 800 O m; unit II is the other resistivity end-member with resistivities smaller than 10 O m representing fractured basalt invaded by seawater. Unit III is a near-surface unit and unit IV a buried unit with resistivity values that range between 50 and 200 O m. These units might correspond to weathered zones or fractured basalts saturated with freshwater. The top of unit II images the saltwater wedge to distances approximately 9 km inland and its slope is in perfect agreement with predictions based on the hydraulic gradient observed in one borehole and the density contrast between Profile A

Elevation (m)

N

Profile A

600

Profile B IV

400 South

200

III

I I

I

0

II

II

−200

III

400

North

Profile B

IV

West

200

East I

I 0 −200

II 0

5000

II 10 000

15 000

20 000

Profile distance (m)

25 000 10

30 000

100 1000 Resistivity (Ω m)

Figure 3 Two inversion models of cross sections of Santa Cruz Island, Galapagos Islands. The positions of the south–north and west–east profiles across the island are shown in the inset over a background of near-surface average resistivity showing extent of mapped area. The four units of hydrological interest are: (I) high-resistivity unsaturated basalts; (II) seawater intrusion wedge underlying the brackish basal aquifer; (III) near-surface, low-resistivity units consisting of colluvial deposits; and (IV) internal, low-resistivity unit of saturated basalts overlying an impermeable substratum. From figure 3 in D’Ozouville N, Auken E, Sorensen K, et al. (2008) Extensive perched aquifer and structural implications revealed by 3D resistivity mapping in a Galapagos volcano. Earth and Planetary Science Letters 269: 518–522.

Hydrogeophysics

salt- and freshwater. Of significant hydrological interest is unit IV that displays electrical resistivities similar to those of freshwater-saturated basalts on other islands. It forms an internal low-resistivity zone that is present only in the upper section of the southern side of the volcano. This wedge-shaped unit covers 50 km2 and it has a thickness that varies between 10 and 80 m. It is quasi-parallel to the topography and coincides with the area of maximum precipitation. D’Ozouville et al. (2008) interpret unit IV as being composed of a similar basalt as unit I, but underlain by an impermeable layer that prohibits further downward percolation.

2.15.5.1.2 High-resolution GPR imaging of alluvial deposits When surface conductivity is insignificant and pore water salinity is reasonably low, one of the primary tools in nearsurface (up to tens of meters or so) hydrogeophysical studies is ground-penetrating radar (Davis and Annan, 1989). This method can be used to image interfaces of the 3 D water content distribution and can therefore be very useful to gain information about variations in water saturation in the vadose zone (Irving et al., 2009) and porosity in the saturated zone (Beres and Haeni, 1991). GPR can also be used to image fractures (Grasmueck, 1996) or investigate the depositional setting (van Overmeeren, 1998; Beres et al., 1999). The widespread use of GPR is mainly due to its superior vertical resolution (in the order of 0.1–1.0 m depending on the antenna frequency and the velocity of the subsurface) and the very fast data acquisition, which makes it possible to routinely obtain high-quality data at close to walking speed. Gravelly, braided river deposits form many aquifers and hydrocarbon reservoirs. These deposits typically display a hierarchical architecture where permeability varies over a multitude of scales (e.g., Ritzi et al., 2004), since permeability is linked to the sediment texture, geometry, and spatial distribution of sedimentary stata. Detailed characterization of such systems is difficult, but at least their statistical properties need to be known prior to attempting reservoir or aquifer management. In order to improve the understanding of such systems, Lunt et al. (2004) developed a 3 D depositional model of the gravelly braided Sagavanirktok River in northern Alaska. The data used to construct this model were obtained from cores, wireline logs, trenches, and some 90 km of GPR profiles using different antenna frequencies (110, 225, 450, and 900 MHz) with corresponding depths of penetration varying from 7 to 1.5 m. The 17 boreholes only provided limited sampling, and the 1.3 km of destructive trenches provided information down to the water table only. The GPR data provide continuous coverage over the whole thickness of the deposits and provide information about the depositional setting both across stream and along stream over the whole channel-belt width of 2.4 km. The GPR data made it possible to locate channel fill, unit bars, side bar deposits, confluence scour, compound braid bar deposits, and other depositional features that are of importance to understand the depositional setting. Figure 4 displays a comparison between a sedimentary log and a collocated GPR profile. Not only reflections corresponding to large-scale compound bar

417

boundaries, but also certain unit bar boundaries are clearly imaged.

2.15.5.1.3 Subsurface flow architecture delineation using seismic methods The use of both GPR and seismic data sets typically entails the processing of the geophysical measurements into estimates of geophysical properties, such as reflectivity or velocity, followed by a comparison of the attributes with direct measurements often available from wellbores (e.g., lithological boundaries). Figure 4 illustrated a comparison between GPR reflectivity and wellbore lithological information. Although this two-step method often provides useful information and takes advantage of expert knowledge, the qualitative approach can limit our ability to quantify errors associated with the interpretation and it can lead to dramatically different interpretations of subsurface heterogeneity depending on the interpreter and the processing steps employed. To circumvent these limitations, Chen et al. (2010) developed a joint inversion method that simultaneously considers surface seismic refraction travel times and wellbore data sets for delineating watershed architecture that may exert an influence on contaminant plume mobility at the Oak Ridge National Laboratory site in Tennessee. The groundwater at this site includes uranium, nitrate, and other contaminants that emanated from a seepage basin (S-3 ponds, Figure 5). Underlying the seepage basin is weathered and fractured saprolite that overlies bedrock. Flow is expected to preferentially occur through the more intensely fractured and weathered zones. It is impossible to image individual fractures on a 100-m scale. However, because the fractures occur in discrete zones at this site and because the P-wave velocity in weathered and fractured zones should be lower than the surrounding more competent rock (e.g., Mair and Green, 1981, Chen et al., 2006, Juhlin and Stephens, 2006), seismic methods hold potential for aiding in the delineation of preferential flow zones. A Bayesian joint inversion approach was developed and tested at two locations within the watershed to delineate architecture that may be important for controlling plume scale transport. Within the developed framework, the seismic firstarrival times and wellbore information about key interfaces were considered as input. A staggered-grid finite-difference method was used to forward model the full seismic waveform in 2 D with subsequent automated travel-time picking. Seismic slowness and indicator variables of key interfaces are considered as unknown variables in the framework. By conditioning to the seismic travel times and wellbore information, Chen et al. (2010) estimated the probability of encountering key interfaces (i.e., between fill, saprolite, weathered low velocity zone, and consolidated bedrock) as a function of location and depth within the watershed. An example of the results obtained from two surface seismic data sets collected along the watershed reveals the presence of a distinct low velocity zone that is coincident with the trend of the plume axis (Figure 5). This example illustrates how the joint inversion approach can explicitly incorporate wellbore data into the inversion of the seismic travel time data in the estimation of aquifer architecture. Although not shown,

4 3 2 1 0−1−2−3−4−5−6 Grain size (phi)

271.0±24.1 ka

289.0±12.8 ka

Base of channel belt

Unit bar

Unit bar Compound bar Unit bar

Unit bar in channel fill

Unit bar

Compound Unit bar bar

Unit bar in channel fill

Unit bar

Compound bar Unit bar

Unit bar

4 3 2 1 0 -1-2-3-4 -5-6 Grain size (phi)

Medium-scale cross sets

Small-scale sets planar strata

Drifted plant fragment

Legend Open framework gravel Sandy gravel Gravelly sand Sand

Diffraction hyperbola

Core location 10 m East

100

50

200

150

Figure 4 Comparison between a sedimentary log and a GPR profile, see arrow that indicates the core location on the GPR profile. GPR profile has a vertical exaggeration of 5:1. Sedimentary core log shows three compound bar deposits, each comprising two or three unit-bar deposits. The additional information offered by the continuous GPR profiles is evident. Modified from Figure 12 in Lunt IA, Bridge JS, and Tye RS (2004) A quantitative, three-dimensional depositional model of gravelly braided rivers. Sedimentology 51: 377–414.

8

7

6

5

4

3

2

1

0

600 200 core

West

Time (ns)

Hydrogeophysics

419

S-3 NT-2

0

300

Source region

Plume axis ne Low velocityy zo

600

Feet 1800

1200

Ground surface 305

300 Elevation (m)

Ground surface Elevation (m)

295 FW600

GW828

FW602

290 285

FW009

TPB25 FW117 FW112

300 295

FW008

FW007

FW010

290 285

280

100 60 20 80 40 120 Distance along the horizontal direction from the South to the North (m)

0

275 100 110 60 70 30 20 80 90 40 50 10 Distance along the horizontal direction from the South to the North (m)

800

1400

2000

2600

3200

Seismic velocity (m s−1) 600

1400

2000

2700

3400

Seismic velocity (m s−1)

Figure 5 Bottom: Examples of seismic velocity models and subsurface architecture obtained through joint stochastic inversion of wellbore and surface-based seismic refraction data sets, which reveal distinct low velocity zones that are laterally persistent along the plume axis. Top: Superposition of low velocity zone region (shown in purple) on top of plume distribution (shown in pink), suggesting the control of the low velocity zone on plume migration. Modified from Chen J, Hubbard S, Gaines D, Korneev V, Baker G, and Watson D (2010) Stochastic estimation of aquifer geometry using seismic refraction data with borehole depth constraints. Water Resources Research (in press).

the approach also provides estimates of uncertainty about the location of the interfaces.

2.15.5.1.4 Fracture zonation characterization using azimuthal electrical methods A significant body of literature has developed on using azimuthal electrical resistivity soundings to determine anisotropic electrical properties in fractured media (e.g., Taylor and Fleming, 1988; Lane et al., 1995). Electrical anisotropy in such systems is due to preferential fracture orientations, variable aperture distributions with azimuth, or clay-filled fractures. Field data suggest that directions of electrical anisotropy can, under certain conditions, be linked to anisotropy in hydraulic transmissivity (e.g., Taylor and Fleming, 1988), and it could therefore serve as an important data source in hydrogeological applications in fractured rock systems. Watson and Barker (1999) show that many of the electrode configurations that have been employed in past azimuthal resistivity surveys cannot discriminate between electrical anisotropy and heterogeneity. This problem can be avoided by using certain specialized electrode configurations. Unfortunately, data collection is very slow and no inversion for anisotropic parameters using such surveys has been performed to date. Linde and Pedersen (2004) demonstrate how these problems can be avoided by employing a frequency-domain EM method,

namely RMT. Despite these methodological developments to estimate azimuthal electrical anisotropy, it is not guaranteed that electrical anisotropy coincides with preferred hydrological flow directions and any such relationship is likely to be site specific. Wishart et al. (2006, 2008) introduced the azimuthal selfpotential gradient (ASPG) method, which provides data that may be sensitive to dominant hydrological flow directions in fractured media. In ASPG, one electrode is successively moved in steps on the order of 101 on the perimeter of an inner circle while the reference electrode moves with steps of the same size on the perimeter of an outer circle with the same midpoint as the inner circle. If the underground is predominantly anisotropic, the data will display a 1801 symmetry except for data errors, while a 3601 symmetry appears for measurements where lateral heterogeneity dominates. Wishart et al. (2008) applied this technique to four different fractured rock field sites in the New Jersey Highlands and found that three sites showed ASPG responses that compared well with observed fracture patterns at the sites. Figure 6 shows an example from one of the sites where the ASPG data display a significant 1801 symmetry indicating that large-scale fracture anisotropy is responsible for the observed ASPG data. It is also seen that the direction of the anomalies corresponds well with the mapped fracture directions at outcrops within 100 m of the measurement locations. The data from an azimuthal resistivity survey

420

Hydrogeophysics

ARP 320

340

0

20 40

260

mV 80

40

0

220 180

60

300

80

280

100

260 16000 8000

140 200

20 40

80

120

240

0

320 60

300 280

340

ARS

160

Ohm m 0

100 120

240 220

140 200

180

160

Figure 6 Azimuthal self-potential (SP) and resistivity data superimposed on rose diagrams of fracture strike sets mapped at the Wawayanda State Park, New Jersey. From Figure 4 in Wishart DN, Slater LD, and Gates A (2008) Fracture anisotropy characterization in crystalline bedrock using fieldscale azimuthal self potential gradient. Journal of Hydrology 358: 35–45.

using an asymmetric arrow type array (Bolshakov et al., 1997) do not seem to correspond to either of the dominant fracture directions; the data are strongly nonsymmetric, indicating that lateral variations in the electrical conductivity structure dominate. It is reasonable to assume that the positive lobes of the ASPG data indicate groundwater flow directions, but no detailed field evidence is available to confirm this even if regional flow considerations point in this direction. The magnitudes of the SP signals presented by Wishart et al. (2008) are rather large (e.g., up to 300 mV over distances of 36 m) and the variations of ASPG signals with offset are sharp. The large magnitudes can partly be explained by the shallow water table (B0.5 m) and only a few meters of till overlying the highly resistive fractured rock mass. Applications in locations with thicker and more conductive overburden and with a deeper location of the source current will likely result in much smaller magnitude and a less clear-cut interpretation even for identical flow and fracture conditions. The usefulness of this technique in other field-settings needs to be assessed, but it appears that the ASPG technique can be a very rapid method to nonintrusively map preferential flow paths in fractured rock at shallow depths where overburden thickness is thin.

2.15.5.2 Delineation of Anomalous Fluid Bodies Geophysical methods, particularly those collected from the ground surface or from aircrafts (e.g., Paine, 2003), have been successfully used to identify anomalous subsurface fluid bodies, such as contaminant plume boundaries and regions impacted by saltwater intrusion. Here, we illustrate the use of surface electrical approaches for delineating high ionic strength plumes and for characterizing redox gradients associated with contaminant plumes.

fluid, electrical methods are commonly used to delineate subsurface plumes having high ionic strength (e.g., Watson et al., 2005; Adepelumi et al., 2005; Titov et al., 2005). As shown in Equation (5), electrical resistivity responds to porosity and surface conduction (often linked to lithology) as well as to saturation and pore fluid ionic strength. As described by Atekwana et al. (2004), activity of the natural microbial population can also impact the electrical resistivity through facilitating processes such as mineral etching, which appear to be more prevalent at the fringes of organic plumes. If the contrast between the concentration of the groundwater and the plume is great enough so that other contributions are considered to be negligible, electrical methods can be used, at least in the absence of significant clay units, to indicate contrasts in pore water electrical conductivity, or to delineate approximate boundaries of high-ionic-strength plumes. An example of the use of inverted surface electrical resistivity data to delineate a deep (B50 m) nitrate plume at the contaminated Department of Energy (DOE) Hanford Reservation in Washington is given by Rucker and Fink (2007). They collected six ERT transects (each at least 200 m long), inverted the data to estimate the electrical resistivity distribution in the contaminated region, and compared their results with wellbore borehole measurements of pore water electrical conductivity and contaminant concentrations. They found a strong, negative correlation between electrical resistivity and nitrate concentration above a threshold value, which was used with the electrical models to delineate the plume. Figure 7 shows several of the inverted transects as well as the correlation between electrical resistivity and nitrate concentration obtained from co-located electrical and wellbore measurements.

2.15.5.2.1 Electrical resistivity to delineate high-ionicstrength plume boundaries

2.15.5.2.2 SP imaging of redox potentials associated with contaminated plumes

Because of the strong positive correlation between total dissolved solids (TDSs) and the electrical conductivity of the pore

The traditional application of the SP method has been in mineral exploration where large negative SP anomalies are

Hydrogeophysics

421

220

Inverted resistivity (Ω m)

200 180 160 Constant NO3 value

140 120 100

Variable NO3 value

80 60 40

(b)

102

103

104 Nitrate (mg

105

106

l−1)

(a)

Figure 7 (a) Inverted electrical resistivity profiles at the BC crib area of the contaminated Hanford, Washington Reservation, where the low electrical resistivity (high electrical conductivity) regions were interpreted as the plume. (b) Observed relationship between electrical conductivity and nitrate concentration at a single wellbore location. Modified from Rucker DF and Fink JB (2007) Inorganic plume delineation using surface high resolution electrical resistivity at the BC Cribs and Trenches Site, Hanford. Vadose Zone Journal 6: 946–958.

typically associated with mineral veins (Fox, 1830). As an extreme example, Goldie (2002) presents a peak anomaly of  10.2 V associated with highly resistive high-sulfidation oxide gold deposits. The main contribution of such anomalies is thought to be related to electrochemical half-reactions (Sato and Mooney, 1960; Bigalke and Grabner, 1997), even if it has been suggested that some field data contradict this model (Corry, 1985). Naudet et al. (2003, 2004) observed large negative SP anomalies over the Entressen domestic landfill outside Marseille, France. Redox potential, or Eh, indicates the tendency for oxidation–reduction reactions to occur. Strong redox gradients often become established adjacent to contaminant plumes. They found that the residual SP data (Figure 8(a)), where the effects of streaming currents had been filtered out, were strongly correlated with the difference in redox potential between groundwater samples in the contaminated landfill and uncontaminated areas. They invoked an explanation in analogy with the models of Sato and Mooney (1960) and Bigalke and Grabner (1997). To remotely map variations in redox potential, Linde and Revil (2007) developed a linear inversion model where the difference in redox potential is retrieved from the residual SP data assuming a known 1 D electrical resistivity model and a known depth at which electrochemical reactions take place. They created a simplified representation of the electrical conductivity structure of the Entressen landfill based on ERT models and they assumed that source currents are located at the water table. Figure 8(b) shows a comparison between the

simulated and observed residual SP data of Naudet et al. (2003, 2004). Figure 8(c) displays the retrieved redox potentials assuming a known background value outside of the contaminated area. By comparing these estimates and measured redox potentials in the wells (Figure 8(d)), they found that the inversion results can retrieve the measured redox potentials quite well given the simplifying assumptions involved. It should be noted that equally good data fits between the simulated and observed SP data could have been achieved by shifting the depth at which the sources are imposed or by assuming that the vertical dipole sources are distributed over a volume and not over an area (Blakely, 1996). The interpretation of SP data must therefore be treated with caution, and significant prior constraints must be imposed. Nevertheless, it appears that SP mapping and monitoring may provide a cheap and reliable method for monitoring field scale distribution of redox potential at contaminated sites. It is necessary that this approach is tested at other research sites before its applicability can be properly assessed.

2.15.5.3 Hydrological Process Monitoring A particularly powerful component in hydrogeophysics is the use of a suite of geophysical data sets, collected at the same locations as a function of time, to monitor hydrological processes. Such repeated studies are often referred to as time-lapse geophysics, and their advantages compared to static images are

Landfill

0

26

17

5

1846000

−200

Ref. 21 Regional groundwater flow direction

1844000 73

804000

806000

808000

−400 −600

East (m)

(a)

−200

North (m)

Landfill

0

1011 2 26 17 5

−200

Ref. 21 Regional groundwater flow direction

1844000

(c)

−400 −600

73

804000

806000

−200 −400 −600

808000

East (m)

0

Observed residual SP (mV)

(b)

1848000

1846000

0

−600 −400 −200

Inverted redox potential (mV)

North (m)

1011 2

Simulated residual SP (mV)

1848000

Residual self-potential (mV)

Hydrogeophysics

Redox potential (mV)

422

100

21 73

0 −100

17

10

−200

26

11 2

5

−300 −300 −200 −100

(d)

0

100

Measured redox potential (mV)

Figure 8 (a) Residual SP map at the Entressen landfill, in which the black lines indicate the SP profiles (2417 SP measurements). (b) Comparison of simulated SP with the residual SP estimated from the measured SP data. The response of the inverted model fits the residual SP to the estimated standard deviation. (c) Inverted redox potential in the aquifer at Entressen. (d) Comparison of inverted redox potentials in the aquifer with in situ measurements from Entressen (the correlation coefficient is 0.94). Modified from Linde N and Revil A (2007) Inverting self-potential data for redox potentials of contaminant plumes. Geophysical Research Letters 34: L14302 (doi: 10.1029/2007GL030084).

significant for process monitoring. First of all, changes in welldesigned time-lapse inversions are most often primarily related to changes in state variables only (e.g., temperature, pressure, partial saturations of different phases, and the electrical conductivity of the pore fluid) and not to characteristics of the rock matrix itself. Time-lapse imaging has also the advantage that errors in the forward model tend to cancel (e.g., LaBrecque and Yang, 2001; Lien and Mannseth, 2008). It should be noted that subsurface engineered manipulations, such as those associated with environmental remediation, aquifer storage and recovery, and carbon sequestration can indeed alter the physical properties of the material. For example, remediation treatments can induce biogeochemical transformations that in turn alter the pore geometry and ultimately the fluid flow characteristics (Englert et al., 2009; Li et al., 2009). The geophysical responses to such processes are currently under intense study in the research area of biogeophysics (Atekwana et al., 2006; Williams et al., 2005; Chen et al., 2009; Slater et al., 2009), but are not covered in this chapter. In what follows, we present several examples that illustrate the use of time-lapse geophysics, including: GPR to monitor the spatiotemporal distribution of soil water content in agricultural and hillslope settings; the use of GPR to monitor the distribution of saline tracers in fractured rock; and the use of EM methods to monitor seasonal changes in freshwater– seawater dynamics.

2.15.5.3.1 Soil moisture monitoring The vadose zone mediates many of the processes in the hydrological cycle, such as the partitioning of precipitation into infiltration and runoff, groundwater recharge, contaminant transport, plant growth, evaporation, and sensible and latent energy exchanges between the Earth’s surface and its atmosphere. As an example, in catchment hydrology, the readiness of an area to generate surface runoff during storm rainfall is related to its surface storage capacity. Given the predominant effects of soil moisture on the production of crops, soil salinization, carbon cycling, and climate feedback, development of methods for monitoring moisture content over field-relevant scales is desirable (e.g., Vereecken et al., 2008). Equations (5) and (13) indicate that both the dielectric constant and electrical conductivity are sensitive to water content. Because of this sensitivity, geophysical methods that are sensitive to these properties (e.g., GPR and ERT) have been used fairly extensively to monitor the spatiotemporal distribution of soil moisture. As described by Huisman et al. (2003) and Lambot et al. (2008), GPR is commonly used in hydrogeophysical studies to estimate water content. Various GPR waveform components and configurations have been used to estimate water content, including: crosshole radar velocity (Hubbard et al., 1997; Binley et al., 2002), surface ground wave velocity (Grote et al., 2003), subsurface reflection (Greaves et al., 1996; Lunt et al., 2005), and air-launched ground-surface reflection approaches

Hydrogeophysics

(Lambot et al., 2006). An example of the use of time-lapse surface reflection GPR coupled with a Bayesian method to estimate seasonal changes in water content in the root zone of an agricultural site is given by Hubbard et al. (2006). Within a 90 m  220 m section of this agricultural site, a thin (B0.1 m), low-permeability clay layer was identified from borehole samples and logs at a depth of 0.8–1.3 m below ground surface. GPR data were collected several times during the growing season using 100 MHz surface antennas; these data revealed that the thin clay layer was associated with a subsurface channel. Following equations (3) and (15), as the bulk water content in the unit above the GPR reflector increased, the dielectric constant increased, which lowered the velocity and lengthened the two-way travel time to the reflector. As a result, the GPR reflections revealed seasonal changes in the travel time to the clay layer as a function of average root zone moisture content. At the wellbore locations, a site-specific relationship between the dielectric constant and volumetric water content was used with the radar travel times to the clay reflector to estimate the depth-averaged volumetric water content of the soils above the reflector. Compared to average water content measurements from calibrated neutron probe logs collected over the same depth interval, the estimates obtained from GPR reflections at the borehole locations had an average error of 1.8% (Lunt et al., 2005). To assess seasonal variations in the root zone water content between the wellbores, the travel time picks associated with all GPR data sets, the wellbore information about the depths to the clay layer, and the site-specific petrophysical relationship were used within a Bayesian procedure (Hubbard et al., 2006). Figure 9 illustrates the estimated volumetric water content for the zone located above the reflecting clay layer at different times during the year. The figure indicates seasonal variations in mean water content and also that the channel-shaped feature influences

220 m

October 2002

November 2002

423

water content distribution: within this area the soils are consistently wetter than the surrounding soils. The soil moisture variations appeared to play a significant role in the crop performance: crops located within the channel region had consistently higher crop weight relative to the surrounding regions. These results suggest that the two-way GPR reflection travel times can be used to obtain estimates of average soil layer water content when GPR reflectors are present and when sufficient borehole control is available. Several studies have also explored the use of surface ERT data sets for characterizing moisture infiltration and redistribution at the hillslope and watershed scales. For example, Berthold et al. (2004) compared electrical conductivity estimates from surface ERT images with groundwater electrical conductivity measurements to evaluate the roles of wetlands and ponds on depression-focused groundwater recharge within a Canadian wildlife region. The surface electrical data revealed a complex pattern of salt distribution that would have been difficult to understand given point measurements alone. Koch et al. (2009) collected surface electrical profiles over time along 18 transects within a German hillslope environment, and used the images together with conventional measurements to interpret flow pathways and source areas of runoff.

2.15.5.3.2 Saline tracer monitoring in fractured rock using time-lapse GPR methods Hydrogeophysical applications in fractured media are challenging because of the large and discrete variations between the physical properties of the intact rock mass and the fractures (NRC, 1996). Time-lapse imaging of geophysically detectable tracers has been used in recent years to improve the understanding of fracture distribution and connectivity. The best adapted geophysical technique to image individual

April 2003

tu ea lf ne an

0.2

Ch

Ch

Ch

an

an

ne

ne

lf

lf

ea

ea

tu

tu

re

re

re

0.3

0

0.1

Volumetric water content (m3 m−3)

0.4

0

90 m

Figure 9 Plan-view map of average volumetric water content of the top soil layer (o1.5 m below ground surface) at the agricultural study site, estimated using 100 MHz GPR reflection travel-time data and borehole neutron probe data within a Bayesian estimation approach. Color key at right indicates relative volumetric water content, from red (drier) to blue (wetter). Modified from Hubbard S, Lunt I, Grote K, and Rubin Y (2006) Vineyard soil water content: mapping small scale variability using ground penetrating radar. In: Macqueen RW and Meinert LD (eds.) Fine Wine and Terroir – The Geoscience Perspective. Geoscience Canada Reprint Series Number 9 (ISBN 1-897095-21-X; ISSN 0821-381X). St. John’s, NL: Geological Association of Canada.

424

Hydrogeophysics

fractures away from boreholes, up to few tens of meters away from the boreholes, is probably single-hole radar reflection measurements. This method has been shown to be a useful tool in site characterization efforts to determine possible orientations and lengths of fractures in nuclear waste repository laboratories (Olsson et al., 1992) and in characterizing unstable rock masses (Spillmann et al., 2007). By stimulating individual fractures by adding a saline tracer, it is possible to image tracer movement from the surface (Tsoflias et al., 2001; Talley et al., 2005; Tsoflias and Becker, 2008) and in-between boreholes (Day-Lewis et al., 2003, 2006) by investigating how the amplitude of GPR signals varies over time for a given transmitter– receiver geometry while the saline tracer migrates in the rock fractures. One problem with such studies is that the data acquisition time is often comparable to the timescale of the hydrological flow processes in the fractures where fluid flow velocities might be very high, creating large inversion artifacts if the data acquisition time is ignored in the inversion process. Day-Lewis et al. (2002, 2003) present an innovative inversion method for difference-attenuation crosshole GPR data where the data acquisition time is included within the inversion. Synthetic (Day-Lewis et al., 2002) and field-based (Day-Lewis et al., 2003, 2006) inversion results show significant improvements compared with classical time-lapse inversion algorithms. The research of Day-Lewis et al. (2003, 2006) was carried out at the Forest Service East (FSE) well field at the US Geological Survey (USGS) Fractured Rock Hydrology Research Site located near Mirror Lake, New Hampshire. This well field consists of 14 boreholes distributed over an area of 120  80 m2. Saline injection tests were carried out at 45 m depth where four boreholes, with side lengths of approximately 10 m located in a square-like shape seen from above, are hydraulically connected (Hsieh and Shapiro, 1996). These tracer tests were performed using weak-doublet tracer tests, where fluid was pumped out of one borehole at a rate of 3.8 l min1 and water was injected in another borehole at 1.9 l min1. After achieving steady-state flow, the injection fluid was changed from freshwater to a sodium chloride (NaCl) concentration of 50 g l1 NaCl. Injection of freshwater was resumed after 10 min. The electrical conductivity ratio of these two fluids was estimated to be close to 170. A conventional packer system was used in the pumping well, whereas a special PVC packer system that allowed measurements while preventing vertical flow and the saline solution from entering the boreholes was used in the injection well and in two neighboring wells where GPR measurements were also conducted. The energy of a GPR signal that arrives at the receiving antenna depends to a large degree on the electrical conductance of the media in between the transmitting and receiving antenna. It is expected that the magnitude of the signal at the receiving antenna decreases significantly when a saline tracer passes the ray path. Difference-attenuation inversion is a linear problem since electrical conductivity has no significant effect on the actual ray path. Figure 10 displays variations in the ray energy that arrives in the receiver antennas normalized by the ray length for different transmitter and receiver separations during the tracer experiment of Day-Lewis et al. (2003) for a borehole plane roughly perpendicular to the injection and pumping borehole. Figure 12 also shows the corresponding chloride concentration in the pumping well. The geophysical

difference-attenuation data and the chloride data seem to agree qualitatively and a quicker breakthrough in the GPR data is observed because they were acquired over an area halfway between the injection and pumping boreholes. This type of data was later inverted by Day-Lewis et al. (2003) and they showed that it was possible to remotely monitor the tracer movement relatively well given that only three 2 D slices through the 3 D volume could be imaged. It appears that difference-attenuation data might provide the resolution needed to study fluid flow in fractured rock with only limited hydrological point sampling.

2.15.5.3.3 Seasonal changes in regional saltwater dynamics using time-lapse EM methods Falga`s et al. (2009) present one of few published time-lapse hydrogeophysical studies at the km scale (see Ogilvy et al. (2009) and Nguyen et al. (2009) for seawater intrusion studies using ERT). They used a frequency-domain EM method, namely Controlled-Source Audiomagnetotellurics (CSAMT) (Zonge and Hughes, 1991), to monitor freshwater–seawater interface dynamics in the deltaic zone of the Tordera River in northeastern Spain. Monitoring of saltwater intrusion in coastal aquifers is important due to population growth and since most of the World’s population is concentrated along coastal areas. The CSAMT data were collected over an ancient paleochannel that controls seawater intrusion in a part of the delta. During 2 years, a profile of seven soundings was acquired along a 1700-m N–S trending line. Due to agricultural activity, they could not recover previous site locations with accuracy higher than 100 m when performing the repeated measurements. The resulting individually inverted resistivity models are shown in Figure 11 together with a weighted root-meansquare (RMS) data misfit calculated with an assumed error level of 5%. To better distinguish temporal changes, the inversions used the inversion results of the first survey as initial model for the subsequent inversions. The changes of the electrical resistivity models over time clearly indicate saltwater encroachment in the low-resistivity layer at approximately 50 m depth. These dynamic processes are best imaged in the northern part of the profile where the seawater wedge retreated toward the sea from April 2004 until December 2004, followed by progression until August 2005, and finally followed by a new retreat until May 2006. Multilevel sampling of a piezometer (W06) in April 2004 displayed a saltwater content of approximately 8% at a 50-m depth. Additional evidence to support the interpretation of the geoelectrical models in terms of seawater intrusion is offered by the piezometric levels that were the lowest in August 2005 when the seawater intrusion was interpreted to be at its maximum. Another zone displaying seawater intrusion dynamics is shown in a shallow aquifer located in the upper tenths of meters close to the sea located to the South. Even if the study of Falga`s et al. (2009) had certain limitations, namely rather few stations, long periods between measurements, and not identical measurement locations between surveys, it still shows the potential of EM methods to monitor seawater intrusion processes on a scale that is relevant for water-resource planning.

Hydrogeophysics

425

1.8

x = 0.000 m, z = 197.039 m

1.6

1 x = 0.000 m, z = 198.397 m

1.4

0.8 x = 1.408 m, z = 197.039 m

1 0.6 x = 1.408 m, z = 198.397 m

0.8

x = 1.816 m, z = 197.039 m 0.6

0.4

Chloride concentration (g l−1)

Difference attenuation (dB m−1)

1.2

x = 1.816 m, z = 198.397 m 0.4

0.2 0.2 x = 3.224 m, z = 198.397 m

0

0

0

100

200

300

400

500

Time (min) Figure 10 Nodal difference-attenuation histories between two boreholes (in black) and measured chloride concentration (in red). From Day-Lewis FD, Lane JW, and Gorelick SM (2006) Combined interpretation of radar, hydraulic, and tracer data from a fractured-rock aquifer near Mirror Lake, New Hampshire, USA. Hydrogeology Journal 14: 1–14.

EM methods have a higher sensitivity to conductors (e.g., the seawater plume) than the more commonly used ERT method, even if they have a poor resolving power in defining the lower boundary of conductors. This limitation can partly be resolved by combining this type of geophysical data with other types of geophysical data, such as seismic refraction data during the inversion (Gallardo and Meju, 2007).

2.15.5.4 Hydrogeological Parameter or Zonation Estimation for Improving Flow Predictions Developing a predictive understanding of subsurface flow is complicated by the inaccessibility of the subsurface, the disparity of scales across which controlling processes dominate (e.g., Gelhar, 1993), and the sampling bias associated with

different types of measurements (e.g., Scheibe and Chien, 2003). In this section, we describe the use of geophysical methods to improve flow predictions, through improved parametrization of flow and transport models as well as through fully coupled hydrogeophysical inversion. Although the examples provided here have been conducted at the local scale, joint or fully coupled hydrogeophysical inversion at the watershed scale is a research area that we expect to become more advanced in the coming years.

2.15.5.4.1 Hydraulic conductivity and zonation estimation using GPR and seismic methods Several studies have described the use of geophysical data for estimating hydraulic conductivity (e.g., Cassiani et al., 1998; Hyndman et al., 2000; Hubbard et al., 2001; Chen et al., 2001;

426

Hydrogeophysics N-S

W-06

7

6

5

4

3

2

1

100 200

RMS 3.40

M-2

RMS 4.04

M-3

RMS 4.01

0 100 200 300

August 2004

M-1

300

Aprial 2004

0

0

200

Dec. 2004

100

300 0 April 2005

M-5

RMS 3.40

August 2005

200

M-6

RMS 3.07

M-4

300 0 100 200 300 0 100 200 300

Dec. 2005

RMS 3.30

100

0

200 M-7

300

RMS 4.81 0

Resistivity (Ω m) 1

2.5 1.6

6.3 4

16 10

63

m

meters

40 100 25

400

May 2006

100

250

160

630 400 1000

Figure 11 A series of two-dimensional (2D) electrical conductivity models as a function of date, with the triangles indicating survey location and the y-axis depth in m. From Falga`s E, Ledo J, Marcuello A, and Queralt P (2009) Monitoring freshwater–seawater interface dynamics with audiomagnetotellurics data. Near Surface Geophysics 7(5–6): 391–399.

Gloaguen et al., 2001; Slater, 2007; Linde et al., 2008). A few studies have also illustrated the value of geophysically obtained information for improving flow and transport predictions (Scheibe and Chien, 2003; Bowling et al., 2006; Scheibe et al., 2006). One such example is provided by the linked hydrogeophysical-groundwater modeling study performed at the DOE Oyster Site in Virginia. At this site, tomographic data were used together with borehole flowmeter

logs to develop a site-specific petrophysical relationship that linked radar and seismic velocity with hydraulic conductivity. Using a Bayesian approach, a prior probability of hydraulic conductivity was first obtained through geostatistical interpolation (i.e., kriging) of the hydraulic conductivity values obtained at the wellbore location using the flowmeter logs. Within the Bayesian framework, these estimates were then updated using the developed petrophysical relationship and

Hydrogeophysics

estimates of radar and seismic velocity were obtained along the tomographic transects (Figure 12). The method yielded posterior estimates of hydraulic conductivity (and their uncertainties) along the geophysical transects that honored the wellbore measurements (Chen et al., 2001; Hubbard et al., 2001). Examples of mean values of the geophysically obtained hydraulic conductivity estimates are shown in Figure 12, where the transects are oriented parallel and perpendicular to geological strike. The estimates were obtained at the spatial resolution of the geophysical model, which had pixel dimensions of 0.25 m  0.25 m. The geophysically obtained estimates were then used to develop a synthetic aquifer model (Scheibe and Chien, 2003). Other types of data sets were also used to develop other aquifer models, including interpolated core hydraulic conductivity measurements and interpolated flowmeter data. The breakthrough of a bromide tracer through these different aquifer models was simulated and subsequently compared with the breakthrough of the bromide tracer measured at the Oyster site itself (Scheibe and Chien, 2003). Even though this site was fairly homogeneous (the hydraulic conductivity varied over one order of magnitude) and had extensive borehole control (i.e., wellbores every few meters), it was difficult to capture the variability of hydraulic conductivity using borehole data alone with sufficient accuracy to ensure reliable transport predictions. Scheibe and Chien (2003) found that ‘‘conditioning to geophysical interpretations with larger spatial support significantly improved the accuracy and precision of model predictions’’ relative to wellbore-based data sets. This study suggested that the geophysically based methods provided information at a reasonable scale and resolution for understanding field-scale processes. This is an important point, because it is often difficult to take information gained at the laboratory scale or even from discrete wellbore samples and apply it at the field scale. The level of detail shown in the hydraulic conductivity estimates of Figure 12 may not always be necessary to adequately describe the controls on transport; in some cases, defining only contrasts between hydraulic units (Hill, 2006) or the hydraulic zonation may be sufficient to improve flow predictions. Several studies have illustrated the utility of tomographic methods for mapping zonation of lithofacies or hydrologically important parameters. Hyndman and Gorelick (1996) jointly used tracer and seismic tomographic data to map hydrological zonation within an alluvial aquifer. Linde et al. (2006c) used tomographic zonation constraints in the inversion of tracer test data and found that the constraints improved hydrogeological site characterization. Hubbard et al. (2008) used a discriminant analysis approach to estimate hydraulic conductivity zonation at the contaminated Hanford 100 H site, and found that the identified heterogeneity controlled the distribution of remedial amendments injected into the subsurface for bioremediation purposes as well as the subsequent biogeochemical transformations.

2.15.5.4.2 Joint modeling to estimate temporal changes in moisture content using GPR In this example, we illustrate the value of the joint inversion approach for taking advantage of the complementary nature

427

of geophysical and hydrological data and for circumventing some of the obstacles commonly encountered in other types of integration approaches (see Section 2.15.4.2). As was previously discussed, the use of GPR methods for mapping water content distributions in the subsurface is now well established. However, in general, GPR measurements cannot be directly related to the soil hydraulic parameters needed to make hydrological predictions in the vadose zone (such as the permeability and the parameters describing the relative permeability and capillary pressure functions). On the other hand, time-lapse GPR data often contain information that can be indirectly related to the soil hydraulic properties, since these soil hydraulic properties influence the time- and spacevarying changes in water distribution, which in turn affect GPR data. Kowalsky et al. (2004, 2005) illustrated an approach for incorporating time-lapse GPR and hydrological measurements into a hydrological–geophysical joint inversion framework for estimating soil hydraulic parameter distributions. Coupling between the hydrological and GPR simulators was accomplished within the framework of an inverse model (iTOUGH2, Finsterle, 1999). The inversion was performed using a maximum a posteriori (MAP) approach that utilized concepts from the pilot point method. One of the benefits of this approach was that it directly used the GPR travel times rather than radar velocity tomograms, which circumvented some of the problems that were discussed in Section 2.15.4.2. The approach also accounted for uncertainty in the petrophysical function that related water content and dielectric permittivity. The approach was applied to data collected at the 200 East Area of the US Department of Energy (DOE) Hanford site in Washington. The Hanford subsurface is contaminated with significant quantities of metals, radionuclide, and organics; contaminants are located in the saturated as well as in a thick vadose zone. Gaining an understanding of vadose zone hydraulic parameters, such as permeability, is critical for estimating plume infiltration at the site and the ultimate interception with groundwater and the nearby Columbia River. To gain information about the vadose zone hydraulic parameters, an infiltration test was performed by ponding water on the ground surface and subsequently measuring the subsurface moisture distribution over time using neutron probe data collected within wellbores and radar tomographic data collected between boreholes (Figures 13(a) and 13(b)). Because water infiltration behavior is a function of the permeability distribution, the joint inversion procedure could be used with the time-lapse moisture data to estimate log permeability. The inversion procedure was also used to estimate other parameters of the petrophysical relationship, porosity, and the injection rate, none of which were measured precisely at the site. Figure 13(c) illustrates the permeability values estimated from the joint inversion procedure, which have been conditioned to GPR travel times and to the measured hydrological properties. The obtained permeability values were then used to predict fluid flow at future times. The accuracy of predictions for future times was evaluated through comparison with data collected at later times but not used in the inversion. In the first case, inversion was performed using only neutron probe data collected in two wells at three different times. In the second case, inversion was performed

Comparison of three geophysical attributes obtained from tomographics data Radar velocity Seismic velocity Radar attenuation B2

S9

S18

6.4 GPR velocity (cm ns−1)

Cor = 0.68

0 02 0.

5

0 01 0.

5

01 0.

00 0.

0.

00

0

Hydraulic conductivity

1 1.5 2 2.5 3

60

3.5

65

50 81 82

83 84

Velocity (M/NSO)

1675

1025

1075

1125

0.1

Velocity (M/NSO)

0.5 0.2 Velocity (M/NSO)

6.3 6.2 6.1 6 5.9 5.8 −4

0.4

−3

4

−2 Ln (k)

−1

0 Log conducitivity

Data Da ta

4.5 5 5.5

SE M3 0

9

0.001

8

7

6

01

6

5

0.

BAYES: P(m|d)=[P(d|m)P(m)]/P(d) m=model parameters, d=data

Distance (m) downgradient from B2

K cm s−1

1

Prior information

3−Explore posterior distribution

B2

NW

T1

00

1−Define priors

Posterior distribution

0.

Prior distribution

S14 S10 S12 S9

T2

Depth (m below MSL)

2−Define likelihood functions

6

0010

1

0

Flowmeter measurements at S9

6

Estimates of hydraulic conductivity Figure 12 Example of Bayesian approach for integrating disparate data sets for the estimation of hydraulic conductivity distributions, where the mean value of the estimated hydraulic conductivity distributions is shown on the bottom right. Modified from Hubbard S, Chen J, Peterson J, et al. (2001) Hydrogeological characterization of the DOE bacterial transport site in Oyster, Virginia, using geophysical data. Water Resources Research 37(10): 2431–2456.

Hydrogeophysics

429

θ 4

0.12

4

6

6

0.06 0.04

8

Depth (m)

0.08

Depth (m)

0.1 10 12 14

16 −4 −2

0 −5

−5

0 x (m)

(a)

12 14

5

16

8 10

y (m)

0

5 (b)

2 x (m)

4

6

8

4 2 0 −2 −4 −6 y (m) −8

log(k[m2])

Depth (m)

−10 −4

−11

−6

−12

−8

−13

−10 −12 −4

(c)

−14 −2

0

2

x (m)

4

6

8 −8

0 −2 −4 −6 y (m)

2

4 −15

Figure 13 Time-lapse data sets collected during water injection at Hanford site, including (a) interpolated water content inferred from dense neutronprobe measurements and (b) ground-penetrating radar acquisition geometry. Estimates of log permeability (c) obtained using the coupled inversion approach. Modified from Kowalsky et al. (2006).

using GPR data collected at two times in addition to the neutron probe data used in the first case. Compared to predictions made through inversion of only neutron probe data, inclusion of GPR data in the joint inversion resulted in more accurate estimates of water content at later times.

2.15.6 Summary and Outlook This chapter has reviewed several case studies that illustrated how hydrogeophysical methods can be used to: map subsurface architecture, estimate subsurface hydrological properties or state variables, and monitor subsurface processes associated with natural or engineered in situ perturbations to the subsurface system. These and many other studies have now demonstrated that hydrogeophysical approaches can successfully be used to gain insight about subsurface hydrological processes, provide input that improves flow and transport predictions, and provide information over spatial scales that are relevant to the management of water resources and contaminant remediation. Critical to the success of hydrogeophysical studies are several factors: (1) the acquisition of high-quality geophysical data; (2) the availability of

petrophysical relationships that can link geophysical properties to the parameters or processes relevant for the hydrological study; and (3) the use of inversion approaches that allow for reliable and robust estimation of hydrological parameters of interest. Here, we briefly comment on each of these important factors and associated research needs. Section 2.15.2 reviewed many of the geophysical methods that are common or are being increasingly employed in hydrogeophysical studies, including: electrical resistivity, IP, controlled-source inductive EM, SP, GPR, seismic, SNMR, gravity, magnetics, and well logging methods. We stressed that acquisition of high-quality data is critical to a successful hydrogeophysical study. The choice of which geophysical data to invoke for a particular investigation must be made based on the expected sensitivity of the geophysical attribute to the properties associated with the characterization objective (or the contrast of the target properties with the surrounding sediments or rocks). Different geophysical methods perform optimally in different environments and have different resolving capabilities. It is thus necessary, when deciding on which geophysical method to use to consider the general geological setting and the size/depth/contrast magnitude of the characterization target. Although these characteristics

430

Hydrogeophysics

should be considered prior to choosing a method (and ideally considered through synthetic modeling), commonly the performance of a geophysical method cannot be truly assessed until it is tested at a particular field site. This is because factors that influence its performance (such as clay content, depth of particular target, contrast in characterization target properties with surrounding material, and presence of cultural features such as underground pipes) are often not known with sufficient certainty prior to field testing. For this reason, geophysical campaigns are ideally performed in an iterative manner, starting first with reconnaissance campaigns that involve testing the geophysical responses of a few different methods prior to choosing the method for further highresolution investigation. Section 2.15.3 described common petrophysical models that link electrical conductivity, dielectric permittivity, complex conductivity, and SP measurements to hydrological variables, which are commonly based on theoretical considerations or on laboratory- or field-based experiments. Unfortunately, all of these petrophysical model types pose challenges for hydrogeophysical studies. Theoretical models often invoke assumptions or simplifications that deviate from heterogeneous, in situ conditions. Problems with laboratorybased measurements (e.g., Ferre´ et al., 2005) are that it is very difficult to (1) acquire undisturbed samples that adequately represent conditions in the near subsurface and (2) upscale developed relationships from the laboratory to the field scale (Moysey et al., 2005). Application of field-scale relationships (e.g., using co-located hydrogeophysical wellbore data sets; Hubbard et al., 2001) can also be problematic if the petrophysical relationship differs at locations away from the wellbore (Linde et al., 2006c). Finally, because most geophysical attributes are sensitive to more than one property that typically varies substantially in the subsurface, methods must be developed to handle nonuniqueness in geophysical responses to property variations (Hubbard and Rubin, 1997). The development and testing of petrophysical relationships that describe the linkages between field-scale geophysical responses to variably saturated, semi- to unconsolidated, low-pressure materials that typify many of our shallow subsurface environments continues to be a need in hydrogeophysics. Embedded in that need is the development of methods that can adequately handle scale effects, nonuniqueness, and uncertainties associated with petrophysical relationships. The importance of parameter estimation/integration methods that honor available hydrogeological and geophysical data in the interpretation procedure was described in Section 2.15.4. We defined three different parameter estimation processes, namely: (1) direct mapping; (2) integration approaches (geostatistical and Bayesian); and (3) joint inversion or fully coupled hydrogeophysical inversion. Each of these has advantages and limitations, and the decision about which approach to use is a function of the data available, the characterization objective and project budget, and the experience of the interpreter with the different methods. Clearly, the motivation exists to take advantage of the complimentary nature of hydrological and geophysical data and modeling to improve experimental design and interpretation while recognizing that each of these approaches has associated uncertainty. We thus believe that one of the most important

developments in hydrogeophysical research in the coming years will arise from data integration schemes that provide a flexible way to couple different hydrological and geophysical data and model types in a framework that explicitly assesses uncertainty in the final model or model predictions. An important challenge will be the development of methods that disregard models that are inconsistent with our available data and a priori conceptions while retaining a representative subset of models that are consistent with available data. It is expected that joint inversion approaches can provide more significant improvements compared to other approaches, especially when working with time-lapse data and when the hydrological dynamics of the geophysical and hydrological forward responses display strong nonlinearities. Although inversion approaches have been developed to meet some of these criteria, for the most part they have been tested in conjunction with specific research projects and are not generally accessible for use by nonspecialists or flexible enough to be applied to other problems and data sets. An existing need is thus the development of software that will facilitate the transfer of the state-of-the-art inversion algorithms, which allow joint consideration of geophysical and hydrological measurements and phenomena and that provide meaningful assessments of uncertainty, into practice. Related to all three key factors in hydrogeophysical studies (high-quality geophysical data sets, petrophysics, and integration methods) is the need to better advance our capabilities to improve the characterization of subsurface hydrological parameters and processes at the larger watershed scale. The majority of the hydrogeophysical studies that have focused on quantitative hydrological parameter estimation or model coupling have been performed at the local scale (typically with length scales o10 m), where the disparity in measurement support scale between wellbore (direct) measurements and geophysical measurements is smaller and where stationarity of petrophysical relationships can often be reasonably assumed. Although these studies have illustrated the power of hydrogeophysical methods for improving the resolution and understanding of subsurface properties or processes at the local scale, they are often still limited in their ability to inform about behavior that may be most relevant at the larger scales where water resources or environmental contaminants are managed. As described in Section 2.15.4, although a handful of case studies have now illustrated the potential that geophysical methods hold for providing quantitative information over large spatial scales, additional effort is needed to continue to advance this area of watershed hydrogeophysics (Robinson et al., 2008). In particular, there is a great need to develop petrophysical models and integration schemes that permit the coupling of different hydrological and geophysical data and model types within a framework that explicitly assesses uncertainty in the final model or model predictions over watershed- or plume-relevant scales.

Acknowledgments Support for Susan Hubbard was provided by the US Department of Energy, Biological and Environmental Research Program as part of the Oak Ridge Integrated Field Research

Hydrogeophysics

Center (ORIFRC) project and through DOE Contract DEAC0205CH11231 to the LBNL Sustainable Systems Scientific Focus Area. We thank Lee Slater and Se´bastien Lambot whose constructive reviews helped to substantially improve the text.

References Adepelumi AA, Ako BD, Afolabi O, and Arubayi JB (2005) Delineation of contamination plume around oxidation sewage ponds in southwestern Nigeria. Environmental Geology 48: 1137--1146. Ajo-Franklin JB, Minsley BJ, and Daley TM (2007) Applying compactness constraints to differential traveltime tomography. Geophysics 72: R67--R75. Alumbaugh DL and Newman GA (2000) Image appraisal for 2-D and 3-D electromagnetic inversion. Geophysics 65: 1455--1467. Annan AP (2005) GPR methods for hydrogeological studies. In: Rubin Y and Hubbard S (eds.) Hydrogeophysics, ch. 7, pp. 185–214. Dordrecht: Springer. Archie GE (1942) The electrical resistivity log as an aid in determining some reservoir parameters. Transactions of the American Institute of Mining, Metallurgy, and Petroleum Engineering 146: 54--62. Atekwana EA, Werkema DD, Allen JP, et al. (2004) Evidence of microbial enhanced electrical conductivity in hydrocarbon-contaminated sediments. Geophysics Research Letters 31: L2350. Atekwana EA, Werkema DD, and Atekwana EA (2006) Biogeophysics: The effects of microbial processes on geophysical properties of the shallow subsurface. In: Vereecken H, Binley A, Cassiani G, Revil A, and Titov K (eds.) Applied Hydrogeophysics, ch. 6, pp. 161–193. Dordrecht: Springer. Beres M and Haeni FP (1991) Application of ground-penetrating radar methods in hydrogeological studies. Ground Water 29: 375--386. Beres M, Huggenberger P, Green AG, and Horstmeyer H (1999) Using two- and threedimensional georadar methods to characterize glaciofluvial architecture. Sedimentary Geology 129: 1--24. Bernabe´ Y and Revil A (1995) Pore-scale heterogeneity, energy dissipation and the transport properties of rocks. Geophysical Research Letters 22: 1529--1532. Berthold S, Bentley L, and Hayashi M (2004) Integrated hydrogeological and geophysical study of depression-focused groundwater recharge in the Canadian prairies. Water Resources Research 40: W06505. Bigalke J and Grabner EW (1997) The geobattery model: A contribution to large scale electrochemistry. Electrochemica Acta 42: 3443. Binley A, Cassiani G, Middleton R, and Winship P (2002) Vadose zone flow model parameterisation using cross-borehole radar and resistivity imaging. Journal of Hydrology 267: 147--159. Binley A and Kemna A (2005) DC resistivity and induced polarization methods. In: Rubin Y and Hubbard S (eds.) Hydrogeophysics, ch. 5, pp. 129–156. Dordrecht: Springer. Binley A, Slater LD, Fukes M, and Cassiani G (2005) Relationship between spectral induced polarization and hydraulic properties of saturated and unsaturated sandstone. Water Resources Research 41: W12417. Birchak JR, Gardner CG, Hipp JE, and Victor JM (1974) High dielectric constant microwave probes for sensing soil moisture. Proceedings of the IEEE 62: 93--98. Blakely RJ (1996) Potential Theory in Gravity and Magnetic Applications. Cambridge: Cambridge University Press. Bolshakov DK, Modin IN, Pervage EV, and Shevnin VA (1997) Separation of anisotropy and inhomogeneity influence by the spectral analysis of azimuthal resistivity diagrams. Paper presented at 3rd EEGS-ES Meeting, Aarhus, Denmark. Bowling JC, Zheng C, Rodriguez AB, and Harry D (2006) Geophysical constraints on contaminant modeling in a heterogeneous fluvial aquifer. Journal of Contaminant Hydrology 85: 72--88. Brovelli A and Cassiani G (2008) Effective permittivity of porous media: A critical analysis of the complex refractive index model. Geophysical Prospecting 56: 715--727. Brovelli A and Cassiani G (2010) A combination of the Hashin–Shtrikman bounds aimed at modelling electrical conductivity and permittivity of variably saturated porous media. Geophysical Journal International 180: 225--237. Bussian AE (1983) Electrical conductance in a porous medium. Geophysics 48: 1258--1268. Butler J (2005) Hydrogeological methods for estimation of spatial variations in hydraulic conductivity. In: Rubin Y and Hubbard S (eds.) Hydrogeophysics, ch. 2, pp. 23–58. Dordrecht: Springer. Bo¨rner FD and Scho¨n JH (1991) A relation between the quadrature component of electrical conductivity and the specific surface area of sedimentary rocks. Log Analyst 32: 612--613.

431

Carcione JM, Ursin B, and Nordskag JI (2007) Cross-property relations between electrical conductivity and the seismic velocity of rocks. Geophysics 72: E193--E204. Cassiani G, Bohm G, Vesnaver A, and Nicholich R (1998) A geostatistical framework for incorporating seismic tomography auxillary data into hydraulic conductivity estimation. Journal of Hydrology 206(1–2): 58--74. Chapelle FH (2001) Groundwater Microbiology and Geochemistry. New York: Wiley. Chen J, Hubbard S, Gaines D, Korneev V, Baker G, and Watson D (2010) Stochastic estimation of aquifer geometry using seismic refraction data with borehole depth constraints. Water Resources Research (in press). Chen J, Hubbard S, Peterson J, et al. (2006) Development of a joint hydrogeophysical inversion approach and application to a contaminated fractured aquifer. Water Resources Research 42: W06425 (doi: 10.1029/ 2005WR004694). Chen J, Hubbard S, and Rubin Y (2001) Estimating the hydraulic conductivity at the South Oyster Site from geophysical tomographic data using Bayesian techniques based on the normal linear regression model. Water Resources Research 37: 1603--1616. Chen J, Hubbard SS, Williams KH, Pride S, Li L, and Slater L (2009) A state-space Bayesian framework for estimating biogeochemical transformations using timelapse geophysical data. Water Resources Research 45: W08420 (doi: 10.1029/ 2008WR007698). Cole KS and Cole RH (1941) Dispersion and absorption in dielectrics, 1. Alternating current characteristics. Journal of Chemical Physics 9: 341--351. Corry CE (1985) Spontaneous polarization associated with porphyry sulfide mineralization. Geophysics 50: 1020. Dafflon B, Irving J, and Holliger K (2009) Simulated-annealing-based conditional simulation for the local-scale characterization of heterogeneous aquifers. Journal of Applied Geophysics 68: 60--70. Davis JL and Annan AP (1989) Ground-penetrating radar for high-resolution mapping of soil and rock stratigraphy. Geophysical Prospecting 37: 531--551. Day-Lewis FD, Harris JM, and Gorelick SM (2002) Time-lapse inversion of crosswell radar data. Geophysics 67: 1740--1752. Day-Lewis FD and Lane JW, Jr. (2004) Assessing the resolution-dependent utility of tomograms for geostatistics. Geophysical Research Letters 31: L07503. Day-Lewis FD, Lane JW, and Gorelick SM (2006) Combined interpretation of radar, hydraulic, and tracer data from a fractured-rock aquifer near Mirror Lake, New Hampshire, USA. Hydrogeology Journal 14: 1--14. Day-Lewis FD, Lane JW, Jr., Harris JM, and Gorelick SM (2003) Time-lapse imaging of saline-tracer transport in fractured rock using difference-attenuation radar tomography. Water Resources Research 39: 1290. Day-Lewis FD, Singha K, and Binley A (2005) Applying petrophysical models to radar travel time and electrical resistivity tomograms: Resolution-dependent limitations. Journal of Geophysical Research 110: B08206. D’Ozouville N, Auken E, Sorensen K, et al. (2008) Extensive perched aquifer and structural implications revealed by 3D resistivity mapping in a Galapagos volcano. Earth and Planetary Science Letters 269: 518--522. Englert A, Hubbard SS, Williams KH, Li L, and Steefel CI (2009) Feedbacks between hydrological heterogeneity and bioremediation induced biogeochemical transformations. Environmental Science and Technology 43(13): 5197--5204. Ernst JR, Maurer H, Green AG, and Holliger K (2007) Full-waveform inversion of crosshole radar data based on 2-D finite-difference time-domain solutions of Maxwell’s equations. IEEE Transactions on Geoscience and Remote Sensing 45(9): 2807--2828. Everett ME and Meju MA (2005) Near-surface controlled source electromagnetic induction: Background and recent advances. In: Rubin Y and Hubbard S (eds.) Hydrogeophysics, ch. 6, pp. 157–183. Dordrecht: Springer. Falga`s E, Ledo J, Marcuello A, and Queralt P (2009) Monitoring freshwater–seawater interface dynamics with audiomagnetotellurics data. Near Surface Geophysics 7 (5–6): 391--399. Farquharson CG (2008) Constructing piecewise-constant models in multidimensional minimum-structure inversions. Geophysics 73: K1--K9. Ferre´ TPA, Binley A, Geller J, Hill E, and Illangasekare T (2005) Hydrogeophysical methods at the laboratory scale. In: Rubin Y and Hubbard S (eds.) Hydrogeophysics, pp. 441–463. Dordrecht: Springer. Finsterle S (1999) iTOUGH2 User Guide, Lawrence Berkeley National Laboratory LBNL-40040. Fox RM (1830) On the electromagnetic properties of metalliferous veins in the mines of Cornwall. Philosophical Transactions of the Royal Society 130: 399. Freeze RA and Cherry JA (1979) Groundwater. Upper Saddle River, NJ: Prentice Hall. Friedel S (2003) Resolution, stability and efficiency of resistivity tomography estimated from a generalized inverse approach. Geophysical Journal International 153: 305--316.

432

Hydrogeophysics

Gallardo LA and Meju MA (2003) Characterization of heterogeneous near-surface materials by joint 2D inversion of dc resistivity and seismic data. Geophysical Research Letters 30: 1658. Gallardo LA and Meju MA (2004) Joint two-dimensional DC resistivity and seismic travel time inversion with cross-gradient constraints. Journal of Geophysical Research 109: B03311. Gallardo LA and Meju MA (2007) Joint two-dimensional cross-gradient imaging of magnetotelluric and seismic traveltime data for structural and lithological classification. Geophysical Journal International 169: 1261--1271. Gelhar LW (1993) Stochastic Subsurface Hydrology. New York: Prentice Hall. Ghorbani A, Camerlynck C, Florsch N, Cosenza P, and Revil A (2007) Bayesian inference of the Cole–Cole parameters from time- and frequency-domain induced polarization. Geophysical Prospecting 55: 589--605. Gloaguen E, Chouteau M, Marcotte D, and Chapuis R (2001) Estimation of hydraulic conductivity of an unconfined aquifer using cokriging of GPR and hydrostratigraphic data. Journal of Applied Geophysics 47(2): 135--152. Goldie M (2002) Self-potentials associated with the Yanacocha high-sulfidation gold deposit in Peru. Geophysics 67: 684--689. Golub GH and van Loan CF (1996) Matrix Computations, 3rd edn. Baltimore, MD: Johns Hopkins Press. Grasmueck M (1996) 3-D ground-penetrating radar applied to fracture imaging in gneiss. Geophysics 61: 1050--1064. Greaves RJ, Lesmes DP, Lee JM, and Toksov MN (1996) Velocity variations and water content estimated from multi-offset, ground-penetrating radar. Geophysics 61: 683--695. Grote K, Hubbard SS, and Rubin Y (2003) Field-scale estimation of volumetric water content using GPR ground wave techniques. Water Resources Research 39: 1321--1333. Gue´guen Y and Palciauskas V (1994) Introduction to the Physics of Rocks. Princeton, NJ: Princeton University Press. Gu¨nther T, Ru¨cker C, and Spitzer K (2006) Three-dimensional modelling and inversion of dc resistivity data incorporating topography – II. Inversion. Geophysical Journal International 166: 506--517. Hashin Z and Shtrikman S (1962) A variational approach to the theory of the effective magnetic permeability of multiphase materials. Journal of Applied Physics 33: 3125--3131. Hastings WK (1970) Monte Carlo sampling methods using Markov chains and their applications. Biometrika 57: 97--109. Hertrich M (2008) Imaging of groundwater with nuclear magnetic resonance. Progress in Nuclear Magnetic Resonance Spectroscopy 53: 227--248. Hill M (2006) The practical use of simplicity in developing groundwater models. Ground Water 44(6): 775--781. Hinze WJ (1990) The role of gravity and magnetic methods in engineering and environmental studies. In: Ward S (ed.) Geotechnical and Environmental Geophysics: Review and Tutorial, SEG Investigations in Geophysics No. 5, vol. 1, pp. 75–126. Tulsa, OK: Society Of Exploration Geophysicists. Hsieh PA and Shapiro AM (1996) Hydraulic characteristics of fractured bedrock underlying the FSE well field at the Mirror Lake Site, Grafton Country, New Hampshire. In: Morganwalp DW and Aronson DA (eds.) U.S. Geological Survey Toxic Substances Hydrology Program: Proceedings of the Technical Meeting, US Geological Survey Water Resources Investigative Report 94-4015, vol. 1, pp. 127–130, Colorado Springs, CO, USA. 20–24 September 1993. Ho¨rdt A, Blaschlek R, Kemna A, and Zisser N (2007) Hydraulic conductivity estimation from induced polarisation data at the field scale – the Krauthausen case history. Journal of Applied Geophysics 62: 33--46. Hubbard S, Chen J, Peterson J, et al. (2001) Hydrogeological characterization of the DOE bacterial transport site in Oyster, Virginia, using geophysical data. Water Resources Research 37(10): 2431--2456. Hubbard S, Lunt I, Grote K, and Rubin Y (2006) Vineyard soil water content: mapping small scale variability using ground penetrating radar. In: Macqueen RW and Meinert LD (eds.) Fine Wine and Terroir – The Geoscience Perspective. Geoscience Canada Reprint Series Number 9 (ISBN 1-897095-21-X; ISSN 0821-381X). St. John’s, NL: Geological Association of Canada. Hubbard S, Peterson JE, Majer EL, et al. (1997) Estimation of permeable pathways and water content using tomographic radar data. The Leading Edge of Exploration 16(11): 1623--1628. Hubbard S and Rubin Y (1997) Ground penetrating radar assisted saturation and permeability estimation in bimodal systems. Water Resources Research 33(5): 971--990. Hubbard SS, Williams K, Conrad M, et al. (2008) Geophysical monitoring of hydrological and biogeochemical transformations associated with Cr(VI) biostimulation. Environmental Science and Technology (doi: 10.1021/es071702s)

Huisman S, Hubbard SS, Redman D, and Annan P (2003) Monitoring soil water content with ground-penetrating radar: A review. Vadose Zone Journal 2: 476--491. Huisman JA, Rings J, Vrugt JA, Sorg J, and Vereecken H (2010) Hydraulic properties of a model dike from coupled Bayesian and multi-criteria hydrogeophysical inversion. Journal of Hydrology 380: 62--73. Hyndman DW and Gorelick SM (1996) Estimating lithologic and transport properties in three dimensions using seismic and tracer data: The Kesterson aquifer. Water Resources Research 32(9): 2659--2670. Hyndman DW, Harris JM, and Gorelick SM (2000) Inferring the relationship between seismic slowness and hydraulic conductivity in heterogeneous aquifers. Water Resources Research 36(8): 2121--2132. Irving J, Knight R, and Holliger K (2009) Estimation of the lateral correlation structure of subsurface water content from surface-based ground-penetrating radar reflection images. Water Resources Research 45: W12404. Jackson D (1976) Most squares inversion. Journal of Geophysical Research 81: 1027--1030. Johnson DL, Koeplik J, and Schwartz LM (1986) New pore-size parameter characterizing transport in porous media. Physical Review Letters 57: 2564--2567. Juhlin C and Stephens MB (2006) Gently dipping fracture zones in Paleoproterozoic metagranite, Sweden: Evidence from reflection seismic and cored borehole data and implications for the disposal of nuclear waste. Journal of Geophysical Research 111: B09302. Kalscheuer T and Pedersen LB (2007) A non-linear truncated SVD variance and resolution analysis of two-dimensional magnetotelluric models. Geophysical Journal International 169: 435--447. Keller GV (1987) Rock and mineral properties. In: Nabighian MN (ed.) Electromagnetic Methods in Applied Geophysics, Vol. 1 – Theory, Investigations in Geophysics 3, pp. 13–51. Tulsa, OK: Society of Exploration Geophysicists (SEG). Kemna A, Binley A, Ramirez A, and Daily W (2000) Complex resistivity tomography for environmental applications. Chemical Engineering Journal 77: 11--18. Keys SW (1989) Borehole Geophysics Applied to Ground-Water Investigations. Dublin, OH: National Water Well Association. Knight R (1991) Hysteresis in the electrical resistivity of partially saturated sandstone. Geophysics 56: 2139--2147. Kobr M, Mares S, and Paillet F (2005) Geophysical well logging: Borehole geophysics for hydrogeological studies: Principles and applications. In: Rubin Y and Hubbard S (eds.) Hydrogeophysics, ch. 10, pp. 291–332. Dordrecht: Springer. Koch K, Wenninger J, Uhlenbrook S, and Bonell M (2009) Joint interpretation of hydrological and geophysical data: Electrical resistivity tomography results from a process hydrological research site in the Black Forest Mountains, Germany. Hydrological Processes 23: 1501--1513. Kowalsky MB, Finsterle S, Peterson J, et al. (2005) Estimation of field-scale soil hydraulic and dielectric parameters through joint inversion of GPR and hydrological data. Water Resources Research 41: W11425. Kowalsky MB, Finsterle SA, and Rubin Y (2004) Estimating flow parameter distributions using ground-penetrating radar and hydrological measurements during transient flow in the vadose zone. Advances in Water Resources 27(6): 583--599. Krause P, Naujoks M, Fink M, and Kroner C (2009) The impact of soil moisture changes on gravity residuals obtained with a superconducting gravimeter. Journal of Hydrology 373: 151--163. LaBrecque DJ and Yang X (2001) Difference inversion of ERT data: A fast inversion method for 3-D in situ monitoring. Journal of Environmental and Engineering Geophysics 6: 83--89. Lambot S, Binley A, Slob E, and Hubbard S (2008) Ground penetrating radar in hydrogeophysics. Vadose Zone Journal 7: 137--139 (doi: 10.2136/vzj2007.0180) Lambot S, Rhebergen J, van den Bosch I, Slob EC, and Vanclooster M (2004b) Measuring the soil water content profile of a sandy soil with an off-ground monostatic ground penetrating radar. Vadose Zone Journal 3: 1063--1071. Lambot S, Slob E, Rheberger J, Lopera O, Jadoon KZ, and Vereecken H (2009) Remote estimation of the hydraulic properties of a sand using full-waveform integrated hydrogeophysical inversion of time-lapse, off-ground GPR data. Vadose Zone Journal 8: 743--754. Lambot S, Slob EC, van den Bosch I, Stockbroeckx B, and Vanclooster M (2004a) Modeling of ground penetrating radar for accurate characterization of subsurface electric properties. IEEE Transactions on Geoscience and Remote Sensing 42: 2555--2568. Lambot S, Weihermu¨ller L, Huisman JA, Vereecken H, Vanclooster M, and Slob EC (2006) Analysis of air-launched ground-penetrating radar techniques to measure the soil surface water content. Water Resources Research 42: W11403 (doi: 10.1029/2006WR005097). Leroy P Jr. and Revil P (2004) A triple-layer model of the surface electrochemical properties of clay minerals. Journal of Colloid and Interface Science 270: 371--380.

Hydrogeophysics Leroy P and Revil A (2009) A mechanistic model for the spectral induced polarization of clay materials. Journal of Geophysical Research 114: B10202 (doi: 10.1029/ 2008JB006114). Leroy P, Revil A, Kemna A, Cosenza P, and Ghorbani A (2008) Complex conductivity of water-saturated packs of glass beads. Journal of Colloid and Interface Sciences 321: 103--117. Lesmes DP and Friedman SP (2005) Relationships between the electrical and hydrogeological properties of rocks and soils. In: Rubin Y and Hubbard SS (eds.) Hydrogeophysics, pp. 87--128. Dordrecht: Springer. Li L, Steefel CI, Williams KH, Wilkins MJ, and Hubbard SS (2009) Mineral transformation and biomass accumulation during uranium bioremediation at Rifle, Colorado. Environmental Science and Technology 43(14): 5429--5435. Lien M and Mannseth T (2008) Sensitivity study of marine CSEM data for reservoir production monitoring. Geophysics 73: F151--F163. Linde N (2009) A comment on ‘‘Characterization of multiphase coupling using a bundle of capillary tubes model’’ by M. D. Jackson (Journal of Geophysical Research, 113: B04201). Journal of Geophysical Research 114: B06209. Linde N, Binley A, Tryggvason A, Pedersen LB, and Revil A (2006a) Improved hydrogeophysical characterization using joint inversion of cross-hole electrical resistance and ground-penetrating radar traveltime data. Water Resources Research 42: W12404. Linde N, Chen J, Kowalsky MB, and Hubbard S (2006b) Hydrogeophysical parameter estimation approaches for field scale characterization. In: Vereecken H, Binley A, Cassiani G, Revil A, and Titov K (eds.) Applied Hydrogeophysics, ch. 2, pp. 9–44. Dordrecht: Springer. Linde N, Finsterle S, and Hubbard S (2006c) Inversion of tracer test data using tomographic constraints. Water Resources Research 42: W04410. Linde N, Jougnot D, Revil A, et al. (2007) Streaming current generation in two-phase flow conditions. Geophysical Research Letters 34: L03306. Linde N and Pedersen LB (2004) Evidence of electrical anisotropy in limestone formations using the RMT technique. Geophysics 69: 909--916. Linde N and Revil A (2007) Inverting self-potential data for redox potentials of contaminant plumes. Geophysical Research Letters 34: L14302 (10.1029/ 2007GL030084). Linde N, Tryggvason A, Peterson J, and Hubbard S (2008) Joint inversion of crosshole radar and seismic traveltimes. Geophysics 73: G29--G37. Lunt IA, Bridge JS, and Tye RS (2004) A quantitative, three-dimensional depositional model of gravelly braided rivers. Sedimentology 51: 377--414. Lunt IA, Hubbard SS, and Rubin Y (2005) Soil moisture content estimation using ground-penetrating radar reflection data. Journal of Hydrology 307: 254--269. Mair JA and Green AG (1981) High-resolution seismic-reflection profiles reveal fracture-zones within a homogeneous granite batholith. Nature 294: 439--442. Maurer H, Holliger K, and Boerner DE (1998) Stochastic regularization: Smoothness or similarity? Geophysical Research Letters 25: 2889--2892. Mavko G, Mukerji T, and Dvorkin J (1998) The Rock Physics Handbook. Cambridge: Cambridge University Press. McLaughlin D and Townley LR (1996) A reassessment of the groundwater inverse problem. Water Resources Research 32: 1131--1161. Menke W (1984) Geophysical Data Analysis–Discrete Inverse Theory. New York: Academic Press. Minsley BJ, Sogade J, and Morgan FD (2007) Three-dimensional source inversion of self-potential data. Journal of Geophysical Research 112: B02202. Molz F, Boman GK, Young SC, and Waldrop WR (1994) Borehole flowmeters: Field applications and data analysis. Journal of Hydrology 163: 347--371. Mosegaard K and Tarantola A (1995) Monte-Carlo sampling of solutions to inverse problems. Journal of Geophysical Research 100: 12431--12447. Moysey S, Singha K, and Knight R (2005) A framework for inferring field-scale rock physics relationships through numerical simulation. Geophysical Research Letters 32: L08304. Naudet V, Revil A, Bottero JY, and Begassat P (2003) Relationship between selfpotential (SP) signals and redox conditions in contaminated groundwater. Geophysical Research Letters 30: 2091. Naudet V, Revil A, Rizzo E, Bottero JY, and Begassat P (2004) Groundwater redox conditions and conductivity in a contaminant plume from geoelectrical investigations. Hydrology and Earth Systems Sciences 8: 8--22. Nguyen F, Kemna A, Antonsson A, et al. (2009) Characterization of seawater intrusion using 2D electrical imaging. Near Surface Geophysics 7: 377--390. NRC (1996) Rock Fractures and Fluid Flow. Washington, DC: National Academy Press. Ogilvy RD, Meldrum PI, Kuras I, et al. (2009) Automated monitoring of coastal aquifers with electrical resistivity tomography. Near Surface Geophysics 7: 367--375. Oldenburg DW and Li Y (1999) Estimating the depth of investigation in dc resistivity and IP surveys. Geophysics 64: 403--416.

433

Olsson O, Falk L, Forslund O, Lundmark L, and Sandberg E (1992) Borehole radar applied to the characterization of hydraulically conductive fracture zones in crystalline rock. Geophysical Prospecting 40: 109--142. Paine JG (2003) Determining salinization extent, identifying salinity sources, and estimating chloride mass using surface, borehole, and airborne electromagnetic induction methods. Water Resources Research 39(3): 3–1–3–10. Parker RL (1994) Geophysical Inverse Theory. Princeton, NJ: Princeton University Press. Pedersen LB, Qian W, Dynesius L, and Zhang P (1994) An airborne tensor VLF system – from concept to realization. Geophysical Prospecting 42: 863--883. Poldini E (1938) Geophysical exploration by spontaneous polarization methods. Mining Magazine, London 59: 278--282. Pollock D and Cirpka OA (2010) Fully coupled hydrogeophysics inversion of synthetic salt tracer experiments. Water Resources Research 46: W07501. Pride S (1994) Governing equations for the coupled electromagnetics and acoustics of porous media. Physical Review B 50: 15678--15696. Purvance D and Andricevic R (2000) On the electrical-hydraulic conductivity correlation in aquifers. Water Resources Research 36: 2905--2913. Revil A and Cathles LM (1999) Permeability of shaly sands. Water Resources Research 35: 651--662. Revil A, Cathles LM, Losh S, and Nunn JA (1998) Electrical conductivity in shaly sands with geophysical applications. Journal of Geophysical Research 103: 23925--23936. Revil A and Glover PWJ (1997) Theory of ionic-surface electrical conduction in porous media. Physical Review B 55: 1757--1773. Revil A and Glover PWJ (1998) Nature of surface electrical conductivity in natural sands, sandstones, and clays. Geophysical Research Letters 25: 691--694. Revil A and Leroy P (2004) Constitutive equations for ionic transport in porous shale’s. Journal of Geophysical Research 109: B03208. Revil A and Linde N (2006) Chemico-electromechanical coupling in microporous media. Journal of Colloid and Interface Sciences 302: 682--694. Revil A, Naudet V, Nouzaret J, and Pessel M (2003) Principles of electrography applied to self-potential electrokinetic sources and hydrogeological applications. Water Resources Research 39(5): 1114. Ritzi RW, Dai ZX, Dominic DF, and Rubin YN (2004) Spatial correlation of permeability in cross-stratified sediment with hierarchical architecture. Water Resources Research 40: W03513. Robinson DA, Binley A, Crook N, et al. (2008) Advancing process-based watershed hydrological research using near-surface geophysics: A vision for, and review of, electrical and magnetic geophysical methods. Hydrological Processes 22: 3604--3635. Rubin Y and Hubbard S (2005) Hydrogeophysics, Water and Science Technology Library, vol. 50. Dordrecht: Springer. Rucker DF and Fink JB (2007) Inorganic plume delineation using surface high resolution electrical resistivity at the BC Cribs and Trenches Site, Hanford. Vadose Zone Journal 6: 946--958. Sassen DS and Everett ME (2009) 3D polarimetric GPR coherency attributes and full- waveform inversion of transmission data for characterizing fractured rock. Geophysics 74: J23--J34 (doi: 10.1190/1.3103253). Sato M and Mooney HM (1960) The electrochemical mechanism of sulfide selfpotentials. Geophysics 25: 226--249. Saunders JH, Herwanger JV, Pain CC, Worthington MH, and de Oliveira CRE (2005) Constrained resistivity inversion using seismic data. Geophysical Journal International 160: 785--796. Scheibe T, Fang Y, Murray CJ, et al. (2006) Transport and biogeochemical reactions of metals in a physically and chemically heterogeneous aquifer. Geosphere 2(4): 220--235 (doi: 10.1130/GES00029.1). Scheibe TD and Chien YJ (2003) An evaluation of conditioning data for solute transport prediction. Ground Water 41(2): 128--141. Schwartz LM, Sen PN, and Johnson DL (1989) Influence of rough surfaces on electrolytic conduction in porous media. Physical Review B 40: 2450--2458. Scho¨n JH (1996) Physical Properties of Rocks – Fundamentals and Principles of Petrophysics. Amsterdam: Elsevier. Scott JBT and Barker RD (2003) Determining pore-throat size in Permo-Triassic sandstones from low-frequency electrical spectroscopy. Geophysical Research Letters 30: 1450. Sen PS, Goode PA, and Sibbit A (1988) Electrical conduction in clay bearing sandstones at low and high salinities. Journal of Applied Geophysics 63: 4832--4840. Sen PN, Scala C, and Cohen MH (1981) A self-similar model for sedimentary rocks with application to the dielectric constant of fused glass beads. Geophysics 46: 781--795. Shokri N, Lehmann P, and Or D (2009) Characteristics of evaporation from partially wettable porous media. Water Resources Research 45: W02415.

434

Hydrogeophysics

Sill WR (1983) Self-potential modeling from primary flows. Geophysics 48: 76--86. Singha K and Gorelick SM (2005) Saline tracer visualized with three-dimensional electrical resistivity tomography: Field-scale spatial moment analysis. Water Resources Research 41: W05023. Slater L (2007) Near surface electrical characterization of hydraulic conductivity: From petrophysical properties to aquifer geometries – a review. Surveys in Geophysics 28: 167--197. Slater L, Atekwana E, Brantley S, et al. (2009) Exploring the geophysical signatures of microbial processes. Eos Transactions AGU 90: 83. Slater L and Lesmes DP (2002) Electrical-hydraulic relationships observed for unconsolidated sediments. Water Resources Research 38: 1213. Slater L, Ntarlagniannis D, Personna Y, and Hubbard S (2007) Pore-scale spectral induced polarization (SIP) signatures associated with FeS biomineral transformations. Geophysical Research Letters 34: L21404 (doi: 10.1029/ 2007GL031840). Slater L, Ntarlagiannis D, and Wishard D (2006) On the relationship between induced polarization and surface area in metal–sand and clay–sand mixtures. Geophysics 71: A1--A5. Sørensen KI and Auken E (2004) SkyTEM – a new high-resolution helicopter transient electromagnetic system. Exploration Geophysics 35: 191--199. Spillmann T, Maurer H, Willenberg H, et al. (2007) Characterization of an unstable rock mass based on borehole logs and diverse borehole radar data. Journal of Applied Geophysics 61: 16--38. Steeples D (2005) Shallow seismic methods. In: Rubin Y and Hubbard S (eds.) Hydrogeophysics, ch. 7, pp. 215–252. Dordrecht: Springer. Stummer P, Maurer H, and Green AG (2004) Experimental design: Electrical resistivity data sets that provide optimum subsurface information. Geophysics 69: 120--139. Styles P, McGrath R, Thomas E, and Cassidy NJ (2005) The use of microgravity for cavity characterization in Karstic terrains. Quarterly Journal of Engineering Geology and Hydrogeology 38(2): 155--169. Suski B, Revil A, Titov K, et al. (2006) Monitoring of an infiltration experiment using the self-potential method. Water Resources Research 42: W08418. Talley J, Baker GS, Becker MW, and Beyrle N (2005) Four dimensional mapping of tracer channelization in subhorizontal bedrock fractures using surface ground penetrating radar. Geophysical Research Letters 32: L04401. Tarantola A (2005) Inverse Problem Theory and Methods for Model Parameter Estimation. Philadelphia, PA: Society for Industrial and Applied Mathematics (SIAM). Taylor RW and Fleming AH (1988) Characterizing jointed systems by azimuthal resistivity techniques. Ground Water 26: 464--474. Telford WM, Geldart LP, and Sheriff RE (1990) Applied Geophysics. Cambridge: Cambridge University Press. Titov KV, Levitski A, Knoosavski PK, Tarasov AV, Ilyin YT, and Bues MA (2005) Combined application of surface geoelectrical methods for groundwater-flow modeling: A case history. Geophysics 70(5): H21--H31. Topp GC, Davis JL, and Annan AP (1980) Electromagnetic determination of soil water content: Measurements in coaxial transmission lines. Water Resources Research 16: 574--582. Tsoflias GP and Becker MW (2008) Ground-penetrating-radar response to fracturefluid salinity: Why lower frequencies are favorable for resolving salinity changes. Geophysics 73: J25--J30.

Tsoflias GP, Halihan T, and Sharp JM Jr. (2001) Monitoring pumping test response in a fractured aquifer using ground-penetrating radar. Water Resources Research 37: 1221--1229. US DOE (2010) Complex Systems Science for Subsurface Fate and Transport, Report from the August 2009 Workshop, DOE/SC-0123, U.S. Department of Energy Office of Science (www.science.doe.gov/ober/BER_workshops.html). van Overmeeren RA (1998) Radar facies of unconsolidated sediments in The Netherlands: A radar stratigraphy interpretation method for hydrogeology. Journal of Applied Geophysics 40: 1--18. Vereecken H, Binley A, Cassiani G, Revil A, and Titov K (2006) Applied Hydrogeophysics, NATO Science Series, Earth and Environmental Science, vol. 71. Dordrecht: Springer. Vereecken H, Huisman JA, Bogena H, Vanderborght J, Vrugt JA, and Hopmans JW (2008) On the value of soil moisture measurements in vadose zone hydrology: A review. Water Resources Research 44: W00D06. Viezzoli A, Christiansen AV, Auken E, and Sørensen K (2008) Quasi-3D modeling of airborne TEM data by spatially constrained inversion. Geophysics 73: F105--F113. Vrugt JA, ter Braak CJF, Diks CGH, et al. (2009) Accelerating Markov chain Monte Carlo simulation by differential evolution with self-adaptive randomized subspace sampling. International Journal of Nonlinear Sciences and Numerical Simulation 10: 273--290. Watson DB, Doll WE, Gamey TJ, Sheehan JR, and Jardine PM (2005) Plume and lithologic profiling with surface resistivity and seismic tomography. Ground Water 43(2): 169--177. Watson KA and Barker RD (1999) Differentiating anisotropy and lateral effects using azimuthal resistivity offset Wenner soundings. Geophysics 64: 739--745. Waxman MH and Smits LJM (1968) Electrical conductivities in oil-bearing shaly sands. Transactions AIME 243: 107--122. Wilkinson PB, Meldrum PI, Chambers JE, Kuras O, and Ogilvy R (2006) Improved strategies for the automatic selection of optimized sets of electrical resistivity tomography measurement configurations. Geophysical Journal International 167: 1119--1126. Williams KH, Kemna A, Wilkins M, et al. (2009) Geophysical monitoring of microbial activity during stimulated subsurface bioremediation. Environmental Science and Technology (doi 10.1021/es900855j). Williams KH, Ntarlagiannis D, Slater LD, Dohnalkova A, Hubbard SS, and Banfield JF (2005) Geophysical imaging of stimulated microbial biomineralization. Environmental Science and Technology 39(19): 7592--7600. Wishart DN, Slater LD, and Gates A (2008) Fracture anisotropy characterization in crystalline bedrock using field-scale azimuthal self potential gradient. Journal of Hydrology 358: 35--45. Wishart DN, Slater LD, and Gates AE (2006) Self potential improves characterization of hydraulically-active fractures from azimuthal geoelectrical measurements. Geophysical Research Letters 33: L17314. Yaramanci U, Kemna A, and Vereecken H (2005) Emerging technologies in hydrogeophysics. In: Rubin Y and Hubbard S (eds.) Hydrogeophysics, ch. 16, pp. 467–486. Dordrecht: Springer. Zhdanov MS (2009) New advances in regularized inversion of gravity and electromagnetic data. Geophysical Prospecting 57: 463--478. Zonge KL and Hughes LJ (1991) Controlled source audio-frequency magnetotellurics. In: Nabighian MN (ed.) Electromagnetic Methods in Applied Geophysics, Vol. 2 – Applications, pp. 713--810. Tulsa, OK: Society of Exploration Geophysicists.

2.16 Hydrological Modeling DP Solomatine, UNESCO-IHE Institute for Water Education and Delft University of Technology, Delft, The Netherlands T Wagener, The Pennsylvania State University, University Park, PA, USA & 2011 Elsevier B.V. All rights reserved.

2.16.1 2.16.1.1 2.16.1.2 2.16.1.3 2.16.2 2.16.2.1 2.16.3 2.16.4 2.16.5 2.16.6 2.16.6.1 2.16.6.2 2.16.6.2.1 2.16.6.2.2 2.16.6.2.3 2.16.6.3 2.16.6.4 2.16.7 2.16.7.1 2.16.7.2 2.16.7.3 2.16.7.4 2.16.7.5 2.16.8 2.16.8.1 2.16.8.2 2.16.8.2.1 2.16.8.2.2 2.16.9 References

Introduction What Is a Model History of Hydrological Modeling The Modeling Process Classification of Hydrological Models Main types of Hydrological Models Conceptual Models Physically Based Models Parameter Estimation Data-Driven Models Introduction Technology of DDM Definitions Specifics of data partitioning in DDM Choice of the model variables Methods and Typical Applications DDM: Current Trends and Conclusions Analysis of Uncertainty in Hydrological Modeling Notion of Uncertainty Sources of Uncertainty Uncertainty Representation View at Uncertainty in Data-Driven and Statistical Modeling Uncertainty Analysis Methods Integration of Models Integration of Meteorological and Hydrological Models Integration of Physically Based and Data-Driven Models Error prediction models Integration of hydrological knowledge into DDM Future Issues in Hydrological Modeling

2.16.1 Introduction Hydrological models are simplified representations of the terrestrial hydrological cycle, and play an important role in many areas of hydrology, such as flood warning and management, agriculture, design of dams, climate change impact studies, etc. Hydrological models generally have one of two purposes: (1) to enable reasoning, that is, to formalize our scientific understanding of a hydrological system and/or (2) to provide (testable) predictions (usually outside our range of observations, short term vs. long term, or to simulate additional variables). For example, catchments are complex systems whose unique combinations of physical characteristics create specific hydrological response characteristics for each location (Beven, 2000). The ability to predict the hydrological response of such systems, especially stream flow, is fundamental for many research and operational studies. In this chapter, the main principles of and approaches to hydrological modeling are covered, both for simulation (process) models that are based on physical principles (conceptual

435 435 435 436 437 437 438 439 440 443 443 444 444 444 445 446 448 448 448 449 449 450 450 451 451 452 452 452 452 453

and physically based), and for data-driven models. Our intention is to provide a broad overview and to show current trends in hydrological modeling. The methods used in data-driven modeling (DDM) are covered in greater depth since they are probably less widely known to hydrological audiences.

2.16.1.1 What Is a Model A model can be defined as a simplified representation of a phenomenon or a process. It is typically characterized by a set of variables and by equations that describe the relationship between these variables. In the case of hydrology, a model represents the part of the terrestrial environmental system that controls the movement and storage of water. In general terms, a system can be defined as a collection of components or elements that are connected to facilitate the flow of information, matter, or energy. An example of a typical system considered in hydrological modeling is the watershed or catchment. The extent of the system is usually defined by the control volume or modeling domain, and the overall

435

436

Hydrological Modeling 

X0

Inputs Outputs State variables

I

X

O Model structure Initial state

X2 = F (X1, ,I1,) O2 = G (X1, ,I1,)

Parameters

Figure 1 Schematic of the main components of a dynamic mathematical model. I, inputs; O, outputs; X, state variables; X0, initial states; and y, parameters.

modeling objective in hydrology is generally to simulate the fluxes of energy, moisture, or other matter across the system boundaries (i.e., the system inputs and outputs). Variables or state variables are time varying and (space/time) averaged quantities of mass/energy/information stored in the system. An example would be soil moisture content or the discharge in a stream [L3/T]. Parameters describe (usually time invariant) properties of the specific system under study inside the model equations. Examples of parameters are hydraulic conductivity [L/T] or soil storage capacity [L]. A dynamic mathematical model has certain typical elements that are discussed here briefly for consistency in language (Figure 1). Main components include one or more inputs I (e.g., precipitation and temperature), one or more state variables X (e.g., soil moisture or groundwater content), and one or more model outputs O (e.g., stream flow or actual evapotranspiration). In addition, a model typically requires the definitions of initial states X0 (e.g., is the catchment wet or dry at the beginning of the simulation) and/or the model parameters y (e.g., soil hydraulic conductivity, surface roughness, and soil moisture storage capacity). Hydrological models (and most environmental models in general) are typically based on certain assumptions that make them different from other types of models. Typical assumptions that we make in the context of hydrological modeling include the assumption of universality (i.e., a model can represent different but similar systems) and the assumption of physical realism (i.e., state variables and parameters of the model have a real meaning in the physical world; Wagener and Gupta, 2005). The fact that we are dealing with real-world environmental systems also carries certain problems with it when we are building models. Following Beven (2009), these problems include the fact that it is often difficult to (1) make measurements at the scale at which we want to model; (2) define the boundary conditions for time-dependent processes; (3) define the initial conditions; and (4) define the physical, chemical, and biological characteristics of the modeling domain.

2.16.1.2 History of Hydrological Modeling Hydrological models applied at the catchment scale originated as simple mathematical representations of the input-response behavior of catchment-scale environmental systems through

parsimonious models such as the unit hydrograph (for flow routing) (e.g., Dooge, 1959) and the rational formula (for excess rainfall calculation) (e.g., Dooge, 1957) as part of engineering hydrology. Such single-purpose event-scale models are still widely used to estimate design variables or to predict floods. These early approaches formed a basis for the generation of more complete, but spatially lumped, representations of the terrestrial hydrological cycle, such as the Stanford Watershed model in the 1960s (which formed the basis for the currently widely used Sacramento model (Burnash, 1995)). This advancement enabled the continuous time representation of the rainfall–runoff relationship, and models of this type are still at the heart of many operational forecasting systems throughout the world. While the general equations of models (e.g., the Sacramento model) are based on conceptualizing plot (or smaller) scale hydrological processes, their spatially lumped application at the catchment scale means that parameters have to be calibrated using observations of rainfall– runoff behavior of the system under study. Interest in predicting land-use change leads to the development of more spatially explicit representations of the physics (to the best of our understanding) underlying the hydrological system in form of the Systeme Hydrologique Europeen (SHE) model in the 1980s (Abbott et al., 1986). The latter is an example of a group of highly complex process-based models whose development was driven by the hope that their parameters could be directly estimated from observable physical watershed characteristics without the need for model calibration on observed stream flow data, thus enabling the assessment of land cover change impacts (Ewen and Parkin, 1996; Dunn and Ferrier, 1999). At that time, these models were severely constrained by our lack of computational power – a constraint that decreases in its severity with increases in computational resources with each passing year. Increasingly available high-performance computing enables us to explore the behavior of highly complex models in new ways (Tang et al., 2007; van Werkhoven et al., 2008). This advancement in computer power went hand in hand with new strategies for process-based models, for example, the use of triangular irregular networks (TINs) to vary the spatial resolution throughout the model domain, that have been put forward in recent years; however, more testing is required to assess whether previous limitations of physically based models have yet been overcome (e.g., the lack of full coupling of processes or their calibration needs) (e.g., Reggiani et al., 1998, 1999, 2000, 2001; Panday and Huyakorn, 2004; Qu and Duffy, 2007; Kollet and Maxwell, 2006, 2008).

2.16.1.3 The Modeling Process The modeling process, that is, how we build and use models is discussed in this section. For ease of discussion, the process is divided into two components. The first component is the model-building process (i.e., how does a model come about), whereas the second component focuses on the modeling protocol (i.e., a procedure to use the model for both operational and research studies). The model-building process requires (at least implicitly) that the modeler considers four different stages of the model

Hydrological Modeling

(see also Beven, 2000). The first stage is the perceptual model. This model is based on the understanding of the system in the modeler’s head due to both the interaction with the system and the modeler’s experience. It will, generally, not be formalized on paper or in any other way. This perceptual model forms the basis of the conceptual model. This conceptual model is a formalization of the perceptual model through the definition of system boundaries, inputs–states–outputs, connections of system components, etc. It is not to be mistaken with the conceptual type of models discussed later. Once a suitable conceptual model has been derived, it has to be translated into mathematical form. The mathematical model formulates the conceptual model in the form of input (–state)–output equations. Finally, the mathematical model has to be implemented as computer code so that the equation can be solved in a computational model. Once a suitable model has been built or selected from existing computer codes, a modeling protocol is used to apply this model (Wagener and McIntyre, 2007). Modeling protocols can vary widely, but generally contain some or most of the elements discussed below (Figure 2). A modeling protocol – at its simplest level – can be divided into model identification and model evaluation parts. The model identification part mainly focuses on identifying appropriate parameters (one set or many parameter sets if uncertainty in the identification process is considered), while the latter focuses on understanding the behavior and performance of the model. The starting point of the model identification part should be a detailed analysis of the data available. Beven (2000) provided suggestions on how to assess the quality of data in

the context of hydrological modeling. This is followed by the model selection or building process. The model-building process has already been outlined previously. In many cases, it is likely that an existing model will be selected though, either because the modeler has extensive experience with a particular model or because he/she has applied a model to a similar hydrological system with success in the past. The universality of models, as discussed above, implies that a typical hydrological model can be applied to a range of systems as long as the basic physical processes of the system are represented within the model. Model choice might also vary with the intended modeling purpose, which often defines the required spatio-temporal resolution and thus the degree of detail with which the system has to be modeled. Once a model structure has been selected, parameter estimation has to be performed. Parameters, as defined above, reflect the local physical characteristics of the system. Parameters are generally derived either through a process of calibration or by using a priori information, for example, of soil or vegetation characteristics. For calibration, it is necessary to assess how closely simulated and observed (if available) output time series match. This is usually done by the use of an objective function (sometimes also called loss function or cost function), that is, a measure based on the aggregated differences between observed and simulated variables (called residuals). The choice of objective function is generally closely coupled with the intended purpose of the modeling study. Sometimes this problem is posed as a multiobjective optimization problem. Methods for calibration (parameter estimation) are covered later in Section 2.16.5. Further, the model

Model identification Data analysis Model selection/building Boundary conditions Objective function defination Revise model

Parameter estimation Order can vary

Validation/verification Sensitivity analysis Further (diagnostic) evaluation Uncertainty analysis Model prediction Model evaluation

Figure 2 Schematic representation of a typical modeling protocol.

437

Data assimilation

438

Hydrological Modeling

should be evaluated with respect to whether it provides the right result for the right reason. Parameter estimation (calibration) is followed by the model evaluation, including validation (checking model performance on an unseen data set, thus imitating model operation), sensitivity, and uncertainty analysis. A comprehensive framework for model evaluation (termed diagnostic evaluation) is proposed by Gupta et al. (2008). One tool often used in such an evaluation is sensitivity analysis, which is the study of how variability or uncertainty in different factors (including parameters, inputs, and initial states) impacts the model output. Such an analysis is generally used either to assess the relative importance of model parameters in controlling the model output or to understand the relative distributions of uncertainty from the different factors. It can therefore be part of the model identification as well as the model evaluation component of the modeling protocol. The subsequent step of uncertainty analysis – the quantification of the uncertainty present in the model – is increasingly popular. It usually includes the propagation of the uncertainty into the model output so that it can be considered in subsequent decision making (see Section 2.16.7). When a model is put into operation, the data progressively collected can be used to update (improve) the model parameters, state variables, and/or model predictions (outputs), and this process is referred to as data assimilation. One aspect needs mentioning here. Due to the lack of information about the modeled process, a modeler may decide not to try to build unique (the most accurate) model, but rather consider many equally acceptable model parametrizations. Such reasoning has led to a Monte-Carlo-like method of uncertainty analysis called Generalised Likelihood Uncertainty Estimator (GLUE) (Beven and Binley, 1992), and to research into the development of the (weighted) ensemble of models, or multimodels (see e.g., Georgakakos et al., 2004).

2.16.2 Classification of Hydrological Models 2.16.2.1 Main types of Hydrological Models A vast number of hydrological model structures has been developed and implemented in computer code over the last few decades (see, e.g., Todini (1988) for a historical review of rainfall–runoff modeling). It is therefore helpful to classify these structures for an easier understanding of the discussion. Many authors present classification schemes for hydrological models (see, e.g., Clarke, 1973; Todini, 1988; Chow et al., 1988; Wheater et al., 1993; Singh, 1995b; and Refsgaard, 1996). The classification schemes are generally based on the following criteria: (1) the extent of physical principles that are applied in the model structure and (2) the treatment of the model inputs and parameters as a function of space and time. According to the first criterion (i.e., physical process description), a rainfall–runoff model can be attributed to two categories: deterministic and stochastic (see Figure 3). A deterministic model does not consider randomness; a given input always produces the same output. A stochastic model has outputs that are at least partially random. Deterministic models can be classified based on whether the model represents a lumped or distributed description of the considered catchment area (i.e., second criterion) and whether the description of the hydrological processes is empirical, conceptual, or more physically based (Refsgaard, 1996). With respect to deterministic models, we will distinguish three classes: (1) data-driven (also called data-based, metric, empirical, or black box models), (2) conceptual (also called parametric, explicit soil moisture accounting or gray box models), and (3) physically based (also called physics-based, mechanistic, or white box models) models. The two latter classes are sometimes referred to as simulation (or process) models. Figure 4 provides some guidelines on estimation of structure and parameters for various types of deterministic models. Note that the distinction between deterministic and stochastic models is not clear-cut. In many modeling studies, it is

Rainfall−runoff model

Deterministic

Randomness

Stochastic

Simulation models Data-driven

Conceptual

Physically based Physical insight

Data requirement Figure 3 Classification of hydrological models based on physical processes. Adapted from Refsgaard JC (1996) Terminology, modelling protocol and classification of hydrological model codes. In: Abbott MB and Refsgaard JC (eds.) Distributed Hydrological Modelling, pp. 17–39. Dordrecht: Kluwer.

Hydrological Modeling

439

Basis for estimation of Classification

Data

Data-based models

Structure

Parameters

Data

Data

Modeling strategy

Top-down

Black box

Conceptual models

Data

A priori Gray box

Physics

Physics-based models

A priori

A priori

Bottom-up

White box Figure 4 Estimation of structure and parameters for various types of deterministic models.

assumed that the modeled variables are not deterministic, but still more developed apparatus of deterministic modeling is used. To account for stochasticity, additional uncertainty analysis is conducted assuming probability distributions for at least some of the variables and parameters involved.

2.16.3 Conceptual Models Conceptual modeling uses simplified descriptions of hydrological processes. Such models use storage elements as the main building component. These stores are filled through fluxes such as rainfall, infiltration, or percolation, and emptied through processes such as evapotranspiration, runoff, and drainage. Conceptual models generally have a structure that is specified a priori by the modeler, that is, it is not derived from the observed rainfall–runoff data. In contrast to empirical models, the structure is defined by the modeler’s understanding of the hydrological system. However, conceptual models still rely on observed time series of system output, typically stream flow, to derive the values of their parameters during the calibration process. The parameters describe aspects such as the size of storage elements or the distribution of flow between them. A number of real-world processes are usually aggregated (in space and time) into a single parameter, which means that this parameter can therefore often not be derived directly from field measurements. Conceptual models make up the vast majority of models used in practical applications. Most conceptual models consider the catchment as a single homogeneous unit. However, one common approach to consider spatial variability is the segmentation of the catchment into smaller subcatchments, the so-called semidistributed approach.

One typical example of a conceptual model – Hydrologiska Byra˚ns Vattenbalansavdelning (HBV) – (Bergstro¨m, 1976) as rainfall–runoff model is given below. The HBV model was developed at the Swedish Meteorological and Hydrological Institute (Hydrological Bureau Water balance section). The model was originally developed for Scandinavian catchments, but has been applied in more than 30 countries all over the world (Lindstro¨m et al., 1997). A schematic diagram of the HBV model (Lindstro¨m et al., 1997) is shown in Figure 5. The model of one catchment comprises subroutines for snow accumulation and melt, soil moisture accounting procedure, routines for runoff generation, and a simple routing procedure. The soil moisture accounting routine computes the proportion of snowmelt or rainfall P (mm h1 or mm d1) that reaches the soil surface, which is ultimately converted to runoff. If the soil is dry (i.e., small value of SM/CF), the recharge R, which subsequently becomes runoff, is small as a major portion of the effective precipitation P is used to increase the soil moisture. Whereas if the soil is wet, the major portion of P is available to increase the storage in the upper zone. The runoff generation routine transforms excess water R from the soil moisture zone to runoff. The routine consists of two conceptual reservoirs. The upper reservoir is a nonlinear reservoir whose outflow simulates the direct runoff component from the upper soil zone, while the lower one is a linear reservoir whose outflow simulates the base flow component of the runoff. The total runoff Q is computed as the sum of the outflows from the upper and the lower reservoirs. The total runoff is then smoothed using a triangular transformation function. Input data are observations of precipitation and air temperature, and estimates of potential evapotranspiration. The

440

Hydrological Modeling

SF − snow RF − rain EA − evapotranspiration SP − snow cover

SF

P − effective precipitation

RF

R − recharge

EA

SM − soil moisture CF − capillary transport

SP

UZ − storage in upper reservoir PERC − percolation

SM

LZ − storage in lower reservoir R

Q0 − fast runoff component

CF Q0

Q1 − slow runoff component Q − total runoff

UZ PERC LZ

Q1

Q = Q0 + Q1 Transform function

Figure 5 Schematic representation of the HBV-96 model with routines for snow, soil, and runoff response. Modified from Lindstro¨m G, Johansson B, Persson M, Gardelin M, and Bergstro¨m S (1997) Development and test of the distributed HBV-96 hydrological model. Journal of Hydrology 201: 272– 228.

time step is usually 1 day, but it is possible to use shorter time steps. The evaporation values used are normally monthly averages, although it is possible to use the daily values. Air temperature data are used for calculations of snow accumulation and melt. It can also be used to adjust potential evaporation when the temperature deviates from normal values, or to calculate potential evaporation. Note that the software IHMS-HBV allows for linking several lumped models and thus making it possible to build separate models for sub-basins, which are integrated, so that the overall model is the semi-distributed model. The HBV model is an example of a typical lumped conceptual model. Other examples of such models differ in the details of describing the catchment hydrology. The following examples can be mentioned: Sugawara’s tank model (Sugawara, 1995), Sacramento model (Burnash, 1995), Xinanjiang model (Zhao and Liu, 1995), and Tracer Aided Catchment (TAC) model (Uhlenbrook and Leibundgut, 2002).

2.16.4 Physically Based Models Physically based models (e.g., Freeze and Harlan, 1969; Beven, 1996, 1989, 2002; Abbott et al., 1986; Calver, 1988) use much more detailed and rigorous representations of physical processes and are based on the laws of conservation of mass, momentum, and energy. They became practically applicable in 1980s, as a result of improvements in computer power. The hope was that the degree of physical realism on which these models are based would be sufficient to relate their parameters, such as soil moisture characteristic and unsaturated zone hydraulic conductivity functions for subsurface flow or

friction coefficients for surface flow, to physical characteristics of the catchment (Todini, 1988), thus eliminating the need for model calibration. However, mechanistic models suffer from high data demand, scale-related problems (e.g., the measurement scales differ from the simulation model (parameter) scales), and from over-parametrization (Beven, 1989). One consequence of the problems of scale is that (at least not all of) the model parameters cannot be derived through measurements; physically based models structures, therefore, still require calibration, usually of a few key parameters (Calver, 1988; Refsgaard, 1997; Madsen and Jacobsen, 2001). The expectation that these models could be applied to ungauged catchments has, therefore, not yet been fulfilled (Parkin et al., 1996; Refsgaard and Knudsen, 1996). They are typically rather applied in a way that is similar to conceptual models (Beven, 1989), thus demanding continued research into new approaches to merge these models with data. Physically based models often use spatial discretizations based on grids, triangular irregular networks, or some type of hydrologic response unit (e.g., Uhlenbrook et al., 2004). A typical model of this kind is, for example, a physically based model based on triangular irregular networks – the Penn State Integrated Hydrologic Model (PIHM) (Qu and Duffy, 2007); its simplified structure is presented in Figure 6. Such models are therefore particularly appropriate when a high level of spatial detail is important, for example, to estimate local levels of soil erosion or the extent of inundated areas (Refsgaard and Abbott, 1996). However, if the main interest simply lies in the estimation of stream flow at the catchment scale, then simpler conceptual or data-driven models often perform well and the high complexity of physically based models is not required (e.g., Loague and Freeze, 1985; Refsgaard and Knudsen, 1996). Regarding the results of

Hydrological Modeling

441

Solar radiation Interception

Precipitation

Transpiration Overland flow

Capillary lift

Infiltration Recharge

Groundwater flow

Bedrock Saturated Unsaturated zone zone

Evaporation Precipitation

Evaporation

Overland weir flow Lateral flow Downstream flow

Infiltration

Unsaturated zone

Recharge

Saturated zone Groundwater flow Bedrock Figure 6 Schematic representation of the PIHM model, an example of a TIN-based physically based hydrological model (Qu and Duffy, 2004).

a comprehensive experiment to compare lumped and distributed models, the reader is referred to Reed et al. (2004). This experiment has shown that due to difficulties in calibrating distributed models, in many cases, conceptual models are in fact more accurate in reproducing the resulting catchment stream flow than the distributed ones.

2.16.5 Parameter Estimation Many, if not most, rainfall–runoff model structures currently used to simulate the continuous hydrological response can be classified as conceptual, if this classification is based on two criteria (Wheater et al., 1993): (1) the model structure is specified prior to any modeling being undertaken and (2) (at least some of) the model parameters do not have a direct

physical interpretation, in the sense of being independently measurable, and have to be estimated through calibration against observed data. Calibration is a process of parameter adjustment (automatic or manual), until catchment and model behavior show a sufficiently (to be specified by the hydrologist) high degree of similarity. The similarity is usually judged by one or more objective functions accompanied by visual inspection of observed and calculated hydrographs (Gupta et al., 2005). The choice of such objective functions has itself been the subject of extensive research over many years. Traditionally, measures based on the mean squared error (MSE) criterion were used, for example, root mean squared error (RMSE) or Nash–Sutcliff efficiency (NSE). Appearance of the squared errors in many formulations is the result of an assumption of the normality (Gaussian distribution) of model errors, and

442

Hydrological Modeling

using the principle of maximum likelihood to derive the error function. Calibration includes the process of finding a set of parameters providing the minimum of RMSE or the maximum value of NSE. For more information on the formulations of these and other error function, the reader is referred to, for example, Gupta et al. (1998). In a recent paper, Gupta et al. (2009) showed certain deficiencies of MSE-based objective functions and suggested possible remedies. Often a single measure may not be enough to capture all the aspects of the system response that the model is supposed to reproduce, and several criteria (objective functions) have to be considered simultaneously so that multiobjective optimization algorithms have to be used. Examples of such multiple objectives include the RMSE calculated separately on low and high flows, timing errors, and error in reproducing the water balance. The models constituting the Pareto in criteria space should be seen as the best models, since there are no models better than these on all criteria. If a single model is to be selected from this set, it is done either by a decision maker (who would use some additional criteria that are difficult to formalize), or by measuring the distance of the models to the ideal point in criteria space, or by using the (weighted) sum of objective functions values. This section covers single-objective optimization; for the use of multiobjective methods, the reader is directed to the papers by Gupta et al. (1998), Khu and Madsen (2005), and Tang et al. (2006) (with the subsequent discussion). Hydrological model structures of the continuous watershed response (mainly stream flow) became feasible in the 1960s. They were usually relatively simple lumped, conceptual mathematical representations of the (perceived to be important) hydrological processes, with little (if any) consideration of issues such as identifiability of the parameters or information content of the watershed response observations. It became quickly apparent that the parameters of such models could not be directly estimated through measurements in the field, and that some sort of adjustment (fine-tuning) of the parameters was required to match simulated system responses with observations (e.g., Dawdy and O’Donnell 1965). Adjustment approaches were initially based on manual perturbation of the parameter values and visual inspection of the similarity between simulated and observed time series. Over the years, a variety of manual calibration procedures have been developed, some having reached very high levels of sophistication allowing hydrologists to achieve very good performing and hydrologically realistic model parameters and predictions, that is, a well-calibrated model (Harlin, 1991; Burnash, 1995). This hydrological realism is still a problem for most automated procedures as discussed in van Werkhoven et al. (2008). Necessary conditions for a hydrological model to be well calibrated are that it exhibits (at least) the following three characteristics (Wagener et al. 2003; Gupta et al. 2005):

1. the input–state–output behavior of the model is consistent with the measurements of watershed behavior; 2. the model predictions are accurate (i.e., they have negligible bias) and precise (i.e., the prediction uncertainty is relatively small); and

3. the model structure and behavior are consistent with a current hydrological understanding of reality. This last characteristic is often ignored in operational settings, where the focus is generally on useful rather than realistic models. This will be an adequate approach in many cases, but will eventually lead to limitations of potential model uses. This problem is exemplified in the current attempts to modeling watershed residence times and flow paths (McDonnell, 2003). This aspect of the hydrologic system, though often not crucial for reliable quantitative flow predictions, is however relevant for many of today’s environmental problems, but cannot be simulated by many of the currently available models. The high number of nonlinearly interacting parameters present in most hydrological models makes manual calibration a very labor-intensive and a difficult process, requiring considerable experience. This experience is time consuming to acquire and cannot be easily transferred from one hydrologist to the next. In addition, manual calibration does not formally incorporate an analysis of uncertainty, as is required in a modern decision-making context. The obvious advantages of computer-based automatic calibration procedures began to spark interest in such approaches as soon as computers became more easily available for research. In automatic calibration, the ability of a parameter set to reproduce the observed system response is measured (summarized) by means of an objective function (also sometimes called loss or cost function). As discussed above, this objective function is an aggregated measure of the residuals, that is, the differences between observed and simulated responses at each time step. An important early example of automatic calibration is the dissertation work by Ibbitt (1970) in which a variety of automated approaches were applied to several watershed models of varying complexity (see also Ibbitt and O’Donnell, 1971). The approaches were mainly based on local-search optimization techniques, that is, the methods that start from a selected initial point in the parameter space and then walk through it, following some predefined rule system, to iteratively search for parameter sets that yield progressively better objective function values. Ibbitt (1970) found that it is difficult to conclude when the best parameter set has been found, because the result depends both on the chosen method and on the initial starting parameter set. The application of local-search calibration approaches to all but the most simple watershed models has been largely unsuccessful. In reflection of this, Johnston and Pilgrim (1976) reported the failure of their 2-year quest to find an optimal parameter set for a typical conceptual rainfall–runoff (RR) model. Their honesty in reporting this failure ultimately led to a paradigm shift as researchers started to look closely at the possible reasons for this lack of success. The difficulty of the task at hand, in fact, only became clear in the early 1990s when Duan et al. (1992) conducted a detailed study of the characteristics of the response surface that any search algorithm has to explore. Their studies showed that the specific characteristics of the response surface, that is, the (n þ 1)-dimensional space of n model parameters and an objective function, of hydrological models give rise to conditions that make it extremely difficult for local optimization strategies to be successful. They listed the following characteristics commonly associated with the response surface of a

Hydrological Modeling

typical hydrological model:

• • • • •

it contains more than one main region of attraction; each region of attraction contains many local optima; it is rough with discontinuous derivatives; it is flat in many regions, particularly in the vicinity of the optimum, with significantly different parameter sensitivities; and its shape includes long and curved ridges.

Concluding that optimization strategies need to be powerful enough to overcome the search difficulties presented by these response surface characteristics, Duan et al. (1992) developed the shuffled complex evolution (SCE-UA) global optimization method (UA, University of Arizona). The SCE-UA algorithm has since been proved to be highly reliable in locating the optimum (where one exists) on the response surfaces of typical hydrological models. However, in a follow-up paper, Sorooshian et al. (1993) used SCE-UA to show that several different parameter combinations of the relatively complex Sacramento model (13 free parameters) could be found which produced essentially identical objective function values, thereby indicating that not all of the parameter uncertainty can be resolved through an efficient global optimizer (see discussion in Wagener and Gupta (2005)). Similar observations of multiple parameter combinations producing similar performances have also been made by others (e.g., Binley and Beven, 1991; Beven and Binley, 1992; Spear, 1995; Young et al., 1998; Wagener et al., 2003). Part of this problem had been attributed to overly complex models for the information content of the system response data available, usually stream flow (e.g., Young, 1992, 1998). It is worth mentioning that practically any direct search optimization algorithm can be used for model calibration. The reason of using direct search (i.e., the search based purely on calculation of the objective function values for different points in the search space) is that for most calibration problems computation of the objective function gradients is not possible, so the efficient gradient-based search cannot be used. (Another name for this class of algorithms is global optimization algorithms since they are focused on finding the global minimum rather than a local one.) For example, in many studies a popular genetic algorithm (GA) is used. If a model is simple and fast running, then it is really not important how efficient the optimization algorithm is. Here, efficiency is measured by the number of the model runs needed by an optimization algorithm to find a more-or-less accurate estimate of the parameter vector leading to the minimum value of the model error. However, if a model is computationally complex, as is the case for physically based and distributed models, efficiency of the optimization algorithm used becomes an issue. With this in mind, the so-called adaptive cluster-covering algorithm (ACCO) was developed (Solomatine, 1995; Solomatine et al., 1999, 2001), and it was shown that on a number of calibration problems it is more efficient than GA and several other algorithms. A large number of other algorithms has been applied to hydrological models, including a multialgorithm genetically adaptive method (AMALGAM) (Vrugt and Robinson, 2007) and epsilon-NSGA-II (NSGA, nondominated sorting genetic

443

algorithm; Tang et al., 2006), and different algorithms have come out as most effective or most efficient depending on the study. A range of algorithms for model calibration can be obtained from the Hydroarchive website. As mentioned above (Figure 2), once parameters are estimated, the model has to be validated, that is, the degree to which a model is an accurate representation of the modeled process has to be determined. Case of ungauged basins. A different problem has to be solved in the case of the so-called ungauged basins, that is, watersheds for which none or insufficiently long observations of the hydrological response variable of interest (usually stream flow) are available. The above-discussed strategy of model calibration cannot be used under those conditions. Early attempts to model ungauged catchments simply used the parameter values derived for neighboring catchments where stream flow data were available, that is, a geographical proximity approach (e.g., Mosley, 1981; Vandewiele and Elias, 1995). However, this seems to be insufficient since nearby catchments can even be very different with respect to their hydrological behavior (Post et al., 1998; Beven, 2000). Others propose the use of parameter estimates directly derived from, among others, soil properties such as porosity, field capacity, and wilting point (to derive model storage capacity parameters); percentage forest cover (evapotranspiration parameters); or hydraulic conductivities and channel densities (time constants) (e.g., Koren et al., 2000; Duan et al., 2001; Atkinson et al., 2002). The main problem here is that the scale at which the measurements are made (often from small soil samples) is different from the scale at which the model equations are derived (often laboratory scale) and at which the model is usually applied (catchment scale). The conceptual model parameters represent the effective characteristics of the integrated (heterogeneous) catchment system (e.g., including preferential flow), which are unlikely to be easily captured using small-scale measurements since there is generally no theory that allows the estimation of the effective values within different parts of a heterogeneous flow domain from a limited number of small-scale or laboratory measurements (Beven, 2000). It seems unlikely that conceptual model parameters, which describe an integrated catchment response, usually aggregating significant heterogeneity (including the effect of preferential flow paths, different soil and vegetation types, etc.), can be derived from catchment properties that do not consider all influences on water flow through the catchment. Further fine-tuning of these estimates using locally observed flow data is needed because the physical information available to estimate a priori parameters is not adequate to define local physical properties of individual basins for accurate hydrological forecasts (Duan et al., 2001). However, useful initial values might be derived in this way (Koren et al., 2000). The advantages of this approach are that the assumed physical basis of the parameters is preserved and (physical) parameter dependence can be accounted for, as shown by Koren et al. (2000). Probably, the most common apprnoach to ungauged modeling is to relate model parameters and catchment characteristics in a statistical manner (e.g., Jakeman et al., 1992; Sefton et al., 1995; Post et al., 1998; Sefton and Howarth, 1998; Abdullah and Lettenmaier, 1997; Wagener et al., 2004;

444

Hydrological Modeling

Merz and Bloschl, 2004; Lamb and Kay, 2004; Seibert, 1999; Lamb et al., 2000; Post and Jakeman, 1996; Fernandez et al., 2000), assuming that the uniqueness of each catchment can be captured in a distinctive combination of catchment characteristics. The basic methodology is to calibrate a specific model structure, here called the local model structure, to as large a number of (gauged) catchments as possible and derive statistical (regression) relationships between (local) model parameters and catchment characteristics. These statistical relationships, here called regional models, and the measurable properties of the ungauged catchment can then be used to derive estimates of the (local) model parameters. This procedure is usually referred to as regionalization or spatial generalization (e.g., Lamb and Calver, 2002). While this approach has been widely applied, it still does not constrain existing uncertainty sufficiently in many cases (Wagener and Wheater, 2006). Recent approaches also used regionalized information about stream flow characteristics to further reduce this uncertainty (Yadav et al., 2007; Zhang et al., 2008). It seems as if the most promising strategies for the future lie in combining as much information as possible to reduce predictive uncertainty, rather than relying on a single approach.

2.16.6 Data-Driven Models 2.16.6.1 Introduction Along with the physically based and conceptual models, the empirical models based on observations (experience) are also popular. Such models involve mathematical equations that have been assessed not from the physical process in the catchment but from analysis of data – concurrent input and output time series. Typical examples here are the unit hydrograph method and various statistical models – for example, linear regression, multilinear, ARIMA, etc. During the last decade, the area of empirical modeling received an important boost due to developments in the area of machine learning (ML). It can be said that it now entered a new phase and deserves a special name – DDM. DDM is based on the analysis of all the data characterizing the system under study. A model can then be defined based on connections between the system state variables (input, internal and output variables) with only a limited number of assumptions about the physical behavior of the system. The methods used nowadays can go much further than the ones used in conventional empirical modeling: they allow for solving prediction problems, reconstructing highly nonlinear functions, performing classification, grouping of data, and building rule-based systems. It is worth mentioning that among some hydrologists there is still a certain skepticism about the use of DDM. In their opinion, such models do not relate to physical principles and mathematical reasoning, and view building models from data sets as a purely computational exercise. This is true, and indeed DDM cannot be a replacement of process-based modeling, but should be used in situations where data-driven models are capable of generating improved forecasts of hydrological variables. There are cases where the traditional statistical models (typically linear regression or ARIMA-class models) are

accurate enough, and there is no need of using sophisticated methods of ML. Some of the concerns of this nature are discussed, for example, by Gaume and Gosset (2003), See et al. (2007), Han et al. (2007), and Abrahart et al., 2008. Abrahart and See (2007) also addressed some of these problems, however, demonstrated that the existing nonlinear hydrological relationships, which are so important when building flow forecasting models, are effectively captured by a neural network, the most widely used DDM method. In this respect, positioning of data-driven models is important: they should be seen as complementary to process-based simulation models; they cannot explain reality but could be effective predictive tools.

2.16.6.2 Technology of DDM 2.16.6.2.1 Definitions One may identify several fields that contribute to DDM: statistical methods, ML, soft computing (SC), computational intelligence (CI), data mining (DM), and knowledge discovery in databases (KDDs). ML is the area concentrating on the theoretical foundations of learning from data and it can be said that it is the major supplier of methods for DDM. SC is emerging from fuzzy logic, but many authors attribute to it many other techniques as well. CI incorporates two areas of ML (neural networks and fuzzy systems), and, additionally, evolutionary computing that, however, can be better attributed to the field of optimization than to ML. DM and KDDs used, in fact, the methods of ML and are focused typically at large databases being associated with banking, financial services, and customer resources management. DDM can thus be considered as an approach to modeling that focuses on using the ML methods in building models that would complement or replace the physically based models. The term modeling stresses the fact that this activity is close in its objectives to traditional approaches to modeling, and follows the steps traditionally accepted in (hydrological) modeling. Examples of the most common methods used in data-driven hydrological modeling are linear regression, ARIMA, artificial neural networks (ANNs), and fuzzy rulebased systems (FRBSs). Such positioning of DDM links to learning which incorporates determining the so far unknown mappings (or dependencies) between a system’s inputs and its outputs from the available data (Mitchell, 1997). By data, we understand the known samples (data vectors) that are combinations of inputs and corresponding outputs. As such, a dependency (mapping or model) is discovered (induced), which can be used to predict (or effectively deduce) the future system’s outputs from the known input values. By data, we usually understand a set K of examples (or instances) represented by duple /xk, ykS, where k ¼ 1,y, K, vector xk ¼ {x1,y, xn}k, vector yk ¼ {y1,y, ym}k, n ¼ number of inputs, and m ¼ number of outputs. The process of building a function (or mapping, or model) y ¼ f (x) is called training. If only one output is considered, then m ¼ 1. (In relation to hydrological and hydraulic models, training can be seen as calibration.) In the context of hydrological modeling, the inputs and outputs are typically real numbers (xk, ykAZRn), so the main

Hydrological Modeling

learning problem solved in hydrological modeling is numerical prediction (regression). Note that the problems of clustering and classification are rare but there are examples of it as well (see, e.g., Hall and Minns, 1999; Hannah et al., 2000; Harris et al., 2000). As already mentioned, the process of building a data-driven model follows general principles adopted in modeling: study the problem, collect data, select model structure, build the model, test the model, and (possibly) iterate. There is, however, a difference with physically based modeling: in DDM not only the model parameters but also the model structure are often subject to optimization. Typically, simple (or parsimonious) models are valued (as simple as possible, but no simpler). An example of such parsimonious model could be a linear regression model versus a nonlinear one, or a neural network with the small number of hidden nodes. Such models would automatically emerge if the so-called regularization is used: the objective function representing the overall model performance includes not only the model error term, but also a term that increases in value with the increase of model complexity represented, for example, by the number of terms in the equation, or the number of hidden nodes in a neural network. If there is a need to build a simple replica of a sophisticated physically based hydrological model, DDM can be used as well: such models are called surrogate, emulation, or metamodels (see, e.g., Solomatine and Torres, 1996; Khu et al., 2004). They can be used as fast-working approximations of complex models when speed is important, for example, in solving the optimization or calibration problems.

2.16.6.2.2 Specifics of data partitioning in DDM Obviously, data analysis and preparation play an important role in DDM. These steps are considered standard by the experts in ML but are not always given proper attention by hydrologists building or using such models. Three data sets for training, cross-validation, and testing. Once the model is trained (but before it is put into operation), it has to be tested (or verified) by calculating the model error (e.g., RMSE) using the test (or verification) data set. However, during training often there is a need to conduct tests of the model that is being built, so yet another data set is needed – the crossvalidation set. This set serves as the representative of the test set. As a model gradually improves as a result of the training process, the error on the training data will be gradually decreasing. The cross-validation error will also be first decreasing, but as the model starts to reproduce the training data set better and better, this error will start to increase (effect of over fitting). This typically means that the training should be stopped when the error on cross-validation data set starts to increase. If these principles are respected, then there is a hope that the model will generalize well, that is, its prediction error on unseen data will be small. (Note that the test data should be used only to test the final model, but not to improve (optimize) the model.) One may see that this procedure is more complex than the standard procedure of the hydrological model calibration – when no data are allocated for cross-validation, and, worse, often the whole data set is used to calibrate the model.

445

Note that in an important class of ML models – support vector machines (SVMs) – a different approach is taken: it is to build the model that would have the best generalization ability possible without relying explicitly on the cross-validation set (Vapnik, 1998). In connection to the issues covered above, there are two common pitfalls, especially characteristic of DDM applications where time series are involved, that are worth mentioning here. The desired properties of the three data sets. It is desired that the three sets are statistically similar. Ideally, this could be automatically ensured by the fact that data sets are sufficiently large and sampled from the same distribution (typical assumption in machine and statistical learning). However, in reality of hydrological modeling, such situations are rare, so normally a modeler should try to ensure at least some similarity in the distributions, or, at least, similar ranges, mean and variance. Statistical similarity can be achieved by careful selection of examples for each data set, by random sampling data from the whole data set, or employing an optimization procedure resulting in the sets with predefined properties (Bowden et al., 2002). One of the approaches is to use the 10-fold validation method when a model is built 10 times, trained each time on 9/10th of the whole set of available data and validated on 1/10th (number of runs is not necessarily 10). A version of this method is the leave-one-out method when K models are built using K  1 examples and not using one (every time different). The modeler is left with 10 or K trained models, so the resulting model to be used is either one of these models, or an ensemble of all the built models, possibly with the weighted outputs. Strictly speaking, for generation of the statistically similar training data sets for building a series of similar but different models, one should typically rely on the well-developed statistical (re)sampling methods such as bootstrap originated by Tibshirani in the 1970s (see Efron and Tibshirani, 1993) where (in its basic form) K data are randomly selected from K original data. For many hydrologists, there could be a visualization (or even a psychological) problem. If one of these procedures is followed, the data will not be always contiguous: it would not be possible to visualize a hydrograph when the model is fed with the test set. There is nothing wrong with such a model if the time structure of all the data sets is preserved. Such models, however, may be rejected by practitioners, since they are so different from the traditional physically based models that always generate contiguous time series. A possible solution here is to consider the hydrological events (i.e., contiguous blocks of data), to group the data accordingly, and to try to ensure the presence of statistically similar events in all the three data sets. This is all possible of course, if there is enough data. In the situations when the data set is not large enough to allow for building all three sets of substantial size, modelers could be forced not to build cross-validation set at all with the hope that the model trained on training set would perform well on the test set as well. An alternative could be performing 10-fold cross-validation but it is somehow rarely used.

446

Hydrological Modeling

2.16.6.2.3 Choice of the model variables Apart from dividing the data into several subsets, data preparation also includes the selection of proper variables to represent the modeled process, and, possibly, their transformation (Pyle, 1999). A study on the influence of different data transformation methods (linear, logarithmic, and seasonal transformations, histogram equalization, and a transformation to normality) was undertaken by Bowden et al. (2003). On a (limited) case study (forecasting salinity in a river in Australia 14 days ahead), they found that the model using the linear transformation resulted in the lowest RMSE and more complex transformations did not improve the model. Our own experience shows that it is sometimes also useful to apply the smoothing filters to reduce the noise in the hydrological time series. Choice of variables is an important issue, and it has to be based on taking the physics of the underling processes into account. State variables of data-driven models have nothing to do with the physics, but their inputs and outputs do have. In DDM, the physics of the process is introduced mainly via the justified and physically based choice of the relevant input variables. One may use visualization to identify the variables relevant for predicting the output value. There are also formal methods that help in making this choice more justified, and the reader can be directed to the paper by Bowden et al. (2005) for an overview of these. Mutual information which is based on Shannon’s entropy (Shannon, 1948) is used to investigate linear and nonlinear dependencies and lag effects (in time series data) between the variables. It is the measure of information available from one set of data having knowledge of another set of data. The average mutual information (AMI) between two variables X and Y is given by

AMI ¼

X i;j

 PXY ðxi ; yj Þlog2

PXY ðxi ; yj Þ PX ðxi ÞPY ðyj Þ

 ð1Þ

where PX(x) and PY(y) are the marginal probability density functions (PDFs) of X and Y, respectively, and PXY(x,y) the joint PDFs of X and Y. If there is no dependence between X and Y, then by definition the joint probability density PXY(x,y) would be equal to the product of the marginal densities (PX(x) PY(y)). In this case, AMI would be zero (the ratio of the joint and marginal densities in Equation (1) being 1, giving the logarithm a value of 0). A high value of AMI would indicate a strong dependence between two variables. Accurate estimate of the AMI depends on the accuracy of estimation of the marginal and joint probabilities density in Equation (1) from a finite set of examples. The most widely used approach is estimation of the probability densities by histogram with the fixed bin width. More stable, efficient, and robust probability density estimator is based on the use of kernel density estimation techniques (Sharma, 2000). It is our hope that the adequate data preparation and the rational and formalized choice of variables will become a standard part of any hydrological modeling study.

2.16.6.3 Methods and Typical Applications Most hydrological modeling problems are formulated as simulation of forecasting of real-valued variables. In

terminology of machine (statistical) learning, this is a regression problem. A number of linear and (sometimes) nonlinear regression methods have been used in the past. Most of the methods of ML can also be seen as sophisticated nonlinear regression methods. Many of them, instead of using very complex functions, use combinations of many simple functions. During training, the number of these functions and the values of their parameters are optimized, given the functions’ class. Note that ML methods typically do not assume any special kind of distribution of data, and do not require the knowledge of such distribution. Multilayer perceptron (MLP) is a device (mathematical model) that was originally referred to as an ANN (Haykin, 1999). Later ANN became a term encompassing other connectionist models as well. MLP consists of several layers of mutually interconnected nodes (neurons), each of which receives several inputs, calculates the weighted sum of them, and then passes the result to a nonlinear squashing function. In this way, the inputs to an MLP model are subjected to a multiparameter nonlinear transformation so that the resulting model is able to approximate complex input–output relationships. Training of MLP is in fact solving the problem of minimizing the model error (typically, MSE) by determining the optimal set of weights. MLP ANNs are known to have several dozens of successful applications in hydrology. The most popular application was building rainfall–runoff models: Hsu et al. (1995), Minns and Hall (1996), Dawson and Wilby (1998), Dibike et al. (1999), Abrahart and See (2000), Govindaraju and Rao (2001), Coulibaly et al. (2000), Hu et al. (2007), and Abrahart et al. (2007b). They were also used to model river stage–discharge relationships (Sudheer and Jain, 2003; Bhattacharya and Solomatine, 2005). ANNs were also used to build surrogate (emulation, meta-) models for replicating the behavior of hydrological and hydrodynamic models: in model-based optimal control of a reservoir (Solomatine and Torres, 1996), calibration of a rainfall–runoff model (Khu et al., 2004), and in multiobjective decision support model for watershed management (Muleta and Nicklow, 2004). Most theoretical problems related to MLP have been solved, and it should be seen as a quite reliable and wellunderstood method. Radial basis functions (RBFs) could be seen as a sensible alternative to the use of complex polynomials. The idea is to approximate some function y ¼ f(x) by a superposition of J functions F(x, s), where s is a parameter characterizing the span or width of the function in the input space. Functions F are typically bell shaped (e.g., a Gaussian function) so that they are defined in the proximity to some representative locations (centers) wj in n-dimensional input space and their values are close to zero far from these centers. The aim of learning here is to find the positions of centers wj and the parameters of the functions f(x). This can be accomplished by building an RBF neural network; its training allows the identification of these unknown parameters. The centers wj of the RBFs can be chosen using a clustering algorithm, the parameters of the Gaussian can be found based on the spread (variance) of data in each cluster, and it can be shown that the weights can be found by solving a system of linear equations. This is done for a certain number of RBFs, with the

Hydrological Modeling

exhaustive optimization run across the number of RBFs in a certain range. The areas of RBF networks applications are the same as those of MLPs. Sudheer and Jain (2003) used RBF ANNs for modeling river stage–discharge relationships and found out that on the considered case study RBF ANNs were superior to MLPs; Moradkhani et al. (2004) used RBF ANNs for predicting hourly stream flow hydrograph for the daily flow for a river in USA as a case study, and demonstrated their accuracy if compared to other numerical prediction models. In this study, RBF was combined with the self-organizing feature maps used to identify the clusters of data. Nor et al. (2007) used RBF ANN for the same purpose, however, for the hourly flow and considering only storm events in the two catchments in Malaysia as case studies. Regression trees and M5 model trees. These models can be attributed simultaneously to (piece-wise) linear regression models, and to modular (multi)models. They use the following idea: progressively split the parameter space into areas and build in each of them a separate regression model of zero or first order (Figure 7). In M5 trees models in leaves are first order (linear). The Boolean tests ai at nodes have the form xioC and are used to progressively split the data set. The index of the input variable i and value C are chosen to minimize the standard deviation in the subsets resulting from the split. Mn are models built for subsets filtered down to a given tree leaf. The resulting model can be seen as a set of linear models being specialized on the certain subsets of the training set – belonging to different regions of the input space. M5 algorithm to build such model trees was proposed by Quinlan (1992).

Training data set

a1 New instance

a2

M3

a4

M1

a3

M2

M4

M5

Output

Figure 7 Building a tree-like modular model (M5 model tree). Boolean tests ai have the form xioC and split data set during training. Mn are linear regression models built on data subsets, and applied to a new instance input vector in operation.

447

Combination of linear models was already used in dynamic hydrology in the 1970s (e.g., multilinear models by Becker and Kundzewicz (1987)). Application of the M5 algorithm to build such models adds rigor to this approach and makes it possible to build the models automatically and to generate a range of models of different complexity and accuracy. MTs are often almost as accurate as ANNs, but have some important advantages: training of MTs is much faster than ANNs, it always converges, and the results can be easily understood by decision makers. An early (if not the first) application of M5 model trees in river flow forecasting was reported by Kompare et al. (1997). Solomatine and Dulal (2003) used M5 model tree in rainfall– runoff modeling of a catchment in Italy. Stravs and Brilly (2007) used M5 trees in modeling the precipitation interception in the context of the Dragonja river basin case study. Genetic programming (GP) and evolutionary regression. GP is a symbolic regression method in which the specific model structure is not chosen a priori, but is a result of the search (optimization) process. Various elementary mathematical functions, constants, and arithmetic operations are combined in one function and the algorithm tries to construct a model recombining these building blocks in one formula. The function structure is represented as a tree and since the resulting function is highly nonlinear, often nondifferentiable, it is optimized by a randomized search method – usually a GA. Babovic and Keijzer (2005) gave an overview of GP applications in hydrology. Laucelli et al. (2007) presented an application of GP to the problem of forecasting the groundwater heads in the aquifer in Italy; in this study, the authors also employed averaging of several models built on the data subsets generated by bootstrap. One may limit the class of possible formulas (regression equations), allowing for a limited class of formulas that would a priori be reasonable. In evolutionary regression (Giustolisi and Savic, 2006), a method similar to GP, the elementary functions are chosen from a limited set and the structure of the overall function is fixed. Typically, a polynomial regression equation is used, and the coefficients are found by GA. This method overcomes some shortcomings of GP, such as the computational requirements – the number of parameters to tune and the complexity of the resulting symbolic models. It was used, for example, for modeling groundwater level (Giustolisi et al., 2007a) and river temperature (Giustolisi et al., 2007b), and the high accuracy and transparency of the resulting models were reported. FRBSs. Probability is not the only way to describe uncertainty. In his seminal paper, Lotfi Zadeh (1965) introduced yet another way of dealing with uncertainty – fuzzy logic, and since then it found multiple successful applications, especially in application to control problems. Fuzzy logic can be used in combining various models, as done previously, for example, by See and Openshaw (2000) and Xiong et al. (2001), building the so-called fuzzy committees of models (Solomatine, 2006), and also the instrumentarium of fuzzy logic can be used for building the so-called FRBSs which are effectively regression models. FRBS can be built by interviewing human experts, or by processing historical data and thus forming a data-driven model. These

448

Hydrological Modeling

rules are patches of local models overlapped throughout the parameter space, using a sort of interpolation at a lower level to represent patterns in complex nonlinear relationships. The basics of the data-driven approach and its use in a number of water-related applications can be found in Ba´rdossy and Duckstein (1995). Typically, the following rules are considered:

IF x1 is A1;r ANDyAND xn is An;y THEN y is B where {x1,y, xn} ¼ x is the input vector; Aim the fuzzy set; r the index of the rule, r ¼ 1,y,R. Fuzzy sets Air (defined as membership functions with the values ranging from 0 to 1) are used to partition the input space into overlapping regions (for each input these are intervals). The structure of B in the consequent could be either a fuzzy set (then such model is called a Mamdani model) or a function y ¼ f (x), often linear, and then the model is referred to as Takagi–Sugeno–Kang (TSK) model. The model output is calculated as a weighted combination of the R rules’ responses. Output of the Mamdani model is fuzzy (a membership function of irregular shape), so the crisp output has to be calculated by the so-called defuzzification operator. Note that in TSK model, each of the r rules can be interpreted as local models valid for certain regions in the input space defined by the antecedent and overlapping fuzzy sets Air. Resemblance to the RBF ANN is obvious. FRBSs were effectively used for drought assessment (Pesti et al., 1996); reconstruction of the missing precipitation data by a Mamdani-type system (Abebe et al., 2000b); control of water levels in polder areas (Lobbrecht and Solomatine et al., 1999); and modeling rainfall–discharge dynamics (Vernieuwe et al., 2005; Nayak et al., 2005). Casper et al. (2007) presented an interesting study where TSK type of FRBS has been developed using soil moisture and rainfall as input variables to predict the discharge at the outlet of a small catchment, with the special attention to the peak discharge. One of the limitations of FRBS is that the demand for data grows exponentially with an increase in the number of input variables. SVMs. This ML method is based on the extension of the idea of identifying a hyperplane that separates two classes in classification. It is closely linked to the statistical learning theory initiated by V. Vapnik in the 1970s at the Institute of Control Sciences of the Russian Academy of Science (Vapnik, 1998). Originally developed for classification, it was extended to solving prediction problems, and, in this capacity, was used in hydrology-related tasks. Dibike et al. (2001) and Liong and Sivapragasam (2002) reported using SVMs for forecasting the river water flows and stages. Bray and Han (2004) addressed the issue of tuning SVMs for rainfall–runoff modeling. In all reported cases, SVM-based predictors have shown good results, in many cases superseding other methods in accuracy. Instance-based learning (IBL). This method allows for classification or numeric prediction directly by combining some instances from the training data set. A typical representative of IBL is the k-nearest neighbor (k-NN) method. For a new input vector xq (query point), the output value is calculated as the mean value of the k-nearest neighboring examples, possibly weighted according to their distance to xq. Further extensions are known as locally weighted regression (LWR) when the regression model is built on k nearest instances: the training instances are assigned

weights according to their distance to xq and the regression equations are generated on the weighted data. In fact, IBL methods construct a local approximation to the modeled function that applies well in the immediate neighborhood of the new query instance encountered. Thus, it describes a very complex target function as a collection of less complex local approximations, and often demonstrates competitive performance when compared, for example, to ANNs. Karlsson and Yakowitz (1987) introduced this method in the context of water, focusing however only on (single-variate) time-series forecasts. Galeati (1990) demonstrated the applicability of the k-NN method (with the vectors composed of the lagged rainfall and flow values) for daily discharge forecasting and favorably compared it to the statistical autoregressive model with exogenous input (ARX) model, and used the k-NN method for adjusting the parameters of the linear perturbation model for river flow forecasting. Toth et al. (2000) compared the k-NN approach to other time-series prediction methods in a problem of short-term rainfall forecasting. Solomatine et al. (2007) explored a number of IBL methods, tested their applicability in rainfall–runoff modeling, and compared their performance to other ML methods. To conclude the coverage of the popular data-driven methods, it can be mentioned that all of them are developed in the ML and CI community. The main challenges for the researchers in hydrology and hydroinformatics are in testing various combinations of these methods for particular waterrelated problems, combining them with the optimization techniques, developing the robust modeling procedures able to work with the noisy data, and in developing the methods providing the model uncertainty estimates.

2.16.6.4 DDM: Current Trends and Conclusions There are a number of challenges in DDM: development of the optimal model architectures, making models more robust, understandable, and ready for inclusion into existing decision support systems. Models should adequately reflect reality, which is uncertain, and in this respect developing the methods of dealing with the data and model uncertainty is currently an important issue. One of the interesting questions that arise in case of using a data-driven model is the following one: to what extent such models could or should incorporate the expert knowledge into the modeling process. One may say that a typical ML algorithm minimizes the training (cross validation) error seeing it as the ultimate indicator of the algorithms performance, so is purely data-driven – and this is what is expected from such models. Hydrologists, however, may have other consideration when assessing the usefulness of a model, and typically wish to have a certain input to building a model rightfully hoping that the direct participation of an expert may increase the model accuracy and trust in the modeling results. Some of the examples of merging the hydrological knowledge and the concepts of process-based modeling with those of DDM are mentioned in Section 2.16.8.2. Data-driven models are seen by many hydrologists as tools complementary to process-based models. More and more practitioners are agreeing to that, but many are still to be convinced. Research is now oriented toward development of

Hydrological Modeling

the optimal model architectures and avenues for making datadriven models more robust, understandable, and very useful for practical applications. The main challenge is in the inclusion of DDM into the existing decision-making frameworks, while taking into consideration both the system’s physics, expert judgment, and the data availability. For example, in operational hydrological forecasting, many practitioners are trained in using process-based models (mainly conceptual ones) that serve them reasonably well, and adoption of another modeling paradigm with inevitable changes in their everyday practice could be a painful process. Making models capable of dealing with the data and model uncertainty is currently an important issue as well. It is sensible to use DDM if (1) there is a considerable amount of observations available; (2) there were no considerable changes to the system during the period covered by the model; and (3) it is difficult to build adequate processbased simulation models due to the lack of understanding and/or to the ability to satisfactorily construct a mathematical model of the underlying processes. Data-driven models can also be useful when there is a necessity to validate the simulation results of physically based models. It can be said that it is practically impossible to recommend one particular type of a data-driven model for a given problem. Hydrological data are noisy and often of poor quality; therefore, it is advisable to apply various types of techniques and compare and/or combine the results.

2.16.7 Analysis of Uncertainty in Hydrological Modeling 2.16.7.1 Notion of Uncertainty Webster’s Dictionary (1998) defines uncertain as follows: not surely or certainly known, questionable, not sure or certain in knowledge, doubtful, not definite or determined, vague, liable to vary or change, not steady or constant, varying. The noun uncertainty results from the above concepts and can be summarized as the state of being uncertain. However, in the context of hydrological modeling, uncertainty has a specific meaning, and it seems that there is no consensus about the very term of uncertainty, which is conceived with differing degrees of generality (Kundzewicz, 1995). Often uncertainty is defined with respect to certainty. For example, Zimmermann (1997) defined certainty as ‘‘certainty implies that a person has quantitatively and qualitatively the appropriate information to describe, prescribe or predict deterministically and numerically a system, its behaviour or other phenomena.’’ Situations that are not described by the above definition shall be called uncertainty. A similar definition has been given by Gouldby and Samuels (2005): ‘‘a general concept that reflects our lack of sureness about someone or something, ranging from just short of complete sureness to an almost complete lack of conviction about an outcome.’’ In the context of modeling, uncertainty is defined as a state that reflects our lack of sureness about the outcome of a physical processes or system of interest, and gives rise to potential difference between assessment of the outcome and its true value. More precisely, uncertainty of a model output is the

449

state or condition that the output cannot be assessed uniquely. Uncertainty stems from incompleteness or imperfect knowledge or information concerning the process or system in addition to the random nature of the occurrence of the events. Uncertainty resulting from insufficient information may be reduced if more information is available.

2.16.7.2 Sources of Uncertainty Uncertainties that can affect the model predictions stem from a variety of sources (e.g., Melching, 1995; Gupta et al., 2005), and are related to our understanding and measurement capabilities regarding the real-world system under study: 1. Perceptual model uncertainty, that is, the conceptual representation of the watershed that is subsequently translated into mathematical (numerical) form in the model. The perceptual model (Beven, 2001) is based on our understanding of the real-world watershed system, that is, flowpaths, number and location of state variables, runoff production mechanisms, etc. This understanding might be poor, particularly for aspects relating to subsurface system characteristics, and therefore our perceptual model might be highly uncertain (Neuman, 2003). 2. Data uncertainty, that is, uncertainty caused by errors in the measurement of input (including forcing) and output data, or by data processing. Additional uncertainty is introduced if long-term predictions are made, for instance, in the case of climate change scenarios for which as per definition no observations are available. A hydrological model might also be applied in integrated systems, for example, connected to a socioeconomic model, to assess, for example, impacts of water resources changes on economic behavior. Data to constrain these integrated models are rarely available (e.g., Letcher et al., 2004). An element of data processing, that is, uncertainty, is introduced when a model is required to interpret the actual measurement. A typical example is the use of radar rainfall measurements. These are measurements of reflectivity that have to be transformed to rainfall estimates using a (empirical) model with a chosen functional relationship and calibrated parameters, both of which can be highly uncertain. 3. Parameter estimation uncertainty, that is, the inability to uniquely locate a best parameter set (model, i.e., a model structure parameter set combination) based on the available information. The lack of correlation often found between conceptual model parameters and physical watershed characteristics will commonly result in significant prediction uncertainty if the model is extrapolated to predict the system behavior under changed conditions (e.g., land-use change or urbanization) or to simulate the behavior of a similar but geographically different watersheds for which no observations of the variable of interest are available (i.e., the ungauged case). Changes in the represented system have to be considered through adjustments of the model parameters (or even the model structure), and the degree of adjustment has so far been difficult to determine without measurements of the changed system response. 4. Model structural uncertainty introduced through simplifications, inadequacies, and/or ambiguity in the description

450

Hydrological Modeling

of real-world processes. There will be some initial uncertainty in the model state(s) at the beginning of the modeled time period. This type of uncertainty can usually be taken care of through the use of a warm-up (spin-up) period or by optimizing the initial state(s) to fit the beginning of the observed time series. Errors in the model (structure and parameters) and in the observations will also commonly cause the states to deviate from the actual state of the system in subsequent time periods. This problem is often reduced using data assimilation techniques as discussed later. Figure 8 presents how different sources of the uncertainty might vary with model complexity. As the model complexity (and the detailed representation of the physical process) increases, structural uncertainty decreases. However, with the increasing complexity of model, the number of inputs and parameters also increases and consequently there is a good chance that input and parameter uncertainty will increase. Due to the inherent trade-off between model structure uncertainty and input/parameter uncertainty, for every model there is the optimal level of model complexity where the total uncertainty is minimum.

prediction intervals consist of the upper and lower limits between which a future uncertain value of the quantity is expected to lie with a prescribed probability. The endpoints of a prediction interval are known as the prediction limits. The width of the prediction interval gives us some idea about how uncertain we are about the uncertain entity. Although useful and successful in many applications, probability theory is, in fact, appropriate for dealing with only a very special type of uncertainty, namely random (Klir and Folger, 1988). However, not all uncertainty is random. Some forms of uncertainty are due to vagueness or imprecision, and cannot be treated with probabilistic approaches. Fuzzy set theory and fuzzy measures (Zadeh, 1965, 1978) provide a nonprobabilistic approach for modeling the kind of uncertainty associated with vagueness and imprecision. Information theory is also used for representing uncertainty. Shannon’s (1948) entropy is a measure of uncertainty and information formulated in terms of probability theory. Another broad theory of uncertainty representation is the evidence theory introduced by Shafer (1976). Evidence theory, also known as Dempster–Shafer theory of evidence, is based on both the probability and possibility theory. In hydrological modeling, the primary tool for handling uncertainty is still probability theory, and, to some extent, fuzzy logic.

2.16.7.3 Uncertainty Representation For many years, probability theory has been the primary tool for representing uncertainty in mathematical models. Different methods can be used to describe the degree of uncertainty. The most widely adopted methods use PDFs of the quantity, subject to the uncertainty. However, in many practical problems the exact form of this probability function cannot be derived or found precisely. When it is difficult to derive or find PDF, it may still be possible to quantify the level of uncertainty by the calculated statistical moments such as the variance, standard deviation, and coefficient of variation. Another measure of the uncertainty of a quantity relates to the possibility to express it in terms of the two quantiles or prediction intervals. The Minimum uncertainty

Uncertainty

Total uncertainty

Structure uncertainty

Input and parameter uncertainty

2.16.7.4 View at Uncertainty in Data-Driven and Statistical Modeling In DDM, the sources of uncertainty are similar to those for other hydrological models, but there is an additional focus on data partitioning used for model training and verification. Often data are split in a nonoptimal way. A standard procedure for evaluating the performance of a model would be to split the data into training set, cross-validation set, and test set. This approach is, however, very sensitive to the specific sample splitting (LeBaron and Weigend, 1994). In principle, all these splitting data sets should have identical distributions, but we do not know the true distribution. This causes uncertainty in prediction as well. The prediction error of any regression model can be decomposed into the following three sources (Geman et al., 1992): (1) model bias, (2) model variance, and (3) noise. Model bias and variance may be further decomposed into the contributions from data and training process. Furthermore, noise can also be decomposed into target noise and input noise. Estimating these components of prediction error (which is however not always possible) helps to compute the predictive uncertainty. The terms bias and variance come from a well-known decomposition of prediction error. Given N data points and M models, the decomposition is based on the following equality: N X M N 1 X 1X ðti  yij Þ2 ¼ ðti  yi Þ2 NM i¼1 j¼1 N i¼1

þ Model complexity Figure 8 Dependency of various sources of uncertainty on the model complexity.

N X M 1 X ð yi  yij Þ2 NM i¼1 j¼1

ð2Þ

where ti is the ith target, yij the ith output of the jth model, and P  yi ¼ 1=M M j¼1 yij the average model output calculated for input i.

Hydrological Modeling

The left-hand-side term of Equation (2) is the well-known MSE. The first term on the right-hand side is the square of bias and the last term is the variance. The bias of the prediction errors measures the tendency of over- or under-prediction by a model, and is the difference between the target value and the model output. From Equation (2), it is clear that the variance does not depend on the target, and measures the variability in the predictions by different models.

2.16.7.5 Uncertainty Analysis Methods Once the uncertainty in a model is acknowledged, it should be analyzed and quantified with the ultimate aim to reduce the impact of uncertainty. There is a large number of uncertainty analysis methods published in the academic literature. Pappenberger et al. (2006) provided a decision tree to help in choosing an appropriate method for a given situation. Uncertainty analysis process in hydrological models varies mainly in the following: (1) type of hydrological models used; (2) source of uncertainty to be treated; (3) the representation of uncertainty; (4) purpose of the uncertainty analysis; and (4) availability of resources. Uncertainty analysis has comparatively a long history in physically based and conceptual modeling (see, e.g., Beven and Binley, 1992; Gupta et al., 2005). Uncertainty analysis methods in all the above cases should involve: (1) identification and quantification of the sources of uncertainty, (2) reduction of uncertainty, (3) propagation of uncertainty through the model, (4) quantification of uncertainty in the model outputs, and (6) application of the uncertain information in decision-making process. A number of methods have been proposed in the literature to estimate model uncertainty in rainfall–runoff modeling. Reviews of various methods of uncertainty analysis on hydrological models can be found in, for example, Melching (1995), Gupta et al. (2005), Montanari (2007), Moradkhani and Sorooshian (2008), and Shrestha and Solomatine (2008). These methods are broadly classified into several categories (most of them result in probabilistic estimates): 1. analytical methods (see, e.g., Tung, 1996); 2. approximation methods, for example, first-order second moment method (Melching, 1992); 3. simulation and sampling-based (Monte Carlo) methods leading to probabilistic estimates that may also use Bayesian reasoning (Kuczera and Parent, 1998; Beven and Binley, 1992); 4. methods based on the analysis of the past model errors and either using distribution transforms (Montanari and Brath, 2004) or building a predictive ML of uncertainty (Shrestha and Solomatine, 2008; Solomatine and Shrestha, 2009); and 5. methods based on fuzzy set theory (e.g., Abebe et al., 2000a; Maskey et al., 2004). Analytical and approximation methods can hardly be applicable in case of using complex computer-based models. Here, we will present only the most widely used probabilistic methods based on random sampling – Monte Carlo simulation method. The GLUE method (Beven and Binley, 1992), widely used in hydrology, can be seen as a particular case of

451

MC approach. MC simulation is a flexible and robust method capable of solving a great variety of problems. In fact, it may be the only method that can estimate the complete probability distribution of the model output for cases with highly nonlinear and/or complex system relationship (Melching, 1995). It has been used extensively and also as a standard means of comparison against other methods for uncertainty assessment. In MC simulation, random values of each of uncertain variables are generated according to their respective probability distributions and the model is run for each of the realizations of uncertain variables. Since we have multiple realizations of outputs from the model, standard statistical technique can be used to estimate the statistical properties (mean, standard deviation, etc.) and empirical probability distribution of the model output. MC simulation method involves the following steps: 1. randomly sample uncertain variables Xi from their joint probability distributions; 2. run the model y ¼ g(xi) with the set of random variables xi; 3. repeat the steps 1 and 2 s times, storing the realizations of the outputs y1, y2,y, ys; and 4. from the realizations y1, y2,y, ys, derive the cdf and other statistical properties (e.g., mean and standard deviation) of Y. When MC sampling is used, the error in estimating PDF is inversely proportional to the square root of the number of runs s and, therefore, decreases gradually with s. As such, the method is computationally expensive, but can reach an arbitrarily level of accuracy. The MC method is generic, invokes fewer assumptions, and requires less user input than other uncertainty analysis methods. However, the MC method suffers from two major practical limitations: (1) it is difficult to sample the uncertain variables from their joint distribution unless the distribution is well approximated by a multinormal distribution (Kuczera and Parent, 1998) and (2) it is computationally expensive for complex models. Markov chain Monte Carlo (MCMC) methods such as Metropolis and Hastings (MH) algorithm (Metropolis et al., 1953; Hastings, 1970) have been used to sample parameter from its posterior distribution. In order to reduce the number of samples (model simulations) necessary in MC sampling , more efficient Latin Hypercube sampling has been introduced (McKay et al., 1979). Further, the following methods in this row can be mentioned: Kalman filter and its extensions (Kitanidis and Bras, 1980), the DYNIA approach (Wagener et al., 2003), the BaRE approach (Thiemann et al., 2001), the SCEM-UA algorithm (Vrugt et al., 2003), and the DREAM algorithm (Vrugt et al., 2008b), a version of the MCMC scheme. Most of the probabilistic techniques for uncertainty analysis treat only one source of uncertainty (i.e., parameter uncertainty). Recently, attention has been given to other sources of uncertainty, such as input uncertainty or structure uncertainty, as well as integrated approach to combine different sources of uncertainty. The research shows that input or structure uncertainty is more dominant than the parameter uncertainty. For example, Kavetski et al. (2006) and Vrugt et al. (2008a), among others, treat input uncertainty in hydrological

452

Hydrological Modeling

modeling using Bayesian approach. Butts et al. (2004) analyzed impact of the model structure on hydrological modeling uncertainty for stream flow simulation. Recently, new schemes have emerged to estimate the combined uncertainties in rainfall–runoff predictions associated with input, parameter, and structure uncertainty. For instance, Ajami et al. (2007) used an integrated Bayesian multimodel approach to combine input, parameter, and model structure uncertainty. Liu and Gupta (2007) suggested an integrated data assimilation approach to treat all sources of uncertainty. Regarding the sources of uncertainty, Monte-Carlo-type methods are widely used for parameter uncertainty, Bayesian methods and/or data assimilation can be used for input uncertainty and Bayesian model averaging method is suitable for structure uncertainty. The appropriate uncertainty analysis method also depends on whether the uncertainty is represented as randomness or fuzziness. Similarly, uncertainty analysis methods for real-time forecasting purposes would be different from those used for design purposes (e.g., when estimating design discharge hydraulic structure design). It should be noted that the practice of uncertainty analysis and the use of the results of such analysis in decision making are not yet widely spread. Some possible misconceptions are stated by Pappenberger and Beven (2006): a) uncertainty analysis is not necessary given physically realistic models, b) uncertainty analysis is not useful in understanding hydrological and hydraulic processes, c) uncertainty (probability) distributions cannot be understood by policy makers and the public, d) uncertainty analysis cannot be incorporated into the decisionmaking process, e) uncertainty analysis is too subjective, f) uncertainty analysis is too difficult to perform and g) uncertainty does not really matter in making the final decision.

Some of these misconceptions however have explainable reasons, so the fact remains that more has to be done in bringing the reasonably well-developed apparatus of uncertainty analysis and prediction to decision-making practice.

2.16.8 Integration of Models 2.16.8.1 Integration of Meteorological and Hydrological Models Water managers demand much longer lead times in the hydrological forecasts. Forecasting horizon of hydrological models can be extended if along with the (almost) real-time measurements of precipitation (radar and satellite images, gauges), their forecasts are used. The forecasts can come only from the numerical weather prediction (NWP; meteorological) models. Linking of meteorological and hydrological models is currently an adopted practice in many countries. One of the examples of such an integrated approach is the European flood forecasting system (EFFS), in which development started in the framework of EU-funded project in the beginning of the 2000s. Currently, this initiative is known as the European flood alert system (EFAS), which is being developed by the EC Joint Research Centre (JRC) in close collaboration with several European institutions. EFAS aims at developing a 4–10-day inadvance EFFS employing the currently available medium-range

weather forecasts. The framework of the system allows for incorporation of both detailed models for specific basins as well as a broad scale for entire Europe. This platform is not supposed to replace the national systems but to complement them. The resolution of the existing NWP models dictates to a certain extent the resolution of the hydrological models. LISFLOOD model (Bates and De Roo, 2000) and its extension module for inundation modeling LISFLOOD-FP have been adopted as the major hydrological response model in EFAS. This is a rasterized version of a process-based model used for flood forecasting in large river basins. LISFLOOD is also suitable for hydrological simulations at the continental scale, as it uses topographic and land-use maps with a spatial resolutions up to 5 km. It should be mentioned that useful distributed hydrological models that are able to forecast floods at meso-scales have grid sizes from dozens of meters to several kilometers. At the same time, the currently used meteorological models, providing the quantitative precipitation forecasts, have mesh sizes from several kilometers and higher. This creates an obvious inconsistency and does not allow to realize the potential of the NWP outputs for flood forecasting – see, for example, Bartholmes and Todini (2005). The problem can partly be resolved by using downscaling (Salathe, 2005; Cannon, 2008), which however may bring additional errors. As NWP models use more and more detailed grids, this problem will be becoming less and less acute. One of the recent successful software implementations of allowing for flexible combination of various types of models from different suppliers (using XML-based open interfaces) and linking to the real-time feeds of the NWP model outputs is the Delft-FEWS (FEWS, flood early warning system) platform of Deltares (Werner, 2008). Currently, this platform is being accepted as the integrating tool for the purpose of operational hydrological forecasting and warning in a number of European countries and in USA. The other two widely used modeling systems (albeit less open in the software sense) that are also able to integrate meteorological inputs are (1) the MIKE FLOOD by DHI Water and Environment, based on the hydraulic/hydrologic modeling system MIKE 11 and (2) FloodWorks by Wallingford Software.

2.16.8.2 Integration of Physically Based and Data-Driven Models 2.16.8.2.1 Error prediction models Consider a model simulating or predicting certain water-related variable (referred to as a primary model). This model’s outputs are compared to the recorded data and the errors are calculated. Another model, a data-driven model, is trained on the recorded errors of the primary model, and can be used to correct errors of the primary model. In the context of river modeling, this primary model would be typically a physically based model, but can be a data-driven model as well. Such an approach was employed in a number of studies. Shamseldin and O’Connor (2001) used ANNs to update runoff forecasts: the simulated flows from a model and the current and previously observed flows were used as input, and the corresponding observed flow as the target output. Updates of daily flow forecasts for a lead time of up to 4 days were

Hydrological Modeling

made, and the ANN models gave more accurate improvements than autoregressive models. Lekkas et al. (2001) showed that error prediction improves real-time flow forecasting, especially when the forecasting model is poor. Abebe and Price (2004) used ANN to correct the errors of a routing model of the River Wye in UK. Solomatine et al. (2006) built an ANN-based rainfall–runoff model whose outputs were corrected by an IBL model.

2.16.8.2.2 Integration of hydrological knowledge into DDM An expert can contribute to building a DDM by bringing in the knowledge about the expected relationships between the system variables, in performing advanced correlation and mutual information analysis to select the most relevant variables, determining the model structure based on hydrological knowledge (allowed, e.g., by the M5flex algorithm by Solomatine and Siek (2004)), and in deciding what data should be used and how it should be structured (as it is done by most modelers). It is possible to mention a number of studies where an attempt is made to include a human expert in the process of building a data-driven model. For solving a flow forecasting problem, See and Openshaw (2000) built not a single overall ANN model but different models for different classes of hydrological events. Solomatine and Xue (2004) introduced a human expert to determine a set of rules to identify various hydrological conditions for each of which a separate specialized data-driven model (ANN or M5 tree) was built. Jain and Srinivasulu (2006) and Corzo and Solomatine (2007) also applied decomposition of the flow hydrograph by a threshold value and then built the separate ANNs for low and high flow regimes. In addition, Corzo and Solomatine (2007) were building two separate models related to base and excess flow which were identified by the Ekhardt’s (2005) method, and used overall optimization of the resulting model structure. All these studies demonstrated the higher accuracy of the resulting models where the hydrological knowledge and, wherever possible, models were directly used in building data-driven models.

2.16.9 Future Issues in Hydrological Modeling Natural and anthropogenic changes constantly impact the environment surrounding us. Available moisture and energy change due to variability and shifts in climate, and the separation of precipitation into different pathways on the land surface are altered due to wildfires, beetle infestations, urbanization, deforestation, invasive plant species, etc. Many of these changes can have a significant impact on the hydrological regime of the watershed in which they occur (e.g., Milly et al., 2005; Poff et al., 2006; Oki and Kanae, 2006; Weiskel et al., 2007). Such changes to water pathways, storage, and subsequent release (the blue and green water idea of Falkenmark and Rockstroem (2004)) are predicted to have significant negative impacts on water security for large population groups as well as for ecosystems in many regions of the world (e.g., Sachs, 2004). The growing imbalances among freshwater supply, its consumption, and human population will only increase the problem (Vo¨ro¨smarty et al.,

453

2000). A major task for hydrologic science lies in providing predictive models based on sound scientific theory to support water resource management decisions for different possible future environmental, population, and institutional scenarios. But can we provide credible predictions of yet unobserved hydrological responses of natural systems? Credible modeling of environmental change impact requires that we demonstrate a significant correlation between model parameters and watershed characteristics, since calibration data are, by definition, unavailable. Currently, such a priori or regionalized parameters estimates are not very accurate and will likely lead to very uncertain prior distributions for model parameters in changed watersheds, leading to very uncertain predictions. Much work is to be done to solve this and to provide the hydrological simulations with the credibility necessary to support sustainable management of water resources in a changing world. The issue of model validation has to be given much more attention. Even if calibration and validation data are available, the historical practice of validating the model based on calculation of the Nash–Sutcliffe coefficient or some other squared error measure outside the calibration period is inadequate. Often low or high values of these criteria cannot clearly indicate whether or not the model under question has descriptive or predictive power. The discussion on validation has to move on to use more informative signatures of model behavior, which allow for the detection of how consistent the model is with system at hand (Gupta et al., 2008). This is particularly crucial when it comes to the assessment of climate and land-use change impacts, that is, when future predictions will lie outside the range of observed variability of the system response. Another development is expected with respect to modeling technologies, mainly in the more effective merging of data into models. One of the aspects here is the optimal use of data for model calibration and evaluation. In this respect, more rigorous approach adopted in DDM (e.g., use of crossvalidation and optimal data splitting) could be useful. Modern technology allows for accurate measurements of hydraulic and hydrologic parameters, and for more and more accurate precipitation forecasts coming from NWP models. Many of these come in real time, and this permits for a wider use of data-driven models with their combination with the physically based models, and for wider use of updating and data assimilation schemes. With more data being collected and constantly increasing processing power, one may also expect a wider use of distributed models. It is expected that the way the modeling results are delivered to the decision makers and public will also undergo changes. Half of the global population already owns mobile phones with powerful operating systems, many of which are connected to widearea networks, so the Information and communication technology (ICT) for the quick dissemination of modeling results, for example, in the form of the flood alerts, is already in place. Hydrological models will be becoming more and more integrated into hydroinformatics systems that support full information cycle, from data gathering to the interpretation and use of modeling results by decision makers and the public.

454

Hydrological Modeling

References Abbott MB, Bathurst JC, Cunge JA, O’Connell PE, and Rasmussen J (1986) An introduction to the European Hydrological System – Systeme Hydrologique Europeen, SHE. 1. History and philosophy of a physically-based, distributed modelling system. Journal of Hydrology 87: 45--59. Abdulla FA and Lettenmaier DP (1997) Development of regional parameter estimation equations for a macroscale hydrologic model. Journal of Hydrology 197: 230–257. Abebe AJ, Guinot V, and Solomatine DP (2000a) Fuzzy alpha-cut vs. Monte Carlo techniques in assessing uncertainty in model parameters. In: Proceedings of the 4th International Conference on Hydroinformatics. Cedar Rapids, IA, USA. Abebe AJ and Price RK (2004) Information theory and neural networks for managing uncertainty in flood routing. ASCE: Journal of Computing in Civil Engineering 18(4): 373--380. Abebe AJ, Solomatine DP, and Venneker R (2000b) Application of adaptive fuzzy rulebased models for reconstruction of missing precipitation events. Hydrological Sciences Journal 45(3): 425--436. Abrahart B, See LM, and Solomatine DP (eds.) (2008) Hydroinformatics in Practice: Computational Intelligence and Technological Developments in Water Applications. Heidelberg: Springer. Abrahart RJ, Heppenstall AJ, and See LM (2007a) Timing error correction procedure applied to neural network rainfall–runoff modelling. Hydrological Sciences Journal 52(3): 414--431. Abrahart RJ and See L (2000) Comparing neural network and autoregressive moving average techniques for the provision of continuous river flow forecast in two contrasting catchments. Hydrological Processes 14: 2157--2172. Abrahart RJ and See LM (2007) Neural network modelling of non-linear hydrological relationships. Hydrology and Earth System Sciences 11: 1563--1579. Abrahart RJ, See LM, Solomatine DP, and Toth E (eds.) (2007b) Data-driven approaches, optimization and model integration: Hydrological applications. Special Issue: Hydrology and Earth System Sciences 11: 1563–1579. Ajami NK, Duan Q, and Sorooshian S (2007) An integrated hydrologic Bayesian multimodel combination framework: Confronting input, parameter, and model structural uncertainty in hydrologic prediction. Water Resources Research 43: W01403 (doi:10.1029/2005WR004745). Atkinson SE, Woods RA, and Sivapalan M (2002) Climate and landscape controls on water balance model complexity over changing timescales. Water Resources Research 38(12): 101029/2002WR001487. Babovic V and Keijzer M (2005) Rainfall runoff modelling based on genetic programming. In: Andersen MG (ed.) Encyclopedia of Hydrological Sciences, vol. 1, ch. 21. New York: Wiley. Ba´rdossy A and Duckstein L (1995) Fuzzy Rule-Based Modeling with Applications to Geophysical, Biological and Engineering Systems. Boca Raton, FL: CRC Press. Bartholmes J and Todini E (2005) Coupling meteorological and hydrological models for flood forecasting. Hydrology and Earth System Sciences 9: 55--68. Bates PD and De Roo APJ (2000) A simple raster-based model for floodplain inundation. Journal of Hydrology 236: 54--77. Becker A and Kundzewicz ZW (1987) Nonlinear flood routing with multilinear models. Water Resources Research 23: 1043--1048. Bergstro¨m S (1976) Development and application of a conceptual runoff model for Scandinavian catchments. SMHI Reports RHO, No. 7. Norrko¨ping, Sweden: SMHI. Beven K (1989) Changing ideas in hydrology: The case of physically-based models. Journal of Hydrology 105: 157--172. Beven KJ (1996) A discussion on distributed modelling. In: Reefsgard CJ and Abbott MB (eds.) Distributed Hydrological Modelling, pp. 255–278. Kluwer: Dodrecht. Beven K (2000) Rainfall–Runoff Modeling – the Primer. Chichester: Wiley. Beven KJ (2001) Uniqueness of place and process representations in hydrological modeling. Hydrology and Earth System Sciences 4: 203--213. Beven KJ (2002) Towards an alternative blueprint for a physically based digitally simulated hydrologic response modelling system. Hydrological Processes 16: 189–206. Beven KJ (2009) Environmental Modeling: An Uncertain Future? London: Routledge. Beven K and Binley AM (1992) The future role of distributed models: Model calibration and predictive uncertainty. Hydrological Processes 6: 279–298. Bhattacharya B and Solomatine DP (2005) Neural networks and M5 model trees in modelling water level – discharge relationship. Neurocomputing 63: 381--396. Bowden GJ, Dandy GC, and Maier HR (2003) Data transformation for neural network models in water resources applications. Journal of Hydroinformatics 5: 245--258. Bowden GJ, Dandy GC, and Maier HR (2005) Input determination for neural network models in water resources applications. Part 1 – background and methodology. Journal of Hydrology 301: 75--92.

Bowden GJ, Maier HR, and Dandy GC (2002) Optimal division of data for neural network models in water resources applications. Water Resources Research 38(2): 1--11. Bray M and Han D (2004) Identification of support vector machines for runoff modelling. Journal of Hydroinformatics 6: 265--280. Burnash RJC (1995) The NWS river forecast system – catchment model. In: Singh VP (ed.) Computer Models of Watershed Hydrology, pp. 311--366. Highlands Ranch, CO: Water Resources Publications. Butts MB, Payne JT, Kristensen M, and Madsen H (2004) An evaluation of the impact of model structure on hydrological modelling uncertainty for streamflow simulation. Journal of Hydrology 298: 222--241. Calder IR (1990) Evaporation in the Uplands. Chichester: Wiley. Calver A (1988) Calibration, validation and sensitivity of a physically-based rainfall– runoff model. Journal of Hydrology 103: 103–115. Cannon AJ (2008) Probabilistic multisite precipitation downscaling by an expanded Bernoulli-gamma density network. Journal of Hydrometeorology 9(6): 1284--1300. Casper MPO, Gemmar M, Gronz J, and Stu¨ber M (2007) Fuzzy logic-based rainfall–runoff modelling using soil moisture measurements to represent system state. Hydrological Sciences Journal (Special Issue: Hydroinformatics) 52(3): 478--490. Chow V, Maidment D, and Mays L (1988) Applied Hydrology. New York, NY: McGrawHill. Clarke R (1973) A review of some mathematical models used in hydrology, with observations on their calibration and use. Journal of Hydrology 19(1): 1--20. Corzo G and Solomatine DP (2007) Baseflow separation techniques for modular artificial neural network modelling in flow forecasting. Hydrological Sciences Journal 52(3): 491--507. Coulibaly P, Anctil F, and Bobbe B (2000) Daily reservoir inflow forecasting using artificial neural networks with stopped training approach. Journal of Hydrology 230: 244--257. Dawdy DR and O’Donnell T (1965) Mathematical models of catchment behavior. Journal of the Hydraulics Division, American Society of Civil Engineers 91: 113–137. Dawson CW and Wilby R (1998) An artificial neural network approach to rainfall– runoff modelling. Hydrological Sciences Journal 43(1): 47--66. Dibike Y, Solomatine DP, and Abbott MB (1999) On the encapsulation of numericalhydraulic models in artificial neural network. Journal of Hydraulic Research 37(2): 147--161. Dibike YB, Velickov S, Solomatine DP, and Abbott MB (2001) Model induction with support vector machines: Introduction and applications. ASCE Journal of Computing in Civil Engineering 15(3): 208--216. Dooge JCI (1957) The rational method for estimating flood peaks. Engineering 184: 311--313. Dooge JCI (1959) A general theory of the unit hydrograph. Journal of Geophysical Research 64: 241--256. Dunn SM and Ferrier RC (1999) Natural flow in managed catchments: A case study of a modelling approach. Water Research 33: 621–630. Duan Q, Gupta VK, and Sorooshian S (1992) Effective and efficient global optimization for conceptual rainfall–runoff models. Water Resources Research 28: 1015–1031. Duan Q, Schaake J, and Koren V (2001) A priori estimation of land surface model parameters. In: Lakshmi V, Albertson J, and Schaake J (eds.) Land Surface Hydrology, Meteorology, and Climate: Observations and Modelling. American Geophysical Union, Water Science and Application, vol. 3, pp. 77–94. Efron B and Tibshirani RJ (1993) An Introduction to the Bootstrap. New York: Chapman and Hall. Ekhardt K (2005) How to construct recursive digital filters for baseflow separation. Hydrological Processes 19: 507--515. Ewen J and Parkin G (1996) Validation of catchment models for predicting land-use and climate change impacts. 1: Method. Journal of Hydrology 175: 583–594. Falkenmark M and Rockstroem J (2004) Balancing Water for Humans and Nature. The New Approach in Ecohydrology. (ISBN 1853839272).. Sterling, VA: Stylus Publishing. Fernandez W, Vogel RM, and Sankarasubramanian S (2000) Regional calibration of a watershed model. Hydrological Sciences Journal 45(5): 689–707. Freeze RA and Harlan RL (1969) Blueprint for a physically-based, digitally-simulated hydrologic response model. Journal of Hydrology 9: 237–258. Galeati G (1990) A comparison of parametric and non-parametric methods for runoff forecasting. Hydrological Sciences Journal 35(1): 79--94. Gaume E and Gosset R (2003) Over-parameterisation, a major obstacle to the use of artificial neural networks in hydrology? Hydrology and Earth System Sciences 7(5): 693--706. Geman S, Bienenstock E, and Doursat R (1992) Neural networks and the bias/variance dilemma. Neural Computation 4(1): 1--58.

Hydrological Modeling

Georgakakos KP, Seo D-J, Gupta H, Schaake J, and Butts MB (2004) Towards the characterization of streamflow simulation uncertainty through multimodel ensembles. Journal of Hydrology 298: 222--241. Giustolisi O, Doglioni A, Savic DA, and di Pierro F (2007a) An evolutionary multiobjective strategy for the effective management of groundwater resources. Water Resources Research 44: W01403. Giustolisi O, Doglioni A, Savic DA, and Webb BW (2007b) A multi-model approach to analysis of environmental phenomena. Environmental Modelling Systems Journal 22(5): 674--682. Giustolisi O and Savic DA (2006) A symbolic data-driven technique based on evolutionary polynomial regression. Journal of Hydroinformatics 8(3): 207--222. Gouldby B and Samuels P (2005) Language of risk. Project definitions. FLOODsite Consortium Report T32-04-01. http://www.floodsite.net (accessed April 2010) Govindaraju RS and Rao RA (eds.) (2001) Artificial Neural Networks in Hydrology. Dordrecht: Kluwer. Gupta HV, Beven KJ, and Wagener T (2005) Model calibration and uncertainty estimation. In: Andersen M (ed.) Encyclopedia of Hydrological Sciences, pp. 2015--2031. New York, NY: Wiley. Gupta HV, Kling H, Yilmaz KK, and Martinez GF (2009) Decomposition of the mean squared error and NSE performance criteria: Implications for improving hydrological modelling. Journal of Hydrology 377: 80--91. Gupta HV, Sorooshian S, and Yapo PO (1998) Toward improved calibration of hydrological models: Multiple and noncommensurable measures of information. Water Resources Research 34(4): 751--763. Gupta HV, Wagener T, and Liu Y (2008) Reconciling theory with observations: Elements of a diagnostic approach to model evaluation. Hydrological Processes 22(18): 3802--3813 (doi:10.1002/hyp.6989). Hall MJ and Minns AW (1999) The classification of hydrologically homogeneous regions. Hydrological Sciences Journal 44: 693--704. Han D, Kwong T, and Li S (2007) Uncertainties in real-time flood forecasting with neural networks. Hydrological Processes 21(2): 223--228. Hanna SR (1988) Air quality model evaluation and uncertainty. Journal of the Air Pollution Control Association 38: 406--442. Hannah DM, Smith BPG, Gurnell AM, and McGregor GR (2000) An approach to hydrograph classification. Hydrological Processes 14: 317--338. Harlin J (1991) Development of a process oriented calibration scheme for the HBV hydrological model. Nordic Hydrology 22: 15--36. Harris NM, Gurnell AM, Hannah DM, and Petts GE (2000) Classification of river regimes: A context for hydrogeology. Hydrological Processes 14: 2831--2848. Hastings WK (1970) Monte Carlo sampling methods using Markov chains and their applications. Biometrika 57(1): 97--109. Haykin S (1999) Neural Networks: A Comprehensive Foundation. New York: McMillan. Hsu KL, Gupta HV, and Sorooshian S (1995) Artificial neural network modelling of the rainfall–runoff process. Water Resources Research 31(10): 2517--2530. Hu T, Wu F, and Zhang X (2007) Rainfall–runoff modeling using principal component analysis and neural network. Nordic Hydrology 38(3): 235--248. Ibbitt RP (1970) Systematic parameter fitting for conceptual models of catchment hydrology. PhD dissertation, Imperial College of Science and Technology, University of London. Ibbitt RP and O’Donnell T (1971) Fitting methods for conceptual catchment models. Journal of the Hydraulics Division, American Society of Civil Engineers 97(9): 1331–1342. Jain A and Srinivasulu S (2006) Integrated approach to model decomposed flow hydrograph using artificial neural network and conceptual techniques. Journal of Hydrology 317: 291--306. Jakeman AJ, Hornberger GM, Littlewood IG, Whitehead PG, Harvey JW, and Bencala KE (1992) A systematic approach to modelling the dynamic linkage of climate, physical catchment descriptors and hydrological response components. Mathematics and Computers in Simulation 33: 359–366. Johnston PR and Pilgrim DH (1976) Parameter optimization for watershed models. Water Resources Research 12(3): 477–486. Karlsson M and Yakowitz S (1987) Nearest neighbour methods for non-parametric rainfall runoff forecasting. Water Resources Research 23(7): 1300--1308. Kavetski D, Kuczera G, and Franks SW (2006) Bayesian analysis of input uncertainty in hydrological modeling: 1. Theory. Water Resources Research 42: W03407 (doi:10.1029/2005WR004368). Khu ST and Madsen H (2005) Multiobjective calibration with Pareto preference ordering: An application to rainfall–runoff model calibration. Water Resources Research 41: W03004 (doi:10.1029/2004WR003041). Khu S-T, Savic D, Liu Y, and Madsen H (2004) A fast evolutionary-based metamodelling approach for the calibration of a rainfall–runoff model. In: Pahl-Wostl C, Schmidt S, Rizzoli AE, and Jakeman AJ (eds.) Transactions of the 2nd Biennial

455

Meeting of the International Environmental Modelling and Software Society, vol. 1, pp. 141--146. Manno, Switzerland: iEMSs. Kitanidis PK and Bras RL (1980) Real-time forecasting with a conceptual hydrologic model: Analysis of uncertainty. Water Resources Research 16(6): 1025--1033. Klir G and Folger T (1988) Fuzzy Sets, Uncertainty, and Information. Englewood Cliffs, NJ: Prentice Hall. Kollet S and Maxwell R (2006) Integrated surface–groundwater flow modeling: A freesurface overland flow boundary condition in a parallel groundwater flow model. Advances in Water Resources 29: 945--958. Kollet S and Maxwell R (2008) Demonstrating fractal scaling of baseflow residence time distributions using a fully-coupled groundwater and land surface model. Geophysical Research Letters 35(7): L07402 (doi:10.1029/2008GL033215). Kompare B, Steinman F, Cerar U, and Dzeroski S (1997) Prediction of rainfall runoff from catchment by intelligent data analysis with machine learning tools within the artificial intelligence tools. Acta Hydrotechnica 16/17: 79--94 (in Slovene). Koren VI, Smith M, Wang D, and Zhang Z (2000) Use of soil property data in the derivation of conceptual rainfall–runoff model parameters. 15th Conference on Hydrology, Long Beach, American Meteorological Society, USA, Paper 2.16. Kuczera G and Parent E (1998) Monte Carlo assessment of parameter uncertainty in conceptual catchment models: The Metropolis algorithm. Journal of Hydrology 211: 69--85. Kundzewicz Z (1995) Hydrological uncertainty in perspective. In: Kundzewicz Z (ed.) New Uncertainty Concepts in Hydrology and Water Resources, pp. 3--10. Cambridge: Cambridge University Press. Lamb R and Calver A (2002) Continuous simulation as a basis for national flood frequency estimation. In: Littlewood I (ed.) Continuous River Flow Simulation: Methods, Applications and Uncertainties, pp. 67–75. BHS–Occasional Papers, No. 13, Wallingford. Lamb R, Crewett J, and Calver A (2000) Relating hydrological model parameters and catchment properties to estimate flood frequencies from simulated river flows. BHS 7th National Hydrology Symposium, Newcastle-upon-Tyne, vol. 7, pp. 357–364. Lamb R and Kay AL (2004) Confidence intervals for a spatially generalized, continuous simulation flood frequency model for Great Britain. Water Resources Research 40: W07501 (doi:10.1029/2003WR002428). Laucelli D, Giustolisi O, Babovic V, and Keijzer M (2007) Ensemble modeling approach for rainfall/groundwater balancing. Journal of Hydroinformatics 9(2): 95--106. LeBaron B and Weigend AS (1994) Evaluating neural network predictors by bootstrapping. In: Proceedings of the International Conference on Neural Information Processing (ICONIP’94). Seoul, Korea. Lekkas DF, Imrie CE, and Lees MJ (2001) Improved non-linear transfer function and neural network methods of flow routing for real-time forecasting. Journal of Hydroinformatics 3(3): 153--164. Letcher RA, Jakeman AJ, and Croke BFW (2004) Model development for integrated assessment of water allocation options. Water Resources Research 40: W05502. Lindstro¨m G, Johansson B, Persson M, Gardelin M, and Bergstro¨m S (1997) Development and test of the distributed HBV-96 hydrological model. Journal of Hydrology 201: 272–228. Liong SY and Sivapragasam C (2002) Flood stage forecasting with SVM. Journal of American Water Resources Association 38(1): 173--186. Liu Y and Gupta HV (2007) Uncertainty in hydrologic modeling: Toward an integrated data assimilation framework. Water Resources Research 43: W07401 (doi:10.1029/ 2006WR005756). Loague KM and Freeze RA (1985) A comparison of rainfall–runoff techniques on small upland catchments. Water Resources Research 21: 229–240. Lobbrecht AH and Solomatine DP (1999) Control of water levels in polder areas using neural networks and fuzzy adaptive systems. In: Savic D and Walters G (eds.) Water Industry Systems: Modelling and Optimization Applications, pp. 509--518. Baldock: Research Studies Press. Madsen H and Jacobsen T (2001) Automatic calibration of the MIKE SHE integrated hydrological modelling system. 4th DHI Software Conference, Helsingør, Denmark, 6–8 June 2001. McDonnell JJ (2003) HYPERLINK "http://www.cof.orst.edu/cof/fe/watershd/pdf/2003/ McDonnellHP2003.pdf" Where does water go when it rains? Moving beyond the variable source area concept of rainfall–runoff response. Hydrological Processes 17: 1869–1875. Maskey S, Guinot V, and Price RK (2004) Treatment of precipitation uncertainty in rainfall–runoff modelling: A fuzzy set approach. Advances in Water Resources 27(9): 889--898. McKay MD, Beckman RJ, and Conover WJ (1979) A comparison of three methods for selecting values of input variables in the analysis of output from a computer code. Technometrics 21(2): 239--245. Melching CS (1992) An improved first-order reliability approach for assessing uncertainties in hydrological modelling. Journal of Hydrology 132: 157--177.

456

Hydrological Modeling

Melching CS (1995) Reliability estimation. In: Singh VP (ed.) Computer Models of Watershed Hydrology, pp. 69--118. Highlands Ranch, CO: Water Resources Publications. Merz B and Blo¨schl G (2004) Regionalisation of catchment model parameters. Journal of Hydrology 287(1–4): 95–123. Metropolis N, Rosenbluth AW, Rosenbluth MN, and Teller AH (1953) Equations of state calculations by fast computing machines. Journal of Chemical Physics 26(6): 1087--1092. Milly PCD, Dunne KA, and Vecchia AV (2005) Global patterns of trends in streamflow and water availability in a changing climate. Nature 438: 347--350. Minns AW and Hall MJ (1996) Artificial neural network as rainfall–runoff model. Hydrological Sciences Journal 41(3): 399--417. Mitchell TM (1997) Machine Learning. New York: McGraw-Hill. Montanari A (2007) What do we mean by uncertainty? The need for a consistent wording about uncertainty assessment in hydrology. Hydrological Processes 21(6): 841--845. Montanari A and Brath A (2004) A stochastic approach for assessing the uncertainty of rainfall–runoff simulations. Water Resources Research 40: W01106 (doi:10.1029/ 2003WR002540). Moradkhani H, Hsu KL, Gupta HV, and Sorooshian S (2004) Improved streamflow forecasting using self-organizing radial basis function artificial neural networks. Journal of Hydrology 295(1): 246–262. Moradkhani H and Sorooshian S (2008) General review of rainfall–runoff modeling: Model calibration, data assimilation, and uncertainty analysis. In: Sorooshian S, Hsu K, Coppola E, Tomassetti B, Verdecchia M, and Visconti G (eds.) Hydrological Modelling and the Water Cycle, pp. 1--24. Heidelberg: Springer. Mosley MP (1981) Delimitation of New Zealand hydrologic regions. Journal of Hydrology 49: 173–192. Muleta MK and Nicklow JW (2004) Joint application of artificial neural networks and evolutionary algorithms to watershed management. Journal of Water Resources Management 18(5): 459--482. Nayak PC, Sudheer KP, and Ramasastri KS (2005) Fuzzy computing based rainfall–runoff model for real time flood forecasting. Hydrological Processes 19: 955--968. Neuman SP (2003) Maximum likelihood Bayesian averaging of uncertain model predictions. Stochastic Environmental Research and Risk Assessment 17(5): 291–305. Nor NA, Harun S, and Kassim AH (2007) Radial basis function modeling of hourly streamflow hydrograph. Journal of Hydrologic Engineering 12(1): 113--123. Oki T and Kanae S (2006) Global hydrological cycles and world water resources. Science 313(5790): 1068--1072. Panday S and Huyakorn PS (2004) A fully coupled physically-based spatiallydistributed model for evaluating surface/subsurface flow. Advances in Water Resources 27: 361–382. Pappenberger F and Beven KJ (2006) Ignorance is bliss: Or seven reasons not to use uncertainty analysis. Water Resources Research 42: W05302 (doi:10.1029/ 2005WR004820). Pappenberger F, Harvey H, Beven K, Hall J, and Meadowcroft I (2006) Decision tree for choosing an uncertainty analysis methodology: A wiki experiment. Hydrological Processes 20: 3793--3798. Parkin G, O’Donnell G, Ewen J, Bathurst JC, O’Connell PE, and Lavabre J (1996) Validation of catchment models for predicting land-use and climate change impacts. 1. Case study for a Mediterranean catchment. Journal of Hydrology 175: 595--613. Pesti G, Shrestha BP, Duckstein L, and Boga´rdi I (1996) A fuzzy rule-based approach to drought assessment. Water Resources Research 32(6): 1741--1747. Poff NL, Bledsoe BP, and Cuhaciyan CO (2006) Hydrologic variation with land use across the contiguous United States: Geomorphic and ecological consequences for stream ecosystems. Geomorphology 79: 264--285. Post DA and Jakeman AJ (1996) Relationships between physical attributes and hydrologic response characteristics in small Australian mountain ash catchments. Hydrological Processes 10: 877–892. Post DA, Jones JA, and Grant GE (1998) An improved methodology for predicting the daily hydrologic response of ungauged catchments. Environmental Modeling and Sofware 13: 395–403. Pyle D (1999) Data Preparation for Data Mining. San Francisco, CA: Morgan Kaufmann. Qu Y and Duffy CJ (2007) A semi-discrete finite-volume formulation for multi-process watershed simulation. Water Resources Research 43: W08419 (doi:10.1029/ 2006WR005752). Quinlan JR (1992) Learning with continuous classes. In: Adams N and Sterling L (eds.) AI’92 Proceedings of the 5th Australian Joint Conference on Artificial Intelligence, pp. 343–348. World Scientific: Singapore.

Reed S, Koren V, Smith M, et al. (2004) Overall distributed model intercomparison project results. Journal of Hydrology 298(1–4): 27--60. Refsgaard JC (1996) Terminology, modelling protocol and classification of hydrological model codes. In: Abbott MB and Refsgaard JC (eds.) Distributed Hydrological Modelling, pp. 17--39. Dordrecht: Kluwer. Refsgaard JC (1997) Validation and intercomparison of different updating procedures for real-time forecasting. Nordic Hydrology 28(2): 65–84. Refsgaard JC and Abbott MB (1996) The role of distributed hydrological modeling in water resources management. In: Distributed Hydrologic Modeling, chap. 1. Dordrecht, The Netherlands: Kluwer Academic Publishers. Refsgaard JC and Knudsen J (1996) Operational validation and intercomparison of different types of hydrological models. Water Resources Research 32(7): 2189--2202. Reggiani P, Hassanizadeh EM, Sivapalan M, and Gray WG (1999) A unifying framework for watershed thermodynamics: Constitutive relationships. Advances in Water Resources 23(1): 15--39. Reggiani P, Sivapalan M, Hassanizadeh M, and Gray WG (2001) Coupled equations for mass and momentum balance in a stream network: Theoretical derivation and computational experiments. Proceedings of the Royal Society of London. Series A. Mathematical, Physical and Engineering Sciences 457: 157--189. Reggiani P, Sivapalan M, and Hassanizadeh SM (1998) A unifying framework for watershed thermodynamics: Balance equations for mass, momentum, energy and entropy, and the second law of thermodynamics. Advances in Water Resources 22(4): 367--398. Reggiani P, Sivapalan M, and Hassanizadeh SM (2000) Conservation equations governing hillslope responses: Exploring the physical basis of water balance. Water Resources Research 36(7): 1845--1863. Sachs JS (2004) Sustainable development. Science 304(5671): 649 (doi: 10.1126/ science.304.5671.649). Salathe EP (2005) Downscaling simulations of future global climate with application to hydrologic modelling. International Journal of Climatology 25: 419--436. See LA, Solomatine DP, Abrahart R, and Toth E (2007) Hydroinformatics: Computational intelligence and technological developments in water science applications – editorial. Hydrological Sciences Journal 52(3): 391--396. See LM and Openshaw S (2000) A hybrid multi-model approach to river level forecasting. Hydrological Sciences Journal 45(4): 523--536. Sefton CEM and Howarth SM (1998) Relationships between dynamic response characteristics and physical descriptors of catchments in England and Wales. Journal of Hydrology 211: 1–16. Sefton CEM, Whitehead PG, Eatherall A, Littlewood IG, and Jakeman AJ (1995) Dynamic response characteristics of the Plynlimon catchments and preliminary analysis of relationships to physical catchment descriptors. Environmetrics 6: 465–472. Seibert J (1999) Regionalisation of parameters for a conceptual rainfall–runoff model. Agricultural and Forest Meteorology 98–99: 279–293. Shafer G (1976) Mathematical Theory of Evidence. Princeton, NJ: Princeton University Press. Shamseldin AY and O’Connor KM (2001) A non-linear neural network technique for updating of river flow forecasts. Hydrology and Earth System Sciences 5(4): 557--597. Shannon CE (1948) A mathematical theory of communication. Bell System Technical Journal 27: 379--423. and 623–656. Sharma A (2000) Seasonal to interannual rainfall probabilistic forecasts for improved water supply management: Part 3 – a nonparametric probabilistic forecast model. Journal of Hydrology 239(1–4): 249--258. Shrestha DL and Solomatine D (2008) Data-driven approaches for estimating uncertainty in rainfall–runoff modelling. International Journal of River Basin Management 6(2): 109--122. Singh VP (1995b) Watershed modeling Computer Models of Watershed Hydrology, pp. 1--22. Highlands Ranch, CO: Water Resources Publication. Solomatine D, Dibike YB, and Kukuric N (1999) Automatic calibration of groundwater models using global optimization techniques. Hydrological Sciences Journal 44(6): 879--894. Solomatine D and Shrestha DL (2009) A novel method to estimate total model uncertainty using machine learning techniques. Water Resources Research 45: W00B11 (doi:10.1029/2008WR006839). Solomatine DP (1995) The use of global random search methods for models calibration Proceedings of the 26th Congress of the International Association for Hydraulic Research, vol. 1, pp. 224--229. London: Thomas Telford. Solomatine DP (2006) Optimal modularization of learning models in forecasting environmental variables. In: Voinov A, Jakeman A, and Rizzoli A (eds.) Proceedings of the iEMSs 3rd Biennial Meeting: ‘‘Summit on Environmental Modelling and

Hydrological Modeling

Software’’. Burlington, VT, USA, July 2006. Manno, Switzerland: International Environmental Modelling and Software Society. Solomatine DP and Dulal KN (2003) Model tree as an alternative to neural network in rainfall–runoff modelling. Hydrological Sciences Journal 48(3): 399--411. Solomatine DP, Maskey M, and Shrestha DL (2006) Eager and lazy learning methods in the context of hydrologic forecasting Proceedings of the International Joint Conference on Neural Networks. Vancouver, BC: IEEE. Solomatine DP, Shrestha DL, and Maskey M (2007) Instance-based learning compared to other data-driven methods in hydrological forecasting. Hydrological Processes 22(2): 275--287 (doi:10.1002/hyp.6592). Solomatine DP and Siek MB (2004) Flexible and optimal M5 model trees with applications to flow predictions In: Liong, Phoon, and Babovic (eds.) Proceedings of the 6th International Conference on Hydroinformatics. Singapore: World Scientific. Solomatine DP and Torres LA (1996) Neural network approximation of a hydrodynamic model in optimizing reservoir operation. In: Muller A (ed.) Proceedings of the 2nd International Conference on Hydroinformatics, pp. 201--206. Rotterdam: Balkema. Solomatine DP, Velickov S, and Wust JC (2001) Predicting water levels and currents in the North Sea using chaos theory and neural networks. In: Li, Wang, Pettitjean, and Fisher (eds.) Proceedings of the 29th IAHR Congress. Beijing, China. Solomatine DP and Xue Y (2004) M5 model trees and neural networks: Application to flood forecasting in the upper reach of the Huai River in China. ASCE Journal of Hydrologic Engineering 9(6): 491--501. Sorooshioan S, Duan Q, and Gupta VK (1993) Calibration of rainfall–runoff models: Application of global optimization to the Sacramento Soil Moisture Accounting Model. Water Resources Research 29: 1185–1194. Spear RC (1998) Large simulation models: Calibration, uniqueness, and goodness of fit. Environmental Modeling and Sofware 12: 219–228. Stravs L and Brilly M (2007) Development of a low-flow forecasting model using the M5 machine learning method. Hydrological Sciences–Journal–des Sciences Hydrologiques (Special Issue: Hydroinformatics) 52(3): 466--477. Sudheer KP and Jain SK (2003) Radial basis function neural network for modeling rating curves. ASCE Journal of Hydrologic Engineering 8(3): 161--164. Sugawara M (1995) Tank model. In: Singh VP (ed.) Computer Models of Watershed Hydrology, pp. 165--214. Highlands Ranch, CO: Water Resources Publication. Tang Y, Reed P, Van Werkhoven K, and Wagener T (2007) Advancing the identification and evaluation of distributed rainfall–runoff models using global sensitivity analysis. Water Resources Research 43: W06415.1--W06415.14 (doi:10.1029/ 2006WR005813). Tang Y, Reed P, and Wagener T (2006) How effective and efficient are multiobjective evolutionary algorithms at hydrologic model calibration? Hydrology and Earth System Sciences 10: 289--307. Thiemann M, Trosser M, Gupta H, and Sorooshian S (2001) Bayesian recursive parameter estimation for hydrologic models. Water Resources Research 37(10): 2521--2536. Todini E (1988) Rainfall–runoff modeling – past, present and future. Journal of Hydrology 100(1–3): 341--352. Toth E, Brath A, and Montanari A (2000) Comparison of short-term rainfall prediction models for real-time flood forecasting. Journal of Hydrology 239: 132--147. Tung Y-K (1996) Uncertainty and reliability analysis. In: Mays LW (ed.) Water Resources Handbook, pp. 7.1--7.65. New York, NY: McGraw-Hill. Uhlenbrook S and Leibundgut Ch (2002) Process-oriented catchment modeling and multiple-response validation. Hydrological Processes 16: 423--440. Uhlenbrook S, Roser S, and Tilch N (2004) Hydrological process representation at the meso-scale: The potential of a distributed, conceptual catchment model. Journal of Hydrology 291(3–4): 278--296. Vandewiele GL and Elias A (1995) Monthly water balance of ungauged catchments obtained by geographical regionalization. Journal of Hydrology 170(1–4,8): 277–291. Van Werkhoven K, Wagener T, Tang Y, and Reed P (2008) Understanding watershed model behavior across hydro-climatic gradients using global sensitivity analysis. Water Resources Research 44: W01429 (doi:10.1029/2007WR006271). Vapnik VN (1998) Statistical Learning Theory. New York: Wiley. Vernieuwe H, Georgieva O, De Baets B, Pauwels VRN, Verhoest NEC, and De Troch FP (2005) Comparison of data-driven Takagi–Sugeno models of rainfall-discharge dynamics. Journal of Hydrology 302(1–4): 173--186. Vo¨ro¨smarty CJ, Green P, Salisbury J, and Lammers RB (2000) Global water resources: Vulnerability from climate change and population growth. Science 289: 284–288 (doi:10.1126/science.289.5477.284). Vrugt J, ter Braak C, Clark M, Hyman J, and Robinson B (2008a) Treatment of input uncertainty in hydrologic modeling: Doing hydrology backward with Markov chain Monte Carlo simulation. Water Resources Research 44: W00B09 (doi:10.1029/ 2007WR006720). Vrugt JA, Gupta HV, Bouten W, and Sorooshian S (2003) A shuffled complex evolution metropolis algorithm for optimization and uncertainty assessment of hydrologic

457

model parameters. Water Resources Research 39(8): 1201 (doi:10.1029/ 2002WR001642). Vrugt JA and Robinson BA (2007) Improved evolutionary optimization from genetically adaptive multimethod search. Proceedings of the National Academy of Sciences 104(3): 708--711. Vrugt JA, ter Braak CJF, Gupta HV, and Robinson BA (2008b) Equifinality of formal (DREAM) and informal (GLUE) Bayesian approaches in hydrologic modeling? Stochastic Environmental Research and Risk Assessment 23: 1011--1026 (doi:10.1007/s00477-008-0274-y). Wagener T and Gupta HV (2005) Model identification for hydrological forecasting under uncertainty. Stochastic Environmental Research and Risk Assessment 19: 378--387 (doi:10.1007/s00477-005-0006-5). Wagener T and McIntyre N (2007) Tools for teaching hydrological and environmental modeling. Computers in Education Journal XVII(3): 16–26. Wagener T, McIntyre N, Lees MJ, Wheater HS, and Gupta HV (2003) Towards reduced uncertainty in conceptual rainfall–runoff modelling: Dynamic identifiability analysis. Hydrological Processes 17(2): 455--476. Wagener T and Wheater HS (2006) Parameter estimation and regionalization for continuous rainfall–runoff models including uncertainty. Journal of Hydrology 320(1–2): 132--154. Wagener T, Wheater HS, and Gupta HV (2004) Rainfall–Runoff Modelling in Gauged and Ungauged Catchments. 332pp. London: Imperial College Press. Weiskel PK, Vogel RM, Steeves PA, DeSimone LA, Zarriello PJ, and Ries KG III (2007) Water-use regimes: Characterizing direct human interaction with hydrologic systems. Water Resources Research 43: W04402 (doi:10.1029/2006WR005062). Wheater HS, Jakeman AJ, and Beven KJ (1993) Progress and directions in rainfall– runoff modelling. In: Jakeman AJ, Beck MB, and McAleer MJ (eds.) Modelling Change in Environmental Systems, pp. 101--132. Chichester: Wiley. Werner M (2008) Open model integration in flood forecasting. In: Abrahart B, See LM, and Solomatine DP (eds.) Practical Hydroinformatics: Computational Intelligence and Technological Developments in Water Applications, p. 495. Heidelberg: Springer. Widen-Nilsson E, Halldin S, and Xu C-Y (2007) Global water-balance modelling with WASMOD-M: Parameter estimation and regionalization. Journal of Hydrology 340: 105--118. Xiong LH, Shamseldin AY, and O’Connor KM (2001) A non-linear combination of the forecasts of rainfall–runoff models by the first-order Takagi–Sugeno fuzzy system. Journal of Hydrology 245(1–4): 196--217. Yadav M, Wagener T, and Gupta HV (2007) Regionalization of constraints on expected watershed response behavior for improved predictions in ungauged basins. Advances in Water Resources 30: 1756--1774 (doi:10.1016/ j.advwatres.2007.01.005). Young PC (1992) HYPERLINK "http://eprints.lancs.ac.uk/21104/" Parallel processes in hydrology and water quality: A unified time-series approach. Water and Environment Journal 6(6): 598–612. Young PC (1998) Data-based mechanistic modelling of environmental, ecological, economic and engineering systems. Environmental Modeling and Sofware 13: 105–122. Zadeh LA (1965) Fuzzy sets. Information and Control 8(3): 338--353. Zadeh LA (1978) Fuzzy set as a basis for a theory of possibility. Fuzzy Sets and Systems 1(1): 3--28. Zhang Z, Wagener T, Reed P, and Bushan R (2008) Ensemble streamflow predictions in ungauged basins combining hydrologic indices regionalization and multiobjective optimization. Water Resources Research 44: W00B04 (doi:10.1029/ 2008WR006833). Zhao RJ and Liu XR (1995) The Xinanjiang model. In: Singh VP, et al. (eds.) Computer Models of Watershed Hydrology, pp. 215--232. Littelton, CO: Water Resources Publication. Zimmermann H-J (1997) A fresh perspective on uncertainty modeling: Uncertainty vs. uncertainty modeling. In: Ayyub BM and Gupta MM (eds.) Uncertainty Analysis in Engineering and Sciences: Fuzzy Logic, Statistics, and Neural Network Approach, pp. 353--364. Heidelberg: Springer.

Relevant Websites http://www.deltares.nl Deltares. http://www.dhigroup.com DHI; DHI software. http://efas.jrc.ec.europa.eu European Commission Joint Research Centre. http://www.sahra.arizona.edu SAHRA; Hydroarchive.

2.17 Uncertainty of Hydrological Predictions A Montanari, University of Bologna, Bologna, Italy & 2011 Elsevier B.V. All rights reserved.

2.17.1 Introduction 2.17.2 Definitions and Terminology 2.17.2.1 Probability 2.17.2.2 Randomness 2.17.2.3 Random Variable 2.17.2.4 Stochastic Process 2.17.2.5 Stationarity 2.17.2.6 Ergodicity 2.17.2.7 Uncertainty 2.17.2.8 Global Uncertainty and Individual Uncertainties 2.17.2.9 Uncertainty Assessment 2.17.2.10 Probabilistic Estimate/Estimation/Assessment of Uncertainty (Probabilistic Uncertainty) 2.17.2.11 Nonprobabilistic Estimate/Estimation/Assessment of Uncertainty (Nonprobabilistic Uncertainty) 2.17.2.12 Confidence Band 2.17.2.13 Equifinality 2.17.2.14 Behavioral Model 2.17.3 Classification of Uncertainty and Reasons for the Presence of Uncertainty in Hydrology 2.17.3.1 Inherent Randomness 2.17.3.2 Model Structural Uncertainty 2.17.3.3 Model Parameter Uncertainty 2.17.3.4 Data Uncertainty 2.17.3.5 Operation Uncertainty 2.17.4 Uncertainty Assessment 2.17.5 Classification of Approaches to Uncertainty Assessment 2.17.5.1 Research Questions about Uncertainty in Hydrology 2.17.5.2 An Attempt of Classification 2.17.6 Assessment of the Global Uncertainty of the Model Output 2.17.6.1 Analytical Methods 2.17.6.2 The Generalized Likelihood Uncertainty Estimation 2.17.6.3 The Bayesian Forecasting System 2.17.6.4 Techniques Based on the Statistical Analysis of the Model Error 2.17.6.5 Bayesian Model Averaging 2.17.6.6 Machine Learning Techniques 2.17.7 Assessment of Data Uncertainty 2.17.7.1 Precipitation Uncertainty 2.17.7.2 River Discharge Uncertainty 2.17.8 Assessment of Parameter Uncertainty 2.17.8.1 The MOSCEM-UA Method 2.17.8.2 The AMALGAM Method 2.17.9 Assessment of Model Structural Uncertainty 2.17.10 Uncertainty Assessment as a Learning Process 2.17.11 Conclusions Acknowledgments References

2.17.1 Introduction Hydrological modeling is receiving increasing attention from researchers and practitioners. The increasing availability of mathematical tools and computing power together with an improved understanding of the dynamics of hydrological processes has favored the continuous development of new

459 460 460 460 461 461 461 461 461 461 462 462 462 462 462 462 462 463 463 463 464 464 464 465 465 465 466 466 466 467 469 470 470 471 471 472 473 473 474 474 474 475 475 475

modeling approaches in the past few decades (see Chapter 2.16 Hydrological Modeling). Hydrological modeling is an attractive option today for solving many practical problems of environmental engineering, flood protection, water resource management, and applied hydrology in general. Setting up a hydrological model in order to solve a practical problem requires the application of proper procedures of

459

460

Uncertainty of Hydrological Predictions

model identification, parameter calibration, hypothesis testing, model testing (also called model validation), and uncertainty assessment. The above procedures are often strictly related and are the subject of an increasing research activity by hydrologists. In particular, uncertainty estimation is very much related with parameter calibration and model validation. It consists of a verification of the hydrological model appropriateness and performances finalized to providing a quantitative assessment of its reliability. As a matter of fact, uncertainty estimation in hydrological surface and subsurface modeling is today one of the most important subfields of hydrology, according to the numerous contributions in recent scientific literature. Uncertainty reduction is also one of the main goals of the Prediction in Ungauged Basins (PUB) initiative promoted by the International Association of Hydrological Sciences. While quantitative uncertainty assessment in hydrology is often considered a relatively new topic, it is worth noting that hydrologists were aware of uncertainty and used to deal with it because the first hydrological studies and applications were carried out. In particular, empirical techniques were used to compensate for the lack of information about model reliability. For instance, hydrologists are well used to adopt safety factors or allowance for freeboard, which are usually set basing on consensus, expert opinion, and empirical evidence. These safety factors were the first and very useful tools to take into account inherent uncertainty and imperfect knowledge of hydrological processes in hydrological design. However, expert knowledge is by itself subjective and referred to specific contexts and situations. The call for a generalized and systematic approach to uncertainty estimation in hydrology is the motivation for the renewed interest in the past few years. One of the reasons why uncertainty assessment in hydrology was not much investigated on theoretical basis until the recent past is that hydrological modeling itself is a relatively young discipline. In fact, the first hydrological models were the rational formula proposed by Kuichling (1889) (although the principles of the method were introduced by Mulvaney (1851)), and the unit hydrograph model proposed by Sherman (1932). Most of the hydrological models that are used today were proposed after the 1960s. The interest in new techniques for uncertainty assessment was stimulated by Spear and Hornberger (1980), who introduced the generalized sensitivity analysis methodology, also known as regional sensitivity analysis. Their work inspired the development of the generalized likelihood uncertainty estimation (GLUE; Beven and Binley, 1992; see Section 2.17.6.2), which works under the hypothesis that different sets of model parameters/ structures may be equally likely as simulators of the real system. In the 1990s the emerging need for reliable techniques for uncertainty estimation, for the multitude of modeling situations and approaches that are experienced in hydrology, stimulated the development of many methods (for a long, though still incomplete, list one can refer to Liu and Gupta (2007) and Matott et al. (2009)). Another reason limiting the use of uncertainty assessment methods is that the transfer of the know-how about uncertainty in hydrology from scientists to end-users was and still is, difficult, notwithstanding the relevant research activity mentioned above. Pappenberger and Beven (2006) provided

an extensive analysis of this issue. A relevant problem today is that uncertainty assessment in hydrology suffers from the lack of a coherent terminology and a systematic approach. The result of this situation is that it is extremely difficult (if not impossible) to obtain a coherent picture of the available methods. This lack of clarity is an example of linguistic uncertainty (Regan et al., 2003). Therefore, much is still to be done to reach a coherent treatment of the topic. Quantitative uncertainty assessment in conditions of data scarcity is a very difficult task, if not impossible, in some cases. Usually, uncertainty estimation in applied scientific modeling is dealt with by comparing the model output with observed data, by borrowing concepts from statistics. According to this procedure, model reliability is quantified in a probabilistic framework. However, statistical testing becomes not as reliable in situations of data scarcity and therefore the use of statistical concepts for uncertainty assessment in hydrology sometimes may not be appropriate. This is one of the reasons why hydrologists are looking for different procedures that can be complementary or alternative to statistics. Moreover, uncertainty in hydrology might arise from limited knowledge (epistemic uncertainty, see Section 2.17.3) or from natural variability. In the former case, we deal with uncertainties that might not be aleatory in nature. They can be treated with statistical methods (e.g., the BATEA method, see Section 2.17.7.1), but many authors question the validity of statistics in this case and prefer nonstatistical approaches. These procedures are generally conceived in order to allow incorporation of expert knowledge in a theoretically based framework. They are characterized by a certain degree of subjectivity, which needs to be reduced as much as possible in order to allow their application in situations where knowledge is lacking. Therefore, different philosophies and approaches for quantifying the reliability of hydrological models were recently proposed. As a result, an active debate recently began about the relative advantages of each of them. Such debate in many cases assumed a philosophical behavior, because the philosophy underlying each method is one of the main subjects of the discussion. On the one hand, such a debate stimulated additional developments and insights in itself; on the other, it is still not clear which approach is most appropriate given the needs of the user. For this reason, the hydrologic scientific community still calls for more pragmatism in uncertainty estimation. On the one hand, hydrology is a science where uncertainty is very significant. Progress in monitoring techniques, process understanding, and modeling will certainly reduce uncertainty in the future but will never eliminate it. On the other hand, hydrologists are in charge of providing design variables that play a fundamental role in water engineering, civil protection, and water resource management. Therefore, it is clear that the efficient real world use of an uncertain design variable should necessarily be based on uncertainty assessment. This chapter aims at presenting a comprehensive introduction to the subject of uncertainty assessment in hydrology. After presenting a brief glossary and a discussion about the reasons for the presence of uncertainty in hydrology, a review of the most-used approaches to uncertainty assessment is presented.

Uncertainty of Hydrological Predictions

2.17.2 Definitions and Terminology There is currently a linguistic uncertainty affecting the topic of uncertainty assessment in hydrology (Regan et al., 2003; Beven, 2009), meaning that an agreed terminology is still lacking. Some basic definitions are provided in the following.

2.17.2.1 Probability Probability can be defined in different ways. In fact, probability is currently interpreted according to two broad and distinguished views. The classical frequentist view of probability defines the probability of an event occurring in a particular trial as the frequency with which it occurs in a long sequence of similar trials. In a Bayesian or subjectivist view, the probability of an event is dependent upon the state of information available and this information can include expert opinion. Probability theory forms the basis of classical statistics, which has estimators based on a likelihood function that represents how likely an observed data sample is for a given model and parameter set.

2.17.2.2 Randomness Randomness is a term that is used within science with different meanings. In statistics, and hydrology as well, a random process is such that its outcome cannot be predicted deterministically. Randomness does not imply lack of knowledge about the process dynamics or impossibility to set up a deterministic model for it. However, if a deterministic model can be set up for a process, randomness implies that such a model cannot perfectly predict the process outcome. For instance, in the case of a roulette wheel, if the geometric and dynamic behaviors of the system are perfectly known, then the number on which the ball will stop would be a certainty. However, one is fully aware that even a small imperfection in the description of the geometry of the system and/or its initial conditions makes the outcome of the experiment unpredictable. A probabilistic description can thus be more useful than a deterministic one for analyzing the pattern of outcomes of repeated rolls of a roulette wheel. Physicists face the same situation in kinetic theory of gases, where the system, while deterministic in principle, is very complex so that only a statistical description of its properties is feasible. Another example is the experiment of dropping balls into a spiked sieve. Here, the geometry of the system is perfectly known as well as the initial and boundary conditions. However, once a ball is dropped in the sieve, it is impossible to predict deterministically its trajectory, because no one can predict which way the ball will follow after hitting a spike. However, the distribution of the balls at the bottom of the sieve is well known to be Gaussian. In this case, the full comprehension of the geometry and dynamics of the system does not allow one to set up a deterministic description, while a stochastic description can provide a satisfactory model. Actually, one cannot exactly predict the number of balls in each bar, but the probabilistic prediction will have a small uncertainty.

461

An important discovery of the twentieth-century physics was the random character of all physical processes that occur at subatomic scales and are governed by the laws of quantum mechanics. This means that probability theory is required to describe nature. This type of interpretation was questioned by many scientists, as the famous quote by Albert Einstein, from a letter to Max Born, clearly testifies: ‘‘I am convinced that He does not play dice.’’ A similar controversy currently occurs in hydrology (for an interesting discussion, see Koutsoyiannis et al. (2009)). The trend toward the so-called physically based models induced in the last few decades the inspiration to pursue a completely deterministic description of hydrological systems, through a better understanding of the internal dynamics of hydrological processes. However, such deterministic description is so complicated that only a probabilistic treatment is possible. This does not mean that knowledge is unuseful. On the contrary, it allows one to set up a plausible probabilistic description of the random outcome.

2.17.2.3 Random Variable A random variable maps all possible outcomes from a random event into the real numbers. As such, it is affected by uncertainty and cannot be deterministically predicted. Random variables can assume discrete and continuous values.

2.17.2.4 Stochastic Process A stochastic process can be defined as a collection of random variables. For instance, if we assume that the river flow at time t is a random variable, then the time series of river flow observations during an assigned observation period is a realization of a stochastic process. While a deterministic process gives only one possible value of its output under assigned initial and boundary conditions (as it is the case, e.g., for the solution of an ordinary differential equation), the output of a stochastic process is affected by some uncertainty that is described by the corresponding probability distributions. This means that there are many possible paths for the evolution of the process, with some of them being more likely and others less. A stochastic process can assume discrete or continuous values. Although the random variables of a stochastic process may be independent, in most commonly considered situations in hydrology, they exhibit statistical correlations. A stochastic process can include a deterministic representation but always includes a random component which makes its output uncertain.

2.17.2.5 Stationarity A stochastic process is strictly stationary when the joint probability distribution of an arbitrary number of its random variables does not change when shifted in time or space. As a result, parameters such as the statistics of the process also do not change over time or position. Stationarity is a property of the mathematical representation of the system, or an ensemble of outcomes from a repeatable experiment, and therefore does not constitute an actual property of the natural process itself. This latter follows just one trajectory and therefore its outcome is unique, because nature and life do not

462

Uncertainty of Hydrological Predictions

enable repeatability. Stationarity is a property that is used in statistics in order to make inference about the physical process and therefore does not imply any assumption on the natural process itself. It is interesting to mention that the opposite of stationarity is nonstationarity, which implies that the above statistics change accordingly to deterministic functions of time, where deterministic means that the above-mentioned functions should be known independently of the data and should apply to any time, past, present, and future (Papolulis, 1991). Conversely, if the above functions are random (i.e., realizations of stationary stochastic processes), then the process is stationary. The concept of stationarity is a way to find invariant properties in complex natural systems. In view of what was anticipated above, it is important to note that stationarity does not imply that the statistics of a realization of a process are constant in time. Actually, such statistics are affected by sampling variability and therefore they certainly change after a time shift. The crucial issue is to detect if such a change exists in the process and can be expressed through a deterministic function of time. Recently, the scientific literature presented contributions stating that stationarity is dead because of hydrological change and climate change. Actually, stationarity is an assumtpion and therefore can hardly be dead.

2.17.2.6 Ergodicity A stochastic process is said to be ergodic if its statistical properties can be deduced from a single, sufficiently long sample (realization) of the process.

2.17.2.7 Uncertainty Uncertainty can be defined as an attribute of information (Zadeh, 2005; Montanari, 2007). In the context of hydrology, uncertainty is generally meant to be a quantitative indication of reliability for a given hydrological quantity, either observed or inferred by using models. The indication of reliability can be provided by estimating the error affecting the quantity or the expected range of variability (due to uncertainty) for the quantity itself. Uncertainty can be broadly grouped into two major categories, namely, aleatory and epistemic uncertainty (see Section 2.17.3), and can be inferred by using probabilistic or nonprobabilistic methods.

2.17.2.8 Global Uncertainty and Individual Uncertainties Global uncertainty can be defined as the discrepancy between the model output and the true value of the corresponding variable. Different uncertainties can compensate each other in the formation of the global uncertainty; for instance, parameter errors can compensate, at least in part, for data errors and model structural errors. These different uncertainties are termed individual uncertainties and are specifically referred to with a terminology which recall their causal origin, such as parameter uncertainty and model structural uncertainty (see Section 2.17.3.2 for an extended description). The terms above are not formally defined and therefore some linguistic uncertainty is present. For instance, the terms parameter

uncertainty, input uncertainty, and model structural uncertainty should be used to indicate the uncertainty affecting the model parameters, input, and structure, respectively. Hereafter this is the meaning that will be used in this chapter. However, these terms are sometimes used to indicate the part of uncertainty in the model output that is caused by imperfect parameters, input, and model structure, respectively. While global uncertainty is relatively easy to estimate a posteriori, for instance, by computing the difference between the model output and the corresponding observed variable (under the assumption that this latter is correct), the identification of the contribution of individual uncertainties above is impossible, unless assumptions are introduced or independent observations are available (see Section 2.17.4). This means that it is usually difficult, if not impossible, to assess whether the model performance is affected by, say, a parameter error rather than a model structural error.

2.17.2.9 Uncertainty Assessment In what follows, we refer to uncertainty assessment to mean a quantitative evaluation of uncertainty affecting a hydrological variable, parameter, or model. Uncertainty estimation and uncertainty quantification will be considered synonymous with uncertainty assessment, which is different from uncertainty analysis and uncertainty modeling. The former is a preliminary step of uncertainty assessment aimed at identifying the reasons for the presence of uncertainty and the nature of uncertainty itself, while the latter term refers to the tools that are used for uncertainty assessment.

2.17.2.10 Probabilistic Estimate/Estimation/Assessment of Uncertainty (Probabilistic Uncertainty) We will use the term probabilistic estimate of uncertainty to mean that uncertainty estimation for a given hydrological quantity has been carried out consistently with formal probability theory. In the probabilistic approach, uncertainties are characterized by the probabilities associated with events. Therefore, if one refers to the output of a hydrological model, the related probability distribution should actually provide an estimate of the frequency with which the true values fall within a given range.

2.17.2.11 Nonprobabilistic Estimate/Estimation/ Assessment of Uncertainty (Nonprobabilistic Uncertainty) Nonprobabilistic methods to uncertainty estimation in hydrology are frequently applied. Nonprobabilistic methods are various generalizations of probability theory that have emerged since the 1950s, including random set theory, evidence theory, fuzzy set theory, and possibility theory (Jacquin and Shamseldin, 2007). In particular, fuzzy set theory and possibility theory have received considerable attention from hydrologists, because much human reasoning about hydrological systems is possibilistic rather than strictly probabilistic. We reason about whether a given scenario could happen, without necessarily endeavoring to attach probabilities to the likelihood of it happening, particularly in situations of very scarce information.

Uncertainty of Hydrological Predictions

2.17.2.12 Confidence Band A range around an estimated quantity that encompasses the true value with a probability 1  a, where a is the significance level and 1  a is the confidence level. It is worth pointing out that the terminology is sometimes ambiguous. Some authors use the term confidence band or confidence interval when referring to the distribution of estimates that cannot be observed (e.g., a model parameter), while the term prediction interval is used when referring to the distribution of future values. Moreover, some authors indicate with the term tolerance interval a range in the observations that encompasses a 1  a proportion of the population of the related random variable. For more details, the reader is reffered to Hahn and Meeker (1991). Figure 1 shows an example of confidence bands computed with the meta-Gaussian approach (Montanari and Brath, 2004; see Section 2.17.6.2) for river flow simulations referred to the Samoggia River at Calcara (Italy). It is interesting to note that the shape of the confidence bands themselves provides indications about the goodness of the fit provided by the model. Moreover, the skew in the prediction distribution results indicates that a systematic error is likely to be present.

2.17.2.13 Equifinality

River flow (m3 s−1)

Equifinality implies that in a system interacting with its environment a given end state can be reached by more than one potential mean. The term is due to von Bertalanffy (1968), the founder of general systems theory. The idea of equifinality suggests that similar results may be achieved with different initial conditions, different model parameters, and different model structures. In hydrology the concept of equifinality was introduced by Beven (1993) as an unavoidable effect of the presence of uncertainty. For an extended discussion, see Beven

Observed Simulated 95% confidence bands

110 100 90 80 70 60 50 40 30 20 10 0 0

(a)

(2006a). Equifinality leads to the idea of multimodeling solutions in hydrology (see Section 2.17.6.5).

2.17.2.14 Behavioral Model Within the context of equifinality, a behavioral model is one that provides an acceptable simulation of observed natural processes. In a multimodel approach, the collection of behavioral models provides a means for assessing the uncertainty of their output (see Section 2.17.6.5).

2.17.3 Classification of Uncertainty and Reasons for the Presence of Uncertainty in Hydrology There have been many attempts presented by the literature to classify uncertainty in hydrology. The proposed solutions were not always in agreement because, given the uncertain nature of hydrological processes, it is sometimes impossible to unambiguously decipher the reason for the presence of errors. It is generally agreed that uncertainties can be grouped into two major categories: (1) natural variability (also called structural uncertainty, aleatory, external, objective, inherent, random, irreducible, or stochastic uncertainty) and (2) knowledge uncertainty (also called epistemic, functional, internal, reducible, or subjective uncertainty (Table 1 in NRC, 2000; Koutsoyiannis et al., 2009; Hall and Solomatine, 2008). These two categories have different ramifications. In fact, the global uncertainty of a given model or variable may be characterized in three ways: purely structural, partly epistemic and partly structural, and purely epistemic (Cullen and Frey, 1999). When evaluating model performances and when possible, the different types of uncertainty should be separated (Cullen and Frey, 1999; Hoffman and Hammonds, 1994; Nauta, 2000; Sonich-Mullin, 2001). However, this is not always possible and therefore epistemic uncertainty and natural variability are often dealt with in an integrated fashion. Other classifications were proposed. According to the causes for the presence of uncertainty in hydrology (which nevertheless are not always identifiable), one may identify the following categories: (1) inherent randomness (the geometry 200

Observed river flow (m3 s−1)

In a more general context, we will refer to nonprobabilistic uncertainty when the estimation is carried out by using other approaches than formal probabilistic ones. This category includes probabilistic methods where some of the underlying assumptions are relaxed.

463

175 150 125 100 75 50 25 0

10

20

30

40

Hours (from 4 p.m. of 06/23/1995)

50

0 (b)

25

50

75

100

125

150

175

200

Simulated river flow (m3 s−1)

Figure 1 (a) Example of confidence bands computed with the meta-Gaussian approach (Montanari and Brath (2004); see Section 2.17.6.2) for a flood event occurred in the Samoggia River at Calcara (Italy) in 1995. (b) Example of confidence bands computed with the meta-Gaussian approach (Montanari and Brath (2004); see Section 2.17.6.2) drawn on a scatterplot of observed versus simulated hourly river flows for the Samoggia River at Calcara (Italy) during the years 1995–97.

464

Uncertainty of Hydrological Predictions

Table 1

Uncertainty assessment methods in hydrology, along with their classification (see Section 2.17.5) and purpose (see Sections 45.6–45.10)

Assessment method

Classification

Type of uncertainty estimated

AMALGAM BATEA BFS BMA DYNIA GLUE

Nonprobabilistic, parameter estimation Probabilistic, parameter estimation, uncertainty assessment, sensitivity analysis Probabilistic, Bayesian Probabilistic, multimodel Nonprobabilistic, identifiability analysis Nonprobabilistic (when an informal likelihood is used), parameter estimation, uncertainty assessment, sensitivity analysis Probabilistic, parameter estimation, uncertainty assessment, sensitivity analysis

Parameter Precipitation induced Global Global Parameter Global, parameter, data, structural

IBUNE Machine learning Meta-Gaussian MOSCEM-UA SCE-UA

Nonprobabilistic

Global, precipitation induced, model structure induced Usually global, in principle all

Probabilistic, data analysis Nonprobabilistic, parameter estimation, sensitivity analysis Probabilistic, parameter estimation

Global Parameter Parameter

Classification is ambiguous in some cases; it distinguishes between probabilistic and nonprobabilistic methods, as well as among the seven categories introduced by Matott et al. (2009) (see Section 2.17.5.2). AMALGAM, a multialgorithm genetically adaptive method for multiobjective optimization; BATEA, Bayesian total error analysis; BFS, Bayesian forecasting system; BMA, Bayesian multimodel analysis; DYNIA, dynamic identifiability analysis; GLUE, generalized likelihood uncertainty estimation; IBUNE, integrated Baysian uncertainty estimator; MOSCEM-UA, multiobjective shuffled complex evolution University of Arizona; SCE-UA, shuffled complex evolution university of Arizona. References for the methods are in the text.

of the control volumes, the weather, etc.); (2) model structural uncertainty that reflects the inability of a model to represent precisely the true behavior of the system; (3) model parameter uncertainty; and (4) data uncertainty. When using models to make engineering or management decisions about hydrologic systems we also have to deal with (5) operation uncertainties (associated with construction and maintenance; Loucks and Van Beek, 2005). The above sources of uncertainty are briefly discussed in the following. It is generally agreed that uncertainty in hydrology cannot be eliminated, no matter if it is epistemic in nature or induced by inherent randomness. For instance, rainfall inputs to a catchment might be highly structured, with different structures in different events that lead to nonrandom errors in estimates of areal rainfall. This type of error can be reduced with new measurement techniques but cannot be fully removed.

2.17.3.1 Inherent Randomness Inherent randomness is one of the main reasons for the presence of uncertainty and is a intrinsic behavior of hydrological processes. For instance, a deterministic description of subsurface flow paths is impossible. Different soils and rocks, irregular macropores, faults and cracks with their heterogeneous patterns in both space and time, combined with two phase flows, varying wetting fronts, form a extremely complex system, for which a deductive description is impossible (Koutsoyiannis et al., 2009). Inherent randomness emerges also from meteorology, variability of the surface flow paths, and so on. It is not only related to a coarse description of the system (that also induces uncertainty, which is nevertheless epistemic at least in part; in fact, it could be reduced by an increased capability to monitor the processes at finer spatial and temporal scales) but rather related to the effective impossibility to describe deterministically the inherent variability of the process. Inherent randomness has been long

discussed by the hydrologic community in the recent past (Koutsoyiannis et al., 2009; Koutsoyiannis, 2009). It has been argued that dynamical systems theory has well shown that uncertainty can emerge even from an insignificant perturbation of the initial conditions of a pure, simple, and fully known deterministic (chaotic) dynamics.

2.17.3.2 Model Structural Uncertainty In the ideal situation in which perfect input data are available and model parameters are perfect, model structural uncertainty is defined as the uncertainty in the model output induced by the inhability of the hydrological model to perfectly reproduce the dynamics of hydrological systems. This means that the model output would still be uncertain even in the ideal situation in which no other uncertainties are present. Model structural uncertainty can be induced by imperfect model structure or lack of computational power. If the reason for the presence of uncertainty is an incorrect selection of the model or the computational tools, then model structural uncertainty is epistemic; on the other hand, the effective impossibility to describe the system with a mathematical model induces the presence of irreducible uncertainty. Given that model structural uncertainty is epistemic at least in part, the search for improved modeling tools has been the main focus of the hydrologic scientific community in the last few decades.

2.17.3.3 Model Parameter Uncertainty Model parameter uncertainty is the result of the lack of a sufficiently extended database of good quality, or the inefficiency of the optimization algorithm and/or the related objective function, which induce parameter estimates to be significantly uncertain even if a perfect model and a perfect knowledge of the system were available. This is a relevant

Uncertainty of Hydrological Predictions

problem in all the hydrological applications and motivates the intense efforts that were dedicated to parameter estimation and related uncertainty assessment (e.g., Ibbitt and O’Donnell, 1974; Alley, 1984; Kleisenn et al., 1990; Gan and Burges, 1990; Duan et al., 1992; Brath et al., 2004; Vrugt et al., 2003a; 2003b; Vrugt and Robinson, 2007). Parameter estimation can be coupled with sensitivity analysis and model diagnostic to identify the most sensitive parameters in periods of model failures, thus gaining insights into the reasons for model inadequacy (Sieber and Uhlenbrook, 2005). To this end, Wagener et al. (2003) proposed the dynamic identifiability analysis (DYNIA). Usually, an objective function is used to calibrate the model parameters to observed data. Independently of the objective function and the tools employed to optimize the model parameters, most hydrological models suffer from the existence of multiple optima of the objective function itself and the presence of high interaction or correlation among the parameters. These problems make parameter calibration uncertain even when a relatively large database is at disposal (Kuczera and Mroczkowski, 1998). Parameter uncertainty also arises when the parameters are not calibrated but rather estimated on the basis of field surveys or expert knowledge, for instance, while defining land-cover parameters. Parameter uncertainty can be epistemic at least in part.

2.17.3.4 Data Uncertainty Data uncertainty is an emerging problem that is gaining renewed attention by hydrologists in the recent past (see, e.g., Di Baldassarre and Montanari, 2009; Dottori et al., 2009; Koussis, 2009; Petersen-Øverleir and Reitan, 2009). In fact, even modern technologies cannot avoid the presence of a significant approximation in observations of, say, rainfall, river flows (for both low and high flows), and so forth. Data uncertainty emerges from limitation of the monitoring techniques (instrumentation error, rating curve approximations, etc.) or variability of the spatial and temporal distribution of the observed hydrological variables (spatial variability of rainfall, time variability of streamflow, etc.). It follows that hydrological models are optimized against imperfect data and therefore an error is induced in hydrological simulations. Data uncertainty has both epistemic and aleatory components and therefore it is particularly important how observation errors are treated. Some authors claim that treating data error with purely statistical approaches may induce overconditioning in hydrological modeling (Beven, 2006a).

2.17.3.5 Operation Uncertainty Operation uncertainty arises when hydrological models are used in the real world. In fact, it is well known that in realtime applications uncertainties of different nature are present that do not affect off-line exercises. Often the data and the initial and boundary conditions cannot be preliminary checked, the computational time might become a relevant constraint, the end-users operate under stress and therefore the human error becomes more likely, there is a weak ability to identify decision criteria, communication becomes difficult, and so forth. Operation uncertainty is difficult to assess, is

465

rarely considered by researchers, and represents an emerging awareness among hydrological modelers and end-users. As a matter of fact, the identification of the most suitable model should be carried out in view of operation uncertainty as well. Data assimilation can be used to constrain uncertainty during model application.

2.17.4 Uncertainty Assessment It is well known that uncertainty assessment in hydrology is a topical issue. Already in 1905, W.E. Cooke, who was issuing daily weather forecasts in Australia, stated: ‘‘It seems to me that the condition of confidence or otherwise form a very important part of the prediction, and ought to find expression.’’ Uncertainty assessment in hydrology involves the analysis of multiple sources of error, the main ones being outlined in Section 2.17.3. The contribution of these latter to the formation of the global uncertainty cannot be quantified independently, unless (1) one is willing to introduce subjective assumptions about the nature of the individual error components or (2) independent observations are available for estimating each source of error. As an example for the latter solution, the reader is referred to Winsemius et al. (2006, 2008) where gravity and evaporation measurements are used to constrain the water balance and the land surface parameters, respectively, for a rainfall–runoff model. However, in some hydrological applications it is not necessary to separate different sources of error. For this reason in many cases, uncertainty is assessed in an aggregated solution, therefore quantifying global uncertainty.

2.17.5 Classification of Approaches to Uncertainty Assessment This section aims to propose a classification of uncertainty assessment methods in hydrology. Classifying the methods is useful to clarify their behavior and operational purpose. However, it should be premised that such a classification might be subjective, because some methods lend themselves to different interpretations of their nature and scope.

2.17.5.1 Research Questions about Uncertainty in Hydrology The uncertain nature of hydrology has pushed hydrologists to raise many questions related to uncertainty assessment. The most urgent ones are those related to quantifying the reliability of the output variables of hydrological models (forecasts, simulations, etc.). Hydrological simulation is often used in real-time prediction systems for natural hazards or for assessing long-term effects of climate change or for assessing the reliability of proposed water resource management strategies. In these cases, quantifying the uncertainty of the hydrological model response is extremely important from a societal point of view. Uncertainty assessment in hydrology includes additional research issues. Among them, there is the call for assessing the uncertainty of observed data, model parameters, and model structure. These issues are also significant for gaining further

466

Uncertainty of Hydrological Predictions

insight into the dynamics of hydrological processes. Indeed, to identify the most appropriate model is a means to provide support to hydrological theory. Therefore, uncertainty assessment became strongly related to parameter estimation, multiobjective optimization, model identification, model building, model diagnostics, model averaging, data collection, and information theory in general. All topics in this list have gained the attention of researchers in recent years and are often allocated under the one umbrella of uncertainty assessment in hydrology. Indeed, it would be helpful for end-users to formally identify such subtopics and the related research questions.

2.17.5.2 An Attempt of Classification The traditional way of dealing with uncertainty in science is through statistics and probability (see, e.g., Montanari et al., 2009) but, as mentioned above, nonprobabilistic approaches to uncertainty analysis are also popular in hydrology. In some cases, it is not easy to classify an approach as either probabilistic or not. In fact, there are some methods that are based on probability theory, but in real-word applications simplifying assumptions are often introduced which finally lead to a nonprobabilistic estimation of the likelihood of a given scenario. Such assumptions are introduced in order to overcome operational problems, for instance, due to lack of enough data to support a statistical application. The decision to use probabilistic or nonprobabilistic methods is currently the most controversial issue in hydrologic uncertainty analysis. This debate has raised the very relevant question about the capability of probabilistic and nonprobabilistic methods to correctly infer the frequency properties of hydrological simulations and predictions (see, e.g., Beven, 2006a; Montanari, 2005, 2007; Mantovan and Todini, 2006; Beven et al., 2007, 2008). Criticism about probabilistic methods is focused on the concern that for many data sets it is not clear if the assumptions of classical statistics (e.g., stationarity) can be justified. The main reason for criticism of nonprobabilistic methods is that they are subjective and not necessarily coherent from a statistical point of view (see, e.g., the criticism of Mantovan and Todini (2006) with respect to GLUE). Moreover, on known problems for which the data do support the necessary probabilistic assumptions, probabilistic and nonprobabilistic methods provide different answers (e.g., Stedinger et al., 2008). The suitability of probabilistic versus nonprobabilistic methods and the difference in their response are dictated by the knowledge that the user has about the structure of the error model. Using a correctly based inference should lead to similar results in uncertainty assessment. Conversely, some authors claim that with unknown error structure it is dangerous to rely on statistical methods based on simple assumptions about the nature of the errors themselves. There is an increasing consensus about the opportunity to use probabilistic approaches, as a way to efficiently summarize the information content of the data, when sufficient information is available to support statistical hypotheses with appropriate statistical tests (Montanari et al., 2009). Conversely, data scarcity calls for expert knowledge to support uncertainty assessment. Above all, data scarcity calls for the

integration of different types of information, within a framework that is unavoidably subjective, given that the information itself is often soft. Besides the above, additional classifications were recently proposed for uncertainty assessment methods. For instance, Matott et al. (2009) identified seven categories of models: (1) Data analysis methods, including analytical and statistical procedures for evaluating the accuracy of data. These include also parametrization of probability distributions. (2) Identifiability analysis, aiming at detecting data inadequacy and suggesting model improvements. (3) Parameter estimation methods, quantifying uncertain model parameters. (4) Uncertainty analysis techniques, meaning methods to propagate sources of uncertainty through the model to generate probability distributions for the model output. These methods include approximation and sampling methods. (5) Sensitivity analyses, investigating to what extent different sources of variation in the input of a mathematical model affect the variation of the output. Sensitivity analysis aims at identifying what source of uncertainty weights more on the model output (see, e.g., Van Griensven et al. (2006); Go¨tzinger and Ba´rdossy, 2008). Sensitivity analysis and uncertainty estimation are well distinguished. Their results can be comparable, because a probability distribution of model outputs corresponding to different inputs can be similar to the analogous distribution derived through the analysis of probabilistic uncertainty. This similarity of results has originated a confusion of terms in some applications. (6) Multimodel analysis, consisting of generating multiple possible outputs accordingly to different models, parameters, and boundary conditions. (7) Bayesian methods, which were previously defined (this category could be joined with category 4 above). The seven categories above are not strictly separated, meaning that a method can belong to more than one of them. Another classification for uncertainty assessment methods for the model output was recently proposed by Shrestha and Solomatine (2008) who consider the following categories: (1) analytical methods, using derived distribution methods to compute the probability distribution function of the model output; (2) approximation methods, providing only the moments of the distribution of the uncertain output variable; (3) simulation and sampling–based methods, estimating the full distribution of the model output via simulation; (4) Bayesian methods, which combine Bayes’ theorem and various simulation approaches to either estimate or update the probability distribution function of the parameters of the model and consequently estimate the uncertainty of the model output; (5) methods based on the analysis of the model errors, such as the meta-Gaussian approach described in Section 2.17.6.4; and (6) fuzzy-theory-based methods, providing a nonprobabilistic approach for modeling the kind of uncertainty associated with vagueness and imprecision. Whatever approach is chosen to uncertainty assessment, the end-user should be made fully aware of the assumptions and drawbacks of the method that is being used. The presence of subjectivity should be clearly stated and the limitations of the underlying hypotheses, both in the probabilistic and nonprobabilistic approaches, clearly described and discussed. An appropriate terminology should also be used to make the meaning of the provided confidence bands clear. Whenever a

Uncertainty of Hydrological Predictions

subjective method is adopted, the user should be made aware that the uncertainty bands reflect user belief instead of providing a frequentist assessment of the probability of the true value to fall between them. Appropriate use of the methods being proposed by the scientific community, depending on the user needs and data availability, would allow us to successfully reach a better communication between scientists and end-users. It is as important to communicate uncertainty as communicate the assumptions on which an assessment has been based.

2.17.6 Assessment of the Global Uncertainty of the Model Output Assessment of the global uncertainty for the model output is by far the application that is most frequently presented by the hydrological literature, as a means for quantifying model reliability and providing end-users with operational indications. Several methods are available to this end, ranging from statistically based to subjective approaches.

2.17.6.1 Analytical Methods The most direct method to assess the uncertainty of a system output is to derive its statistics from a knowledge of the statistical properties of the system itself and the input data (Langley, 2000). However, this approach may be limited by two main problems: first, the derivation of the statistics of the output can imply significant mathematical and numerical difficulties; and, second, the statistical properties of the system and the input may not be known in detail. The first difficulty has stimulated the development of a first type of uncertainty assessment technique, namely, the approximate analytical methods. An example is the asymptotic reliability analysis, like the first-order reliability method (FORM) and second-order reliability method (SORM). Examples of applications in hydrology are given by Melching (1992) and Vrugt and Bouten (2002). Point estimate methods are an interesting option too, in view of their computational efficiency (Tsai and Franceschini, 2005). The second problem mentioned above may be even more difficult to deal with. For instance, the definition of the statistics of the system is a delicate step of the uncertainty assessment method recently proposed by Huard and Mailhot (2006) in a hydrological context.

2.17.6.2 The Generalized Likelihood Uncertainty Estimation GLUE was introduced by Beven and Binley (1992), who were inspired by the generalized sensitivity analysis methodology proposed by Spear and Hornberger (1980). GLUE rejects the concept of an optimum model and parameter set and assumes that, prior to input of data into a model, all models and parameter sets have an equal likelihood of being acceptable. The acceptance of the existence of multiple likely models has been called equifinality (Beven, 1993) to suggest that this should be accepted as a generic problem in hydrological modeling rather than simply

467

reflecting the problem of identifying the true model in the face of uncertainty. GLUE is performed by first selecting different modeling options (different hydrological models and different parameters). In order to reduce the computational requirements of the procedure, it might be necessary to limit the dimension of the sample space of the parameters and models. Then, a high number N of simulation is generated by sampling the model and parameter spaces accordingly to a prior probability distribution. In the absence of prior knowledge, uniform sampling can be used. By increasing N one increases the probability of trying all of the most relevant solutions. The different models are then run for each of the parameter sets and the model output is then compared to a record of observed data (e.g., for observed hydrographs or annual maximum peak flows, see Cameron et al. (1999); another interesting example is given by Blazkova and Beven (2009)). The performance of each trial is assessed via likelihood measures, either formal or informal. This includes rejecting some parameter sets as nonbehavioral. For instance, the Nash and Sutcliffe (1970) efficiency can be used as informal likelihood measure of the simulation of a continuous hydrograph. All parameter sets that lead to obtaining an efficiency above a subjective threshold are retained. Finally, likelihood weighted uncertainty bounds are calculated depending on the likelihood (Freer et al., 1996). For instance, the calculated likelihoods can be rescaled to produce a cumulative sum of 1.0, thereby obtaining informal weights. A cumulative distribution function of simulated discharges is then constructed using the rescaled weights. Linear interpolation is used to extract the discharge estimates corresponding to cumulative probabilities of a/2, 0.5, and 1.0  a/2. This allows 100(1  a)% uncertainty bounds to be derived, in addition to a median simulation. If either (or both) the likelihood measure or the procedure for computing the rescaled weights is informal, the probabilities computed with GLUE do not possess the classical frequentist meaning. Therefore, strictly speaking, it is inappropriate to refer to them with the term probability and many authors classify GLUE as a nonprobabilistic approach. Conversely, if formal statistical procedures are used, GLUE assumes the behavior of a probabilistic methodology. For extended discussions, the reader is referred to Beven et al. (2008) and Stedinger et al. (2008). GLUE could be applied in principle even in the absence of observed historical data, in those real-world applications in which the likelihood measure is estimated on the basis of expert knowledge. GLUE is highly computationally demanding, especially if the number of significant model parameters is high. This problem may prevent the application of GLUE when dealing with complex models. Beven (2006a) formally introduced a different procedure for the identification of behavioral models, by following previous practical experiments by Pappenberger and Beven (2004) and Page et al. (2007). A recent interesting application is presented by Liu et al. (2009). In this approach, limits of acceptability are preliminarily identified for the model output or selected performance measures. All the models that meet the limits of acceptability are retained so that an envelope of behavioral model simulations can be identified. Finally, a likelihood weighted cumulative density function for the

468

Uncertainty of Hydrological Predictions

hydrological model and aims at estimating the global uncertainty of the forecast, which is considered to be caused by: (1) precipitation uncertainty, which is dominant and quantified by the probability distribution of the future rainfall specified by the PQPF and (2) hydrologic uncertainty, which is the aggregate of all uncertainties arising from sources other than precipitation uncertainty. In particular, it aggregates the model uncertainty and parameter uncertainty. The BFS has three structural components: the precipitation uncertainty processor (PUP; Kelly and Krzysztofowicz, 2000), the hydrologic uncertainty processor (HUP; Krzysztofowicz and Kelly, 2000), and the integrator (INT; Krzysztofowicz, 2001b). Figure 2 reports a sketch of the BFS structure adapted from Krzysztofowicz (2002). The PUP has the purpose of mapping precipitation uncertainty to output uncertainty under the hypothesis that there is no hydrologic uncertainty. This involves running the hydrological model for a set of specified quantiles of the probability distribution of the future rainfall. The HUP quantifies hydrologic uncertainty under the hypothesis that there is no precipitation uncertainty. Finally, the INT integrates the two uncertainties in order to produce a PRSF. For extended details on the PUP and INT, the interested reader is invited to refer to Kelly and Krzysztofowicz (2000) and Krzysztofowicz (2001b, 2002). Next, we provide a brief description of the HUP for the purpose of illustrating the meta-Gaussian approach adopted by BFS. Let hn denote the true river stage on day n, counting from day n ¼ 0 when the forecast is issued. At the forecast time the actual river stage on day n is unknown and thus uncertain.

model output can be computed as previously in GLUE so that simulation quantiles can be estimated (see also Blazkova and Beven, 2009). There are many other variants of GLUE; for example, Tolson and Shoemaker (2008) and Mugunthan and Shoemaker (2006) combined optimization methods with a nonprobabilistic GLUE-like approach to increase computational efficiency of nonprobabilistic uncertainty analysis. The hydrological literature presented many applications to GLUE to numerous hydrological problems, including rainfall– runoff modeling (Cameron et al., 1999), groundwater modeling (Christensen, 2003), inundation modeling (Aronica et al., 1998, 2002), and urban water-quality modeling (Freni et al., 2008, 2009).

2.17.6.3 The Bayesian Forecasting System The Bayesian Forecasting System (BFS) was proposed by Krzysztofowicz (1999, 2001a, 2002), Krzysztofowicz and Kelly (2000), and Krzysztofowicz and Herr (2001). The purpose is to produce a probabilistic river stage forecast (PRSF) based on a probabilistic quantitative precipitation forecasting (PQPF) as an input to a hydrological model that is in charge of simulating the response of a river basin to precipitation. It can be adapted to produce a probabilistic river discharge forecast. The BFS assumes that the dominant source of uncertainty derives from the imperfect knowledge of the future precipitation, so that it can be assumed that all other sources of uncertainty play a minor role. The system can work with any

PQPF

Deterministic inputs

Deterministic hydrologic model

Precipitation amounts Model river stages

Precipitation uncertainty processor (PUP)

Hydrologic uncertainty processor (HUP) Marginal distributions

Output distribution parameters Posterior distribution parameters Observed river stage Integrator INT Real-time processing

PRSF

Prior simulation Figure 2 Sketch of the Bayesian forecasting system. PQPF and PRSF are probabilistic quantitative precipitation forecasting and probabilistic river stage forecasting, respectively. Adapted from Krzysztofowicz R (2002) Bayesian system for probabilistic river stage forecasting. Journal of Hydrology 268: 16–40.

Uncertainty of Hydrological Predictions

Therefore, it is treated as a random variable which we refer to with the symbol Hn. Let sn be an estimate of Hn from the hydrologic model based on all the input variables and the true precipitation amount. Estimate sn is treated as a realization of the random variable Sn. One would observe hn ¼ sn if there were no hydrologic uncertainty (let us remark that HUP is developed under the assumptions that there is no precipitation uncertainty). The presence of hydrologic uncertainty gives rise to a probability distribution of the actual river stage Hn, conditional on a realization of the model river stage Sn ¼ sn. Therefore, we can treat Hn as a random variable whose probability distribution is conditioned on the corresponding realization of the model river stage Sn ¼ sn. The purpose of the HUP is to provide an estimate for such probability distribution. This is achieved by applying a Bayesian technique. First of all, Hn and Sn are transformed to the Gaussian variables Wn and Xn respectively, by applying the standard normal quantile transform (NQT; see Kelly and Krzysztofowicz, 1997) and assuming that all conditional and joint densities are Gaussian. Then, it is assumed that the actual river stage process is well represented by a Markov stochastic process of order 1 and strictly stationary. This allows one to derive the prior probability distribution gn(wn|w0) of wn conditional on W0 ¼ w0. The prior density is derived under the assumption that the following normal linear equation applies in the transformed domain:

Wn ¼ rWn1 þ Yn

ð1Þ

where r is a parameter and Yn is a random variable stochastically independent of Wn1 and normally distributed with mean zero and variance s2(Yn). Therefore, the probability distribution of wn is Gaussian with mean equal to rwn1 and variance s2(Yn). The recursive application of the above derivation allows one to estimate gn(wn|w0). Subsequently, a probability distribution of the normalized model river stage xn conditioned on wn and w0 is built, and is denoted as fn(xn|wn, w0). This is derived under the hypothesis that the stochastic dependence between the transformed variables is governed by the following normal linear equation:

Xn ¼ an Wn þ dn W0 þ bn þ Fn

ð2Þ

in which an, bn, and dn are parameters and Fn is a variate stochastically independent of (Wn,W0) and normally distributed with mean zero and variance s2(Fn). It follows that the probability distribution fn(xn|wn, w0) is Gaussian with mean and variance (see Krzysztofowicz and Kelly, 2000):

EðXn jWn ¼ wn ; W0 ¼ w0 Þ ¼ an wn þ dn w0 þ bn

ð3Þ

VarðXn jWn ¼ wn ; W0 ¼ w0 Þ ¼ sn 2 ðFÞ

ð4Þ

The coefficients r, an, bn, and dn are derived by running the hydrological model for an extended record of the model input with a perfect forecast of precipitation amount, thus obtaining joint realizations of the model-actual river stage process that are transformed to the Gaussian probability distribution through the NQT. These joint realizations are used to estimate

469

the parameters of the above regressions (1) and (2). This design of the analysis assures that there is no precipitation uncertainty, but only hydrologic uncertainty. Once gn(wn|w0) and fn(xn|wn, w0) are known, the total probability law allows one to derive the distribution kn(xn|w0) of the transformed model river stage conditioned on w0, while the Bayes theorem allows one to derive the posterior density of wn conditioned on xn and w0, namely:

rðwn jxn ; w0 Þ ¼

f n ðxn jwn ; w0 Þgn ðwn ; w0 Þ kn ðxn jw0 Þ

ð5Þ

where

kðxn jw0 Þ ¼

Z

N

f n ðxn jwn ; w0 Þgn ðwn jw0 Þdwn

ð6Þ

N

Finally, the inverse of the NQT allows one to derive the posterior density of hn conditioned on sn and h0. Such distribution allows one to quantify the hydrologic uncertainty. More details are provided by Krzysztofowicz and Kelly (2000). Despite a theoretical development that may appear complicated, the BFS has the advantage of being easy to apply and allowing rapid implementation in real time. However, it was conceived for estimating the uncertainty of forecasted variables only.

2.17.6.4 Techniques Based on the Statistical Analysis of the Model Error Several methods for uncertainty assessment were proposed based on the statistical analysis of the model error. Accordingly, the model error is treated as a stochastic process for which realizations are obtained by performing off-line simulations which are matched with the corresponding observations. Of course, observed data are themselves uncertain and therefore the model reliability analysis could not be correct in absolute terms (in the ideal situation of a perfect model, if we compared its response with uncertain output observations, that we assumed to be correct, we would wrongly conclude that the model is uncertain). However, in any case, from a practical point of view, the difference between the model response and what we measure in the field gives an important information for the sake of inferring reality based on the model output (Refsgaard et al., 2006). A technique for global uncertainty assessment based on the analysis of the model error is the meta-Gaussian approach proposed by Montanari and Brath (2004) for the case of hydrological simulations and extended by Montanari and Grossi (2008) for hydrological forecasting. Next, the latter methodology is presented, therefore making reference to realtime flood forecasting systems. The meta-Gaussian approach is probabilistic. In order to estimate the uncertainty of a hydrological forecast, it is assumed that the forecast error is a stationary and ergodic stochastic process, denoted with the symbol E(t). Its statistical properties are inferred by analyzing a past realization eobs(t) ¼ Qobs(t)  Qpred(t) that it is assumed to be available, where Qobs(t) and Qpred(t) are true and forecasted river flows, respectively. The use of a meta-Gaussian model is

470

Uncertainty of Hydrological Predictions

then proposed to derive the time-varying probability distribution of the forecast error. In particular, the probability distribution of E(t) is inferred on the basis of its dependence on M selected explanatory random variables. The statistical inference is performed in the Gaussian domain, by preliminarily transforming E(t) and the explanatory variables to the Gaussian probability distribution. The above transformation is operated through the NQT. The probabilistic model for E(t) is built as follows. First of all, it is assumed that positive and negative errors come from two different statistical populations E(þ)(t) and E()(t). Therefore, the probability model for E(t) is given by a mixture of two probability distributions, one for E(þ)(t) and one for E()(t). The mixture is composed such that the area of the probability distribution of E(þ)(t) is equal to the percentage, P(þ), of positive errors over the total sample size of the available past realization eobs(t) of the forecast error. The two realizations e ðþÞobs ðtÞ and e ðÞobs ðtÞ are transformed through the NQT, therefore obtaining the normalized realizations Ne ðþÞobs ðtÞ and Ne ðÞobs ðtÞ. Then, M explanatory variables, X(i)(t) with i ¼ 1,y, M (which should be readily available at the forecast time), are selected in order to explain the variability in time of the marginal statistics of E(þ)(t) and E()(t). The values of such explanatory variables for the realizations e ðþÞobs ðtÞ and e ðÞobs ðtÞ above are estimated and then transformed by using the NQT, therefore obtaining the normalized explanatory variables Nx ðiÞobs ðtÞ with i ¼ 1,y, M. In the Gaussian domain, it is assumed that the forecast error can be expressed as a linear combination of the selected explanatory variables. Let us focus on the positive error. The linear combination can be expressed through the following relationship: ðþÞ

ðþÞ

Ne ðþÞ ðtj Þ ¼ C1 Nxð1Þ ðtj Þ þ C2 Nxð2Þ ðtj Þ ðþÞ

þ ? þ CM NxðMÞ ðtj Þ þ eðþÞ ðtj Þ

ð7Þ

where e ðþÞ ðtj Þ is an outcome of a homoscedastic and Gaussian random variable and tj is an assigned time step. An analogous relationship holds for Ne()(t). It is assumed that positive and negative errors are conditioned by the same explanatory variables, but the fit of the linear regression (7) leads to a different set of coefficient values. Such coefficients are estimated by inserting in (7) the past realizations of transformed (i) forecast error, Ne(þ) obs (t), and explanatory variables, Nxobs(t), and then by identifying the coefficient values that lead to the best fit (for instance by minimizing the sum of the squares of e ðþÞ ðtj Þ). The goodness of the fit provided by (7) can be verified by drawing a normal probability plot and a residual plot for e ðþÞ ðtj Þ as in Montanari and Brath (2004). Once the linear regression (7) has been calibrated, for positive and negative errors, the probability distribution of the transformed positive forecast error can be easily derived for real-time and real-world applications. Such distribution is Gaussian and is expressed by the following relationship:

Ne ðþÞ ðtj ÞB G½m½NeðþÞ ðtj Þ; s½NeðþÞ ðtj Þ

ð8Þ

where ‘B’ means equality in probability distribution and G indicates the Gaussian distribution whose parameters

are given by ðþÞ

ðþÞ

m½Ne ðþÞ ðtj Þ ¼ C1 Nxð1Þ ðtj Þ þ C2 Nxð2Þ ðtj Þ ðþÞ

þ ? þ CM NxðMÞ ðtj Þ s½Ne ðþÞ ðtj Þ ¼ s½eðþÞ ðtj Þ

ð9Þ ð10Þ

Analogous relationships (from (8) to (10)) hold for the negative error. Therefore, the confidence bands for the transformed forecast and an assigned significance level can be straightforwardly derived. In detail, the upper confidence band of the transformed forecast at the a significance level is given by the 1  aX(2  P(þ)) quantile of the Gaussian distribution given by (8), (9), and (10). Given that P(þ) can be arbitrarily close to 0, in the technical computation one may obtain values greater than 1 of aX(2  P(þ)). This means that the probability of getting a positive forecast error is small enough to make equal to 0 the width of the upper confidence band at the a significance level. For instance, if P(þ) ¼ 0.5 and a ¼ 10%, the transformed upper confidence band is given by the well-known relationship: ðþÞ

Ne90% ðtj Þ ¼ m½NeðþÞ ðtj Þ þ 1:96s½NeðþÞ ðtj Þ

ð11Þ

Finally, by applying back the NQT one obtains the confidence bands for the assigned significance level in the untransformed domain. The reason why positive and negative errors are treated separately is that a good fit is frequently not achieved through the linear regression (7) when the errors are pooled together. In fact, in this case, it appears that the NQT is not effective in making the errors homoscedastic and therefore the assumption of linearity does not hold. The reason for this result is that the NQT is not efficient in assuring homoscedasticity if the mean of the model error is not significantly changing across the range of the error itself, as it often happens when dealing with hydrological models. By treating positive and negative errors separately, the problem disappears and the assumptions of the linear regression are met. Finally, it is important to note that the only assumption made about the sign of the future forecast error is that it has a probability equal to P(þ) to be positive. Therefore, no inference is made on the sign of the forecast error on the basis of the explanatory variables. Figure 3 shows the confidence bands computed with the meta-Gaussian approach for the forecast with 1-h lead time of two flood events occurred on the Toce River at Candoglia, in Italy.

2.17.6.5 Bayesian Model Averaging Bayesian model averaging (BMA, Hoeting et al., 1999) is a statistical way of postprocessing model output ensembles to derive predictive probability density functions for hydrological variables. It represents the predictive probability distribution as a weighted average of the individual predictive probabilities of each model, where the weights are posterior probabilities of the models themselves and reflect the models’ relative contributions to predictive skill over a training period. The combination of multiple models is an important component of

Uncertainty of Hydrological Predictions

Observed Simulated 95% confidence bands

Observed Simulated 95% confidence bands

1200

2400 River discharge (m3 s−1)

River discharge (m3 s−1)

471

1000 800 600 400 200 0

2000 1600 1200 800 400 0

0

20

40

60

80

100

120

140

Hours from 0.00 of 3 November, 1994

0

10

20

30

40

50

60

70

Hours from 14.00 of 19 September, 1999

Figure 3 95% confidence bands computed with the meta-Gaussian approach for the forecast with 1-h lead time of two flood events occurred on the Toce River at Candoglia, on 3 November 1994 (left) and 19 September 1999 (right). From Montanari A and Grossi G (2008) Estimating the uncertainty of hydrological forecasts: A statistical approach. Water Resources Research 44: W00B08 (doi:10.1029/2008WR006897).

model validation (Burnham and Anderson, 2002). Multimodeling solutions are often applied in real time forecasting (ensemble forecasting). See, for instance, the activity carried out in the framework of the HEPEX Project (Franz et al., 2005; Zappa et al., 2008). BMA is applied in a Bayesian framework. Let M ¼ {Mi; i ¼ 1,2,y, N} be a set of N hydrological models for obtaining the vector zˆ of hydrological variables. Given a set of data, D, the posterior probability Pr(zˆ|D) of zˆ is obtained through the BMA according to the law of total probability:

PrðˆzjDÞ ¼ EM ½PrðˆzjMi ; DÞ ¼

N X

PrðˆzjMi ; DÞPrðMi jDÞ ð12Þ

i¼1

where Prðˆz; DÞ is the posterior probability of zˆ for the given data set D, PrðˆzjMi ; DÞ is the posterior probability of zˆ for given data set D and model Mi, PrðMi ; DÞ is the posterior model probability for model Mi, and EM is the expectation operator over simulation models. Essentially, Equation (12) says that the probability distribution given by the model ensemble for the output variable is a weighted mixture of the individual distributions given by each model, where the weights are the posterior model probabilities. Therefore, Equation (12) presupposes that individual probability distributions for the output from each model, conditioned on the model itself and the available data set, are available. According to Bayes’ rule, the posterior model weight is

PrðDjMi ÞPrðMi Þ PrðMi ; DÞ ¼ PN i¼1 PrðDjMi ÞPrðMi Þ

ð13Þ

where PrðDjMi Þ is the marginal model likelihood function for model Mi, Pr(Mi) is the prior model probability for model Mi, P and p PrðMi Þ ¼ 1. A uniform distribution can be assumed for the priors if better information is not available. Equation P (13) implies the total model weight p PrðMi jDÞ ¼ 1. The marginal model likelihood function PrðDjMi Þ plays an important role in the determination of the degree of importance for each model, given the same data set. For noninformative

model priors, higher posterior model weights reflect better agreement between results and observed data. According to Equation (12), the law of total expectation allows one to obtain the means of the predicted zˆ over the models for given data D:

EðˆzjDÞ ¼ EM ½EjˆzjMðpÞ ; D ¼

X

E½ˆzjMðpÞ ; DPrðMðpÞ jDÞ ð14Þ

p

where E is the expectation operation over zˆ. Analogous relationships allow one to obtain the covariance matrix of the predicted zˆ , therefore allowing a quantification of uncertainty. For more details, and an application that refers to the prediction of groundwater heads, the reader is referred to Li and Tsai (2009). There are plenty of applications of BMA in hydrology (see, for instance, Ajami et al. (2006), Duan et al. (2007), Zhang et al. (2009), Reggiani et al. (2009), and Li and Tsai (2009)). Figure 4 shows confidence bands computed with BMA for simulations of river flows obtained with the soil and water assessment tool (SWAT) model in the Yellow River Headwater Basin (from Zhang et al., 2009) BMA tends to be computationally demanding and relies heavily on prior information about models. Neuman (2003) proposed a maximum likelihood version (MLBMA) of BMA to render it computationally feasible and to allow dealing with cases where reliable prior information is lacking (Ye et al., 2004). BMA is also used within the Integrated Bayesian Uncertainty Estimator (IBUNE) proposed by Ajami et al. (2007).

2.17.6.6 Machine Learning Techniques In the recent past, there has been an increased interest about machine learning technique for global uncertainty assessment (see, for instance, Shrestha et al., 2009; Solomatine and Shrestha, 2009; Hall and Solomatine, 2008). These methods are frequently used as a mean to approximate complex models for uncertainty assessment, therefore obtaining a less computationally intensive approach.

472

Uncertainty of Hydrological Predictions 3000

66.7% interval Observed data

2000

Streamflow (cm s)

1000 0 3000 2000

5

10 15 20 25 30 35 40 45 50 55 60 90% interval Observed data

1000 0

5 10 15 20 25 30 35 40 45 50 55 60 Monthly flow from January 1986 to December 1990

Figure 4 Confidence bands computed with BMA for simulations of river flows obtained with the SWAT model in the Yellow River Headwater Basin. From Zhang X, Srinivasan R, and Bosch D (2009) Calibration and uncertainty analysis of the SWAT model using genetic algorithms and Bayesian model averaging. Journal of Hydrology 374: 307–317.

Machine learning is concerned with the design and development of algorithms that allow computers to learn based on data, such as from sensor data or databases. A major focus of machine learning research is to automatically learn to recognize complex patterns and make intelligent decisions based on data. Machine learning techniques include, among others, approaches that have been widely used in hydrology, such as neural networks, nearest-neighbor methods, and statistical methods.

2.17.7 Assessment of Data Uncertainty Among the sources of uncertainty in hydrology, data uncertainty is often believed to play a marginal role. Accordingly, only a few attempts have been made to quantify the effects of uncertainty in observations on hydrological modeling (see, for instance, Clarke, 1999). Different types of observations are currently used in hydrological modeling. The most common applications usually refer to precipitation as input and river flows as output data, although very often solar radiation, temperature, wind speed, soil moisture, groundwater levels, geomorphological features, land use, and others are also employed. Some of the above variables are affected by a limited uncertainty with respect to the others. In particular, uncertainty in precipitation and river flow is often considered to be dominant, because of the spatial variability of rainfall and snowfall on the one hand, and the errors in the determination of the rating curve on the other. The presence of uncertainty in input and output data induces two types of problems to hydrologists: the first is related to its estimation (to what extent the observed data are uncertain?), whereas the second is connected to accounting for such uncertainty in hydrological modeling.

2.17.7.1 Precipitation Uncertainty Hydrologists are well aware that a multitude of problems and research issues are related to precipitation uncertainty, which

are connected to precipitation monitoring and prediction (Chua and Bras, 1982; Gottschalk and Jutman, 1982). Precipitation monitoring is carried out through direct measurements (raingauges and snowgauges) or remote sensing (satellite, radar, and microsensors). The uncertainty of gauge measurements is typically limited and therefore the estimation error of the precipitation field is mainly induced by spatial variability. When remote-sensed data are used, spatial variability is generally better estimated but the uncertainty in point measurements is relevant. There is a large body of literature about uncertainty assessment for precipitation, starting from the pioneeristic work of Thiessen (1911). Uncertainty of gauge measurements of precipitation has been the subject of numerous case studies (see, for instance, Morrissey et al., 1995; Brath et al., 2004). These studies proved that the estimation error of mean areal precipitation significantly depends on the climatic conditions, the spatial structure of the precipitation itself, the morphology of the catchment, and the gauging network. The task of quantifying remotely sensed precipitation uncertainty has proved to be difficult. A fundamental problem is the lack of a term of comparison (Habib and Krajewski, 2002). Numerous studies compared remotely sensed and gauged data and showed significant disagreement. For instance, Austin (1987) found that for individual storms, radar and raingauge measurements can differ of a factor of 2 or more. In a more recent investigation, Brandes et al. (1999) found that radar-to-gauge ratios of storm totals were in the range of about 0.7–1.9. These differences become even more significant when satellite versus raingauge comparisons are carried out. In general, estimating precipitation uncertainty is a difficult task and no general rule exists. Integrating different monitoring techniques is certainly a potentially valuable solution. Turning to the purpose of accounting for precipitation uncertainty in hydrological modeling, different methods were proposed by the literature. The BFS described in Section 2.17.6.3 is a relevant example. GLUE can be applied as well, by introducing an input error model and then generating many different realizations of the input data themselves, with which one can derive a likelihood weighted model output. Kavetski et al. (2002, 2006a, 2006b) introduced the Bayesian total error analysis (BATEA), that is, a method for explicitly accounting for sampling and measurement uncertainty in both input and output data. In view of the inability to build a formal and sufficiently representative input error model in many real-world applications, BATEA is based on the use of vague error models, with the awareness that such an approach can cause a degeneration of the reliability of the inference equations. The basic working hypothesis of BATEA consists of assuming that the input uncertainty is multiplicative Gaussian and independent of each storm, even though its general framework allows alternative uncertainty models. The multiplier approach assumes that the storm depth is the only quantity in error, whereas the rainfall pattern is correct up to a multiplicative constant mi, such that d0i ¼ mi di , where d0i and di are the observed and true precipitation depths for the ith storm. Accordingly, the parameter vector of the hydrologic model is extended to include the parameters of the uncertainty

Uncertainty of Hydrological Predictions

models. Parameter inference is then carried out within a Bayesian framework, which requires indentifying a prior distribution of the model parameters that is subsuquently updated by using Bayes’ theorem in view of the available observations. The above treatment of input uncertainty implies that the dimensionality of the parameter vector is increased with latent variables, whose number depends on the sample size of the observed data (and the type of error model assumed). Moreover, if both the input and the output data are observed with large uncertainty, the utility of any parameter estimation methodology becomes questionable. Finally, one might be concerned that rainfall multipliers can possibly interact with other sources of error and therefore separation of errors in BATEA is conditional on other sources of uncertainty. For instance, classic underprediction by a hydrological model after a long dry period can be compensated by increasing the rainfall multiplier. A similar treatment for precipitation uncertainty is used in IBUNE (Ajami et al., 2007).

2.17.7.2 River Discharge Uncertainty Pelletier (1987) reviewed 140 publications dealing with uncertainty in the determination of the river discharge, thereby providing an extensive summary. He referred to the case where the river discharge at a given cross section is measured by using the velocity–area method, that is,

Q obs ðtÞ ¼ AðtÞ  vðtÞ

ð15Þ

where t is the sampling time, Q obs ðtÞ the measured river discharge, A(t) the cross-sectional area of the river, and v(t) the velocity of the river flow averaged over the cross section. Errors in Q obs ðtÞ are originated by uncertainties in both A(t) and v(t), which in turn are originated by uncertainty in the current meter, variability of the river flow velocity over the cross section, and uncertainty in the estimation of the cross-section geometry. Pelletier (1987) highlighted that the overall uncertainty in a single determination of river discharge, at the 95% confidence level, can vary in the range (8–20%), mainly depending on the exposure time of the current meter, the number of sampling points where the velocity is measured, and the value of v(t). Another interesting contribution was provided by the European ISO EN Rule 748 (1997) that quantified the expected errors in the determination of the river discharge with the velocity–area method. The conclusions were similar to those of Pelletier (1987). In some cases, including the usual practice in many countries of Europe, river discharge values are estimated by using the rating curve method, which is very easy to apply. According to the rating curve method, observations of river stage are converted into river discharge by means of a rating curve, which is preliminarily estimated by using observations collected using the velocity–area method. Hence, an additional error is induced by imperfect estimation of the rating curve. Di Baldassarre and Montanari (2009) proposed a model for estimating the error affecting river flow observations derived by the rating curve method. The model aims at taking into account the main sources of uncertainty within a

473

simplified approach. The most important assumptions underlying the model are as follows. (1) The uncertainty induced by imperfect observation of the river stage is negligible. This is consistent with the fact that these errors are usually very limited (around 1–2 cm; e.g., Schmidt, 2002; Pappenberger et al., 2006) and therefore of the same order of magnitude as standard topographic errors. (2) The geometry of the river is assumed to be invariant, which means that the rating curve changes in time only because of seasonal variation of roughness (see below). This assumption has been made because the uncertainty induced by possible variations of the river geometry is heavily dependent on the considered case study and no general rule can be suggested. However, it is worth noting that, using this assumption, the study neglects one of the most relevant sources of uncertainty that may affect river discharge observations where relevant sediment transport and erosion processes are present. (3) Uncertainty in the rating curve derives from the following causes: errors in the river discharge measurements that are used to calibrate the rating curve itself; interpolation and extrapolation error of the rating curve; unsteady flow conditions; and seasonal changes of roughness. The uncertainty affecting the river discharge measurements was estimated by Di Baldassarre and Montanari (2009) according to the guidelines reported by the European ISO EN Rule 748 (1997), which lead to an estimate of about 5–6% when the measurements are collected in ideal conditions. This outcome matches the indications reported in Leonard et al. (2000) and Schmidt (2002). The remaining sources of uncertainty were evaluated by Di Baldassarre and Montanari (2009) by developing a numerical simulation study for a 330-km reach of the Po River (Italy). The study focused on 17 cross sections and found that the estimation of river discharge using the rating curve method is affected by an increasing error for increasing river discharge values. At the 95% confidence level, the error ranges from 6.2% to 42.8% of the observation, with an average value of 25.6%. Furthermore, the uncertainty induced by the extrapolation of the rating curve is dominating the other errors in high flow conditions. In fact, previous contributions in hydrology (e.g., Rantz et al., 1982) do not recommend extrapolating rating curves beyond a certain range. Nevertheless, several hydrological applications are unavoidably based on flood flow observations (e.g., calibration and validation of rainfall–runoff models, flood frequency analysis, and boundary conditions of flood inundation models) and therefore one needs to extrapolate the rating curve beyond the measurement range (Pappenberger et al., 2006). The above analysis proved that river flow uncertainty can indeed be very significant and therefore should be accounted for in practical applications. An interesting opportunity is offered by the application of GLUE according to the limits of acceptability concept (Blazkova and Beven, 2009). Once the uncertainty in the river flows is estimated, it is possible to fix limits of acceptability for the observed river flows, and the models that do not respect them can be rejected as nonbehavioral. The collection of the behavioral outputs allows the user to obtain an envelope of likely model simulations. The above approach is nonprobabilistic.

474

Uncertainty of Hydrological Predictions

2.17.8 Assessment of Parameter Uncertainty Calibration in hydrology is increasingly done automatically, while manual calibration (through trial and error procedures) is used only when dealing with complex models requiring high computational costs. Parameter calibration techniques lead to either a single solution or multiple solutions (i.e., parameter sets). The approaches leading to a single solution are basically optimization problems, while techniques leading to multiple likely solutions can serve as tools for uncertainty assessment. Many search algorithms have been successfully devised and applied to automatically find the optimal parameter set for hydrological models, which can be subdivided into local, global, and hybrid search techniques (Duan et al., 1992; Mugunthan and Shoemaker, 2006; Tolson and Shoemaker, 2008; Thyer et al., 2009; Tonkin and Doherty, 2009). Approaches for multiple-solution parameter estimation can be broadly divided into two categories: importance sampling and Markov chain Monte Carlo (MCMC) sampling (Kuczera and Parent, 1998). With this approach full parameter distributions rather than simple point estimates can be obtained. Methods based on importance sampling aim to identify a set of behavioral model parameter configurations according to a selected objective function. Then, parameter distributions are estimated using a weighted combination of the behavioral parameter sets. GLUE is perhaps the most-used method based on importance sampling. MCMC parameter estimation incorporates importance sampling into a procedure for evaluating conditional probability distributions. Prior parameter distributions are selected (for instance, by assigning a uniform distribution or a distribution derived through expert knowledge) and the sampler evolves them into posterior distributions that are estimated by using the observed data. Thus, multiple-solution approaches can be used to assess parameter uncertainty. A relevant example within this respect is the SCEM-UA algorithm by Vrugt et al. (2003a). Once the uncertainty in the parameters is known, simulation approaches can be applied to estimate the related uncertainty induced in the model output. An example is given by Thorsen et al. (2001) who assessed the uncertainty in simulations of nitrate leaching induced by using model parameters obtained from databases at the European level. End-users frequently experience the case where multiple or competing objectives need to be optimized. According to this need, numerous multiobjective optimization algorithms have been devised, with numerous developments in the recent past (see, for instance, Zhang et al., 2008). Relevant examples are the MOSCEM-UA and AMALGAM methods (Vrugt et al., 2003b; Vrugt and Robinson, 2007). These two methods are briefly described in the following.

2.17.8.1 The MOSCEM-UA Method Multiobjective calibration problems can be dealt with by defining more than one optimization criteria (objective functions) that correspond to different performance measures of the selected model. Then, a multicriteria optimization method can be used to identify the set of nondominated, efficient, or Pareto optimal solutions (Gupta et al., 1998). The

Pareto solutions represent tradeoffs among the different performance measures that are often conflicting. As such, moving from one solution to another results in the improvement of one objective and deterioration in one or more others. A simple way to deal with multiobjective calibration is to weigh the different criteria into a single objective function and to run a large number of independent single-criteria optimization runs using different values for the weights (Madsen, 2000). This method is simple to implement, but has the drawback that a complete single-objective optimization is to be solved to obtain each discrete Pareto solution. Moreover, maintaining the independence of the various criteria will allow the user to analyze the tradeoffs among the different criteria, therefore enabling an improved understanding of the limitations of the model structure. MOSCEM-UA (Vrugt et al., 2003b) is an effective and efficient MCMC sampler, which is capable of generating a fairly uniform approximation of the Pareto frontier within a single optimization run. The algorithm is closely related to the SCEM-UA algorithm (Vrugt et al., 2003a). In addition, MOSCEM-UA uses a newly developed, improved concept of Pareto dominance, thereby also containing the single-criteria solutions at the extremes of the Pareto solution set. For more details, the interested reader is invited to refer to Vrugt et al. (2003b). The ensemble of the models lying on the Pareto frontier allows the user to identify an envelope of model outputs corresponding to the nondominated solutions.

2.17.8.2 The AMALGAM Method AMALGAM (Vrugt and Robinson, 2007) is a follow-up of MOSCEM-UA and is specifically designed to take full advantage of the power of distributed computer networks. AMALGAM runs multiple different search strategies simultaneously for population evolution and adaptively updates the weights of these individual methods based on their reproductive success. This ensures a fast, reliable, and computationally efficient solution to multiobjective optimization problems.

2.17.9 Assessment of Model Structural Uncertainty Model structural uncertainty is induced by inadequateness of the hydrological model to represent the hydrological system. This situation is also characterized by reduced model identifiability, because the imperfectness of the modeling solutions makes many of them potentially suboptimal, regardless of the different values of the selected performance measure. In the presence of model, structural uncertainty a performance measure becomes less effective and therefore the highest of its value does not necessarily identify the best model. For instance, a performance measure that lays emphasis on floods may be biased toward a (imperfect) model that could not be as reliable in reproducing the low flows. This is the reason why multiobjective calibration is frequently applied in hydrology. A statistical and rigorous evaluation of model structural uncertainty is not possible in practical hydrological applications, at least because it should necessarily be performed with perfect data. The literature proposed approximate

Uncertainty of Hydrological Predictions

techniques for estimating the uncertainty in the model output induced by model structural uncertainty. The most popular of them are based on multimodel analysis. In fact, the variability of the response provided by different models, if other uncertainty sources are negligible, provides indications on the uncertainty induced by a wrong model structure. Multimodel analysis is based on the use of many different plausible models that may consider, for instance, alternative processes and alternative simplified approximations. An example of model combination is the BMA presented in Section 2.17.6.5. Another quantitative approach for performing multimodel application was presented by Burnham and Anderson (2002) and Ye et al. (2008). It is implemented by assigning performance scores and importance weights to each candidate model with which the ensemble of model outputs can be constructed basing on the importance of each model. Multimodel applications can be performed also by applying GLUE, with which different models can be considered and evaluated according to a single likelihood measure or one or more likelihood measures. An example of application of GLUE with different modeling solutions is provided by Rojas et al. (2009). From a practical point of view, the above techniques are often applied for assessing the global uncertainty in the model output instead of model structural uncertainty only, because it is impossible to carry out such techniques in the absence of data and parameter uncertainty. As such, the combination of different models with uncertain parameters and uncertain data bases does not allow one to separate the above sources of uncertainty, unless one makes heavy assumptions (see, e.g., the IBUNE method; Ajami et al. (2007); see also Clark et al. (2008)).

2.17.10 Uncertainty Assessment as a Learning Process Uncertainty assessment is an effective mean to quantitatively assess model reliability and therefore perform model diagnostic and evaluation. These latter, in turn, provide indications about the model ability to simulate hydrology at a given place and therefore about the correctness of our understanding of the hydrological processes at that place. Thus, uncertainty assessment plays a fundamental role in the learning process. In the past, the learning process was mainly linked to parameter estimation for a given model. The optimal parameter values, actually, provide information about the conditions of the system. Treating modeling more explicitly as a learning process allows one to follow a new approach to this problem based on a methodology that will link models, databases, and parameters with the areas of interest, thereby providing information on the dominant hydrological processes (see Beven (2007); applications are presented in Montanari et al. (2006), Fenicia et al. (2008), and Schoups et al. (2008)). This is part of the downward modeling approach that recently gained increased attention within the context of PUB. One of the most exciting future perspectives is the possibility to implement many different models as a process of

475

learning about specific places (Beven, 2007). The representation will be uncertain so that this learning process should be implemented within a framework of uncertainty estimation. Indeed, uncertainty estimation, providing quantitative information about model reliability, if coupled with a multimodel approach, could provide indications about the dominant hydrological processes and their dynamics. Within this framework it is also necessary to set up a mechanism for model rejection (e.g., the model providing the simulations presented in Figure 1 could be rejected because it is too biased). There is the potential problem that model rejection is not by default embedded in uncertainty assessment methods. In particular, it is not embedded in statistical approaches, which in many cases do not assess the motivation for the presence of uncertainty. Model rejection is often based on expert knowledge, which is subjective but indeed necessary in the context of a learning process (see, for instance, Merz and Blo¨schl (2008a, 2008b)). This implies that the use of statistical methods for uncertainty assessment in a learning process should be based on including in the statistical representation the available information about the underlying physical process (for an extended discussion, see Koutsoyiannis (2009)).

2.17.11 Conclusions Uncertainty assessment in hydrology is a relevant practical problem and still a research challenge. The limited extension of hydrological databases and the complexity of hydrological processes, whose dynamics and domains are to a great extent nonobservable, make the interpretation of the results of hydrological modeling studies not easy. The intense research activity recently done on uncertainty resulted in the development of many new techniques for uncertainty assessment, which differ in behavior and scope. It is essential to formally define a terminology and make clear the prerogatives of each method in order to make clear to end-users the meaning of uncertainty in hydrology and convey them a useful information. In order to provide a contribution to this end, we provide in Table 1 a brief summary of the most-used uncertainty assessment methods, including those presented here, by also providing an attempt of classification and by specifying their purpose. Uncertainty assessment in hydrology will represent a research challenge for a long time to come. Uncertainty is an inherent property of hydrological processes which in principle will not prevent gaining a much better understanding of how water flows downstream. Uncertainty in hydrology should not be viewed as a limitation to be eliminated but rather as a intrinsic feature that needs to be properly and objectively quantified, whenever possible, with scientific method, that is, through the collection of data by means of observation and experimentation, and the formulation and testing of hypotheses. Communicating uncertainty to end-users should not undermine their confidence in models (Beven, 2006b; Pappenberger and Beven, 2006; Faulkner et al., 2007), but rather increase it through an improved perception of the underlying natural processes and an increased awareness of model

476

Uncertainty of Hydrological Predictions

reliability. Uncertainty does not mean lack of knowledge or lack of modeling capability but that the predicted value of a hydrological variable is uncertain. A proper estimation of uncertainty is the way forward to a reliable hydrological design and therefore a proper management of the environment and water resources.

Acknowledgments The author is grateful to Demetris Koutsoyiannis, Jasper Vrugt, Keith Beven, Simone Castiglioni, an anonymous referee and the Editor Stefan Uhlenbrook for providing very useful comments on the text. The support of the Italian Government, through the National Research Project ‘‘Uncertainty estimation for precipitation and river discharge data. Effects on water resources planning and flood risk management’’ is also acknowledged.

References Ajami NK, Duan Q, Gao X, and Sorooshian S (2006) Multi-model combination techniques for hydrological forecasting: Application to distributed model intercomparison project results. Journal of Hydrometeorology 8: 755--768. Ajami NK, Duan Q, and Sorooshian S (2007) An integrated hydrologic Bayesian multimodel combination framework: Confronting input, parameter, and model structural uncertainty in hydrologic prediction. Water Resources Research 43: W01403 (doi:10.1029/2005WR004745). Alley WM (1984) On the treatment of evapotranspiration, soil moisture accounting and aquifer recharge in monthly water balance models. Water Resources Research 20: 1137--1149. Aronica G, Bates PD, and Horritt MS (2002) Assessing the uncertainty in distributed model predictions using observed binary pattern information within GLUE. Hydrological Processes 16: 2001--2016. Aronica G, Hankin B, and Beven KJ (1998) Uncertainty and equifinality in calibrating distributed roughness coefficients in a flood propagation model with limited data. Advances in Water Resources 22: 349--365. Austin PM (1987) Relation between measured radar reflectivity and surface rainfall. Monthly Weather Review 115: 1053--1070. Beven KJ (1993) Prophesy, reality and uncertainty in distributed hydrological modeling. Advances in Water Resources 16: 41--51. Beven KJ (2006a) A manifesto for the equifinality thesis. Journal of Hydrology 320: 18--36. Beven KJ (2006b) On undermining the science? Hydrological Processes 20: 3141--3146. Beven KJ (2007) Towards integrated environmental models of everywhere: Uncertainty, data and modeling as a learning process. Hydrology and Earth System Sciences 11: 460--467. Beven KJ (2009) Environmental Modelling: An Uncertain Future?: An Introduction to Techniques for Uncertainty Estimation in Environmental Prediction. London: Taylor and Francis. Beven KJ and Binley AM (1992) The future of distributed models: Model calibration and uncertainty prediction. Hydrological Processes 6: 279--298. Beven KJ, Smith PJ, and Freer JE (2007) Comment on ‘‘Hydrological forecasting uncertainty assessment: Incoherence of the GLUE methodology’’ by Pietro Mantovan and Ezio Todini. Journal of Hydrology 338: 315--318. Beven KJ, Smith PJ, and Freer JE (2008) So just why would a modeller choose to be incoherent? Journal of Hydrology 354: 15--32. Blazkova S and Beven KJ (2009) A limits of acceptability approach to model evaluation and uncertainty estimation in flood frequency estimation by continuous simulation: Skalka catchment, Czech Republic. Water Resources Research 45: W00B16 (doi:10.1029/2007WR006726). Brandes EA, Vivekanandan J, and Wilson JW (1999) A comparison of radar reflectivity estimates of rainfall from collocated radars. Journal of Atmospheric Oceanic Technology 16: 1264--1272. Brath A, Montanari A, and Toth E (2004) Analysis of the effects of different scenarios of historical data availability on the calibration of a spatially-distributed hydrological model. Journal of Hydrology 291: 272--288.

Burnham KP and Anderson DR (2002) Model Selection and Multimodel Inference, 2nd edn. New York: Springer. Cameron DS, Beven KJ, Tawn J, Blazkova S, and Naden P (1999) Flood frequency estimation by continuous simulation for a gauged upland catchment (with uncertainty). Journal of Hydrology 219: 169--187. Christensen S (2003) A synthetic groundwater modelling study of the accuracy of GLUE uncertainty intervals. Nordic Hydrology 35: 45--59. Chua S-H and Bras RL (1982) Optimal estimations of mean areal precipitation in regions of orographic influence. Journal of Hydrology 57: 23--48. Clark MP, Slater AG, Rupp DE, et al. (2008) Framework for understanding structural errors (FUSE): A modular framework to diagnose differences between hydrological models. Water Resources Research 44: W00B02 (doi:10.1029/2007WR006735). Clarke RT (1999) Uncertainty in the estimation of mean annual flood due to rating curve indefinition. Journal of Hydrology 222: 185--190. Cullen AC and Frey HC (1999) Probabilistic Techinques in Exposure Assessment: A Handbook for Dealing with Variability and Uncertainty in Model and Inputs. New York: Plenum. Di Baldassarre G and Montanari A (2009) Uncertainty in river discharge observations: A quantitative analysis. Hydrology and Earth System Sciences 13: 913--921. Dottori F, Martina MLV, and Todini E (2009) A dynamic rating curve approach to indirect discharge measurement. Hydrology and Earth System Sciences 13: 847--863. Duan Q, Ajami NK, Gao X, and Sorooshian S (2007) Multi-model ensemble hydrologic prediction using Bayesian model averaging. Advances in Water Resources 30: 1371--1386. Duan Q, Sorooshian S, and Gupta VK (1992) Effective and efficient gobal optimization for conceptual rainfall-runoff models. Water Resources Research 28: 1015--1031. European ISO EN Rule 748 (1997) Measurement of liquid flow in open channels – velocity-area methods, Reference number ISO 748:1997 (E), International Standard. Faulkner H, Parker D, Green C, and Beven KJ (2007) Developing a translational discourse to communicate uncertainty in flood risk between science and the practitioner. AMBIO 36: 692--704. Fenicia F, Savenije HHG, Matgen P, and Pfister L (2008) Understanding catchment behavior through stepwise model concept improvement. Water Resources Research 44: W01402 (doi:10.1029/2006WR005563). Franz K, Ajami N, Schaake J, and Buizza R (2005) Hydrologic ensemble prediction experiment focuses on reliable forecasts. Eos 86: 239. Freer J, Beven KJ, and Ambroise B (1996) Bayesian uncertainty in runoff prediction and the value of data: An application of the GLUE approach. Water Resources Research 32: 2163--2173. Freni G, Mannina G, and Viviani G (2008) Uncertainty in urban stormwater quality modelling: The effect of acceptability threshold in the GLUE methodology. Water Research 42: 2061--2072. Freni G, Mannina G, and Viviani G (2009) Uncertainty assessment of an integrated urban drainage model. Journal of Hydrology 373: 392--404. Gan TY and Burges SJ (1990) An assessment of a conceptual rainfall-runoff model’s ability to represent the dynamics of small hypothetical catchments 2. Hydrologic responses for normal and extreme rainfall. Water Resources Research 26: 1605--1619. Gottschalk L and Jutman T (1982) Calculation of areal means of meteorologic variables for watersheds. 7th Nordic Hydrological Conference. NHP-Report No. 2, pp. 720– 736. Go¨tzinger J and Ba´rdossy A (2008) Generic error model for calibration and uncertainty estimation of hydrological models. Water Resources Research 44: W00B07 (doi:10.1029/2007WR006691). Gupta HV, Sorooshian S, and Yapo PO (1998) Toward improved calibration of hydrologic models: Multiple and noncommensurable measures of information. Water Resources Research 34: 751--763. Habib E and Krajewski WF (2002) Uncertainty analysis of the TRMM ground-validation radar-rainfall products: Application to the TEFLUN-B field campaign. Journal of Applied Meteorology 41: 558--572. Hahn GJ and Meeker WQ (1991) Statistical Intervals; A Guide for Practitioners. New York: Wiley. Hall J and Solomatine DP (2008) A framework for uncertainty analysis in flood risk management decisions. International Journal of River Basin Management 6: 85--98. Hoeting JA, Madigan D, Raftery AE, and Volinsky CT (1999) Bayesian model averaging: A tutorial. Statistical Science 14: 382--401 (doi:10.1214/ss/ 1009212519). Hoffman FO and Hammonds JS (1994) Propagation of uncertainty in risk assessments: The need to distinguish between uncertainty due to lack of knowledge and uncertainty due to variability. Risk Analysis 14: 707--712.

Uncertainty of Hydrological Predictions

Huard D and Mailhot A (2006) A Bayesian perspective on input uncertainty in model calibration: Application to hydrological model ‘‘abc’’. Water Resources Research 42: W07416 (doi:10.1029/2005WR004661). Ibbitt RP and O’Donnell T (1974) Designing conceptual catchment models for automatic fitting methods. In: Mathematical Models in Hydrology Symposium, IAHS-AISH, Publication No. 2, pp. 461–475. Jacquin AP and Shamseldin AY (2007) Development of a possibilistic method for the evaluation of predictive uncertainty in rainfall-runoff modeling. Water Resources Research 43: W04425. doi:10.1029/2006WR005072. Kavetski D, Franks SW and Kuczera G (2002) Confronting input uncertainty in environmental modelling. In: Calibration of Watershed Models. Duan Q, Sorooshian S, Gupta H, Rosseau H, and Turcotte R (eds.). Water Science and Application, vol. 6, pp. 49–68. Washington, DC: AGU. Kavetski D, Kuczera G, and Franks SW (2006a) Bayesian analysis of input uncertainty in hydrological modeling: 1. Theory. Water Resources Research 42: W03407 (doi:10.1029/2005WR004368). Kavetski D, Kuczera G, and Franks SW (2006b) Bayesian analysis of input uncertainty in hydrological modeling: 2. Application. Water Resources Research 42: W03408 (doi:10.1029/2005WR004376). Kelly KS and Krzysztofowicz R (1997) A bivariate meta-Gaussian density for use in hydrology. Stochastic Hydrology and Hydraulics 11: 17--31. Kelly KS and Krzysztofowicz R (2000) Precipitation uncertainty processor for probabilistic river stage forecasting. Water Resources Research 36: 2643--2653. Kleisenn FM, Beck MB, and Weather HS (1990) The identifiability of conceptual hydrochemical models. Water Resources Research 26: 2979--2992. Koussis A (2009) Comment on ‘‘A dynamic rating curve approach to indirect discharge measurement’’ by Dottori et al. (2009). Hydrology and Earth System Sciences Discussions 6: 7429--7437. Koutsoyiannis D (2009) A random walk on water. Hydrology and Earth System Sciences Discussions 6: 6611--6658. Koutsoyiannis D, Makropoulos C, Langousis A, et al. (2009) HESS opinions: ‘‘Climate, hydrology, energy, water: Recognizing uncertainty and seeking sustainability’’. Hydrology and Earth System Sciences 13: 247--257. Krzysztofowicz R (1999) Bayesian theory of probabilistic forecasting via deterministic hydrologic model. Water Resources Research 35: 2739--2750. Krzysztofowicz R (2001a) The case for probabilistic forecasting in hydrology. Journal of Hydrology 249: 2--9. Krzysztofowicz R (2001b) Integrator of uncertainties for probabilistic river stage forecasting: Precipitation-dependent model. Journal of Hydrology 249: 69--85. Krzysztofowicz R (2002) Bayesian system for probabilistic river stage forecasting. Journal of Hydrology 268: 16--40. Krzysztofowicz R and Herr HD (2001) Hydrologic uncertainty processor for probabilistic river stage forecasting: Precipitation-dependent model. Journal of Hydrology 249: 46--68. Krzysztofowicz R and Kelly KS (2000) Hydrologic uncertainty processor for probabilistic river stage forecasting. Water Resources Research 36: 3265--3277. Kuczera G and Mroczkowski M (1998) Assessment of hydrologic parameter uncertainty and the worth of multiresponse data. Water Resources Research 34: 1481--1489. Kuczera G and Parent E (1998) Monte Carlo assessment of parameter uncertainty in conceptual catchment models: The Metropolis algorithm. Journal of Hydrology 211: 69--85. Kuichling E (1889) The relation between rainfall and the discharge of sewers in populous districts. Transactions ASCE 20: 1--60. Langley RS (2000) Unified approach to probabilistic and possibilistic analysis of uncertain systems. Journal of Engineering Mechanics 126: 1163--1172. Leonard J, Mietton M, Najib H, and Gourbesville P (2000) Rating curve modelling with Manning’s equation to manage instability and improve extrapolation. Hydrological Sciences Journal 45: 739--750. Li X and Tsai FT-C (2009) Bayesian model averaging for groundwater head prediction and uncertainty analysis using multimodel and multimethod. Water Resources Research 45: W09403 (doi:10.1029/2008WR007488). Liu Y, Freer J, Beven KJ, and Matgen P (2009) Towards a limits of acceptability approach to the calibration of hydrological models: Extending observation error. Journal of Hydrology 367: 93--103. Liu Y and Gupta HV (2007) Uncertainty in hydrologic modeling: Toward an integrated data assimilation framework. Water Resources Research 43: W07401 (doi:10.1029/ 2006WR005756). Loucks DP and Van Beek E (2005) Water resources systems planning and management an introduction to methods, models and applications. In: Studies and Reports in Hydrology, United Nations Educational, Scientific and Cultural Organization (ISBN 92-3-103998-9). Paris, France: UNESCO. Madsen H (2000) Automatic calibration of a conceptual rainfall-runoff model using multiple objectives. Journal of Hydrology 235: 276--288.

477

Mantovan P and Todini E (2006) Hydrological forecasting uncertainty assessment: Incoherence of the GLUE methodology. Journal of Hydrology 330: 368--381. Matott LS, Babendreier JE, and Purucker ST (2009) Evaluating uncertainty in integrated environmental models: A review of concepts and tools. Water Resources Research 45: W06421 (doi:10.1029/2008WR007301). Melching CS (1992) An improved first-order reliability approach for assessing uncertainties in hydrologic modeling. Journal of Hydrology 132: 157--177. Merz R and Blo¨schl G (2008a) Flood frequency hydrology: 1. Temporal, spatial, and causal expansion of information. Water Resources Research 44: W08432, doi:10.1029/2007WR006744. Merz R and Blo¨schl G (2008b) Flood frequency hydrology: 2. Combining data evidence. Water Resources Research 44: W08433, doi:10.1029/2007WR006745. Montanari A (2005) Large sample behaviors of the generalized likelihood uncertainty estimation (GLUE) in assessing the uncertainty of rainfall-runoff simulations. Water Resources Research 41: W08406 (doi:10.1029/2004WR003826). Montanari A (2007) What do we mean by ‘uncertainty’? The need for a consistent wording about uncertainty assessment in hydrology. Hydrological Processes 21: 841--845. Montanari A and Brath A (2004) A stocastic approach for assessing the uncertainty of rainfall-runoff simulations. Water Resources Research 40: W01106 (doi:10.1029/ 2003WR002540). Montanari A and Grossi G (2008) Estimating the uncertainty of hydrological forecasts: A statistical approach. Water Resources Research 44: W00B08 (doi:10.1029/ 2008WR006897). Montanari A, Shoemaker CA, and Van de Giesen NC (2009) Uncertainty assessment in surface and subsurface hydrology: An overview. Water Resources Research 45: W00B00 (doi:10.1029/2009WR008471). Montanari L, Sivapalan M, and Montanari A (2006) Investigation of dominant hydrological processes in a tropical catchment in a monsoonal climate via the downward approach. Hydrology and Earth System Sciences 10: 769--782. Morrissey ML, Maliekal JA, Greene JS, and Wang J (1995) The uncertainty of simple spatial averages using rain gauge networks. Water Resources Research 31: 2011--2017. Mugunthan P and Shoemaker CA (2006) Assessing the impacts of parameter uncertainty for computationally expensive groundwater models. Water Resources Research 42: W10428 (doi:10.1029/2005WR004640). Mulvaney TJ (1851) On the use of self-registering rain and flood gauges in making observations of the relations of rainfall and of flood discharges in a given catchment. Proceeding of the Institute of Civil Engineers of Ireland 4: 18--31. Nash JE and Sutcliffe JV (1970) River flow forecasting through conceptual models. 1: A discussion of principles. Journal of Hydrology 10: 282--290. Nauta MJ (2000) Separation of uncertainty and variability in quantitative microbial risk assessment models. International Journal of Food Microbiology 57: 9--18. Neuman SP (2003) Maximum likelihood Bayesian averaging of alternative conceptualmathematical models. Stochastic Environmental Resources Risk Assessment 17: 291--305. NRC (National Research Council) (2000) Risk Analysis and Uncertainty in Flood Damage Reduction Studies. Washington, DC: National Academy Press. Page T, Beven KJ, Freer J, and Neal C (2007) Modelling the chloride signal at Plynlimon, Wales, using a modified dynamic TOPMODEL incorporating conservative chemical mixing (with uncertainty). Hydrological Processes 21: 292--307. Papolulis A (1991) Probability, Random Variables, and Stochastic Processes, 3rd edn. New York: McGraw-Hill. Pappenberger F and Beven KJ (2004) Functional classification and evaluation of hydrographs based on Multicomponent Mapping (Mx). International Journal of River Basin Management 2: 89--100. Pappenberger F and Beven KJ (2006) Ignorance is bliss: Or seven reasons not to use uncertainty analysis. Water Resources Research 42: W05302 (doi:10.1029/ 2005WR004820). Pappenberger F, Matgen P, Beven KJ, Henry JB, Pfister L, and de Fraipont P (2006) Influence of uncertain boundary conditions and model structure on flood inundation predictions. Advances in Water Resources 29: 1430--1449. Pelletier MP (1987) Uncertainties in the determination of river discharge: A literature review. Canadian Journal of Civil Engineering 15: 834--850. Petersen-Øverleir A and Reitan T (2009) Accounting for rating curve imprecision in flood frequency analysis using likelihood-based methods. Journal of Hydrology 366: 89--100. Rantz SE, et al. (1982) Measurement and Computation of Streamflow, US Geological Survey, Water Supply Paper 2175. http://water.usgs.gov/pubs/wsp/wsp2175/ (accessed March 2010). Refsgaard JC, van der Sluijs JP, Brown J, and Van der Keur P (2006) A framework for dealing with uncertainty due to model structure error. Advances in Water Resources 29: 1586--1597.

478

Uncertainty of Hydrological Predictions

Regan HM, Akcakaya HR, Ferson S, Root KV, Carroll S, and Ginzburg LR (2003) Treatments of uncertainty and variability in ecological risk assessment of singlespecies populations. Human Ecology Risk Assessment 9: 889--906. Reggiani P, Renner M, Weerts AH, and van Gelder PAHJM (2009) Uncertainty assessment via Bayesian revision of ensemble streamflow predictions in the operational river Rhine forecasting system. Water Resources Research 45: W02428 (doi:10.1029/2007WR006758). Rojas R, Batelaan O, Feyen L, and Dassargues A (2009) Assessment of conceptual model uncertainty for the regional aquifer Pampa del Tamarugal – North Chile. Hydrology and Earth System Sciences Discussion 6: 5881--5935. Schmidt AR (2002) Analysis of stage-discharge relations for open channel flow and their associated uncertainties. PhD Thesis, University of Illinois. Schoups G, Van de Giesen NC, and Savenije HHG (2008) Model complexity control for hydrologic prediction. Water Resources Research 44: W00B03 (doi:10.1029/ 2008WR006836). Sherman LK (1932) Streamflow from rainfall by the unit-graph method. Engineering News Record 108: 501--505. Shrestha DL, Kayastha N, and Solomatine DP (2009) A novel approach to parameter uncertainty analysis of hydrological models using neural networks. Hydrology and Earth System Sciences 13: 1235--1248. Shrestha DL and Solomatine DP (2008) Data-driven approaches for estimating uncertainty in rainfall-runoff modelling. International Journal of River Basin Management 6: 109--122. Sieber A and Uhlenbrook S (2005) Sensitivity analyses of a distributed catchment model to verify the model structure. Journal of Hydrology 310: 216--235. Solomatine DP and Shrestha DL (2009) A novel method to estimate model uncertainty using machine learning techniques. Water Resources Research 45: W00B11 (doi:10.1029/2008WR006839). Sonich-Mullin C (2001) Harmonizing the incorporation of uncertainty and variability into risk assessment: An international approach. Human and Ecological Risk Assessment 7: 7--13. Spear RC and Hornberger GM (1980) Eutrophication in Peel Inlet: II. Identification of critical uncertainties via generalized sensitivity analysis. Water Research 14: 43--49. Stedinger JR, Vogel RM, Lee SU, and Batchelder R (2008) Appraisal of the generalized likelihood uncertainty estimation (GLUE) method. Water Resources Research 44: W00B06 (doi:10.1029/2008WR006822). Thiessen AH (1911) Precipitation averages for large areas. Monthly Weather Report 39: 1082--1084. Thorsen M, Refsgaard JC, Hansen S, Pebesma E, Jensen JB, and Kleeschulte S (2001) Assessment of uncertainty in simulation of nitrate leaching to aquifers at catchment scale. Journal of Hydrology 242: 210--227. Thyer M, Renard B, Kavetski D, Kuczera G, Franks SW, and Srikanthan S (2009) Critical evaluation of parameter consistency and predictive uncertainty in hydrological modeling: A case study using Bayesian total error analysis. Water Resources Research 45: W00B14 (doi:10.1029/2008WR006825). Tolson BA and Shoemaker CA (2008) Efficient predicition uncertainty approximation in the calibration of environmental simulation models. Water Resources Research 44: W04411 (doi:10.1029/2007WR005869). Tonkin M and Doherty J (2009) Calibration-constrained Monte Carlo analysis of highly parameterized models using subspace techniques. Water Resources Research 45: W00B10 (doi:10.1029/2007WR006678). Tsai CW and Franceschini S (2005) Evaluation of probabilistic point estimate methods in uncertainty analysis for environmental engineering applications. Journal of Environmental Engineering 131: 387–395.

Van Griensven A, Meixner T, Grunwald S, Bishop T, Di luzio M, and Srinivasan R (2006) A global sensitivity analysis tool for the parameters of multi-variable catchment models. Journal of Hydrology 324: 10--23. Von Bertalanffy L (1968) General System Theory. New York: George Braziller. Vrugt JA and Bouten W (2002) Validity of first-order approximations to describe parameter uncertainty in soil hydrologic models. Soil Science Society of America Journal 66: 1740--1752. Vrugt JA, Gupta HV, Bastidas LA, Bouten W, and Sorooshian S (2003b) Effective and efficient algorithm for multiobjective optimization of hydrologic models. Water Resources Research 39: 1214 (doi:10.1029/2002WR001746). Vrugt JA, Gupta HV, Bouten W, and Sorooshian S (2003a) A Shuffled Complex Evolution Metropolis algorithm for optimization and uncertainty assessment of hydrologic model parameters, Water Resources Research 39: 1201 (doi:10.1029/ 2002WR001642). Vrugt JA and Robinson BA (2007) Improved evolutionary optimization from genetically adaptive multimethod search. Proceedings of the National Academy of Science of the United States of America 104: 708--711. Wagener T, McIntyre N, Lees MJ, Wheater HS, and Gupta HV (2003) Towards reduced uncertainty in conceptual rainfall-runoff modelling: Dynamic identifiability analysis. Hydrological Processes 17: 455--476. Winsemius HC, Savenije HHG, and Bastiaanssen WGM (2008) Constraining model parameters on remotely sensed evaporation: Justification for distribution in ungauged basins? Hydrology and Earth System Sciences 12: 1403--1413. Winsemius HC, Savenije HHG, Gerrits AMJ, Zapreeva EA, and Klees R (2006) Comparison of two model approaches in the Zambezi river basin with regard to model reliability and identifiability. Hydrology and Earth System Sciences 10: 339--352. Ye M, Meyer PD, and Neumann SP (2008) On model selection criteria in multimodel analysis. Water Resources Research 44: W03428 (doi:10.1029/2008WR006803). Ye M, Neuman SP, and Meyer PD (2004) Maximum likelihood Bayesian averaging of spatial variability models in unsaturated fractured tuff. Water Resources Research 40: W05113 (doi:10.1029/2003WR002557). Zadeh LA (2005) Toward a generalized theory of uncertainty (GTU)––an outline. Information Sciences 172: 1--40. Zappa M, Rotach MV, Arpagaus M, et al. (2008) MAP D5-PHASE: Real-time demonstration of hydrological ensemble prediction systems. Atmospheric Science Letters 9: 80--87. Zhang X, Srinivasan R, and Bosch D (2009) Calibration and uncertainty analysis of the SWAT model using genetic algorithms and bayesian model averaging. Journal of Hydrology 374: 307--317. Zhang Z, Wagener T, Reed P, and Bhushan R (2008) Reducing uncertainty in predictions in ungauged basins by combining hydrologic indices regionalization and multiobjective optimization. Water Resources Research 44: W00B04 (doi:10.1029/2008WR006833).

Relevant Websites http://education.mit.edu Gaussian Distribution. http://www.itia.ntua.gr Presentation: Hurst-Kolmogorov dynamics and uncertainity. http://www.agu.org Special issue on uncertainty.

2.18 Statistical Hydrology S Grimaldi, Universita` degli Studi della Tuscia, Viterbo, Italy S-C Kao, Oak Ridge National Laboratory, Oak Ridge, TN, USA A Castellarin, Universita` degli Studi di Bologna, Bologna, Italy S-M Papalexiou, National Technical University of Athens, Zographou, Greece A Viglione, Technische Universita¨t Wien, Vienna, Austria F Laio, Politecnico di Torino, Torino, Italy H Aksoy and A Gedikli, Istanbul Technical University, Istanbul, Turkey & 2011 Elsevier B.V. All rights reserved.

2.18.1 2.18.2 2.18.2.1 2.18.2.1.1 2.18.2.1.2 2.18.2.1.3 2.18.2.1.4 2.18.2.1.5 2.18.2.1.6 2.18.2.1.7 2.18.2.1.8 2.18.2.2 2.18.3 2.18.3.1 2.18.3.1.1 2.18.3.1.2 2.18.3.1.3 2.18.3.1.4 2.18.3.1.5 2.18.3.1.6 2.18.3.1.7 2.18.3.1.8 2.18.3.1.9 2.18.3.2 2.18.3.2.1 2.18.3.2.2 2.18.3.2.3 2.18.3.3 2.18.4 2.18.4.1 2.18.4.2 2.18.4.2.1 2.18.4.2.2 2.18.4.3 2.18.4.3.1 2.18.4.3.2 2.18.5 2.18.5.1 2.18.5.2 2.18.5.3 2.18.6 2.18.6.1 2.18.6.2 2.18.6.2.1 2.18.6.2.2 2.18.6.2.3 2.18.6.2.4 2.18.6.2.5

Introduction Analysis and Detection of Nonstationarity in Hydrological Time Series The Common Nonstationarity Analysis Methods Randomness test Detection of trend Simple regression on time Mann–Kendall test Spearman rank order correlation test Detection of shifts (segmentation) t-Test Mann–Whitney test A New Method of Segmentation Extreme Value Analysis: Distribution Functions and Statistical Inference Probability Distributions for Extreme Events Normal distribution Lognormal distribution Exponential distribution Gamma distribution Pearson type 3 distribution Log-Pearson type 3 distribution Extreme value distributions Generalized Pareto distribution Generalized logistic distribution Parameter Estimation Methods Method of moments Method of L-moments Method of the maximum-likelihood and Bayesian methods Model Verification: Goodness-of-Fit Tests IDF Curves Definition of IDF Curves and Clarifications Empirical Methods Parameter estimation Application in a real-world data set Theoretically Consistent Methods Parameter estimation Application in a real-world data set Copula Function for Hydrological Application Concepts of Dependence Structure and Copulas Copulas in Hydrologic Applications Remarks on Copulas and Future Research Regional Frequency Analysis Index-Flood Procedure, Extensions and Evolutions Classical Regionalization Approach Estimation of the index flood Estimation of the regional dimensionless quantile Homogeneity testing Choice of a frequency distribution Estimation of the regional frequency distribution

479 480 480 480 480 480 481 481 482 482 482 483 484 484 484 485 485 485 485 486 486 488 488 488 488 489 490 490 492 493 494 494 495 496 497 499 500 500 503 505 505 505 506 506 507 507 508 508

479

480

Statistical Hydrology

2.18.6.2.6 2.18.6.3 References

Validation of the regional model Open Problems and New Advances

2.18.1 Introduction Hydrological phenomena such as precipitation, floods, and droughts are inherently random by nature. Due to the complexity of the hydrologic system, these physical processes are not fully understood and reliable deterministic mathematical models are still to be developed. Therefore, in order to provide useful analyses for designing hydraulic facilities and infrastructures, statistical approaches have been commonly adopted. In literature and in the practical hydrological applications, many statistical methods are considered with different aims. Simulation, forecasting, uncertainty analysis, spatial interpolation, and risk analysis are some of the most important ones. The use of statistical analyses is strongly related to the data availability and to the quality of observations. Particular emphasis is given to the case of ungauged area where the statistical approach is particularly important to develop hydrological analyses without direct observations (the relevance of this issue is well documented by the Decade on Prediction in Ungauged Basins (PUB) promoted by the International Association of Hydrological Sciences (IAHS, Sivapalan et al., 2003). This chapter describes some statistical topics widely used in hydrology. Among the large number of subjects available in literature, the attention is focalized on some of them particularly useful either for innovative hydrological analyses or for an appropriate application of common procedures. Many statistical methods are strongly affected by specific conditions to be verified on the available data set. Indeed, for instance, complex procedures, used for different important applications, usually need a very common and simple hypothesis: the stationarity. This condition, simple to define but very difficult to verify, probably is the most important in statistical hydrology. For this reason, the first section of this chapter provides a short review of this topic and a detailed description of the segmentation method that is a promising procedure for time series trend detection. Another primary topic, described here, is the univariate extreme value (EV) analysis. The EV approach is the widest used in hydrology (i.e., for the derivation of return levels for extreme rainfall and flood estimates) and it should be carefully and correctly applied in order to avoid dangerous underor overestimation of the analyzed design variables (rainfall, runoff). With this aim in the second section, a detailed distribution functions used with hydrological variables are described; moreover, the approaches to develop the parameter estimation and the goodness-fit-test steps are reviewed. Since rainfall is the most-observed hydrological phenomenon, a peculiar section is included in this chapter providing an update description of EV-IDF (intensity–duration–frequency) procedure. IDF curves are an invaluable tool in hydrology having a crucial role in the safe and efficient design of major

508 510 511

or minor infrastructures (e.g., water dams, urban hydraulic works, flood design, etc.) that affect human lives. IDF curves are in use almost for a century, and the many different forms and methods proposed and studied through the years underline their importance. During all those years, IDF curves have evolved from purely empirical forms to theoretically more consistent, while today, their study still remains an active field of research. In this text, some of the most commonly used forms and techniques have been presented and applied in a real world data set. The search of the literature and the application presented here reveals that some commonly used techniques and forms of IDF curves may result in underestimating the rainfall intensity, especially for large return periods, and thus should be used with caution. More advanced forms and estimation procedures are described and compared to the most commonly used ones in practical applications. Until now the efforts of hydrologists were primarily devoted to analyze single parameters (flood peak, rainfall intensity, etc.), not because it is not important to consider other variables (i.e., flood duration and flood volume, or rainfall duration and volume, etc.) but because of the absence of a flexible approach to jointly analyze these different but useful variables. However, this is now finally possible, thanks to the relatively recent introduction of copula function. This statistical and mathematical method is quickly evolving and numerous applications are described in literature. Since this approach is promising and it could change and improve many hydrological procedures, a specific section on copula function is considered in this chapter, providing an updated review useful for hydrological applications. As mentioned at the beginning of this section, the ungauged basin is a sensitive problem. Most of the little basins (o150 km2) are characterized by poor hydrological observations (usually few raingauges are available) that stimulated an intense research on statistical methods for regional frequency analysis. Therefore, in the last section, it is essential to include a review and a specific description of this important topic. This chapter is written by a group of researcher members of the Statistics in Hydrology – STAHY Working Group recently launched by the International Association of Hydrological Sciences – IAHS with the purpose of sharing knowledge and stimulating research activities on statistical hydrology.

2.18.2 Analysis and Detection of Nonstationarity in Hydrological Time Series Hydrological time series used in water resources planning studies are very often supposed to meet the stationary hypothesis. Under steady-state natural conditions, time series exhibit regular fluctuations around a mean value; however,

Statistical Hydrology

2.18.2.1 The Common Nonstationarity Analysis Methods

when the natural conditions change markedly, they may form trends or exhibit jumps. Hydrological data series frequently show this type of significant nonstationarity due to several reasons (human activities, climate change, etc.). A random process is an indexed family (xt)tAI of random variables, which may be discrete time if I is a set of integers. A discrete random time process x ¼ (x1,x2, y , xn) is said to be stationary if, for every k and n, the distribution of xkþ1,xkþ2, y , xkþn is the same as the distribution of x1,x2, y , xn (Baseville and Nikiforov, 1993). In other words, a random process or variable is said to be strictly stationary if its statistical properties do not vary with time, and hence independent of changes of time origin. Trends, jumps/shifts, seasonality/periodicity, or nonrandomness in a hydrological time series can be referred to as components of the time series. Presence of these components makes the time series nonstationary. Indeed, nonstationarity is under the effect of persistency and scaling issues (Koutsoyiannis, 2006). Hydrological time series frequently exhibit nonstationary behavior, for example; flow and precipitation or rainfall stay below or above the mean long-term average (Rao and Yu, 1986), although they are generally assumed to be stationary at annual scale. When the time interval used is shorter than a year (month, week, or day), the stationarity assumption in the hydrological time series then becomes nonvalid simply because of the annual cycle of the Earth around the Sun. Trends in a time series can result from gradual natural and human-induced disruptive and evolutionary changes in the environment, whereas a jump may result from sudden catastrophic natural events (Haan, 2002). Any change in the time series is most reliable if it is detected by statistical tests and also has physical and historical evidences (Salas et al., 1980). Therefore, it is considered an important issue to identify (detect), describe (test), and remove these components.

A number of parametric and nonparametric tests have been suggested in literature for the detection of trend and jumps, and for checking randomness. These tests are considered to be important for scientific purposes as well as for practicing hydrologists. In what follows, a combination of the above-mentioned tests has been briefly described.

2.18.2.1.1 Randomness test An adapted version of a simple nonparametric run test, reported by Adeloye and Montaseri (2002), is given below. The test consists of the following steps (Figure 1): 1. The median of the observation is determined. 2. Each data item is examined to find out if it exceeds the median. If a data item exceeds the median, this is defined as a case of success, S, if not, this is defined as a case of failure, F. Cases that are exactly equal to the median are excluded. 3. Successes and failures are counted and denoted by n1 and n2, respectively. 4. The total number of runs (R) in the data set is determined. A run is a continuous sequence of successes until it is interrupted by a failure or vice versa. 5. The test statistics is computed by

  2n1 n2 R 1 n1 þ n2 z ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2n1 n2 ð2n1 n2  n1  n2 Þ ðn1 þ n2 Þ2 ðn1 þ n2  1Þ

t = t +1

R

xt = x50

R

N

z=

N F

xt > x50

Y

n2 = n2 + 1

S

−1

2n1n2 (2n1n2 − n1 − n2) (n1 + n2)2 (n1 + n2 −1)

N

N R Figure 1 Randomness test.

2n1n2 n1 + n2

n1 = n1 + 1

t z/2

Y

Reject H0

H0: The sequence of Ss and Fs is random.

482

Statistical Hydrology

6. Critical values of the standard normal distribution are obtained for the chosen significance level, a, and denoted by 7za/2. 7. Computed statistics z is compared to the critical values 7za/2. H0 is rejected if zo  za/2 or z4za/2.

2.18.2.1.4 Mann–Kendall test The Mann–Kendall test checks the existence of a trend without specifying if the trend is linear or nonlinear. It is widely reported as in Libiseller and Grimwall (2002). The univariate statistics for monotone trend in a time series xt (t ¼ 1, 2, y , n) is defined as

2.18.2.1.2 Detection of trend A number of parametric and nonparametric trend detection tests are available in the literature (Berryman et al., 1988; Cluis et al., 1989; Helsel and Hirsch, 1992; Salas, 1993; Fanta et al., 2001; Yue et al., 2002; Burn and Elnur, 2002; Adeloye and Montaseri, 2002; Xiong and Guo, 2004; Koutsoyiannis, 2006). Among these, one parametric and two nonparametric tests are supplied below.

2.18.2.1.3 Simple regression on time The simple linear trend line between the variable (x) and time (t) can be written as

xt ¼ a þ bt

S¼

X

sgnðxi  xj Þ

ð4Þ

jo i

where

8 > < 1; sgnðxÞ ¼ 0; > : 1;

if x4 0 if x ¼ 0 if xo 0

ð5Þ

If no ties are present and the values of x1, x2, y , xn are randomly ordered, the test statistics has expectation zero and variance

ð2Þ

VðSÞ ¼

nðn  1Þð2n þ 5Þ 18

ð6Þ

where a and b are parameters of the regression model. A linear trend exists when the null hypothesis that b ¼ 0 is rejected. The null hypothesis is rejected if the test statistics, Tc, satisfies

In the case of presence of tied groups, equations are modified (Salas, 1993).

 pffiffiffiffiffiffiffiffiffiffiffiffi   n2  Tc ¼  pffiffiffiffiffiffiffiffiffiffiffiffiffiffi4 T1a=2; v r 1  r2

The Spearman rank order correlation nonparametric test is used to investigate the existence of a trend that might be found in the time series. The step-by-step explanation of the test for a time series xt (t ¼ 1, y , n) observed in time t (Figure 2) is as follows:

ð3Þ

where r is the cross-correlation coefficient between the variable x (x1, x2, y , xn) and time t ¼ 1, 2, y , n, and T1a/2, v is the 1  a/2 quantile of the Student t distribution with v ¼ n  2 degrees of freedom.

t=1

2.18.2.1.5 Spearman rank order correlation test

1. Ranks, Rxt, are assigned to xt, such that the rank 1 is assigned to the largest xt and the rank n to the least xt. Where there are ties in the xt, then a rank equal to the average of

x1< x2 20 per year 100 0 7

99

Rsm Am20 = 0.33 d−1 Kn20 = 1.0 mgN I−1

40

Influent ammonia concs

30

Rsm = For different influent ammonia concentrations

20

10

KnT ðbAT þ 1=Rs Þ Na ¼ Nae ¼ mAmT  ðbAT þ 1=Rs Þ

1

ðmgN l Þ

ð128Þ 0

From Equation (128), the ammonia concentration (Na) in the reactor and effluent (Nae) are independent of the specific yield coefficient (YA) and the influent ammonia concentration (Nai). Using mAm20 ¼ 0.33 d1 and Kn20 ¼1.0 mgN l1 at 20 1C, and taking bAT ¼ 0.04 d1 (Table 10), a plot of Equation (128) with Nae versus sludge age Rs is given in Figure 24. At long sludge ages Nae is very low and remains so until the sludge age is lowered to about 4 days. Below 4 days, Nae increases rapidly and in terms of Equation (128) can exceed the influent FSA concentration, Nai. This clearly is not possible so the limit of validity of Equation (128) is Na ¼ Nai. Substituting Nai for Na in Equation (128) and solving for Rs give the minimum sludge age for nitrification, Rsm below which theoretically, nitrification cannot be achieved, that is,

Rsm ¼

1 ½mAmT =ð1 þ ðKnT =Nai ÞÞ  bAT

ð129Þ

This minimum sludge age varies slightly with the magnitude of Nai (Figure 24) – higher Nai gives a slightly lower Rsm. The effect of Nai on RSm is very small because the magnitude of KnT is very small relative to Nai (o5%). So for Nai 420 mgN l1 (rarely will it be lower than this), and noting that Kn20B1 mgN l1, then KnT/Nai is negligibly small with respect to 1 (o5%). So substituting zero for KnT/Nai in Equation (129) yields

Rsm ¼

1 mAmT  bAT

ðdaysÞ

ð130Þ

For all practical purposes, taking into account the uncertainty in mAm, Equation (130) adequately defines the minimum sludge age for nitrification. Conceptually, Equation (130)

0

2

4 6 Sludge age (days)

8

10

Figure 24 Effluent ammonia concentration vs. sludge age for the steady-state nitrification model.

states that if the net nitrifier multiplication rate (inverse of the net maximum specific growth rate, mAm  bA) is slower than the harvesting rate of the nitrifiers via the sludge waste flow rate, then the nitrifiers cannot be sustained in the system and nitrification cannot take place. At sludge ages lower than the minimum for nitrification, nitrifiers are washed out of the system and so are called washout sludge ages. This concept of washout can be applied to any group of organisms in a bioreactor, and defines the sludge age below which the bioprocess will not take place because the organisms mediating this process are not sustained in the system. The virtually constant value for Rsm insofar as the influent FSA concentration is concerned (for the fixed values of mAmT and bAT) and the rapid decrease in effluent FSA concentration at sludge ages slightly longer than RSm is due to the very low Monod half saturation concentration for the nitrifiers (Kn20). This feature causes that in a particular plant, as the sludge age is increased, once Rs4Rsm, a high efficiency of nitrification will be observed, provided the FSA is the growth limiting nutrient for the ANOs, that is, all other requirements such as oxygen are met. Consequently, under steady-state conditions with increasing sludge age, kinetically, one would expect an AS system either not to nitrify at all, or, if it nitrifies, to nitrify virtually completely depending on whether the sludge age is

466

Biological Nutrient Removal

shorter or longer than the minimum (Rsm), respectively. Conversely, as sludge age decreases, one would expect an AS system to nitrify virtually completely and then quite suddenly cease to nitrify depending on whether the sludge age is shorter or longer than the minimum (Rsm), respectively. This behavior sometimes occurs in full-scale AS systems, where for many years the system has nitrified virtually completely, and suddenly one winter it stops nitrifying and produces very high effluent FSA concentrations. Provided the oxygen supply is not limiting, what happens in these situations is that over the years, the organic (COD) load on the system has increased and in order to maintain the reactor VSS concentration at the required level, the sludge wastage rate (Qw) has been increased, which reduced the sludge age. Then, coupled with a low winter temperature, the system sludge age falls below the minimum and nitrification ceases. This cannot happen with hydraulic control of sludge age, where a fixed proportion of the reactor volume is wasted daily to establish a constant sludge age. However, the secondary settling tank may become overloaded as the reactor TSS concentration increases with time, depending on the settleability with the AS (see Section 4.14.14). An operator therefore can choose the way an AS system fails with increasing organic loading – it does not have to be with nitrification, and so also with N removal.

4.14.20 Factors Influencing Nitrification From the discussion above, it can be seen that there are a number of factors that affect the nitrification process, the minimum sludge age required to achieve it, and the effluent FSA concentration from the AS system: 1. the magnitude of the kinetic constant mAm20 because this rate can vary considerably in different wastewaters; 2. temperature because it decreases the mAm20 rate and Kn20 coefficient; 3. unaerated zones in the reactor because ANOs are obligate aerobes and can grow only under aerobic conditions; 4. DO concentration because Monod kinetics presumes that FSA is the growth-limiting nutrient implying that the oxygen supply must be adequate; 5. cyclic flow and load conditions because FSA is dissolved and therefore the reactor (and effluent) FSA concentration is affected by the instantaneous actual hydraulic retention time; most FSA not nitrified during the actual hydraulic retention time escapes with the effluent; and 6. pH in the reactor because the mAm20 is strongly suppressed by pH outside the 7–8 range. These six factors are discussed further below.

4.14.20.1 Influent Source The maximum specific growth rate constant mAmT has been observed to be quite specific for the wastewater and also to vary between different batches of the same wastewater source. This specificity is so marked that mnmT should not be classified as a kinetic constant but rather as a wastewater characteristic. The effect appears to be of an inhibitory nature due to some substance(s) in the influent wastewater. It does not appear to

be a toxicity problem because a high efficiency of nitrification can be achieved even with a low mAmT value if the sludge age is increased sufficiently. These inhibitory substances are more likely to be present in municipal wastewater flows having some industrial contribution. In general, the higher the industrial contribution, the lower mAmT tends to be, but the specific chemical compounds that cause the reduction of mAmT have not been clearly defined. A standard temperature of 20 1C has been adopted for reporting mAm rates to take into account the effect of temperature. A range mAm20 values have been reported in the range of 0.30–0.75 d1 for municipal wastewaters. These two limits will have a significant effect on the magnitude of the minimum sludge age for nitrification. Two systems, having these respective mAm20 values, will have Rsm values differing by 250%. Clearly due to the link between the sludge age and mAmT, the latter’s value should always be estimated experimentally for optimal design. In the absence of such a measurement, a low value for mAmT necessarily will need to be selected to ensure that nitrification takes place. If the actual mAm is higher, the sludge age of the system will be longer and the reactor volume larger than necessary. However, the investment in the large reactor is not lost because in the future the plant will be able to treat a higher organic load at a shorter sludge age. Experimental procedures to determine mAm20 are given in the literature (e.g., WRC, 1984). The bn20 rate is taken as constant for all municipal wastewater flows at bn20 ¼ 0.04 d1. Its effect is small so that there is no need to enquire closely into all the factors affecting it. Little information on effects of inhibitory agents on KnT is available; very likely KnT will increase with inhibition.

4.14.20.2 Temperature The mAmT, KnT, and bAT constants are sensitive to temperature with a high-temperature sensitivity for the first two, while the endogenous rate is accepted to have the same low-temperature sensitivity as that for OHOs, viz.,

mAmT ¼ mAm20 ðyn ÞðT20Þ ðd1 Þ

ð131aÞ

KnT ¼ Kn20 ðyn ÞðT20Þ ðmgN l1 Þ

ð131bÞ

bAT ¼ bA20 ðyb ÞðT20Þ ðd1 Þ

ð131cÞ

where yn is the temperature sensitivity for nitrification ( ¼ 1.123) and yb the temperature sensitivity for endogenous respiration for ANOs ¼ 1.029. The effect of temperature on mAmT is particularly strong. For every 6 1C drop in temperature, the mAmT value halves which means that the minimum sludge age for nitrification doubles. Design of systems for nitrification, therefore, should be based on the minimum expected system temperature. The temperature sensitivity of KnT is also strong, doubling for every 6 1C increase in temperature. This does not affect the minimum sludge age for nitrification, but it does affect the effluent FSA concentration – the higher the Kn value, the higher the effluent FSA at Rs b Rsm. However, the faster mAmT rate at the higher temperature compensates for the higher KnT value so that the effluent FSA decreases with increase in temperature.

Biological Nutrient Removal

The effect of unaerated zones on nitrification can be formulated based on the following assumptions: 1. Nitrifiers, being obligate aerobes, grow only in the aerobic zones of a system. 2. Endogenous mass loss of the nitrifiers occurs under both aerobic and unaerated conditions. 3. The proportion of the ANOs in the VSS in the unaerated and aerated zones is the same so that the sludge mass fractions of the different zones of the system reflect the distribution of the nitrifier mass as well. From 1–3 above, it can be shown that if a fraction fxt of the total sludge mass is unaerated (i.e., (1  fxt) is aerated), the effluent ammonia is given by

KnT ðbAT þ 1=Rs Þ Nae ¼ mAmT ð1  f xt Þ  ðbAT þ 1=Rs Þ

ð132Þ

Equation (132) is identical in structure to Equation (128), if one views the effect of the unaerated mass (fxt) as reducing the value of mAmT to mAmT(1  fxt), which conforms with (1) to (3) above. This sludge mass fraction approach is compatible with the nitrification kinetics in the AS kinetic models such as ASM1 and ASM2 (Henze et al., 1987, 1995) and UCTOLD and UCTPHO (Dold et al., 1991; Wentzel et al., 1992). In these models, nitrifier growth takes place only in the aerobic zone and endogenous respiration in all the zones. This sludge mass fraction approach is not compatible with the aerobic sludge age approach, which is used in some ND AS system design procedures (WEF, 1998; Metcalf and Eddy, 1991). In the aerobic sludge age approach, it is assumed that the growth and endogenous processes of the nitrifiers are active only in the aerobic zone, with neither processes active in the unaerated zone(s). This aerobic sludge age approach is not compatible with kinetic models and so significantly different predictions can be expected for the nitrification behavior from the aerobic sludge age-based design procedures and kinetic models. Following the same reasoning as that preceding Equation (132), it can be shown that the minimum sludge age for nitrification Rsm in an ND system having an unaerated mass fraction, fxt, is

Rsm ¼

1 mAmT ð1  f xt Þ  bAT

ð133Þ

Alternatively, if Rs is specified, then the minimum aerobic sludge mass fraction (1  fxm) that must be present for nitrification to take place is found by substituting Rs for Rsm and fxm for fxt in Equation (133) and solving for (1  fxm), that is,

ð1  f xm Þ ¼ ðbAT þ 1=Rs Þ=mAmT

ð134Þ

or equivalently, from Equation (134), the maximum allowable unaerated sludge mass fraction at a sludge age of Rs is

f xm ¼ 1  ðbAT þ 1=Rs Þ=mAmT

ð135Þ

For a fixed sludge age, Rs, the design value for the minimum aerobic sludge mass fraction (1  fxm) should always be significantly higher than that given by Equation (134), because

nitrification becomes unstable and the effluent ammonia concentration increases when the aerated sludge mass fraction decreases to near the minimum value as given by Equation (134) in the same way as when the sludge age (Rs) approaches the minimum for nitrification (Rsm). This situation is exacerbated by cyclic flow and ammonia load conditions (see below). Consequently to ensure low effluent ammonia concentrations, the maximum specific growth rate of nitrifiers must be decreased by a factor of safety, Sf, to give the minimum design aerobic sludge mass fraction; from Equation (134),

ð1  f xm Þ ¼ ðbAT þ 1=Rs Þ=ðmAmT =Sf Þ

ð136aÞ

The corresponding maximum design unaerated sludge mass fraction, from Equation (136a), is

f xm ¼ 1  Sf ðbAT þ 1=Rs Þ=mAmT

ð136bÞ

With the aid of the temperature dependency equations for nitrification (Equation (131)), the maximum unaerated sludge mass fraction (fxm) from Equation (136b) is shown in Figure 25 for Sf ¼ 1.25 and mAm20 rates from 0.25 to 0.50 at 14 1C. This shows that fxm is very sensitive to mAmT. Unless a sufficiently large aerobic sludge mass fraction (1  fxm) is provided, nitrification will not take place and consequently nitrogen removal by denitrification is not possible. In fact, the selection of the maximum unaerated sludge mass fraction to achieve near complete nitrification and a required degree of N removal is the single most important decision that is made in the design of the BNR AS system because it defines the system sludge age and, for a selected reactor MLSS concentration, also the reactor volume. From Equations (132) and (136), it can be shown that at fxm for constant flow and ammonia load (i.e., steady-state conditions)

Nae ¼ KnT =ðSf  1Þ ðmgN l1 Þ

ð137Þ

From Equation (137), if Sf is selected at say 1.25 or greater at the minimum wastewater temperature, the effluent ammonia

0.80 Maximum unaerated sludge mass fraction

4.14.20.3 Unaerated Zones

467

Temperature = 14 °C

Factor of safety = 1.25

Recommended maximum 0.60 50

0.

40 36 0.

0.

0.40

30

0.

25

0.

0.20

Am20

0.00 0

10

20 30 Sludge age (days)

40

Figure 25 Maximum unaerated sludge mass fraction required to ensure nitrification vs. sludge age for maximum specific growth rates of nitrifiers mAm20 of 0.25–0.50 d1 at 14 1C for Sf ¼ 1.25.

468

Biological Nutrient Removal

concentration (Nae) will be lower than 2 mgFSA-N l1 at 14 1C for Kn20 ¼1.0 mgN l1. Although Kn is higher at higher temperature, Nae will decrease with increase in temperature because at constant sludge age, Sf increases with increase in mAmT. Consequently, for design the lower expected temperature should be selected to determine the sludge age and the aerobic mass fraction. If this is done, using say Sf ¼ 1.25, then it can be accepted from Equation (137) that the effluent ammonia concentration is below 2 mgN l1 at the lowest temperature and around 1 mgN l1 at 20 1C. In this way, explicitly calculating Nae with Equation (132) is not necessary because provision for near complete nitrification has been made by the selection of Sf. Clearly, selection of the mAm20 and Sf values has major consequences on the effluent FSA concentration and economics (size) of the ND AS system.

4.14.20.3.1 Maximum allowable unaerated mass fraction The above equations allow the two most important decisions in the design of an NDAS system to be made, the maximum unaerated sludge mass fraction and sludge age to ensure near complete nitrification. Evidently from Figure 25, for mAm20 4 0.50 the unaerated mass fraction at 14 1C can be as high as 0.7 at a sludge age of 40 days. Such a high unaerated mass fraction is apparently also acceptable at RsZ10 days at 20 1C. However, there are additional considerations that constrain the unaerated mass fraction – sludge age selection. 1. Experience with laboratory-scale ND (and NDBEPR) systems has shown that at unaerated mass fractions greater than 0.40, the filamentous bulking can become a problem, in particular at low temperatures (o16 1C). Systems with low unaerated mass fractions of o0.30 show greater tendency for good settling sludges (Musvoto et al., 1994; Ekama and Wentzel, 1999a; Tsai et al., 2003). 2. For design of BNR plants for high N and P removal, the unaerated sludge mass fraction fxm usually needs to be high (440%). If the mAm20 value is low (o 0.40 d1, which will be the usual case in designs where insufficient information on the mAm20 is available), the necessary high fxm magnitudes will be obtained only at long sludge ages (Figure 25). For example, if mAm20 ¼ 0.35 d1, then with Sf ¼ 1.3 at Tmin ¼ 14 1C, an fxm ¼ 0.45 (Equation (136b)) gives a sludge age of 25 days and for fxm ¼ 0.55 a sludge age of 37 days. Long sludge ages require large reactor volumes – increasing Rs from 25 to 37 days increases the reactor volume by 40%, whereas fxm increased only 22%. Also, for the same P content in the sludge mass, the P removal is reduced as the sludge age increases because the mass of sludge wasted daily decreases as the sludge age increases. Consequently, for low mAm20 values, the increase in N and P removal that can be obtained by increasing the unaerated sludge mass fraction above 0.50–0.60 might not be economical due to the large reactor volumes this will require, and might even be counterproductive insofar as it affects P removal. A sludge age of 30 days probably is near the limit of economic practicality which, for low mnm14 ¼ 0.16 values, will limit the unaerated mass fraction to about 0.5. At higher mnm14 values, the sludge ages allowing 50% unaerated mass fractions decrease significantly again indicating the advantages of determining

experimentally the value of mAm20 to check whether a higher value is acceptable. 3. An upper limit to the unaerated mass fraction is evident also from experimental and theoretical modeling of the BNR system. Experimentally at 20 1C with Rs ¼ 20 days, if fxm 40.70, the mass of sludge generated is found to increase sharply. Theoretically, this happens for fxm 4 0.60 at T ¼ 14 1C and Rs ¼ 20 days. The reason is that for such a high fxm, the exposure of the sludge to aerobic conditions becomes insufficient to utilize the adsorbed and enmeshed BPOs. This leads to a decrease in active mass and oxygen demand and a buildup of enmeshed nondegraded organics. When this happens, the system still functions in that the COD is removed from the wastewater, but the degradation of the COD is reduced; the system begins to behave as a contact reactor of a contact-stabilization system, that is, a bio-flocculation with minimal degradation. This critical state occurs at lower fxm as the temperature is decreased and the sludge age is reduced. From the above discussion, it would appear that the unaerated mass fraction should not be increased above an upper limit of about 60%, as indicated in Figure 25, unless there is a specific reason for this (Tsai et al., 2003).

4.14.20.4 DO Concentration High DO concentrations, up to 33 mg l1, do not appear to affect nitrification rates significantly. However, low oxygen concentrations reduce the nitrification rate. Stenstrom and Poduska (1980) have suggested formulating this effect as follows:

mAmO ¼ mAm

O ðd1 Þ KO þ O

ð138Þ

where O is oxygen concentration in liquid (mgO l1), KO the half-saturation constant (mgO l1), mAmo the maximum specific growth rate (d1), and mAO the specific growth rate at DO of O mg l1. The value of KO ranges from 0.3 to 2 mgO l1, that is, at DO values below KO the growth rate will decline to less than half the rate where oxygen is present in adequate concentrations. The wide range of KO probably has arisen because the concentration of DO in the bulk liquid is not necessarily the same as inside the biological floc where the oxygen consumption takes place. Consequently, the value will depend on the floc size, mixing intensity, and oxygen diffusion rate into the floc. Furthermore, in a full-scale reactor the DO will vary over the reactor volume due to the discrete points of oxygen input (with mechanical aeration) and the impossibility of achieving instantaneous and complete mixing. For these reasons, it is not really possible to establish a generally applicable minimum oxygen value – each reactor will have a value specific to the conditions prevailing in it. In nitrifying reactors with bubble aeration a popular DO lower limit, to ensure unimpeded nitrification, is 2 mgO l1 at the surface of the mixed liquor. Under cyclic flow and load conditions the difficulties of ensuring an oxygen supply matching the oxygen demand and a lower limit for the DO concentration are difficult.

Biological Nutrient Removal

4

2

3

2

1

0.0 0

0 0 (a)

Max. effl. FSA/steady-state FSA ratio

Amplitude of influent flow and FSA conc. 1.00 0.75 0.50 0.25

6

Steady-state FSA conc. (mgN I−1)

Max. effl. FSA/steady-state FSA ratio

4

8

1

2

3

4

5

T = 22 °C: Raw sewage

10

5

8

0.8

0.0 Amplitude of influent flow and FSA

6

0.6

1.00 4

0.4

0.75 0.50

2

0.2

0.25

0.0

0 2

6

R s /R sm ratio

1.0

Steady-state effluent FSA

(b)

Steady-state FSA conc. (mgN I−1)

T = 14 °C: Raw sewage Steady-state effluent FSA

10

469

4

6

8

10

12

14

16

R s /R sm ratio

Figure 26 (a) Maximum to steady-state effluent FSA concentration ratio vs. sludge age to minimum sludge age for nitrification ratio for influent flow and ammonia concentration amplitude (in phase) of 0.0 (steady state) 0.25, 0.50, 0.75, and 1.0 at 14 1C. (b) Maximum to steady-state effluent FSA concentration ratio vs. sludge age to minimum sludge age for nitrification ratio for influent flow and ammonia concentration amplitude (in phase) of 0.0 (steady-state) 0.25, 0.50, 0.75, and 1.0 at 22 1C.

Where storm flows are not of long duration, flow equalization is a practical way to facilitate control of the DO concentration in the reactor. In fact, most of the diurnal variations in reactor dissolved concentrations are a direct consequence of diurnal flow variation – negligibly little is due to the kinetic rates of the biological processes, especially at long sludge ages. In the absence of flow equalization, amelioration of the adverse effects of low DO concentration during peak oxygen demand periods occurs by increasing the sludge age to significantly longer than the minimum necessary for nitrification, that is, by effectively increasing Sf.

4.14.20.5 Cyclic Flow and Load It is well known both experimentally and theoretically with simulation models that under cyclic flow and load conditions the nitrification efficiency of the AS system decreases compared with that under steady-state conditions. From simulation studies, during the high flow and/or load period, even though the nitrifiers are operating at their maximum rate, it is not possible to oxidize all the ammonia available, and an increased ammonia concentration is discharged in the effluent. This in turn reduces the mass of nitrifiers formed in the system. Equivalently, the effect of diurnal variation in flow and load is to reduce the system sludge age. The average effluent ammonia concentration from a system under cyclic flow and load conditions is therefore higher than that from the same system under constant flow and load (steady-state conditions). The adverse effect of the diurnal flow variation becomes more marked as the fractional amplitude of the flow and load variation increase and is ameliorated as the safety factor Sf increases. Simulation studies of the diurnal flow effect show a relatively consistent trend between the maximum or average effluent FSA concentrations under diurnal conditions and the steady-state effluent FSA concentration versus the ratio of system sludge age and the minimum sludge age for nitrification (Rs/Rsm). For mAm20 ¼ 0.45 d1 (other constants in

Table 10), Figures 26(a) (for 14 1C) and 26(b) (for 22 1C) show the maximum (average not shown) effluent FSA concentration as a ratio of the steady-state effluent FSA concentration versus the system sludge age as a ratio of the minimum sludge age for nitrification (Rs/Rsm) for a single reactor fully aerobic system receiving cyclic influent flow and FSA load as in-phase sinusoidally varying flow and ammonia concentration, both with amplitudes of 0.25, 0.50, 0.75, 1.00, and 0.0 (steady state). For example, at 14 1C (Figure 26(a)) if the system sludge age is 2 times the minimum for nitrification, the maximum effluent FSA concentration is 8 times the steadystate value. From Figure 26(a), the latter is 0.8 mgN l1 so the maximum is 8  0.8 ¼ 6.4 mgN l1. From Figures 26(a) and 26(b), clearly the greater the diurnal flow variation and the lower the temperature, the higher the maximum (and average) effluent ammonia concentrations. This can be compensated for by increasing Sf, which has the effect of increasing the sludge age or decreasing the unaerated mass fraction of the system. This obviously has an impact on the effluent quality and/or economics of the system. The importance of the selection of mAm cannot be overemphasized. If the value of mAm is selected higher than the actual value, even with a safety factor Sf of 1.25–1.35, the plant is likely to produce a fluctuating effluent ammonia concentration, with reduced mean efficiency in nitrification. Hence, conservative estimates of mAm (low) and Sf (high) are essential for ensuring nitrification and low effluent ammonia concentration.

4.14.20.6 pH and Alkalinity The mAm rate is very sensitive to the pH of the mixed liquor outside the 7–8 range. It seems that the free ammonia (NH3) and nitrous acid (HNO2) act inhibitorily when their respective concentrations increase too high. This happens when the pH increases above 8.5 (increasing (NH3)) or decreases below 7

470

Biological Nutrient Removal

(increasing (HNO2)); optimal nitrification rates are expected for 7opHo8.5 with sharp declines outside this range. From the overall stoichiometric equations for nitrification (Equation (118a)), nitrification releases hydrogen ions which in turn decreases H2CO3* Alkalinity of the mixed liquor. For every 1 mgFSA that is nitrified 2  50/14 ¼ 7.14 mg Alkalinity (as CaCO3) is consumed. Based on equilibrium chemistry of the carbonate system (Loewenthal and Marais, 1977), equations linking the pH with H2CO3* Alkalinity for any dissolved carbon dioxide concentration can be developed. These relationships are plotted in Figure 27. When the H2CO3* Alkalinity falls below about 50 mg l1 as CaCO3 then, irrespective of the carbon dioxide concentration, the pH becomes unstable and decreases to low values. Generally, if nitrification causes the H2CO3*Alkalinity to drop below about 50 mg l1 (as CaCO3), problems associated with low pH will arise at a plant, such as poor nitrification efficiency, effluents aggressive to concrete, and the possibility of development of bulking (poor settling) sludges (Jenkins et al., 1993). For any particular wastewater, the effect of nitrification on pH can be readily assessed, as follows: for example if a wastewater has a H2CO3*Alkalinity of 200 mg l1 as CaCO3 and the expected production of nitrate is 24 mgN l1, then the expected H2CO3*Alkalinity in the effluent will be (200  7.14  24) ¼ 29 mg l1 as CaCO3. From Figure 27, such an effluent will have a pH o7.0. Wastewaters having low Alkalinity (capital A denotes H2CO3* Alkalinity) are often encountered where the municipal supply is drawn from areas underlain with sandstone. A practical approach to treating such wastewaters is to (1) dose lime or better (2) create an anoxic zone(s) to denitrify some or all of the nitrate generated. In contrast to nitrification, denitrification takes up hydrogen ions which is equivalent to generating Alkalinity (see Section 4.14.24.2). By considering nitrate as electron acceptor, it can be shown that for every milligram of nitrate denitrified, there is an increase of 1  50/ 14 ¼ 3.57 mg Alkalinity as CaCO3. Hence, incorporating denitrification in a nitrification system causes the net loss of

10 0.5 1.0 2.0 5.0 10.0

Mixed liquor pH value

8 Carbon dioxide concentration (mg I−1 as CaCO3)

6

Saturation ~ 0.5 mg I−1 as CaCO3

4

Alkalinity to be reduced usually sufficiently to maintain the Alkalinity above 50 mg l1 as CaCO3 and consequently the pH above 7. In the example above, where the Alkalinity in the system is expected to decline to 29 mg l1 as CaCO3, if 50% of the nitrate were denitrified, the gain in Alkalinity would be (0.5  24  3.57) ¼ 43 mg l1 as CaCO3 and will result in an Alkalinity of (29 þ 43) ¼ 72 mg l1 as CaCO3 in the system. In this event the pH will remain above 7. For low Alkalinity wastewaters, it is imperative, therefore, that denitrification be built into nitrifying plants, even if N removal is not required. Incorporation of unaerated zones in the system influences the sludge age of the system at which nitrification takes place so that cognizance must be taken of the effect of an anoxic or unaerated zone in establishing the sludge age of a nitrifying– denitrifying plant (see Section 4.14.20.3). In the AS systems treating reasonably well buffered wastewaters, quantifying the effect of pH on nitrification is not critical because pH reduction can be limited or completely obviated by including anoxic zones, thereby ensuring Alkalinity recovery via denitrification. However, in poorly buffered wastewaters, or wastewaters with high influent N (such as AD liquors), the interaction between the biological processes, pH, and nitrification is the single most important one for the N removal AS system. Hence, it is essential to include the effect of pH on the nitrification rate for such wastewaters to quantify this important interaction. From Equation (121), the specific growth rate of the ANOs (mA) is a function of both mAm and Kn. It was shown above that the minimum sludge age is dominated by the magnitude of mAmT; it is only very weakly influenced by KnT. At RscRsm, the effluent ammonia concentration (Nae), although low, is, relatively speaking, significantly higher for larger KnT values: for example, if KnT increases by a factor of 2, the effluent ammonia concentration will increase correspondingly by the same factor (Equation (132)). Consequently, the value of KnT is significant insofar as it governs the effluent ammonia concentration once nitrification takes place at RscRsm. Several investigations have been made to understand the effect of pH on mAmT. These investigations generally have not separated out the effect of pH on mAmT and KnT so that most data are in effect lumped parameter estimates of mAmT. Almost no information is available on the effect of pH on KnT by itself. Quantitative modeling of the effect of pH on mAm has been hampered by the difficulty of accurately measuring the effects of pH on nitrification. Studies have shown that mAm can be expressed as a percentage of the highest value at optimum pH. Accepting this approach and that mAm is highest and remains approximately constant in the pH range for 7.2opHo8.0 but decreases as the pH decreases below 7.2 (Downing et al., 1964; Loveless and Painter, 1968), So¨temann et al. (2005a) modeled the mA  pH dependency as For 5opHo7.2,

2

mAmpH ¼ mAm7:2 yns ðpH7:2Þ 0 −100

0

100 Alkalinity (mg

200 I−1

300

as CaCO3)

Figure 27 Mixed liquor pH vs. H2CO3* alkalinity for different concentrations of carbon dioxide.

400

ð139aÞ

where yns is the pH sensitivity coefficient (E2.35). Declining mAm values at pH48.0 have been observed and it has been noted that nitrification effectively ceases at a pH of about 9.5 (Malan and Gouws, 1966; Wild et al., 1971; Antoniou et al., 1990). Accordingly, for pH47.2, So¨temann et al.

Biological Nutrient Removal

(2005a) proposed Equation (139b) to model the decline in the mAm from pH 47.2 to 9.5 as a function of mAm7.2 using inhibition kinetics as follows:

mAmpH ¼ mAm7:2 KI

Kmax  pH Kmax þ KII  pH

ð139bÞ

where KI ¼ 1.13, Kmax ¼ 9.5, KIIE0.3. The overall effect of pH on mAm is modeled by combining Equations (139a) and (139b), which is given by Equation (139c) and shown in Figure 28. It can be seen that in the range pH ¼ 7.2–8.3, the change in mAmpH is small, with mAmpH/mAm7.2 40.9:

mAmpH ¼ mAm7:2 2:35ðpH7:2Þ KI

Kmax  pH Kmax þ KII  pH

ð139cÞ

where 2.35(pH7.2) is set ¼ 1 for pH47.2,

KI

Kmax  pH ¼1 Kmax þ KII  pH

for pH o7.2 and mAmpH ¼ 0 for pH49.5. Experimental data from the literature are also shown in Figure 28 to provide some quantitative support for Equation (139c). At low pH (o7.2), data from Wild et al. (1971) and Antoniou et al. (1990) fit the equation reasonably well. Very few data are available for pH48.5, but the few points from Antoniou et al. (1990) show reasonable agreement with Equation (139c). Accordingly, Equation (139c) was accepted to calculate mAmpH in the pH range 5.5–9.5. From Equation (139c), the minimum sludge age for nitrification (Rsm) at different pH and temperature (T) and unaerated mass fraction (fxm) is given by

Rsm ¼ 1=½mApHT ð1  f xm Þ  bnT 

ðdaysÞ

ð140Þ

The problem with nitrification in low alkalinity wastewater is that the pH obtained is not known, because it is interactively 1.2

Fraction Unm/Umm7.2

1 Eq (139b) 0.8 0.6 0.4 Eq (139a) Eq (139b) 0.2 0 4

5

6

7

8

9

10

471

established between the degree of nitrification, loss of alkalinity, pH, and mApHT. To investigate this interaction, the biological kinetic ASM1 model for carbon (C) and nitrogen (N) removal was integrated by So¨temann et al. (2005a) with a two-phase (aqueous-gas) mixed weak acid/base chemistry kinetic model to extend application of ASM1 to situations where an estimate for pH in the biological reactor is important. This integration, which included CO2 (and N2) gas generation by the biological processes and their stripping by aeration, made a number of additions to ASM1, inter alia the above effect of pH on the autotrophic nitrifiers (ANOs). From simulation of a long sludge age ND AS system with incrementally decreasing influent H2CO3* Alkalinity, when the effluent H2CO3* alkalinity fell below about 50 mg l1 as CaCO3, the aerobic reactor pH dropped below 6.3, which severely retarded nitrification and caused the minimum sludge age for nitrification (Rsm) to increase up to the operating sludge age of the system. The simulation confirmed the earlier conclusion that when treating low H2CO3* alkalinity wastewater (1) the minimum sludge age for nitrification (Rsm) varies with temperature and reactor pH and (2) for low effluent H2CO3* alkalinity (o50 mg l1 as CaCO3), nitrification becomes unstable and sensitive to dynamic loading conditions resulting in increases in effluent ammonia concentration, reduced nitrification efficiency, and as a result lower N removal. For effluent H2CO3* alkalinity o50 mg l1, lime should be dosed to the influent to raise the aerobic reactor pH and stabilize nitrification and N removal.

4.14.21 Nutrient Requirements for Sludge Production All live biological material and some unbiodegradable organic compounds contain nitrogen (N) and phosphorus (P). The organic sludge mass (VSS) that accumulates in the biological reactor comprises active organisms (XBH), endogenous residue (XEH), and UPOs (XI), each of which contains N and P. From TKN and VSS tests conducted on AS, it has been found that the N content (as N with respect to VSS, fn, mgN/mgVSS) ranges between 0.09 and 0.12 with an average of about 0.10 mgN/ mgVSS. Similarly, from total P and VSS tests, the P content (as P with respect to VSS, fp, mgP/mgVSS) of AS in purely aerobic and anoxic aerobic systems ranges between 0.01 and 0.03 with an average of about 0.025 mgP/mgVSS. From the steady-state model, the relative proportions of the active organisms (XBH), endogenous residue (XEH), and UPOs (XI) change with sludge age. Yet, it has been found that the fn value of the VSS is relatively constant at 0.10 mgN/ mgVSS. This indicates that the N content of the active organisms (XBH), endogenous residue (XEH), and UPOs (XI) is closely the same; if they were significantly different, it would be observed that fn changes in a consistent manner with sludge age. Likewise, for fully aerobic systems, the P content of the three constituents of AS is approximately similar at 0.025 mgP/mgVSS.

pH Figure 28 Maximum specific growth rate of nitrifiers, as a fraction of the rate at pH 7.2, vs. pH of the mixed liquor. (F), Model; (), Malan and Gouws (1966); ( ), Downing et al. (1964); (  ), Wild et al (1971); and (m), Antoniou et al. (1990).

4.14.21.1 Nitrogen Requirements The mass of N (or P) incorporated into the sludge mass is calculated from a N balance over the completely mixed AS

472

Biological Nutrient Removal

system (Figure 2) under steady-state daily conditions, viz., TKN flux out ¼ TKN flux in TKN flux in ¼ Qi Nti (mgN d1) TKN flux out ¼ TKN flux in Qe and Qw

Noting that Qw þ Qe ¼ Qi and Qw ¼ Vp/Rs yields

Qi Nte ¼ Qi Nti  f n Xv Vp =Rs from which

Nte ¼ Nti  f n MXv =ðRs Qi Þ ðmgN l1 Þ

ð141Þ

where Nte is the effluent TKN concentration (mgN l1). The term fnMXv/(RsQi) is denoted Ns and is the concentration of influent TKN in mgN l1 that is incorporated into sludge mass and removed from the system bound in the particulate sludge mass in the waste flow (Qw): 1

Ns ¼ f n MXv =ðRs Qi Þ ðmgN l

influentÞ

ðmgN l1 Þ

ð143Þ

From Equation (141), under daily average conditions, the concentration of N per liter influent required for incorporation into sludge mass is equal to the N content of the mass of sludge (VSS) wasted per day divided by the influent flow. Substituting Equation (106) relating the mass of sludge (VSS) in the reactor (MXv) to the daily average organic load on the reactor (FSti), cancelling Qi and dividing by Sti yields the concentration of N required per liter influent for sludge production per mgCOD/l organic load on the reactor, viz.,

  ð1  f S0 us  f S0 up ÞYH f S0 up Ns ¼ fn ð1 þ f EH bH Rs Þ þ Sti ð1 þ bH Rs Þ f cv ðmgN=mgCODÞ

Nae ¼ Nai þ Nobsi þ Nobpi  ðNs  Noupi Þ ðmgN l1 Þ

ð142Þ

From the N mass balance, this Ns concentration does not include the N in dissolved form in the waste flow. The soluble TKN concentration in the waste flow is the same as the effluent TKN concentration, Nte, which is soluble N in the form of ammonia (Nae) and unbiodegradable soluble organic N (Nouse). Therefore, from Equation (141), provided nitrifiers are not supported in the AS reactor so that nitrification of ammonia to nitrate does not take place, the effluent TKN concentration Nte is given by

Nte ¼ Nti  Ns

organics (Nobsi and Nobpi) is released as FSA when these organics are broken down. This FSA adds to the FSA in the reactor from the influent. Some of the FSA in the reactor is taken up by the OHOs to form new OHO biomass. Some of the OHO biomass in the reactor is lost via the endogenous respiration process. The N associated with the biodegradable part of the OHO biomass is released back to the FSA pool in the reactor but the N in the unbiodegradable endogenous residue part remains as organic N bound in the endogenous residue VSS. Due to these interactions, it is possible that the effluent FSA concentration from a non-nitrifying AS system is higher than the influent FSA concentration – this occurs when the influent TKN comprises a high biodegradable organic N fraction. If the conditions are favorable for nitrification, the net FSA concentration in the reactor is available for the ANOs for growth with the associated generation of nitrate. Unless taken up for OHO growth or nitrified, the FSA remains as such and exits the system with the effluent. So in the absence of nitrification, the effluent ammonia concentration Nae is given by

ð144Þ

The influent TKN comprises ammonia and N bound in organic compounds of a soluble and particulate and biodegradable and unbiodegradable nature. The unbiodegradable organics, some of which contain organic N, are not degraded in the AS system. The influent unbiodegradable soluble organic N (Nousi) exits the system with the effluent (and waste flow) streams. The UPOs are enmeshed with the sludge mass in the reactor and so the organic N associated with these organics exits the system via the daily waste sludge (VSS) harvested from the system. The N bound in the biodegradable

ð145Þ and the effluent TKN (Nte) concentration by

Nte ¼ Nouse þ Nae

ðmgN l1 Þ

ð146Þ

The same approach is applied for the phosphorus (P) requirement for sludge production. Accepting that the P content of the AS in the fully aerobic system without BEPR is 0.025 mgP/mgVSS, the effluent total P (TP) concentration Pte is given by

Pte ¼ Pti  Ps

ðmgP l1 Þ

ð147Þ

where

Ps MXv f p Ns ¼ fp ¼ Sti Rs Qi f n Sti

ðmgP l1 influentÞ

ð148Þ

4.14.21.2 N (and P) Removal by Sludge Production A plot of Equations (144) and (148) versus sludge age is given in Figure 29 for fn ¼ 0.10 mgN/mgVSS, fp ¼ 0.025 mgP/mgVSS for the example raw and settled wastewaters. It is evident that higher concentrations of TKN and TP are required for sludge production for raw than for settled wastewaters. This is because greater quantities of sludge are produced per mgCOD organic load on the reactor at the same sludge age when treating raw wastewaters (see Section 4.14.13). Also, the N and P requirements decrease as the sludge age increases because net sludge production decreases as sludge age increases. Generally, for sludge ages greater than about 10 days, the N removal from the wastewater attributable to net sludge production is less than 0.025 mgN/mgCOD load on the reactor. As influent TKN/COD ratios for domestic wastewater are in the approximate range 0.07–0.13 (Figure 29), it is clear that only a minor fraction of the influent TKN (A in Figure 29) is removed by incorporation into sludge mass. Additional N removal (B in Figure 29) is obtained by transferring the N from the dissolved form in the liquid phase to the gas phase

Biological Nutrient Removal Nutrient requirements 0.035 Approximate range of influent TKN/COD and P/COD ratios of municipal wastewaters

0.12 0.10

0.030 0.025

0.08

0.020

0.06

0.015 B

0.04

0.010 Raw

0.02

0.005 A

P requirement (mgP/mgCOD)

N requirement (mgN/mgCOD)

0.14

Settled

0.00

0.000 0

5

10 15 20 Sludge age (days)

25

30

Figure 29 Approximate minimum nutrient N and P requirements as mgN l1 influent TKN and mgP l1 influent total P per mgCOD l1 organic load on the activated sludge reactor vs. sludge age for the example raw and settled wastewaters at 20 1C together with influent TKN and TP to COD concentration ratio ranges for municipal wastewater.

by autotrophic nitrification and heterotrophic denitrification, which transforms the nitrate to nitrogen gas in anoxic (nonaerated) reactor(s). The details of heterotrophic denitrification are presented below. From Figure 29, normal P removal by incorporation into biological sludge mass is limited at about 0.006 and 0.004 mgP/mgCOD for raw and settled wastewaters respectively, effecting a TP removal of about 20–25% from average municipal wastewaters. As transformation of dissolved orthoP to a gaseous form is not possible, to increase the P removal from the liquid phase, additional ortho-P needs to be incorporated into the sludge mass. This can be achieved in two ways: (1) chemically and/or (2) biologically. With chemical P removal, iron or aluminum chlorides or sulfates are dosed to the influent (pre-precipitation), to the AS reactor (simultaneous precipitation) or to the final effluent (post-precipitation). The disadvantage of chemical P removal is that it significantly increases (1) the salinity of treated wastewater, (2) the sludge production due to the inorganic solids formed, and (3) the complexity and cost of the WWTP. With biological P removal, the environmental conditions in the biological reactor are designed in such a way that a specific group of heterotrophic organisms (called PAOs) grow in the AS reactor. With the accumulated polyPs, these organisms have a much higher P content than the OHOs, as high as 0.38 mgP/ mgPAOVSS (Wentzel et al., 1990). The more PAOs that grow in the reactor, the higher will be the mean P content of the VSS sludge mass in the reactor and therefore the higher the P removal via the wasted sludge. With a significant mass of PAOs present, the mean P content of the VSS sludge mass can increase from 0.025 mgP/mgVSS in aerobic systems to 0.10– 0.15 mgP/mgVSS in biological N and P removal systems. The advantage of biological P removal over chemical P removal is that (1) the salinity of the treated wastewater is not increased, (2) sludge production is increased only between 10% and 15%, and (3) the system is less complex and costly to operate.

473

A disadvantage of biological P removal is that, being biological, it is less dependable and more variable than chemical P removal. The biological processes which mediate biological N and P removal in AS systems and the different reactor configurations in which these take place are described in Section 4.14.28.

4.14.22 Nitrification Design Considerations The kinetic equations describing the interactions between the FSA and the organic N are complex and have been developed in terms of the growth–death–regeneration approach in AS simulation models such as ASM1 and ASM2. However, for steady-state conditions assuming (1) all the biodegradable organics are utilized in the reactor and (2) a TKN mass balance over the AS system, a simple steady-state nitrification model can be developed from the nitrification kinetics and the N requirements for sludge production considered above. This model is adequate for steady-state design and from it some general response graphs are developed below for the example raw and settled wastewaters. Dynamic system responses can be determined with the simulation models once (1) the AS system has been designed and sludge age, zone and reactor volumes and recycle flows are known and (2) the steady-state concentrations have been calculated to serve as initial conditions for the simulation. In the nitrifying AS system design, the (1) effluent FSA, TKN, and nitrate concentrations and (2) the nitrification oxygen demand need to be calculated.

4.14.22.1 Effluent TKN The filtered effluent TKN (Nte) comprises the FSA (Nae) and the unbiodegradable soluble organic N (Nouse). Once mAm20, fxt, Rs, and Sf have been selected, the equations for these concentrations are: 1. Effluent FSA (Nae). Nae is given by Equation (132), which applies only if Rs4Rsm, which will be the case for Sf41.0. 2. Effluent soluble biodegradable organic nitrogen concentration (Nobse). The biodegradable organics (both soluble and particulate) are broken down by the OHOs releasing the organically bond N as FSA. In the steady-state model, it is assumed that all the biodegradable organics are utilized. Hence, the effluent soluble biodegradable organic N concentration (Nobse) is zero. 3. Effluent soluble unbiodegradable organic nitrogen concentration (Nouse). Being unbiodegradable, this concentration of organic N flows though the AS system with the result that the effluent concentration (Nouse) is equal to the influent concentration (Nousi), that is,

Nouse ¼ Nousi

ð149Þ

where Nousi is the influent soluble unbiodegradable organic nitrogen, mgOrgN-N l1 ¼ fN0 ous Nti, where fN0 ous is the soluble unbiodegradable organic N fraction of the influent TKN (Nti). The two nonzero effluent TKN concentrations (FSA, Nae and OrgN, Nouse) are soluble and so exit with the effluent (and

474

Biological Nutrient Removal

waste flow). The soluble (filtered) TKN in the effluent (Nte) is given by their sum, that is,

Nte ¼ Nae þ Nousi

ðfiltered TKNÞ

ð150Þ

If the effluent sample is not filtered, the effluent TKN will be higher by the concentration of TKN in the effluent VSS, that is,

Nte ¼ Nae þ Nouse þ f n Xve

ðunfiltered TKNÞ

nitrogen required for sludge production per mgCOD applied (from Equation (144)). The nitrification capacity to influent COD concentration ratio (Nc/Sti) of a system can be estimated approximately by evaluating each of the terms in Equation (153) as follows:

•

ð151Þ

where Xve is the effluent VSS concentration (mgVSS l1) and fn the N content of VSS (B0.1 mgOrgN-N/mgVSS).

•

4.14.22.2 Nitrification Capacity From a TKN mass balance over the AS system and Rs 4 Rsm, the concentration of nitrate generated in the system (Nne) with respect to the influent flow is given by the influent TKN (Nti) minus the soluble effluent TKN (Nte) and the concentration of influent TKN incorporated in the sludge wasted daily from the AS system (Ns), that is,

Nne ¼ Nc ¼ Nti  Nte  Ns

ð152Þ

The Ns concentration is determined from the mass of N incorporated in the VSS mass harvested from the reactor per day (Equation (142)). The mass of VSS in the reactor (MXv) does not have to include the VSS mass of nitrifiers because this mass, as mentioned earlier, is negligible (o2–4%). In Equation (152), Nc defines the ‘nitrification capacity’ of the AS system. The nitrification capacity (Nc) is the mass of nitrate produced by nitrification per unit average influent flow, that is, mgNO3-N l1. In Equation (150), the effluent TKN concentration (Nte) depends on the efficiency of nitrification. In the calculation for the maximum unaerated sludge mass fraction (fxm) at a selected sludge age, if the factor of safety (Sf) was selected 41.25 to 1.35 at the lowest expected temperature (Tmin), the efficiency of nitrification be high (495%) and Nae generally will be less than 1–2 mgN l1. Also, with Sf 41.25 at Tmin, Nae will be virtually independent of both the system configuration and the subdivision of the sludge mass into aerated and unaerated mass fractions. Consequently, for design, with Sf41.25, Nte will be around 3–4 mgN l1 provided that there is reasonable assurance that the actual mAm20 value will not be less than the value accepted for design and that there is sufficient aeration capacity so that nitrification is not inhibited by an insufficient oxygen supply. Accepting the calculated fxm and selected sludge age (Rs) at the lower temperature, then at higher temperatures the nitrification efficiency and the factor of safety (Sf) both will increase so that at summer temperatures (Tmax), Nte will be lower, approximately 2–3 mgN l1. Dividing Equation (152) by the total influent COD concentration (Sti) yields the nitrification capacity per mgCOD applied to the biological reactor, Nc/Sti, viz.,

Nc =Sti ¼ Nti =Sti  Nte =Sti  Ns =Sti

ð153Þ

where Nc/Sti is the nitrification capacity per mgCOD applied to the AS system (mgN/mgCOD), Nti/Sti the influent TKN/ COD concentration ratio of the wastewater, and Ns/Sti the

•

Nti/Sti: This ratio is a wastewater characteristic and obtained from the measured influent TKN and COD concentrations – it can range from 0.07 to 0.10 for raw municipal wastewater and 0.10 to 0.14 for settled wastewater. Nte/Sti: Provided the constraint for efficient nitrification is satisfied at the lowest temperature (Tmin), the effluent TKN at Tmin (Nte) will be low at B2–3 mgN l1, that is, for influent COD concentrations (Sti) ranging from 1000 to 500, Nte/Sti will range from 0.005 to 0.010. At Tmax, NteE1– 2 mgN l1 making the Nte/Sti ratio lower. Ns/Sti: Given by Equation (144).

A graphical representation of the relative importance of these three ratios to the nitrification capacity, Nc/Sti, is shown in Figure 30(a) (for 14 1C) and 30(b) (for 22 1C) and were generated by plotting Nc/Sti versus sludge age for selected influent TKN/COD (Nti/Sti) ratios of 0.07, 0.08, and 0.09 for the example raw wastewater and settled wastewater for 40% COD and 15% TKN removal in primary settling, viz., 0.113, 0.127, and 0.141. Also shown are the minimum sludge ages for nitrification at unaerated sludge mass fractions of 0.0, 0.2, 0.4, and 0.6 for the example mAm20 value of 0.45 d1. For a particular unaerated sludge mass fraction, the plotted values of Nc/Sti are valid only at sludge ages longer than the corresponding minimum sludge age. These figures show the relative magnitudes of the three terms that affect the nitrification capacity versus sludge age and temperature. 1. Temperature. To obtain complete nitrification at 14 1C (for a selected fxm), the sludge age required is more than double that at 22 1C. The corresponding nitrification capacities per influent COD at 14 1C show a marginal reduction to those at 22 1C, because sludge production at 14 1C is slightly higher than at 22 1C due to the reduction in endogenous respiration rate of the OHOs. 2. Sludge age. For a selected influent TKN/COD ratio (Nti/Sti), the nitrification capacity (Nc/Sti) increases as the sludge age increases because the N required for sludge production decreases with sludge age, making more FSA available for nitrification. However, the increase is marginal for Rs410 days. 3. Influent TKN/COD ratio (Nti/Sti). Clearly, for both raw and settled wastewater, at any selected sludge age, the nitrification capacity (Nc/Sti) is very sensitive to the influent TKN/COD ratio (Nti/Sti). An increase of 0.01 in Nti/Sti causes equal increase of 0.01 in Nc/Sti. For the same Nti/Sti ratio for raw or settled wastewater, the nitrification capacity (Nc/Sti) for raw wastewater is lower than for settled wastewater because more sludge (VSS) is produced per unit COD load from raw wastewater than from settled wastewater because the unbiodegradable particulate COD fraction (fS’up) in raw water is higher than in settled wastewater. Apart from this difference, an increase in influent TKN/COD ratio will result in an equal increase in nitrate concentration (nitrification capacity) per influent

Biological Nutrient Removal

WW Char fS’us Raw 0.07 Settled 0.12

0.10

0.05 Raw wastewater

0.15

0.0 0.2 0.4 0.6 −1 Unaerated mass fraction 14 °C; UA20 = 45 d bA20 = 0.04 d−1; Sf = 1.25

0.00 0 (a)

fS’up Settled wastewater 0.15 0.141 0.04 0.127 0.113 TKN/COD ratio 0.10 0.09 0.08

Nitrification capacity

Nitrification capacity

0.15

5

10 15 20 Sludge age (days)

25

fs’us 0.07 0.12

fs’up 0.15 0.04

Settled wastewater 0.141 0.127

0.10

0.113 TKN/COD ratio 0.10 0.09 0.08

0.05 Raw wastewater

0.00

30

WW Char Raw Settled

475

0.2 0.6 22 °C; UA20 = 0.45 d−1 0.0 0.4 −1 b Unaerated mass fraction A20 = 0.04 d ; Sf = 1.25 0

5

(b)

10 15 20 Sludge age (days)

25

30

Figure 30 Nitrification capacity per mgCOD applied to the biological reactor vs. sludge age for different influent TKN/COD concentration ratios in the example raw and settled wastewaters at 14 1C (a) and 22 1C (b). Also shown as vertical lines are the minimum sludge ages required to achieve nitrification for Sf ¼ 1.25 for unaerated sludge mass fractions of 0.0, 0.2, 0.4, and 0.6.

COD. This decreases the likelihood, or makes it impossible, to obtain complete denitrification using the wastewater organics as electron donor. This will become clear when denitrification is considered below. Because primary settling increases the influent TKN/COD ratio, N removal via nitrification denitrification is always lower with settled wastewater than with raw wastewater. However, this lower N removal comes with the advantage of a smaller biological reactor and lower oxygen demand resulting significant savings in reactor and oxygenation costs.

4.14.22.3 Mass of Nitrifiers (MXA) and Nitrification Oxygen Demand (FOn) Once nitrification takes place because the sludge age of the system is longer than the minimum required for nitrification, the mass of nitrifiers (MXA, mgVSS) in the reactor is calculated from the flux of nitrate generated (FNne) in the same way as the mass of OHOs (MXBH) was calculated from the flux of biodegradable organics (Equation (86)), viz.,

MXBA ¼ FNne YA Rs =ð1 þ bAT Rs Þ ðmgVSSÞ

ð154Þ

where FNne is the flux of nitrate generated ¼ (Qe þ Qw)Nne ¼ Qi Nne (mgN d1) and Nne is given by Equation (152). The oxygen demand for nitrification is simply 4.57 mgO/ mgN times the flux of nitrate produced, that is,

FOn ¼ 4:57 FNne ¼ OURn Vp

ðmgO d1 Þ

ð155Þ

Table 11 Raw and settled wastewater characteristics required for calculating effluent N concentrations from nitrification AS systems Influent WW characteristic

Sym

Raw

Influent TKN (mgN l1) Influent TKN/COD ratio Influent FSA fraction Unbio sol orgN fraction Unbio partic VSS N content

Nti fns fN0 a fN0 ous fn

60 0.08 0.75 0.03 0.1

Influent pH Influent Alk mg l1 as CaCO3 ANO max spec growth rate Influent flow rate (M l d1)

Alk mAm20 Qi

7.5 200 0.45 15

Seta 51 0.113 0.88 0.034 0.1 7.5 200 0.45 15

a

Settled wastewater (WW) characteristics must be selected/calculated to be consistent with the raw wastewater ones and mass balances over the primary settling tanks, e.g., soluble concentrations must be the same in settled wastewater as in raw wastewater.

organics (COD) removal (see Section 4.14.9.5). The wastewater characteristics for the raw and settled wastewaters for COD removal are listed in Table 7 and the additional characteristics required for nitrification are listed in Table 11. The nitrifier kinetic constants in Table 10, adjusted for wastewater temperatures 14 and 22 1C, were applied. No adjustment to mAm20 for pH was made, that is, an effluent Alkalinity 450 mg l1 as CaCO3 was assumed. Also, it is accepted that all the biodegradable organics are degraded and their N content released as ammonia so the effluent soluble biodegradable organic N concentration (Nobse) is zero.

4.14.23.2 Nitrification Process Behavior

4.14.23 Nitrification Design Example 4.14.23.1 Wastewater Characteristics Design of a nitrification AS system without denitrification is considered below. For the purpose of comparison, the nitrifying AS system is designed for the same wastewater flow and characteristics accepted for the design of the AS system for

From Equation (20a), the unbiodegradable soluble organic nitrogen in the effluent is Nouse ¼ Nousi ¼ 1.8 mgN l1 for raw and settled wastewater The ammonia concentration available for nitrification (Nan) is the influent TKN concentration (Nti) minus the N concentration required for sludge production (Ns) (Equation (142)) and the USO N concentration in the effluent

476

Biological Nutrient Removal

(Nouse), viz.,

Nan ¼ Nti  Ns  Nouse

ðmgN l1 Þ

ð156Þ

If the sludge age of the system is shorter than the minimum required for nitrification (RsoRsm), no nitrification takes place and the effluent nitrate concentration (Nne) is zero. The effluent ammonia concentration (Nae) is equal to the nitrogen available for nitrification (Nan, Equation (156)). If Rs4Rsm for Sf ¼ 1.0, most of the FSA available for nitrification is nitrified to nitrate and the effluent nitrate concentration (Nne) is the difference between Nan (Equation (156)) and the effluent FSA concentration given by Equation (132). For both RsoRsm and Rs4Rsm, the effluent TKN concentration (Nte) is the sum of effluent ammonia and unbiodegradable soluble organic nitrogen concentrations (Nte ¼ Nae þ Nouse). For RsoRsm, no nitrification takes place so the effluent nitrate concentration (Nne) is zero and the effluent ammonia concentration (Nae) is given by Nan (Equation (156)). The nitrifier sludge mass (MXA) and the nitrification oxygen demand (FOn) are both zero because Nne is zero. With increasing sludge age starting from Rs ¼ 0, Nae from Equation (132) is first negative (which is of course impossible) and then 4Nan (which is also not possible). For a sludge age slightly longer than Rsm, the Nae falls below Nan. From this sludge age, nitrification takes place and for further (even small) increases in sludge age, the Nae rapidly decreases to low values (o4 mgN l1). Hence for Rs4Rsm, the effluent ammonia concentration (Nae) is given by Equation (132), the effluent TKN concentration by Nte ¼ Nae þ Nouse, and the effluent nitrate concentration (Nne) by

Nne ¼ Nan  Nae ¼ Nti  Ns  Nte

ðmgN l1 Þ

ð157Þ

With nitrification, the nitrifier biomass and nitrification oxygen demand are given by Equations (154) and (155). Substituting the influent N concentrations for raw and settled wastewaters and the values of the kinetic constants at 14 1C into the above equations, the results at different sludge ages were calculated. In Figure 31(a), the different effluent N concentrations from the system versus sludge age for raw and settled wastewater at 14 1C are shown. In Figure 31(c) are shown the nitrifier sludge mass (as a % of the reactor VSS mass) and nitrification oxygen demand for raw and settled wastewater at 14 1C. Also shown in Figure 31(c) are the carbonaceous and total oxygen demands for raw and settled wastewater at 14 1C. The calculations were repeated for 22 1C and shown in Figures 31(b) and 31(d). Figures 31(a) and 31(b) show that once the sludge age is approximately 25% longer than the minimum required for nitrification, nitrification is virtually complete (for steady-state conditions) and comparing the results for raw and settled wastewater, there is little difference between the nitrification oxygen demand and the concentrations of ammonia, nitrate, and TKN in the effluent. The reasons for this similar behavior are: (1) the primary settling tank removes only a small fraction of the influent TKN and (2) settled wastewater results in lower sludge production, so that the FSA available for nitrification in raw and settled wastewater is nearly the same. Once

nitrification takes place, temperature has relatively little effect on the different effluent N concentrations. However, a change in temperature causes a significant change in the minimum sludge age for nitrification. Considering Figures 31(a) and 31(b), for RsoRsm, the effluent ammonia concentration (Nae) and hence the effluent TKN concentration (Nte) increase with increasing sludge age up to Rsm because Ns decreases for increases in Rs. For Rs4Rsm, Nae decreases rapidly to o2 mgN l1 so that for Rs41.3Rsm, the effluent TKN concentration is o4 mgN l1. The increase in nitrate concentration (Nne) with an increase in sludge age for Rs41.3Rsm is mainly due to the reduction in N required for sludge production (Ns). This is important for BNR systems – increasing the sludge age increases the nitrification capacity (see Figure 30) so more nitrate has to be denitrified to achieve the same N removal. Figures 31(c) and 31(d) show that the nitrification oxygen demand increases rapidly once Rs4Rsm but for Rs41.3Rsm, further increases are marginal irrespective of the temperature or wastewater type. This nitrification oxygen demand represents an increase of 42% and 65% above the COD for the raw and settled wastewater. However, the total oxygen demand for treating settled wastewater is only 75% of that for treating raw wastewater. In order that nitrification can proceed without inhibition by oxygen limitation, it is important that the aeration equipment is adequately designed to supply the total oxygen demand; generally, heterotrophic organism growth takes precedence over nitrifier growth when oxygen supply (or ammonia) becomes insufficient. This is because heterotrophic organisms can grow adequately with DO concentrations of 0.5–1.0 mgO l1, whereas nitrifiers tend to require higher DO concentrations. Just as the effluent FSA concentration rapidly decreases for Rs4Rsm, the nitrifier sludge mass rapidly increases once Rs4Rsm, is slightly higher at 14 1C than at 22 1C due to the lower endogenous respiration rate. Also, because the concentrations of nitrate produced with raw and settled wastewater are closely similar (B40 mgN l1), the nitrifier sludge mass is approximately the same at the same sludge age (B430 kgVSS at 10d sludge age and B900 kgVSS at 30 day sludge age). Because with raw wastewater so much more sludge mass is produced than with settled wastewater, the nitrifier sludge mass is a much smaller proportion of the VSS mass with raw waste water (B1.4% at 10 day sludge age) than with settled wastewater (B3.3% at 10 day sludge age). Comparing the nitrifier sludge mass to the heterotrophic sludge mass, as in Figures 31(c) and 31(d), the nitrifier sludge mass comprises o4% of VSS mass even at high TKN/COD ratios for settled wastewater and so is ignored in the determination of the VSS concentration in the AS reactor treating domestic wastewater. It is worth repeating that primary sedimentation removes only a minor fraction of the TKN but a significant fraction of COD (15% and 40% in this example). Even though the settled wastewater has a lower TKN concentration than the raw wastewater, the effluent nitrate concentration does not reflect this difference. This is because the N removal for sludge production is lower for settled than for raw wastewater. Consequently, the nitrate concentration for settled wastewater is nearly the same as for raw wastewater – for different

(a)

Nitrogen-FSA, TKN, NO3 0.0

0.2

0.4

0.6

0.8

1.0

1.2

0

14 °C

5

0

FOc

5

Rsm

Ns

%Nit

25

25

30

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

30

Raw WW Settled WW

Raw WW Settled WW

FOn FOn

FOc

FOc

FOt

FOt

10 15 20 Sludge age (days)

%Nit

10 15 20 Sludge age (days)

Nte

10

Nouse = 1.8 mg N l−1 Nte = Nae + Nouse

Nne

Influent TKN

Ns

0

Nte

Rsm

20

30

40

50

60

14 °C

% Nitrifier VSS

(b)

(d)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 0

0

5

22 °C Rsm

5

FOn FOn

%Nit

FOt

FOt

25

25

30

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

30

Raw WW Settled WW

22 °C

Raw WW Settled WW

FOc

FOc

10 15 20 Sludge age (days)

%Nit

10 15 20 Sludge age (days)

Ns

10

Nouse = 1.8 mg N l−1 Nte = Nae + Nouse

Nne

Influent TKN

Ns Nte

Rsm

20

30

40

50

60

70

Figure 31 Effluent ammonia (Nae), TKN (Nte), and nitrate (Nne) concentrations and N required for sludge production (Ns) vs. sludge age at 14 1C (a) and 22 1C (b) and nitrification (FOn), carbonaceous (FOc) and total (FOt) oxygen demand in kgO/kgCOD load and % nitrifier VSS mass vs. sludge age at 14 1C (c) and 22 1C (d) for the example raw and settled wastewater.

(c)

Oxygen demand (kgO/kgCOD)

Nitrogen-FSA, TKN, NO3 Oxygen demand (kgO/kgCOD)

70

% Nitrifier VSS

478

Biological Nutrient Removal

wastewater characteristics, it can be higher than raw wastewater. In contrast, the maximum N removal by denitrification using the wastewater organics as electron donor, called the denitrification potential, mainly depends on the influent COD concentration and this concentration is significantly reduced by primary sedimentation. This may result in a situation where it may be possible to obtain near-complete nitrate removal when treating raw wastewater but not when treating settled wastewater. The difference in COD and TKN removal in PSTs therefore has a significant effect on the design of BNR systems.

4.14.24 Biological Denitrification

ages can be in the usual fully aerobic short sludge age range of 3–6 days. Unaerated zones should still be incorporated to derive the benefits of denitrification in the event nitrification does take place. When it does not, the unaerated zone will be anaerobic (no input of DO or nitrate) instead of anoxic, and some BEPR may take place. Because BEPR is not required and therefore not exploited to the full, whether or not it takes place is not important because it does not affect the system behavior very much. With some BEPR, the sludge production will be slightly higher (o5%) per COD load, the VSS/TSS ratio and oxygen demand both somewhat lower (by about 5%). However, BEPR may result in mineral precipitation problems in the sludge treatment facilities if the WAS is anaerobically digested.

4.14.24.1 Interaction between Nitrification and N Removal Nitrification is a prerequisite for denitrification – without it biological N removal is not possible. Once nitrification takes place, N removal by denitrification becomes possible and should be included even when N removal is not required (see Section 4.14.14) by incorporating zones in the reactor that are intentionally unaerated. Because the nitrifiers are obligate aerobes, nitrification does not take place in the unaerated zone(s), so to compensate for this, the system sludge age needs to be increased for situations where nitrification is required. For fully aerobic systems and a wastewater temperature 14 1C, a sludge age of 5–7 days may be sufficient for complete nitrification, taking due consideration of the requirement that the effluent FSA concentration should be low even under cyclic flow and load conditions (Sf41.3). For anoxic – aerobic systems, a sludge age of 15–20 days may be required when a 50% unaerated mass fraction is added (Figure 25). Therefore, for plants where N removal is required, invariably the sludge ages are long due to (1) the uncertainty in the mAm20 value, (2) the need for unaerated zones, and (3) the guarantee of nitrification at the minimum average winter temperature (Tmin). For plants where nitrification is a possibility and not obligatory, uncertainty in the mAm20 value is not important and unaerated zones can be smaller, with the result that sludge Table 12

4.14.24.2 Benefits of Denitrification In the design of fully aerobic systems discussed above, it was suggested that when nitrification is not obligatory but a possibility, unaerated zones should still be incorporated in the system to derive the benefits of denitrification. These benefits include (1) reduction in nitrate concentration which ameliorates the problem of rising sludge from denitrification in the secondary settling tank (Section 4.14.14), (2) recovery of alkalinity (Section 4.14.20.6), and (3) reduction in oxygen demand. With regard to (3), under anoxic conditions, nitrate serves as electron acceptor instead of DO in the degradation of organics (COD) by facultative heterotrophic organisms. The oxygen equivalent of nitrate is 2.86 mgO/mgNO3-N which means that 1 mg NO3-N denitrified to N2 gas has the same electron-accepting capacity as 2.86 mg of oxygen. In nitrification to nitrate, the FSA donates eight electrons (e)/mol, the N changing from an e state of  3 to þ 5. In denitrification to N2, the nitrate accepts 5 e/mol, the N changing from an e state of þ 5 to 0. Because 4.57 mgO/mgFSA-N are required for nitrification, the oxygen equivalent of nitrate in denitrification to N2 is 5/8  4.57 ¼ 2.86 mgO/mgNO3-N (Table 12). Therefore, for every 1 mg NO3-N denitrified to N2 gas in the anoxic zone, during which about 2.86/(1  fcvYH) ¼ 8.6 mgCOD

Comparison of nitrification and denitrification processes in single sludge activated sludge systems Nitrification

Denitrification

Form Function Half-reaction Organisms Environment

Ammonia (NHþ 4) Electron donor Oxidation Autotrophs Aerobic

Nitrate (NO 3) Electron acceptor Reduction Heterotrophs Anoxic

Compound Oxid. no.

NH4 þ 3 Nitrification (oxidation)

N2 0

NO2  þ3

8 e/atom N ¼ 4.57 mgO/mgN Net loss Denitrification (reduction) 5e/atom N ¼ 2.86 mgO/mgN Nitrification: 4.57 mgO/mgNH4-N nitrified to NO3-N Denitrification: 2.86 mgO recovered/mg NO3-N denitrified to N2 gas Therefore, denitrification allows at best 62.5% (5/8 or 2.86/4.57) recovery of the nitrification oxygen demand.

NO3  þ5

Biological Nutrient Removal Effluent TKN liquid phase

Raw wastewater TKN/COD = 0.08 1.0

Oxygen demand (kgO/kgCOD on reactor)

14 °C 22 °C

479

Total incl. nitrification

~5%

N in gas phase ~75%

0.8

N in sludge solid phase ~20%

0.6 Total incl. nitrif. and denit. 0.4

N nitrified (transformed in liquid phase) and possibly denitrified (transferred to gas phase)

Carbonaceous

Possibility of nitrification 0.2

Possibility of denitrification

Figure 33 Exit routes for nitrogen in single sludge nitrification denitrification activated sludge systems.

0.0 0

5

10

15 20 Sludge age (days)

25

30

Figure 32 Carbonaceous, total including nitrification and total including nitrification and denitrification oxygen demand per unit COD load on the biological reactor vs. sludge age for the example* raw wastewater. *Note: All the figures in this part which show the behavior of the various activated sludge system configurations were generated from the example raw and settled wastewater characteristics.

is utilized, 2.86 mg less oxygen needs to be supplied to the aerobic zone. Because the oxygen requirement to form the nitrate from ammonia is 4.57 mgO/mgNO3-N, and 2.86 mgO/mgNO3-N is recovered in denitrification to N2 gas, a maximum of 2.86/4.57 (or 5/8ths) ¼ 0.63 of the nitrification oxygen demand can be recovered. A comparison of the nitrification and denitrification reactions is given in Table 12. Under operating conditions, it is not always possible to denitrify all the nitrate formed with the result that the nitrification oxygen recovery by denitrification is about 50% (see Figure 32). Therefore, once the possibility of nitrification exists, it is always worthwhile to consider including intentional denitrification because of the recovery of alkalinity and oxygen. With regard to oxygen, if the oxygen supply is insufficient to meet the combined carbonaceous and nitrification requirement, areas in the aerobic reactor will become anoxic. Under oxygen limited conditions, the aerobic mass fraction in the aerobic reactor will vary depending on the COD and TKN load on the plant over the day. At minimum load, oxygen supply may be adequate so that nitrification may be complete whereas, at peak load, oxygen supply may be insufficient so that nitrification may cease (partially or completely) and denitrification will take place on the accumulated nitrate. This behavior is exploited in the single-reactor ND configurations such as the ditch or Carousel-type systems.

4.14.24.3 N Removal by Denitrification In biological N removal systems, the N is removed by transfer from the liquid phase to the solid and gas phases. About 20% of the influent N is incorporated in the sludge mass (Figure 33) but the bulk of the N (i.e., about 75% when complete denitrification is possible) is removed by transfer to the gas phase via nitrification and denitrification (Figure 33). In the nitrification step, the N remains in the liquid phase because it is transformed from ammonia to nitrate. In the denitrification step, it is transferred from the liquid to the gas phase and escapes to the atmosphere. When complete denitrification is achieved, a relatively small fraction of the influent TKN (B5%) remains in the liquid phase and escapes as total N (TKN þ nitrate) with the effluent. For aerobic conditions, the problem of the designers is to calculate the mass of oxygen electron acceptor required by the OHOs (and ANOs) for the utilization of the known mass of organic electron donors (organics and ammonia) available. For anoxic conditions, the problem is the opposite. Here, the problem is to calculate the mass of electron donors (COD) that are required to denitrify a known mass of electron acceptors nitrate. If sufficient electron donors (COD) are not available then complete denitrification cannot be achieved. The calculation of the nitrogen removal is essentially a reconciliation of electron acceptors (nitrate) and donors (COD) taking due account of (1) the biological kinetics of denitrification and (2) the system operating parameters (such as recycle ratios and anoxic reactor sizes) under which the denitrification is constrained to take place. The electron donors (or COD or energy) for denitrification can come from two sources: (1) internal or (2) external to the AS system. The former are those present in the system itself, that is, those in the incoming wastewater or generated within the biological reactor by the AS itself; the latter are organics imported to the AS system and specifically dosed into the anoxic zone(s) to promote denitrification, (e.g., methanol,

480

Biological Nutrient Removal

acetate, and molasses; Monteith et al., 1980). Here, the focus is on internal COD sources for denitrification, but the principles and procedures are sufficiently general to be adaptable to include external COD (energy) sources also.

4.14.24.4 Denitrification Kinetics There are three internal organics sources, two from the wastewater and one from the AS sludge mass itself. The two in the wastewater are the two main forms of organics (i.e., RBSO) and slowly biodegradable organics (BPO)). The third is slowly biodegradable organics generated by the biomass itself through death and lysis of organism mass (also known as endogenous mass loss/ respiration). This self-generated BPO is utilized in the same way as the wastewater BPO, but is recognized separately because of its different source and rate of supply to that of the influent. The RBSO and BPO (influent or self-generated) are degraded via different mechanisms by the OHOs. The different RBSO and BPO degradation mechanisms lead to different COD utilization rates. The RBSO comprises small simple dissolved organic compounds that can pass directly through the cell wall into the organism, for example, sugars and short-chain fatty acids. Accordingly, the RBSO can be used at a high rate which does not change significantly whether nitrate or oxygen serves as terminal electron acceptor (Ekama et al., 1996a). Simulation models use the Monod equation to model the utilization of RBSO by OHOs under both aerobic and anoxic conditions. The BPO comprises large particulate or colloidal organic compounds, too large to pass into the organism directly. These organics must be broken down (hydrolyzed) in the slime layer surrounding the organism to smaller components, which then can be transferred into the organism and utilized. The extracellular BPO hydrolysis rate is slow and forms the limiting rate in the utilization of BPO (Section 4.14.5.1.3). This hydrolysis rate is much slower under anoxic conditions than under aerobic conditions – only about one-third (Stern and Marais, 1974, van Haandel et al., 1981). This introduces a reduction factor Z in the BPO hydrolysis rate equation for anoxic conditions (Equation (159) below). Research has indicated that the utilization of RBSO is simultaneous with the hydrolysis of BPO. Also the rate of RBSO utilization is considerably faster (7–10 times) than the rate of BPO hydrolysis so the denitrification rate with influent RBSO is much faster than with BPO. Therefore, the influent RBSO is the preferred organic for denitrification and the higher this concentration in the influent with respect to the total COD, the greater the N removal.

4.14.24.5 Denitrification Systems As a result of the different degradation mechanisms and rates of RBSO and BPO utilization, the position of the anoxic zone in the biological reactor significantly affects the denitrification that can be achieved. There are many different configurations of single sludge ND systems but from the point of view of the source of the organics (electron donors), these can be simplified to two basic types of denitrification or combinations of these. The two basic types utilizing internal organics are: (1) post-denitrification, which utilizes self-generated

endogenous organics and (2) pre-denitrification, which utilizes influent wastewater organics. With post-denitrification (Figure 34(a)), the first reactor is aerobic and the second is unaerated. The influent is discharged to the aerobic reactor where aerobic growth of both the heterotrophic and nitrifying organisms takes place. Provided the sludge age is sufficiently long and the aerobic fraction of the system is adequately large, nitrification will be complete in the first reactor. The mixed liquor from the aerobic reactor passes to the anoxic reactor, also called the secondary anoxic reactor, where it is mixed with stirring. The outflow from the anoxic reactor passes through an SST and the underflow is recycled back to the aerobic reactor. The BPO released by the sludge mass via the death of organisms provides the energy source for denitrification in the anoxic reactor. However, the rate of release of energy is low, so that the rate of denitrification is also low. To obtain a meaningful reduction of nitrate, the anoxic mass fraction of the

Anoxic reactor

Aerobic reactor

Waste flow Settler

Influent

Effluent

Sludge recycle

(a)

s

Anoxic Aerobic reactor reactor Mixed liquor Recycle

Waste flow

a

Settler

Influent

Effluent

I

Sludge recycle

(b)

Primary Aerobic anoxic reactor reactor Mixed liquor Recycle a

s

Secondary anoxic reactor Reaeration reactor Waste flow Settler Effluent

Influent

(c)

Sludge recycle

s

Figure 34 (a) The post-denitrification single sludge biological nitrogen removal system. (b) The modified Ludzack–Ettinger single sludge biological nitrogen removal system proposed by Barnard (1973), including the primary anoxic reactor only. (c) The four-stage Bardenpho single sludge biological nitrogen removal system, including primary and secondary anoxic reactors.

Biological Nutrient Removal

system (i.e., the fraction of the mass of sludge in the system that is in the anoxic reactor) must be large and this may cause, depending on the sludge age, cessation of nitrification. Thus, although theoretically the system has the potential to remove all the nitrate, from a practical point this is not possible because the anoxic mass fraction will need to be so large that the conditions for nitrification cannot be satisfied particularly if the temperatures are low (o15 1C). Furthermore, in the anoxic reactor, ammonia is released through organism death and lysis, some of which passes out with the effluent thereby reducing the total nitrogen removal of the system. To minimize the ammonia content of the effluent, a flash or re-aeration reactor sometimes is placed between the anoxic reactor and the SST. In this reactor, N2 gas is stripped from the mixed liquor to avoid possible sludge buoyancy problems in the SST and the ammonia is nitrified to nitrate to assist with compliance of ammonia standards but it reduces the overall efficiency of the nitrate reduction of the system. For these reasons, post-denitrification has not been widely applied in practice.

4.14.24.5.1 The Ludzack–Ettinger system Ludzack and Ettinger (1962) were the first to propose a single sludge ND system utilizing the biodegradable organics in the influent as organics for denitrification. It consisted of two reactors in series, partially separated from each other. The influent was discharged to the first, or primary anoxic reactor which was maintained in an anoxic state by mixing without aeration. The second reactor was aerated and nitrification took place in it. The outflow from the aerobic reactor passed to the SST and the SST underflow was returned to the aerobic (second) reactor. Due to the mixing action in both reactors, an interchange of the nitrified and anoxic liquors was induced. The nitrate which entered the primary anoxic reactor was denitrified to nitrogen gas. Ludzack and Ettinger reported that their system gave variable denitrification results, probably due to the lack of control of the interchange of the contents between the two reactors. In 1973, Barnard proposed an improvement to the Ludzack– Ettinger system by completely separating the anoxic and aerobic reactors, recycling the underflow from the SST to the primary (first) anoxic reactor and providing a mixed liquor recycle from the aerobic to the primary anoxic reactor (Figure 34(b)). These modifications allowed a significant improvement in control over the system N removal performance of the system with the mixed liquor recycle flow. The RBSO and BPO from the influent stimulated high rates of denitrification in the primary anoxic reactor and much higher reductions of nitrate could be achieved than with post-denitrification, even when the pre-denitrification reactor of this system was substantially smaller than the post-denitrification reactor. In this system, called the Modified Ludzack–Ettinger (MLE) system, complete nitrate removal cannot be achieved because a part of the total flow from the aerobic reactor is not recycled to the anoxic reactor but exits the system with the effluent. To reduce the possibility of flotation of sludge in the SST due to denitrification of residual nitrate, the sludge accumulation in the SST needed to be kept to a minimum. This was achieved by having a high underflow recycle ratio from the SST, equal to the mean influent flow (1:1).

481

4.14.24.5.2 The four-stage Bardenpho system In order to overcome the deficiency of incomplete nitrate removal in the MLE system, Barnard (1973) proposed including a secondary anoxic reactor in the system and called it the fourstage Bardenpho system (Figure 34(c)). Barnard considered that the low concentration of nitrate discharged from the aerobic reactor to the secondary anoxic reactor will be denitrified to produce a relatively nitrate-free effluent. He included a flash or re-aeration reactor to strip the nitrogen gas and to nitrify the ammonia released during the denitrification. Although in concept the Bardenpho system has the potential for complete removal of nitrate, in practice this is not possible except when the influent TKN/COD concentration ratio is quite low, o0.09 mgN/mgCOD for normal municipal wastewater at 14 1C. The low denitrification rate and ammonia release (about 20% of the nitrate denitrified) results is an inefficient use of the secondary anoxic sludge mass fraction. As a result of the competition between the aerated and unaerated sludge mass fractions from the requirement to nitrify, (Section 4.14.20.3) usually it is better to exclude the secondary anoxic (and re-aeration) reactor and enlarge the primary anoxic reactor and increase the mixed liquor recycle ratio.

4.14.25 Denitrification Kinetics 4.14.25.1 Denitrification Rates The denitrification behavior in the primary and secondary anoxic zones is best explained by considering these reactors as plug-flow reactors. However, the explanation is equally valid for completely mixed reactors because the denitrification kinetics are essentially zero order with respect to nitrate concentration (van Haandel et al., 1981; Ekama and Wentzel, 1999b). Owing to the two different kinds of biodegradable organics (RBSO and BPO) in the influent wastewater, the denitrification in the primary anoxic reactor follows two phases (Figure 35(a)) – an initial rapid phase where the rate is defined by the simultaneous utilization of RBSO and BPO (K1 þ K2) and a second slower phase where the specific denitrification rate (K2) is defined by the utilization of only BPO originating from the influent and self-generated by the sludge through organism death and lysis. In the secondary anoxic reactor, only a single slow phase of denitrification takes place (Figure 35(b)), the specific rate (K3) being about two-thirds of the slow rate (K2) in the primary anoxic reactor (Stern and Marais, 1974; van Haandel et al., 1981). In the preceding aerobic reactor all the RBSO and most of the BPO of the influent has been utilized with the result that in the secondary anoxic reactor the only biodegradable COD available is BPO from organism death and lysis; the slow rate of supply of this BPO governs the K3 rate and causes this rate to be slower than the K2 rate. The values of the K rates are given in Table 13. A further specific K rate (K4) has been defined for denitrification in intermittently aerated anoxic aerobic digestion of WAS (Warner et al., 1986). This rate is only two-thirds of the K3 rate in the secondary anoxic reactor (Table 13), but sufficiently high to denitrify all the nitrate generated in aerobic digestion of WAS if the 6 h aeration cycle is 50% anoxic and 50% aerobic. Denitrification in anoxic–aerobic digestion adds

482

Biological Nutrient Removal

NO3−N concentration

NO3−N concentration

K1XBH

K2XBH

1st

(a)

K3XBH

Single phase

Second phase

(b)

Time

Time

Figure 35 Nitrate concentration of vs. time profiles in primary anoxic (a) and secondary anoxic (b) plugflow reactors, showing the three phases of denitrification associated with the K1, K2, and K3 rates. In the primary anoxic the initial rapid rate K1 is attributable to the utilization of the influent RBSO and the second slower rate K2 to the utilization of BPO from the influent wastewater and self-generated by organism death and lysis. In the secondary anoxic reactor, the rate K3 is attributable to the utilization of the self-generated BPO only. Table 13

K denitrification rates and their temperature sensitivity

Symbol

20 1 C

y

14 1 C

22 1 C

Equation

K120a K220a K320a K420a

0.72 0.1 0.1 0

1.2 1.08 1.029 1.029

0.241 0.06 0.06 0.04

1.036 0.118 0.08 0.05

158 159 160 161

a

Units – mgNO3-N/(mgOHOVSS d).

the benefits of denitrification to this system, that is, zero alkalinity consumption, oxygen recovery, improved pH control, reduced chemical dosing (Dold et al., 1985), and additionally a nitrogen free dewatering liquor. This last advantage is significant considering the high N content of WAS compared with primary sludge. The constancy of K1, K2, K3 (and K4) specific denitrification rates under constant flow and load conditions can be explained in terms of the kinetics of RBSO and BPO organics utilization included in the AS simulation models such as ASM1 developed later. The utilization of RBSO organics is modeled with the Monod equation and expressing the K1 rate in terms of this yields

K1 ¼

ð1  f cv YH Þf cv mHm Ss 2:86YH Ks þ Ss

where

Ss E 1 ðmgNO3 -N=ðmgOHOVSS dÞÞ Ks þ Ss

ð158Þ

In the plugflow and completely mixed primary anoxic reactor, the Monod term SS/(KS þ SS) remains close to 1 down to low RBSO concentrations because the half-saturation concentration (KS) is low. Accepting YH ¼ 0.45 mgVSS and

fcv ¼ 1.48 mgCOD/mgVSS yields K1 ¼0.26 mH mgNO3-N/ (mgOHOVSS d). So for the measured K1 ¼0.72 mgNO3-N/ (mgOHOVSS d) (Table 13), the mHm must have been about 2.8 d1. This mHm rate is in the range of mHm rates measured in AS systems. In investigating the kinetics of RBSO utilization in aerobic and anoxic selectors, Still et al. (1996) and Ekama et al. (1996a, b) found mHm values ranged between 1.0 d1 in completely mixed reactor systems and 4.5 d1 selector reactor systems, which yields K1 denitrification rates around 0.26 mgNO3-N/(mgOHOVSS d) for completely mixed type systems and 1.17 mgNO3-N/(mgOHOVSS d) for systems in which a selector effect (high mH) has been stimulated in the OHO biomass. The utilization of BPO is expressed in terms of the activesite surface hydrolysis kinetic formulation, which has the form of a Monod equation, except the variable is the adsorbed BPO to active OHO ratio (Xs/XBH), not the bulk liquid BPO concentration. Hence, the K2, K3 (and K4) rates are given by

K2 ¼ K3 ¼ K4 ¼

ð1  f cv YH Þ ZKh ðXs =XBH Þ 2:86f cv YH ½Kx þ ðXs =XBH Þ mgNO3 -N=ðmgOHOVSS dÞ

ð159Þ

where XS/XBH is progressively lower in primary (K2) secondary (K3) and anoxic–aerobic digestion (K4).

Biological Nutrient Removal

In the constant flow and load primary and secondary anoxic plugflow reactors, the (Xs/XBH) ratio changes very little due to the reduced anoxic hydrolysis rate including the Z. The reason for the K2 being higher than K3 arises from different concentrations of adsorbed BPO relative to the active OHO concentration (Xs/XBH) (Figure 36). In the primary anoxic reactor, the ratio is high because adsorbed BPO originates from the influent and OHO death. In the secondary anoxic, the ratio is lower because BPO originates only from OHO death. For the K2 and K3 denitrification rates, there is no simple relationship between the K rates and the ZKh because the adsorbed BPO to OHO ratio (Xs/XBH) is different in the primary and secondary anoxic reactors (and aerobic digester) and varies somewhat with sludge age and unaerated sludge mass fraction. It was concluded that the K1, K2, K3, and K4 denitrification constants have no direct kinetic significance; their constancy is the result of a combination of kinetic reactions which show little variation with sludge age in the range 10–30 days. Temperature does affect the K rates but once these have been adjusted for temperature, again the K rates show little variation at different sludge ages (van Haandel et al., 1981). It can be concluded both from experimental observation and theoretical kinetic points of view that acceptance of constant K2 and K3 rates is acceptable for steady-state design. This is in fact done to estimate the denitrification potential (Dp) of an anoxic reactor under constant flow and load conditions. With regard to K1, this rate can change significantly because the RBSO utilization rate can change appreciably depending on the mixing regime in the anoxic (or aerobic) reactor (Ekama et al., 1986, 1996a, b and Still et al., 1996). However, its variation does not affect ND design significantly because normally primary anoxic reactors are sufficiently large to allow complete utilization of RBSO even when the utilization rate (mHm) is low. In fact, the denitrification design procedure requires that all the RBSO is utilized in the primary anoxic reactor which introduces a minimum primary anoxic sludge mass fraction (fx1 min) and a minimum a-recycle ratio (amin) to

0.12

Specific denit rates (K )

K2 0.10 0.08 0.06

K3 K4

0.04 0.02 0.00 0.0

0.1

0.2

0.3

0.4

XS /XBH ratio (mgCOD/mgCOD) Figure 36 Specific denitrification rate (K) vs. adsorbed SB organics to OHO biomass concentration ratio (XS/XBH), showing the primary anoxic (K2), secondary anoxic (K3), and anoxic–aerobic digestion (K4) specific denitrification rates.

483

ensure this. These concepts can also be used for anoxic selector reactor design (Ekama et al.,1996a). The simulation model was applied also to anoxic–aerobic digestion of WAS. It was found that the model predicted accurately both aerobic and anoxic–aerobic digester behavior under constant and cyclic flow and load conditions and validated the K4 specific denitrification rate (Warner et al., 1986); no significant adjustment to values of the kinetic constants was necessary.

4.14.25.2 Denitrification Potential The concentration of nitrate (per liter influent flow Qi) that an anoxic reactor can denitrify biologically is called that reactor’s denitrification potential. It is called a potential because whether or not it is achieved depends on the nitrate load on the anoxic reactor(s). If too little nitrate is recycled to the anoxic reactor, all the recycled nitrate will be denitrified and the actual removal of nitrate, that is, denitrification performance, will be lower than the potential. In this case the denitrification is system (or recycle) limited. An increase in the system recycle ratios will increase nitrate load on the anoxic reactor and hence also the denitrification. Once the recycle rates are such that the nitrate load on the anoxic reactor(s) equals the denitrification potential of the reactor, then the system denitrification performance is optimal and the recycle ratios are at their optimum values. At this point the anoxic and aerobic reactor nitrate concentrations are just zero and the lowest possible, respectively. Increasing the recycle rates above the optimum increases the nitrate concentration in the anoxic reactor outflow above zero but this does not improve the denitrification performance because the system has now become biological or kinetics limited. The denitrification potential of the anoxic reactor(s) has been achieved and more nitrate cannot be denitrified by the particular anoxic reactors and wastewater. Indeed, increases in the recycle ratios above the optimum values are uneconomical due to unnecessary pumping costs and introduce unnecessary additional DO into the anoxic reactors which causes an undesirable reduction in denitrification performance and increase in effluent nitrate concentration. The principle of denitrification design therefore hinges around (1) calculating the denitrification potential of the anoxic reactor(s); (2) setting the nitrate load imposed on the anoxic reactor(s) equal to the denitrification potential; and (3) calculating the recycle ratios associated with this condition. The recycle ratios so calculated are the optimum values. From the above discussion, it is clear that critical in the design for denitrification is calculation of the nitrate load and denitrification potential. The nitrate load is calculated from the nitrification capacity, which is the concentration of nitrate per liter influent flow (Qi) generated by nitrification (Section 4.14.22.2, Equation (152)). The nitrification capacity (Nc, mg N l1 influent) was shown above to be approximately proportional to the influent TKN concentration (Nti). The denitrification potential is calculated separately for the utilization of the RBSO and BPO. The RBSO gives rise to a rapid denitrification rate so that it can be assumed that it is all utilized in the primary anoxic reactor. This is in fact an objective in the design. Accordingly, the contribution of the RBSO to the denitrification potential is simply the catabolic component of its

484

Biological Nutrient Removal

electron-donating capacity in terms of nitrate as N. Therefore, in the complete utilization of the influent RBSO, a fixed proportion (1  fcvYH) of the RBSO electrons (catabolic component) will be donated to NO3  reducing it to N2. Thus, knowing the influent RBSO concentration and assuming it is all utilized, the denitrification potential of this RBSO is given by

Dp1RBSO ¼ f Sb0 s Sbi ð1  f cv YH Þ=2:86 ðmgNO3 -N l

1

influentÞ

components of the RBSO and BPO yields the total denitrification potential of primary and secondary anoxic reactors, that is,

Dp1 ¼ Dp1RBSO þ Dp1BPO ¼ f Sb0 s Sbi ð1  f cv YH Þ=2:86 þ Sbi K2 f x1 YH Rs =ð1 þ bH Rs Þ ¼ Sbi ff Sb0 s ð1  f cv YH Þ=2:86 þ K2 f x1 YH Rs =ð1 þ bH Rs Þg ðmgN l1 influentÞ

ð163Þ

ð160Þ Dp3 ¼ Dp3RBSO þ Dp3BPO

where Dp1 RBSO is the denitrification potential of the influent RBSO in primary anoxic reactor, Sbi the influent biodeg. COD (mgCOD l1), fSb0 s the RBSO fraction of Sbi, YH the OHO yield coefficient (0.45 mgVSS/mgCOD), and 2.86 the oxygen equivalent of nitrate. For the BPO, this substrate contributes to denitrification in the primary anoxic reactor and the secondary anoxic reactor. The denitrification potentials for the BPO are formulated on the basis of the K2 and K3 specific denitrification rates, respectively. These K rates are a simplification of the kinetic equations describing the utilization of BPO from the influent and/or from organism death and lysis and have a basis in the fundamental biological kinetics incorporated in the AS simulation models such as ASM1 (Henze et al., 1987). The K rates define the denitrification rate as mgNO3  -N denitrified per day per mgOHOVSS mass in the anoxic reactor. To determine the denitrification potential contributed by the BPO, the mass of OHOVSS produced per liter influent flow and the proportion of this mass in the primary and/or secondary anoxic reactors needs to be calculated and multiplied by the K2 or K3 rates. From the steady-state AS model for organics removal (Section 4.14.9.3), the OHO mass in the system (MXBH) is calculated from the biodegradable COD load (Equation (103)). Of this MXBH mass, a fraction fx1 and/or fx3 is continuously present in the primary and/or secondary anoxic reactors, respectively, that is, fx1 and fx3 are the primary and secondary anoxic sludge mass fractions, respectively. The OHOVSS mass in the primary and/or secondary anoxic reactors per liter influent flow is therefore given by

f x1 MXBH =Qi ¼ f x1 Sbi ðYH ÞRs =ð1 þ bH Rs Þ ðmgOHOVSS l1 influent in primary anoxicÞ

f x3 MXBH =Qi ¼ f x3 Sbi ðYH ÞRs =ð1 þ bH Rs Þ ðmgOHOVSS l1 influent in secondary anoxicÞ

Multiplying these masses by the respective K rates gives the primary and secondary anoxic reactor denitrification potentials attributable to BPO (Dp1BPO, Dp3BPO), viz.,

Dp1BPO ¼ K2 f x1 MXBH =Qi ¼ K2 f x1 Sbi YH Rs =ð1 þ bH Rs Þ ð161Þ Dp3BPO ¼ K3 f x3 Sbi YH Rs =ð1 þ bH Rs Þ

ð162Þ

This approach is valid because the K2 and K3 rates are continuous for the entire sludge residence time in the anoxic reactor(s), provided the nitrate concentration does not decrease to zero (Figure 35). Combining the denitrification potential

¼ 0 þ Sbi K3 f x3 YH Rs =ð1 þ bH Rs Þ ðmgN l1 influentÞ ð164Þ In Equations (163) and (164), the K2, K3, and bH rates are temperature sensitive, decreasing as the temperature decreases. The temperature sensitivity of these rates has been measured and is defined in Tables 13 and 6. From Equations (163) and (164), it can be seen that the denitrification potentials are directly proportional to the biodegradable COD concentration of the wastewater (Sbi). This is expected because in the same way that the oxygen demand is directly related to the COD load, so also is the nitrate demand (which is called the denitrification potential) because both oxygen and nitrate act as electron acceptor for the same organic degradation reactions. For the same size anoxic reactor, the primary anoxic has a much larger denitrification potential (by about 2–3 times) than the secondary anoxic because (1) K2 is larger than K3 and (2) more importantly, the RBSO makes a significant contribution to the denitrification potential in the primary anoxic reactor. For this reason the RBSO needs to be accurately specified to ensure reliable estimates of the N removal that can be achieved. For a normal municipal wastewater with an RBSO fraction (fSb0 s) of about 25% (with respect to biodegradable COD), the RBSO contributes about one-third to half of Dp1 depending on the size of the primary anoxic reactor and temperature. In a system where a high degree of N removal is required, between one-fourth and one-third of the carbonaceous oxygen demand is met by denitrification, which reduces the carbonaceous oxygen demand in the aerobic reactor by the same amount. As mentioned earlier, this reduction represents about half of the oxygen that was required to produce the nitrate by nitrification (see Figure 32). From Equation (164), the influent RBSO contribution to the denitrification potential of the secondary anoxic reactor is zero. This is because all the RBSO is utilized either in the preceding primary anoxic and/or in aerobic reactors. However, the Dp3 RBSO term has been included in Equation (164) in the event an external carbon source such as methanol, acetic acid, or high-strength organic wastewater is dosed into the secondary anoxic reactor to improve the denitrification. The sludge mass fraction approach above is valid because the fraction of the VSS (MXv) or TSS (MXt) masses that is OHO mass (MXBH) is constant for specified wastewater characteristics and sludge age and equal to the active fraction (fatOHO or favOHO – Equations (114) and (115)) and very closely the same in the anoxic and aerobic reactors of the system. Therefore, the anoxic and aerobic sludge mass fractions are the same whether calculated from the VSS, TSS, or OHO masses; for example, in an MLE system with anoxic and aerobic reactor volumes

Biological Nutrient Removal of 3000 and 6000 m3, respectively, one notes that nearly one-third of the OHO, VSS, and TSS masses in the system are in the anoxic reactor, and hence the anoxic sludge mass fraction is 0.33.

4.14.25.3 Minimum Primary Anoxic Sludge Mass Fraction In Equation (163), it is assumed that the initial rapid rate of denitrification is always complete, that is, the actual retention time in the primary anoxic reactor is always longer than the time required to utilize all the influent RBSO. This is because in Equation (163), the denitrification attributable to the influent RBSO is stoichiometrically expressed, not kinetically – it gives the concentration of nitrate the K1 rate removes when allowed sufficient time to reach completion. By considering the actual retention time (say t1) required to complete the first phase of denitrification (Figure 35(a)), and noting that t1(a þ s þ 1) is the minimum nominal hydraulic retention time to achieve this, it can be shown that the minimum primary anoxic sludge mass fraction fx1min to remove all the influent RBSO at a rate of K1 mgNO3-N/(mgOHOVSS d) is

f x1min ¼

f Sb0 s ð1  f cv YH Þð1 þ bHT Rs Þ 2:86K1T YH Rs

ð165Þ

Substituting the values of the kinetic constants into Equation (165), yields fx1mino0.08 for Rs 410 days at 14 1C. This value is much lower than most practical primary anoxic reactors so that Equation (163) will be valid in most cases. Equation (165) also applies to sizing anoxic selectors provided K1 (or mH) is appropriately selected (see Section 4.14.25.1, Equation (158); Ekama et al., 1996a).

4.14.25.4 Denitrification – Influence on Reactor Volume and Oxygen Demand From the design approach to nitrification (Equation (136)) and denitrification (Equations (163) and (164)), it can be seen that the design for N removal is done entirely using sludge mass fractions and does not require the volume of the reactor to be known. The volume of the reactor is obtained in the identical fashion as for the fully aerobic system and follows from the choice of the TSS concentration (Xt) for the reactor (Section 4.14.11). The volume of the reactor so obtained is then subdivided in proportion to the calculated primary and/or secondary anoxic and aerobic sludge mass fractions. Consequently, N removal design is grafted directly into the aerobic system design and for the same design reactor TSS concentration and sludge age, a fully aerobic system and an anoxic–aerobic system for N removal will have the same reactor volume. Research has indicated that there are many factors that influence the mass of sludge generated for a given sludge age and daily average COD load, and alternating anoxic–aerobic conditions is one of them. However, relative to the uncertainty in organic (COD) load and unbiodegradable particulate COD fraction and their daily and seasonal variation, these influences are not large enough from a design point of view to be given specific consideration in the design procedure. From a design point of view, the only significant difference between aerobic and anoxic–aerobic conditions is the oxygen demand

485

and this difference needs to be taken into account for economical design (Figure 32).

4.14.26 Development and Demonstration of Design Procedure It was concluded above that the influent wastewater characteristics that need to be accurately known are the influent TKN/COD ratio and RBSO fraction. These have a major influence on the nitrification capacity and denitrification potential, respectively, and hence on the N removal performance and minimum effluent nitrate concentration that can be achieved by biological denitrification. The effect of these two wastewater characteristics on design will be demonstrated below with numerical examples generated from the example raw and settled wastewaters with different influent TKN concentrations and RBSO fractions. The design of biological N removal is developed and demonstrated below by continuing the calculations with the example raw and settled wastewater characteristics listed in Tables 7 and 11. The only additional characteristic required for the denitrification design is the influent RBSO fraction (fSb0 s), which is 0.25 and 0.385 of the biodegradable COD for the raw and settled wastewaters, respectively. The results obtained so far for the COD removal and nitrification calculations for sludge ages 3–30 days are shown in Figures 14, 15, and 31.

4.14.26.1 Review of Calculations For the raw wastewater characteristics (i.e., fS0 up ¼ 0.15 mgCOD/mgCOD, fS0 us ¼ 0.07 mgCOD/mgCOD, Tmin ¼ 14 1C, Sti ¼ 750 mgCOD l1 – see Table 7) and 20 days sludge age, and accepting the nitrogen content of the volatile solids (fn) to be 0.10 mgN/mgVSS, the nitrogen required for sludge production Ns ¼ 17.0 mgN l1 (Equation (144)). From Section 4.14.23.2, the effluent biodegradable and unbiodegradable soluble organic nitrogen concentrations (Nobse and Nouse) are 0.0 and 1.80 mgN l1, respectively. From Equation (132) the effluent ammonia concentration Nae is 2.0 mgN l1. The effluent TKN concentration (Nte) is the sum of Nouse and Nae (Equation (150)) and hence Nte ¼ 3.8 mgN l1. The nitrification capacity (Nc) is found from Equation (152) and for the example raw wastewater (Nti ¼ 60.0 mgN l1; TKN/COD ¼ 0.08 mgN/mgCOD) at 14 1C is

Nc ¼ 60:0  17:0  3:8 ¼ 39:2 mgN l1 The nitrification oxygen demand, FOn is found from Equation (155), that is,

FOn ¼ 4:57Nc Qi ¼ 4:57  39:2  15  106 mgOd1 ¼ 2687 kgOd1 and the mass of nitrifier VSS in the reactor is given by Equation (154), that is,

MXBA ¼ 0:1  20=ð1 þ 0:034  20Þ  39:2  15 ¼ 702 kgVSS

486

Biological Nutrient Removal

In the design, because it is intended to reduce the nitrate concentration as much as possible, the alkalinity change in the wastewater will be minimized; assuming that 80% of the nitrate formed is denitrified, the H2CO3* alk change ¼ 7.14Nc  3.57 (nitrate denitrified) ¼  7.14  39.2 þ 3.57  0.80  39.2 ¼ 168 mg l1 as CaCO3. With an influent H2CO3* alk of 250 mg l1 as CaCO3 the effluent H2CO3* alk ¼ 250–168 ¼ 82 mg l1 as CaCO3, which, from Figure 27, will maintain a pH above 7 (see Section 4.14.20.6).

4.14.26.2 Allocation of Unaerated Sludge Mass Fraction In nitrogen removal systems, the maximum anoxic sludge mass fraction available for denitrification, fxdm, can be set equal to the maximum unaerated sludge mass fraction fxm at the minimum temperature, that is,

f xdm ¼ f xm

ð166Þ

where fxm is given by Equation (136) for selected Rs, mnmT, and Tmin. This is because for N removal systems, unaerated sludge mass need not be set aside for the anaerobic reactor. In N and P removal systems, some of the unaerated sludge mass (0.12–0.25) needs to be set aside for the anaerobic reactor to stimulate BEPR. This sludge mass fraction, called the anaerobic sludge mass fraction and denoted fxa, cannot be used for denitrification. For BEPR to be as high as possible, no nitrate should be recycled to the anaerobic reactor so that zero denitrification takes place in this reactor. So, for the purposes of this development and demonstration of denitrification behavior, it will be accepted that the maximum unaerated sludge mass fraction available at 20 days sludge age (fxm) is all allocated to anoxic conditions, that is, fxdm ¼ fxm ¼ 0.534.

4.14.26.3 Denitrification Performance of the MLE System 4.14.26.3.1 Optimum recycle ratio a In the MLE system, the anoxic sludge mass fraction is all in the form of a primary anoxic reactor, that is, fx1 ¼ fxdm ¼ fxm. The denitrification potential of the primary anoxic reactor Dp1 is found from Equation (163), that is, for the example raw wastewater at 14 1C and fxm ¼ fxdm ¼ fx1 ¼0.534, Dp1 ¼ 52.5 mgN l1. The only additional wastewater characteristic required to calculate Dp1 is the influent RBSO (Sbsi) concentration or fraction (fSb’s), which for the example raw and settled wastewaters are given in Table 14, that is, 0.25 and 0.385 with respect to the biodegradable COD (Sbi), respectively. In the MLE system, if the nitrate concentration in the outflow of the anoxic reactor is zero, then the nitrate concentration in the aerobic reactor (Nnar) is equal to Nc/ (a þ s þ 1), that is, the nitrification capacity in mgN l1 influent flow diluted by the total (no nitrate containing) flow entering the aerobic reactor which is (a þ s þ 1) times the influent flow, where a and s are the mixed liquor and underflow recycle ratios (with respect to the influent average dry weather flow Qi), respectively. Accepting that there is no denitrification in the secondary settling tank (which needs to be minimized anyway due to the problem of rising sludges), the aerobic reactor and system effluent nitrate concentrations (Nnar and

Table 14 Additional wastewater characteristics required for denitrification (and BEPR) design Wastewater Readily biodegradable soluble organics (RBSO) as.y (1)y.fraction of biodegradable organics (BO, Sbi) COD (fSb0 s) (2)y.fraction of total organics (Sti, COD) (fS0 bs) VFA fraction of biodegradable soluble organics (RBSO), (fSbs0 a)

Raw

Settled

0.25

0.385

0.194

0.324

0.10

0.10

Nne, respectively) are equal and given by

Nne ¼ Nnar ¼ Nc =ða þ s þ 1Þ

ð167Þ

Knowing Nne and Nnar and taking into account DO concentrations in the a and s recycles, that is, Oa and Os mgO l1 respectively, the equivalent nitrate load on the primary anoxic reactor (Nnlp) by the a and s recycles is

    Oa Os a þ Nne þ s Nnlp ¼ Nnar 2:86 2:86 The optimum denitrification (i.e., lowest effluent nitrate concentration) is obtained when the equivalent nitrate load on the anoxic reactor is equal to the denitrification potential of the anoxic reactor (i.e., Dp1 ¼ Nnlp), viz.,

 Dp1 ¼

   Nc Oa Nc Os þ þ aþ s ð168Þ ða þ s þ 1Þ 2:86 ða þ s þ 1Þ 2:86

Solving Equation (168) for a yields the a recycle ratio which exactly loads the primary anoxic reactor to its denitrification potential with nitrate and DO. This a value is the optimum because it results in the lowest Nne, that is,

aopt ¼ ½B þ

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi B 2 þ 4AC=ð2AÞ

ð169Þ

where

A ¼ Oa =2:86 B ¼ Nc  Dp1 þ fðs þ 1ÞOa þ sOs g=2:86 C ¼ ðs þ 1ÞðDpp  sOs =2:86Þ  sNc and

Nnemin ¼ Nneaopt ¼ Nc =ðaopt þ s þ 1Þ ðmgN l1 Þ

ð170Þ

For a ¼ aopt, Equation (170) for Nne is valid and will give the minimum Nne attainable. When a raopt Equation (170) is also valid because the assumption on which Equation (169) is based is valid, that is, NnlprDp1 or equivalently, zero nitrate concentration in the outflow of the anoxic reactor. For a 4aopt this assumption is no longer valid and Nne increases as the a recycle ratio increases due to increasing DO flux entering the anoxic reactor. For a4aopt, Nne is given by the difference between the equivalent nitrate load on the anoxic reactor (which is the sum of the nitrification capacity Nc and the nitrate

Biological Nutrient Removal MLE system (settled) Effluent nitrate and a recycle ratio

MLE system (raw) Effluent nitrate vs. a recycle

20

487

20

14 °C s = 0.5 s = 1.0 s = 2.0 22 °C s = 1.0

15

10

Effluent nitrate (mgN I−1)

Effluent nitrate (mgN I−1)

a-opt (14 °C)

a-opt (14 °C) a-opt (22 °C)

5

14 °C s = 0.5 s = 1.0 s = 2.0

15

22 °C s = 1.0

10

N ne min (14 °C)

5

N ne min (14 °C)

0 0

0 5

10 a Recycle ratio

15

0

5

10 a Recycle ratio

15

Figure 37 Effluent nitrate concentration vs. mixed liquor a recycle ratio for the example raw (a) and settled (b) wastewaters for underflow (s) recycle ratio of 1:1 at 14 1C (bold line) and 22 1C (thin line) and for s ¼ 0.5:1 and 2.0:1 at 14 1C (dashed lines).

equivalent of the oxygen concentration with respect to the influent flow) and the denitrification potential Dp1, viz.,

Nne ¼ Nc þ

aOa sOs þ  Dp1 2:86 2:86

ðmgN l1 Þ

ð171Þ

As Nc, Dp1, Os, and Oa are constants, the increase in Nne with increasing a above aopt is linear with slope Oa/2.86 mgN l1. At a ¼ aopt, Equations (170) and (171) give the same Nne concentrations. Accepting the design sludge age of 20 days, which allows a maximum unaerated sludge mass fraction fxm of 0.534, the denitrification behavior of the MLE system is demonstrated below for the example raw and settled wastewaters at 14 and 22 1C. In the calculations the DO concentrations in the a and s recycles, Oa and Os are 2 and 1 mgO l1, respectively, and the underflow recycle ratio s is 1:1. This s recycle ratio is usually fixed at a value such that satisfactory settling tank operation is obtained. Details of secondary settling tank theory, design, modeling, and operation are discussed by Ekama et al. (1997) and Ekama and Marais (2004). Substituting the values for the nitrification capacity Nc and denitrification potential Dp1 into Equations (169) and (170), the optimum mixed liquor recycle ratio aopt and minimum effluent nitrate concentration Nneaopt are obtained, for example, for the settled wastewater at 14 1C

A ¼ 2=2:86 ¼ 0:70 B ¼ 39:6  40:1 þ fð1 þ 1Þ2 þ 1  1g=2:86 ¼ þ1:52 C ¼ ð1 þ 1Þð40:1  1  1=2:86Þ  1  39:6 ¼ þ39:61 Hence, aopt ¼ 6.5 and Nnemin ¼ 4.7 mgN l1. The calculations for the example raw and settled wastewater at 14 and 22 1C show that for all four cases aopt exceeds 5. Although the calculations include the discharge of DO to the anoxic reactor, a recycle ratios above 5 to 6 are not cost effective. The small decreases in Nne which are obtained for even large increases in a recycle ratio above about 5:1 do not warrant the additional pumping costs.

This is illustrated in Figure 37 which shows Nne versus a recycle ratio for the example raw (Figure 37(a)) and settled (Figure 37(b)) wastewater at 14 and 22 1C plotted from Equations (170) and (171). For the settled wastewater (Figure 37(b)) at 14 1C and s ¼ 1:1, for aoaopt, the anoxic reactor is underloaded with nitrate and DO and as the a recycle increases up to aopt, the equivalent nitrate load increases toward the anoxic reactor’s denitrification potential. Initially, Nne decreases sharply for increases in a, but as a increases the decrease in Nne becomes smaller. At 14 1C with a ¼ aopt ¼ 6.5, the anoxic reactor is loaded to its denitrification potential by the a and s recycles and a Nnemin ¼ Nneaopt ¼ 4.7 mgN l1 is achieved. At a ¼ aopt ¼ 6.5, the greatest proportion of the anoxic reactor’s denitrification potential is used for denitrification and therefore yields the minimum effluent nitrate concentration (Nneaopt). This is shown in Figures 38(a) and 38(b) for the raw and settled wastewaters at 14 1C. For the settled wastewater at 14 1C (Figure 38(b)) at a ¼ aopt ¼ 6.5, 88% of the equivalent nitrate load (i.e. (a þ s)  Nnemin ¼ 35.2 mgN l1 out of a Dp1 ¼ 40.1 mgN l1) is nitrate and therefore 88% of the denitrification potential of the anoxic reactor is utilized for denitrification and 12% for DO removal. The higher the a recycle ratio, the greater the proportion of the denitrification potential is utilized for DO removal. At 14 1C, for a4aopt, the equivalent nitrate load exceeds the denitrification potential and as the a recycle increases so Nne increases due to the increased DO mass flow to the anoxic reactor. From Equation (171), at a ¼ 15, Nne ¼ 10.6 mgN l1 and 27% of the denitrification potential is required to remove DO, leaving only 73% for denitrification (Figures 37(b) and 38(b)). For 14 1C, the plots of Nne versus a at underflow s recycle ratios of 0.5:1 and 2.0:1 are also given in Figure 37 and show that aopt is not significantly different at different s recycle ratios. Also, at low a recycle ratios, changes in s have a significant influence on Nne, but at high a recycle ratios, even significant changes in s do not significantly change Nne. This is because at high a, most of the nitrate is recycled to the anoxic reactor by the a recycle, so that changes in s do not

488

Biological Nutrient Removal MLE system (raw) use of denitrification potential 100

Denitrification potential used for nitrate removal

20

% Denit. potential

% Denit. potential

Denitrification potential used for nitrate removal

0

Denitrification potential used for nitrate removal

Denitrification potential used for nitrate removal

40

a-opt (14 °C) a-prac

0 0

(a)

60

20

a-opt (14 °C)

a-prac

Used DO removal

Unused denitrification potential

80

60 40

100

Used DO removal

Unused denitrification potential

80

MLE system (settled) use of denitrification potential

5

10 a Recycle ratio

15

0 (b)

5

10 a Recycle ratio

15

Figure 38 % Denitrification potential unused, used by dissolved oxygen in the recycles and for denitrification vs. a recycle ratio for the example raw (a) and settled (b) wastewaters for underflow (s) recycle ratio of 1:1 at 14 1C.

significantly change the nitrate load on the anoxic reactor. Hence, for the MLE system, decreases in s can be compensated for by increases in a – it makes little difference which recycle brings the nitrate to the anoxic reactor as long as the anoxic reactor is loaded as closely as practically possible to its denitrification potential in order to minimize Nne. For the settled wastewater at 22 1C and s ¼ 1:1 (Figure 37(b)), Nne versus a is similar to that at 14 1C up to a ¼ 6.5. This is because Nc values at 14 and 22 1C for the example raw and settled wastewater are almost the same (i.e., 39.9 and 41.6 mgN l1 at 14 and 22 1C, respectively). However, at 22 1C, the denitrification potential is significantly higher than at 14 1C (40.1 mgN l1 at 14 1C and 52.4 mgN l1 at 22 1C) so that a higher aopt is required (e.g., 17.9) at 22 1C to load the anoxic reactor to its denitrification potential than at 14 1C. Therefore, at 22 1C, as the a recycle increases above 6.5, Nne continues to decrease until aopt ¼ 17.9 is reached. The increase in a from 6.5 to 17.9 reduces Nne from 4.9 to 2.1, that is, only 2.8 mgN l1. This small decrease in Nne is not worth the large increase in pumping costs from 6.5:1 to 17.9:1 required to produce it. Consequently, for economical reasons, the a recycle ratio is limited at a practical maximum (aprac) of say 5:1, which fixes the lowest practical effluent nitrate concentration (Nneprac) from the MLE system between 5 and 10 mgN l1 depending on the influent TKN/ COD ratio. From the design procedure demonstrated so far, it is clear that the procedure hinges around balancing the equivalent nitrate load with the denitrification potential by appropriate choice of the a recycle ratio: for selected system design parameters (sludge age, anoxic mass fraction, underflow recycle ratio, etc.) and wastewater characteristics (temperature, readily biodegradable COD fraction, TKN/COD ratio, etc.), the denitrification potential of the MLE system is fixed. With all the above fixed, the system denitrification performance is controlled by the a recycle ratio, and this performance is optimum when the a recycle ratio is set at the optimum aopt. For aoaopt, the performance will be below optimum because the equivalent nitrate load is less than the denitrification potential (Figure 38); for a ¼ aopt, the performance is optimal because the equivalent

nitrate load equals the denitrification potential; and for a4aopt, the performance is again suboptimal because now the equivalent nitrate load is greater than the denitrification potential and more than necessary DO is recycled to the anoxic reactor which reduces the denitrification (see Figures 37 and 38). If a practical limit on a is set at say aprac ¼ 5:1 and aopt is significantly higher, then a significant proportion of the anoxic reactor’s denitrification potential is not used (Figure 38). There are two options to deal with this unused denitrification potential: (1) change the design, that is, decrease the sludge age (Rs) and/or unaerated sludge mass fraction (fxm) or (2) leave the system as designed (i.e., Rs ¼ 20 days and fxm ¼ 0.534) and keep the unused denitrification potential in reserve as a factor of safety against changes in wastewater characteristics, such as (1) increased organic load, which will require a reduction in sludge age, (2) increased TKN/COD ratio, which will load the anoxic reactor with nitrate at lower a recycle ratios, or (3) decreased RBSO fraction, which decreases the anoxic reactors denitrification potential.

4.14.26.3.2 The balanced MLE system With option (1) the anoxic sludge mass fraction fx1 is decreased to eliminate the unused denitrification potential. The decrease in fx1 increases the aerobic mass fraction and therefore the factor of safety (Sf) on nitrification. To maintain the same Sf, the sludge age of the system can be reduced to that value at which the lower fx1 is equal to the maximum unaerated sludge mass fraction fxm allowed (i.e., fx1 ¼ fxm) for the selected mAm20 and Tmin. An MLE system with a sludge age (Rs) and influent TKN concentration (Nti) such that fx1 ¼ fxm and aopt ¼ aprac (say 5:1), so that this aprac loads the anoxic reactor exactly to its denitrification potential, is called a balanced MLE system. This approach to design of the MLE system was proposed by van Haandel et al. (1982) and gives the most economical AS reactor design, that is, the lowest sludge age, and therefore the smallest reactor volume, and the highest denitrification with the a recycle ratio fixed at some maximum practical limit. The influent TKN/COD ratio, fxm ¼ fx1, fx1 min, Nne, and %N removal (%Nrem) versus sludge age for balanced

Biological Nutrient Removal MLE system (settled, 14 °C) Design at fixed a -opt = 5:1

MLE system (raw, 14 °C) Design at fixed a-opt = 5:1 Balanced sludge age

0.8

0.12

f x1 = f xm

0.6 0.4

TKN/COD Raw WW

0.08

0.2

Influent TKN/COD ratio

0.14

0.16

1.0 Balanced sludge age

0.14 0.12

0.6

Settled WW f x1 = f xm

0.10

0.4

0.08

0.0 0

5

(a)

10 15 20 Sludge age (days)

25

0.06

30

0.0 0

100

20

25

30

100

75

Balanced sludge age

10

50

Effluent nitrate

5

25

0

0 0

5

10 15 20 Sludge age (days)

25

15

75

Balanced sludge age

10

50 Effluent nitrate

5

25

0

30

0 0

(d)

% Nitrogen removal

%N removal

15

Effluent nitrate (mgN I−1)

%N removal % Nitrogen removal

Effluent nitrate (mgN I−1)

10 15 20 Sludge age (days)

MLE system (settled, 14 °C) Design at fixed a -opt = 5:1

MLE system (raw, 14 °C) Design at fixed a-opt = 5:1

(c)

5

(b)

20

0.2

f x1min

f x1min 0.06

0.8

TKN/COD

Anoxic mass fraction

1.0 Anoxic mass fraction

Influent TKN/COD ratio

0.16

0.10

489

5

10 15 20 Sludge age (days)

25

30

Figure 39 Influent TKN/COD ratio (TKN/COD), maximum unaerated (fxm), primary anoxic (fx1), and minimum primary anoxic (fx1 min) sludge mass fractions (a,b) and effluent nitrate concentration and %N removal (c,d) for balanced MLE systems with a 5:1 practical upper limit to the a recycle ratio for the example raw (a,c) and settled (b,d) wastewaters at 14 1C.

MLE systems for the example raw and settled wastewaters at 14 and 22 1C are shown in Figures 39 and 40, respectively. The sludge age which balances the MLE system for given wastewater characteristics and aprac cannot be calculated directly. It is easiest to calculate the influent TKN concentration for a range of sludge ages and choose the sludge age which matches the wastewater TKN concentration (Nti). The procedure for calculating Nti for a balanced MLE system is as follows: from the design mAm20, Tmin, Sf, and a selected sludge age, fxm is calculated from Equation (136). Provided fxm4fx1 min (Equation (165)), fx1 is set equal to fxm. Knowing fx1 and the wastewater characteristics, Dp1 is calculated from Equation (163). This Dp1 and a selected value for aprac are then substituted into Equation (168), which sets the equivalent nitrate load on the anoxic reactor equal to the denitrification potential and hence aopt equals the selected aprac. With Dp1 and a known, Nc is calculated from Equation (168). Once Nc is known, Nti is calculated from Nti ¼ Nte þ Ns þ Nc (Equation (152)), where Nte ¼ Nouse þ Nae (Equation (150)) and Nae is given by Equation (132) because with Sf fixed the Rs  fxm relationship is fixed. With Nc and Nti known, the effluent nitrate concentration Nne and % nitrogen removal (%Nrem) are

found from Equation (170) and %Nrem ¼ 100[Nti  (Nne þ Nte)]/Nti, respectively. This calculation is repeated for different sludge ages. The shortest sludge age allowed is the one which gives fx1 ¼ fxm ¼ fx1min. In Figure 39, for 14 1C, for the raw wastewater (Figures 39(a) and 39(c)), it can be seen that fx1( ¼ fxm) increases from about 0.09 at 8 days sludge age, at which fxm is just greater than fx1 min, to 0.60 at 26 days sludge age, at which fxm is equal to the upper limit set for it. As fx1 increases so the influent TKN/COD ratio increases from 0.061 at 8 days sludge age to 0.115 at 26 days sludge age. With the increase in TKN/COD ratio, the nitrification capacity Nc increases and hence Nne increases from about 3.2 mgN l1 at 8 days sludge age to 9.3 mgN l1 at 26 days sludge age because the a and s recycle ratios remain at 5:1 and 1:1, respectively (see Equation (170)). The %N removal, which includes the N removed via sludge wastage Ns, decreases marginally from 85% to 82% as the influent TKN/COD ratio and sludge age increase for the balanced MLE system. For the settled wastewater at 14 1C (Figures 39(b) and 39(d)), the influent TKN/COD ratio, fx1 and fx1min results are similar to those for the raw wastewater, that is, for the same

Biological Nutrient Removal MLE system (raw, 22 °C) Design at fixed a-opt = 5:1 1.0

0.6

f x1 = f xm

0.4

0.10 Raw WW

0.2

0.14

Balanced sludge age

0.12

0.0

0.4

0.10

0.2

0.08

100

Effluent nitrate

50

25

Balanced sludge age

0 0

5

10 15 20 Sludge age (days)

25

10 15 20 Sludge age (days)

25

30

MLE system (settled, 22 °C) Design at fixed a-opt = 5:1

20

% Nitrogen removal

75

15

5

5

(b)

0 (c)

0

%N removal

10

0.0

0.06

30

MLE system (raw, 22 °C) Design at fixed a-opt = 5:1

20 Effluent nitrate (mgN I−1)

25

Effluent nitrate (mgN I−1)

(a)

10 15 20 Sludge age (days)

0.6

f x1mm

0.06 5

f x1 = f xm

Settled WW

f x1mm 0

0.8

Anoxic mass fraction

0.8

0.08

1.0 TKN/COD

0.14 0.12

0.16

TKN/COD Influent TKN/COD ratio

Balanced sludge age

Anoxic mass fraction

Influent TKN/COD ratio

0.16

MLE system (settled, 22 °C) Design at fixed a-opt = 5:1

100

%N removal 15

75

10

50 Effluent nitrate

5

25

Balanced sludge age

0

0

30

% Nitrogen removal

490

0 (d)

5

10 15 20 Sludge age (days)

25

30

Figure 40 Influent TKN/COD ratio (TKN/COD), maximum unaerated (fxm), primary anoxic (fx1), and minimum primary anoxic (fx1 min) sludge mass fractions (a,b) and effluent nitrate concentration and %N removal (c,d) for balanced MLE systems with a 5:1 practical upper limit to the a recycle ratio for the example raw (a,c) and settled (b,d) wastewaters at 22 1C.

sludge age approximately the same TKN/COD ratio is found for the balanced MLE system. For the settled wastewater, the Nne is slightly lower, increasing from about 3.2 to 6.7 mgN l1 from 8 to 26 days sludge age; also the %N removal is somewhat lower, around 78% mainly due to the lower N removal via sludge wastage Ns. However, it must be remembered that the TKN/COD ratio and RBSO fraction of a settled wastewater are higher than those of the raw wastewater from which it is produced, viz. TKN/COD ratio 0.113 and 0.080 mgN/mgCOD and RBSO fraction (fSb’s) 0.25 and 0.385 for the example settled and raw wastewaters, respectively. Therefore, at 14 1C, while the raw wastewater can be treated in a balanced MLE system at about 11 days sludge age (Figure 39(a)), the sludge age for the settled wastewater balanced MLE system is about 17 days (Figure 39(b)). A comparison of the balanced MLE systems for the example raw and settled wastewaters is given in Table 15. From Table 15 it can be seen that Nne is less than 1 mgN l1 higher for the settled wastewater but the reactor volume and total oxygen demand significantly lower compared with the

Table 15 Comparison of balanced MLE systems treating the example raw and settled wastewaters at 14 1C Parameter

Raw

Settled

Influent TKN/COD ratio Unaerated mass fraction(fxm) Anoxic mass fraction (fx1) Minimum anoxic fraction a Recycle ratio (aprac ¼ aopt) Sludge age (days) Effluent nitrate (Nne, mgN l1) Effluent TKN (Nte, mgN l1) Reactor vol. at 4.5 gTSS l1 (m3) Carb O2 demand (FOc, kgO d1) Nit O2 demand (FOn, kgO d1) O2 recovered (FOd, kgO d1) Tot. O2 demand (FOtd, kgO d1) %N removal Mass TSS wasted (FXt, kg d1) Active fraction wrt TSS (fatOHO)

0.08 0.306 0.306 0.08 5:1 11 5.1 4.3 9484 6156 2492 1327 7321 84.3 3880 0.316

0.113 0.485 0.485 0.108 5:1 17 5.7 4.1 5264 4251 2685 1437 5499 80.9 1394 0.414

Biological Nutrient Removal

raw wastewater. Therefore, from an AS system point of view, treating settled wastewater would be more economical than treating raw wastewater for a comparable effluent quality. Also, both systems require sludge treatment; for the raw wastewater because 11 days sludge age waste sludge is not stable (high active fraction, favOHO) and for the settled wastewater, the primary sludge needs to be stabilized. The 11 days sludge age waste sludge can be stabilized with anoxic aerobic digestion which allows the N released in digestion to be nitrified and denitrified (Warner et al. 1986; Mebrahtu et al., 2010) and primary sludge can be anaerobically digested to benefit from gas generation. The choice of treating raw or settled wastewater therefore does not depend so much on the effluent quality or the economics of the AS system itself, but on the economics of the whole WWTP, including sludge treatment. Because the minimum wastewater temperature (Tmin) governs the AS system (and sludge treatment) design, the balanced MLE system results for 22 1C are not particularly relevant to the temperate climate regions. However, in equatorial and tropical regions, where wastewater treatment is becoming a matter of increasing concern, high wastewater temperatures are encountered. For this reason and for illustrative purposes also, the balanced MLE results for the raw and settled wastewaters are shown in Figure 40. Compared with 14 1C, the upper limit to fxm ¼ 0.60 is reached already at 7 days sludge age and significantly higher influent TKN/COD ratios can be treated at equal sludge ages. These higher TKN/COD ratios result in higher Nne, which for the raw wastewater increases from 3 to 13 mgN l1 and for the settled wastewater from 3 to 9 mgN l1 for increases in sludge age from 4 to 30 days. If Tmin were 22 1C, the example raw and settled wastewaters could be treated at 3 and 4 days sludge age, respectively, yielding Nne of 5 and 6.5 mgN l1, respectively. This reinforces the conclusion in Section 4.14.24.1 that in equatorial and tropical climates it is highly likely that AS plants will nitrify even at very short sludge ages (1–2 days) and therefore to design for denitrification for operational reasons if not for effluent quality reasons.

4.14.26.3.3 Effect of influent TKN/COD ratio When the unused denitrification potential in the anoxic reactor is kept in reserve as a safety factor (option 2), the sludge age and unaerated (anoxic) mass fraction are not changed. For this situation, it is useful to have a sensitivity analysis to see the influence of changing influent TKN/COD ratio and RBSO fraction on the a recycle ratio and effluent nitrate concentration. Continuing with the design for the example raw and settled wastewaters for fixed sludge age at 20 days and unaerated (anoxic) mass fraction at 0.534, a plot of the optimum a recycle ratio aopt and minimum effluent nitrate concentration Nneaopt for underflow recycle ratios s of 0.5, 1.0, and 2.0 versus influent TKN/COD ratio from 0.06 to 0.16 is given in Figure 41 for the raw (a), (c) and settled (b), (d) wastewaters at 14 1C (a), (b) and 22 1C (c), (d). From Figure 41, it can be seen that as the influent TKN/ COD increases, aopt decreases and Nneaopt increases. The aopt–Nneaopt lines in Figure 41 give the system denitrification performance when the denitrification potential of the anoxic reactor is fully used, that is, the system denitrification

491

performance is equal to its denitrification potential and the nitrate concentration is the lowest possible. Also, large increases in the underflow recycle ratio s (i.e., from 0.50:1 to 1.0:1 or 1.0:1 to 2.0:1) decrease aopt but do not change Nneaopt because the DO in the a and s recycles does not differ much in their influence on the anoxic reactor. Therefore, it matters little which recycle flow brings the nitrate load to the anoxic reactor. As long as the anoxic reactor is closely loaded to its denitrification potential, the same minimum effluent nitrate concentration (Nneaopt) will be obtained at aopt. The aopt–Nneaopt lines therefore give the system denitrification performance when the denitrification potential of the anoxic reactor is fully used (Figure 38(b)), that is, the systems denitrification performance is equal to its potential. A better denitrification performance is not possible – the denitrification is kinetics limited and the biomass (and so also the system) does the best it can (for the given K2 denitrification rate). From Equation (170), the system denitrification performance with increasing influent TKN/COD ratio at a fixed practical operating a recycle ratio (aprac) of 5:1 is also shown in Figure 41 as the aprac and Nneaprac lines. It can be seen that Nneaprac increases linearly with increase in influent TKN/COD ratio. For low influent TKN/COD ratios, aprac is considerably lower than aopt and the system denitrification performance is lower than its denitrification potential. This is evident from Nneaprac being greater than Nneaopt. As the TKN/COD ratio increases, aopt decreases until aopt ¼ aprac ¼ 5.0:1. For the raw wastewater at 14 1C (Figure 41(a)), this happens at an influent TKN/COD ratio of 0.104. This is the influent TKN/COD ratio which balances the MLE system for the selected design conditions, namely, 20 days sludge age, fxm ¼ 0.534 and aprac ¼ 5:1 for the example raw wastewater at 14 1C. For influent TKN/ COD ratios 40.104, the a recycle ratio should be set at aopt, which fully uses the anoxic reactor’s denitrification potential and is now lower than aprac ¼ 5:1. Therefore for aprac set at 5:1, only when the influent TKN/COD ratio is 40.104, is the denitrification potential of the anoxic reactor fully used. This same conclusion can be made from Figure 39(a) at 20 days sludge age, that is, fxm ¼ 0.534 and TKN/COD ratio ¼ 0.104. Therefore for influent TKN/COD ratioso0.104, while aprac oaopt, the system denitrification performance is lower than its denitrification potential because not all the denitrification potential of the anoxic reactor is used. Once the TKN/COD ratio increases above that value which balances the MLE system, aoptoaprac and a should be set at aopt to achieve the lowest effluent nitrate concentration (Nneaopt). For these influent TKN/COD ratios, the denitrification potential of the anoxic reactor is fully used and the system denitrification performance is defined by the aopt–Nneaopt lines. Figure 41 is useful because it combines the system denitrification performance (aprac–Nneaprac lines) and the denitrification potential (aopt–Nneaopt lines) in the same diagram as influent TKN/COD ratio increases for a particular wastewater and system design (Rs ¼ 20 days and fxm ¼ 0.534). The intersection point of the straight Nneaprac line and the curved Nneaopt line (i.e., at aopt ¼ aprac ¼ 5:1) gives the influent TKN/ COD ratio for the balanced MLE system for the selected aprac ¼ 5:1. From Figure 41(a), for the raw wastewater at 14 1C, the MLE system (at 20 days sludge age and fxm ¼ 0.534) with a

Biological Nutrient Removal

10

20 15

a-prac

5

0.06

0.08

0.10

15

25 20 15 s=0.5 s=1.0 s=2.0

a-prac Nne-a-prac Nne-a-opt

a-opt

0 0.06

10 5

0.10

0.12

0.14

s=0.5 s=1.0 s =2.0

a-opt Nne-a-opt

0 0.06

0.08

0.10

0.12

10 5 0

0.14

0.16

Influent TKN/COD ratio MLE system (settled, 22 °C) a Recycle ratio and effluent nitrate

15

30 a-opt 25

5

20

a-prac

s=0.5 s =1.0 s =2.0

a-opt

15 10

Nne-a-prac

5 Nne-a-opt

0

0.16

Influent TKN/COD ratio

Example Settled WW

10

0 0.08

a-prac Nne-a-prac

30

10

5

5

(b)

a-opt

20 15

0.16

MLE system (raw, 22 °C) a Recycle ratio and effluent nitrate Example Raw WW

a Recycle ratio

0.14

25

10

0 0.12

Influent TKN/COD ratio

(a)

(c)

5

a-opt

Nne-a-opt

0

10

s=0.5 s=1.0 s=2.0

Nne-a-prac

a Recycle ratio

a-opt

30

Example Settled WW

a-opt

a Recycle ratio

a Recycle ratio

25

15

Effluent nitrate (mgN I−1)

30

Example Raw WW

Effluent nitrate (mgN I−1)

15

MLE system (settled, 14 °C) a Recycle ratio and effluent nitrate

Effluent nitrate (mgN I−1)

MLE system (raw, 14 °C) a Recycle ratio and effluent nitrate

Effluent nitrate (mgN I−1)

492

0.06 (d)

0 0.08 0.10 0.12 0.14 Influent TKN/COD ratio

0.16

Figure 41 Optimum (aopt) and practical upper limit (aprac ¼ 5:1) a recycle ratios (bold lines) and effluent nitrate concentration at aopt (Nneaopt, bold line) and aprac (Nneaprac, dashed line) vs. influent TKN/COD ratio at underflow (s) recycle ratio of 1:1 for the example raw (a,c) and settled (b,d) wastewaters at 14 1C (a,b) and 22 1C (c,d). The optimum a recycle ratio (aopt) values at underflow recycle ratios of 0.5:1 and 2:1 are also shown (thin lines).

recycle ratio 45:1 can maintain effluent nitrate concentrations below 8.1 (total N 12.4) mgN l1 for influent TKN/COD ratios below 0.104 (78.0 mgN l1). With settled wastewater at 14 1C (Figure 41(b)), the MLE system with a 45:1 can maintain effluent nitrate concentrations below 11.3 (total N 14.9) mgN l1 for influent TKN/COD ratios up to 0.132 (59.4 mgN l1). Similarly, from Figures 41(c) and (d), with raw and settled wastewater at 22 1C, the MLE system with a 45:1 can maintain effluent nitrate concentrations below 6.0 and 8.1 mgN l1 (total N 9.9 and 11.1 mgN l1) for influent TKN/COD ratios up to 0.119 (89.3 mgN l1) and 0.148 (66.6 mgN l1). These results show that the MLE system treating settled wastewater delivers lower Nne (by 2–3 mgN l1) than when treating raw wastewater and at influent TKN/ COD ratios significantly higher. However, it should be noted that (1) the influent TKN concentrations (given above) for the raw wastewater are considerably higher than those for the settled wastewater and (2) a settled wastewater with a TKN/ COD ratio of 0.119 (14 1C) or 0.148 (22 1C) would be produced from a raw wastewater with considerably lower influent TKN/COD ratio than 0.104 (14 1C) and 0.132 (22 1C).

4.14.26.3.4 MLE sensivity diagram In Figure 41, the system denitrification performance at a selected aprac ¼ 5 is combined with the system denitrification potential at a ¼ aopt for varying influent TKN/COD ratio and a single influent RBSO fraction value. This influent TKN/COD ratio sensitivity diagram can be extended by adding the Nneaopt lines for other influent RBSO fractions. A sensitivity analysis of the system at the design stage is useful for evaluating the denitrification performance under varying influent TKN/COD ratio and RBSO fractions. These two wastewater characteristics can vary considerably during the life of the plant and have a major impact on the N removal performance of the system. The denitrification potential and system performance are combined for varying influent TKN/COD ratio and RBSO fraction in Figure 42. For the fixed system design parameters (i.e., Rs ¼ 20 days, fxdm ¼ fxm ¼ 0.534, s ¼ 1.0), the curved (bold) lines give Nneaopt when the anoxic reactor is loaded to its denitrification potential, that is, Nne for a ¼ aopt for varying TKN/COD ratio from 0.06 to 0.16 and RBSO fractions from 0.10 to 0.35 for the example raw and settled wastewaters at

Biological Nutrient Removal MLE system (settled, 14 °C) Effluent nitrate vs. TKN/COD ratio

MLE system (raw, 14 °C) Effluent nitrate vs. TKN/COD ratio 40

20 0.5 1

30 RBCOD fraction 20

2 3

0.10 0.15

4

0.20 0.25 0.30 0.35

10

5

6

Effluent nitrate (mgN I−1)

Effluent nitrate (mgN I−1)

0

15

a recycle ratio 0 0.5 0.10

10

1 2

0.35

4

5 6 8 10

10 a recycle ratio

0.30

0.25

3

RBCOD fraction

0 0.06

0.08

0.10

0.12

0.14

0.16

Influent TKN/COD ratio

(a)

0.06

40

0.20

0 0.5

20

1

0.25 0.30 0.35

2 10

3

Effluent nitrate (mgN I−1)

0.15

15

0.15

0.5

0.20

1 10

0.25 2 3 4

5

5

4

5 6 8 10

8 RBCOD fraction

0

0.16

0.10

a recycle ratio 0

0.10 a recycle ratio

0.10 0.12 0.14 Influent TKN/COD ratio

MLE system (settled, 22 °C) Effluent nitrate vs. TKN/COD ratio

20

30

0.08

(b)

MLE system (raw, 22 °C) Effluent nitrate vs. TKN/COD ratio

Effluent nitrate (mgN I−1)

0.20 0.15

5

8

0

0.30 0.35

6 10

RBCOD fraction

0 0.06

(c)

493

0.08

0.10 0.12 0.14 Influent TKN/COD ratio

0.06

0.16 (d)

0.08

0.10 0.12 0.14 Influent TKN/COD ratio

0.16

Figure 42 Effluent nitrate concentration vs. influent TKN/COD ratio for influent readily biodegradable (RBSO) fractions (fSb0 s) of 0.10, 0.15, 0.20, 0.25, 0.30, and 0.35 and mixed liquor a recycle ratio from 0 to 10 for the example raw (a,c) and settled (b,d) wastewaters at 14 1C (a,b) and 22 1C (c,d).

14 1C (Figures 42(a) and 42(b)) and 22 1C (Figures 42(c) and 42(d)). The same Nneaopt lines are given in Figures 41(a) and 41(c) for the raw wastewater RBSO fraction (fSb0 s) ¼ 0.25. These Nneaopt lines are calculated from Equations (169) and (170). The straight lines in Figure 42 give Nneaprac for fixed a recycle ratios at indicated values ranging from 0.0:1 to 10:1. These straight Nneaprac lines give the system performance for some selected a recycle ratio and are calculated with the aid of Equation (170) from the nitrification capacity value at the given TKN/COD ratio, fixed s recycle ratio at 1.0:1, and the selected a recycle ratio. The Nneaprac lines for a ¼ aprac ¼ 5:1 are

the same as the dotted lines in Figure 41. At the intersection points of the straight Nneaprac and curved Nneaopt lines, the system performance equals the denitrification potential and represents balanced MLE designs, that is, aopt ¼ aprac. For example, for the raw wastewater at 14 1C, at a ¼ 5.0:1 and fSb0 s ¼ 0.25, the TKN/COD ratio needs to be 0.104 to give an optimal design, that is, aopt ¼ 5:1 and at this TKN/COD ratio, Nne ¼ 8.1 mgN l1. This is the TKN/COD ratio that balances the MLE system at Rs ¼ 20 days and fxm ¼ 0.534 (see Figure 39). For TKN/COD ratioso0.104, aopt increases above 5:1, but if a is maintained at 5:1 (i.e., a ¼ aprac ¼ 5:1), then Nne

494

Biological Nutrient Removal

versus TKN/COD ratio is given by the a ¼ 5:1 straight Nneaprac line. For TKN/COD ratios 40.104, aopt decreases below 5:1, and Nne versus TKN/COD ratio is given by the curved Nneaopt (bold) line. The aopt value at a particular TKN/COD ratio is given by the a recycle ratio value of the intersection point between the vertical influent TKN/COD ratio line and the curved Nneaopt line, for example, for the example raw wastewater (fSb0 s ¼ 0.25) at 14 1C (Figure 42(a)) at a TKN/COD ratio of 0.12, aopt ¼ 2:1, and Nne is 16.0 mgN l1. The usefulness of Figure 42 is that it gives a performance evaluation of an MLE system at a specified sludge age and anoxic mass fraction for varying influent TKN/COD ratio and RBSO fraction taking due account of an upper a recycle ratio limit of aprac. For the example raw wastewater at 22 1C with a RBSO fraction (fSb0 s) of 0.10 (Figure 42(c)), the influent TKN/ COD ratio needs to be greater than 0.113 for a to beo6.0:1. If a is fixed at aprac ¼ 6.0:1 and the TKN/COD is o0.113, then the anoxic reactor is underloaded with nitrate and the denitrification potential is not achieved. The system performance for influent TKN/CODo0.113 is given by the straight Nne line for a ¼ 6:1. At influent TKN/COD ¼ 0.113, the straight Nne line for a ¼ 6:1 cuts the curved Nneaopt line, a ¼ aopt ¼ 6:1 and the system performance equals the denitrification potential. If a is maintained at 6:1 for TKN/COD 40.113, then the anoxic reactor is overloaded with nitrate and optimal denitrification is not achieved due to the unnecessarily high DO load on the anoxic reactor (similar to that shown in Figure 37(b) for a 46.7). The a recycle ratio therefore should be reduced to aopt for influent TKN/COD ratios 40.113, where aopt is given by the a value along the curved Nne line, which represents system performance equal to denitrification potential. For example, if the TKN/COD ratio ¼ 0.120, a ¼ aopt ¼ 4:1 and this a recycle ratio loads the anoxic reactor to its denitrification potential giving Nne of 12.0 mgN l1. Therefore, for TKN/COD ratio 40.113, the system performance and Nne is given by the curved Nne line provided the a recycle ratio is set to aopt, which is given by a recycle ratio line which passes through the intersection point of the vertical TKN/COD ratio line and the curved Nne line. From the above, it can be seen that only on the curved Nne line for the particular RBSO fraction is the system performance equal to the denitrification potential; also the aopt that produces this is given by the a recycle ratio line that passes through the intersection point of the vertical TKN/COD ratio line and the curved Nne line. This curved Nne line (for which a ¼ aopt) marks the boundary between underloaded and overloaded conditions in the anoxic reactor. In the domain above the curved Nne line, the anoxic reactor is underloaded (left of aopt in Figures 37 and 38) and the system performance (Nne) for a particular TKN/COD ratio is given by the intersection point of the vertical TKN/COD ratio line and the straight a recycle ratio line. In the domain below, curved Nne line, the anoxic reactor is overloaded (right of aopt in Figures 37 and 38). The Nne values obtained from this domain are not valid, but if the a recycle ratio is reduced to aopt (i.e., the a value of the intersection point of the vertical TKN/ COD ratio line and the curved Nne line), then the Nne value again is valid. Valid Nne system performance values are therefore given in Figure 42 only on or above the curved Nne boundary line.

From Figure 42, it can be seen that for MLE system at the design Rs ¼ 20 days and fxm ¼ 0.534 and a recycle ratio limited at say 5.0:1 for economical reasons, then the system is best suited to treating high TKN/COD ratios, depending on the RBSO fraction: 40.091 for fSb0 s ¼ 0.10 and 40.117 for fSb0 s ¼ 0.35. This is because with only a primary anoxic reactor, the MLE system cannot produce a low effluent nitrate concentration (o4–6 mgN l1) at a recycle ratio limited at 5.0:1. If obtaining low effluent nitrate concentrations is not required at low TKN/COD ratios, then a balanced MLE design can be selected by reducing the sludge age as demonstrated in Figures 39 and 40. If obtaining low effluent nitrate concentrations is important at low (o0.10) TKN/COD ratios, then this can be achieved at high a recyle ratios (aopt4aprac) in MLE systems or at low a recycle ratios by including a secondary anoxic reactor. Incorporation of a secondary anoxic reactor (and a re-aeration reactor for practical reasons – see Section 4.14.24.5) produces the four-stage Bardenpho system (Figure 34(c)). However, because the K3 denitrification rate is so low and needs to be reduced by at least 20% to account for the ammonia released during endogenous denitrification (which is re-nitrified in the re-aeration reactor), the net additional nitrate removal achieved in a secondary anoxic reactor is very low, too low for secondary anoxic reactors to be included in N removal systems, unless the influent TKN/COD ratio is unusually low.

4.14.27 System Volume and Oxygen Demand 4.14.27.1 System Volume Having determined the subdivision of the sludge mass into anoxic and aerobic fractions to achieve the required N removal, the actual sludge mass in the system needs to be calculated to determine the volumes of the different reactors. The mass of sludge, total (MLSS) or volatile (MLVSS), in the system for selected sludge age and wastewater characteristics for N removal system is the same as for fully aerobic (COD removal) systems. The equations given in Section 4.14.9 therefore apply to N removal systems also. For the example raw and settled wastewaters, the design parameters for the MLE system are listed in Table 16. The MLSS mass values in the system at 20 days sludge age and 14 1C are 68168 and 26 422 kgTSS, respectively. Selecting an MLSS concentration of 4500 mg l1 (4 kg m3) (see Section 4.14.11) means that the volume of the system treating raw wastewater is 15148 m3 and that treating settled wastewater is 5871 m3. Because the sludge mass in the N removal systems usually is uniformly distributed in the system, that is, each reactor of the system has the same MLSS concentration, the volume fraction of each reactor is equal to its sludge mass fraction. For the example raw and settled wastewaters at 14 1C, the volume of the anoxic reactors are 0.534 15148 ¼ 8089 m3 and 0.534  5871 ¼ 3135 m3, respectively. The nominal and actual hydraulic retention times of the anoxic and aerobic reactors are calculated from the reactor volumes divided by the nominal (influent) and total flows passing through them (Equation (59) and Table 16). Note that the reactor nominal retention time is a consequence of the mass of sludge generated from the influent COD flux, the selected MLSS concentration, and the sludge mass fraction – the retention time per se has no significance in

Biological Nutrient Removal Table 16 Design details of MLE systems treating the example raw and settled wastewaters at 14 1C at 20 days sludge and 0.534 unaerated sludge mass fraction Parameter

Raw

Settled

Influent TKN/COD ratio Influent RBCOD fraction (fSb0 s) Unaerated mass fraction(fxm) Anoxic mass fraction (fx1) Minimum anoxic fraction a Recycle ratio (apracoaopt) Sludge age (days) Effluent nitrate (Nne, mgN l1) Effluent TKN (Nte, mgN l1) Effluent total N (Nne þ Nte) System vol at 4.5 gTSS l1 (m3) Anoxic volume (m3) System ret time – nom (h) Aerobic ret time – nom (h) Aerobic ret time – actual (h) Anoxic ret time – nom (h) Anoxic ret time – actual (h) Carb. O2 demand (FOc, kgO d1) Nit O2 demand (FOn, kgO d1) O2 recovered (FOd, kgO d1) Tot. O2 demand (FOtd, kgO d1) %N removal Mass TSS wasted (FXt, kg d1) Active fraction wrt TSS (fatOHO)

0.08 0.25 0.534 0.534 0.07 5:1 20 5.6 3.8 9.4 15148 8089 24.2 11.2 1.6 12.9 1.85 6679 2685 1440 7924 84.4 3408 0.23

0.113 0.385 0.534 0.534 0.105 5:1 20 5.7 3.8 9.5 5871 3135 9.4 4.4 0.63 5 0.72 4311 2719 1458 5572 81.4 1321 0.383

kinetics of and design for nitrification and denitrification (see Section 4.14.9.3).

4.14.27.2 Daily Average Total Oxygen Demand The total oxygen demand in a nitrogen removal system is the sum of that required for organic material (COD) degradation and nitrification, less than recovered by denitrification. The daily average oxygen demand for (1) organic material removal (FOc) is given by Equations (111) and (2) nitrification is given by Equation (155). These oxygen demands in the MLE system at 20 days sludge age for the example raw and settled wastewaters at 14 and 22 1C are 9364 and 7030 kgO d1 (Table 16). The oxygen recovered by denitrification (FOd) is given by 2.86 times the nitrate flux denitrified (Section 4.14.24.2) where nitrate flux denitrified is the product of the daily average influent flow Qi and the nitrate concentration denitrified. The nitrate concentration denitrified is given by the difference in the nitrification capacity Nc and the effluent nitrate concentration. Hence,

FOd ¼ 2:86ðNc  Nne ÞQi

ðmgO d1 Þ

495

demand by incorporating ND is only 20% of that required for COD removal only, and (4) the effect of temperature on the total oxygen demand is marginal – less than 3% (see also Figure 32). For the settled wastewater, Table 16 shows that (1) the nitrification oxygen demand is about 63% of that required for COD removal; (2) about 54% of the nitrification oxygen demand can be recovered by denitrification; (3) the additional oxygen demand by incorporating nitrification and denitrification is about 30% of that required for COD removal only, and (4) the effect of temperature on the total oxygen demand is marginal – less than 3% more at the lower temperature. Comparing the total oxygen demand (FOtd) for the raw and settled wastewaters, the total oxygen demand for the latter is about 30% less than that of the former. This saving is possible because primary sedimentation removes 35–45% of the raw wastewater COD. Furthermore, for the settled wastewater, the nitrification oxygen demand is a greater proportion of the total; also, less of the nitrification oxygen demand can be recovered by denitrification compared to the raw wastewater. These effects are due to the higher TKN/COD ratio of the settled wastewater. Knowing the average daily total oxygen demand, (FOtd) the peak total oxygen demand can be roughly estimated by means of a simple design rule (Musvoto et al., 2002). From a large number of simulations with AS model no. 1 (ASM1), it was found that, provided the factor of safety on nitrification (Sf) is greater than 1.25–1.35, the relative amplitude (i.e., (peak  average)/average) of the total oxygen demand variation is a fraction 0.33 of the relative amplitude of the TOD of the influent COD and TKN load (i.e., Qi(Sti þ 4.57Nti)). For example, with the raw wastewater case, if the peak influent TOD flux is obtained at a time of day when the influent flow rate, COD and TKN concentrations are 25 M l d1, 1250 mgCOD l1 and 90 mgN l1, respectively – that is, 25(1250 þ 4.57  90) ¼ 41 532 kgTOD d1, and the average influent TOD flux is 15(750 þ 4.57  60) ¼ 15 363 kgTOD d1, the amplitude of the total influent TOD flux is (41 532 15 363)/15 363 ¼1.70; hence, the amplitude of the total oxygen demand is approximately 0.33 1.70 ¼ 0.56; from Table 16 the average daily total oxygen demand (FOtd) is 7924 kgO d1 and hence the peak oxygen demand is (1 þ0.56)  7924 ¼12 378 kgO d1. As with all simplified design rules, the above rule should be used with discretion and caution, and where possible, the peak total oxygen demand is best estimated by means of the AS simulations models.

4.14.28 Biological Excess Phosphorus Removal ð172Þ

From the denitrification performance of the MLE system in Table 16, the oxygen recovered by denitrification for the example raw and settled wastewaters at 14 1C are 1440 and 1458 kgO d1. For the raw wastewater, Table 16 shows that (1) the nitrification oxygen demand (FOn) is about 40% that required for COD removal (FOc), (2) about 55% of FOn can be recovered by incorporating denitrification, (3) the additional oxygen

4.14.28.1 Introduction Phosphorus is the key element in aquatic environments that limits the growth of aquatic plants and algae controls eutrophication. Unlike nitrogen that can be fixed from the atmosphere which contains about 80% nitrogen gas, phosphorus can only come from upstream of aquatic systems (neglecting atmospheric deposition). Diffuse sources of phosphorus, for example, from agricultural fields, are best controlled by proper fertilization plans, while point sources of

Biological Nutrient Removal

phosphorus, for example, from WWTPs, can be removed by chemical or biological processes. Considering the benefit to aquatic environments, strict regulations are being applied for phosphorus removal from wastewaters. Considering the potential benefits of removing phosphorus biologically rather than chemically, along with organic matter and nitrogen from wastewater, BEPR has stimulated much interest in the study of the biochemical mechanisms, the microbiology of the systems, the process engineering and optimization of plants, and in mathematical modeling. Reviews of the development of BEPR have been regularly published over the years (Marais et al., 1983; Arvin, 1985; Wentzel et al., 1991; Jenkins and Tandoi, 1991; van Loosdrecht et al., 1997; Mino et al., 1998; Blackall et al., 2002; Seviour et al., 2003; Oehmen et al., 2007). This section briefly reviews the mechanisms of BEPR, outlines the practical systems to achieve it, summarizes some of the experimental research that led to the development of BEPR models (both steady state and dynamic kinetic), discusses the impact of anoxic zones for denitrification on BEPR, and sets out guidelines for design of NDBEPR systems. In order not to unduly complicate this, the concepts are presented for strictly aerobic phosphorus accumulating organisms (aerobic PAOs) which can use only oxygen as the electron acceptor for energy production. Considering that some denitrifying PAOs (DPAOs) exist and may have a significant impact on the performance of the process, their influence is discussed where appropriate, but is not included in the models described.

4.14.28.2 Principles of BEPR BEPR is the biological uptake and removal of P by AS systems in excess of the amount that is removed by normal completely aerobic AS systems. This is in excess of the normal P requirements for growth of AS. In the completely aerobic AS system, the amount of P typically incorporated in the sludge mass is about 0.02 mgP/mgVSS (0.015 mgP/mgTSS). By the daily wastage of surplus sludge phosphorus is thus effectively removed. This can give a P removal of about 15–25% of the P in many municipal wastewaters. In an BEPR AS system, the amount of P incorporated in the sludge mass is increased from the normal value of 0.02 mgP/mgVSS to values around 0.06–0.15 mgP/ mgVSS (0.05–0.10 mgP/mgTSS). This is achieved by system design or operational modifications that stimulate, in addition to the OHOs present in AS, the growth of organisms that can take up large quantities of P and store them internally in long chains called polyphosphates (polyPs); generically, these organisms are called phosphate accumulating organisms (PAOs). PAOs can incorporate up to 0.38 mgP/mgVSS (0.17 mgP/ mgTSS). In the biological P removal system both the OHOs, which do not remove P in excess, and the PAOs coexist. The larger the proportion of PAOs that can be stimulated to grow in the system, the greater the P content of the AS and, accordingly, the larger the amount of P that can be removed from the influent. Thus, the challenge in design is to increase the amount of the PAOs relative to the OHOs present in the AS as this will increase the capacity for P-accumulation and thereby high phosphorus removal efficiency. The relative proportion of the two organism groups depends, to a large degree, on the fraction

15 Example settled WW % P of VSS (mgP/mgVSS as %)

496

Settled WW 10 Example Raw WW Raw WW 5

P removal =

%P × VSS mass Sludge age × Q i

0 0

10 20 30 % Bio COD obtained by PAOs

40

Figure 43 Percentage P (mgP/mgVSS  100) in VSS mass vs. the proportion of biodegradable COD mass (as %) obtained by PAOs.

of the influent wastewater biodegradable COD that each organism group obtains. The greater the fraction of PAOs in the mixed liquor, the greater the %P content of the AS and the greater the BEPR (Figure 43). Design and operational procedures are oriented toward maximizing the growth of PAOs. In an appropriately designed BEPR system, the PAOs can make up about 40% of the active organisms present (or 15% of VSS; 11% of TSS), and this system can usually remove about 10–12 mgP per 500 mg influent COD l1. From the first publications reporting enhanced P removal in some AS systems, there has been some controversy as to whether the mechanism is a precipitation of inorganic compounds, albeit perhaps biologically mediated, or biological through formation and accumulation of P compounds in the organisms. The objective here is not to discuss the evidence that supports the biological nature of enhanced P removal, but to briefly describe the theory of biological P removal and to demonstrate how this theory can be used as an aid for the design of biological P removal AS systems. This does not imply that precipitation of P due to chemical changes resulting from biological action (e.g. alkalinity and pH) does not take place. Although inorganic precipitation of P can certainly take place, it would appear that in the treatment of municipal wastewaters by an appropriately designed AS system, within the normal ranges of pH, alkalinity and calcium concentrations in the influent, enhanced P removal is principally mediated by a biological mechanism (Maurer et al., 1999; de Haas et al., 2000). These mechanisms are described below.

4.14.28.3 Mechanism of BEPR 4.14.28.3.1 Background Historically, several research groups have made a number of important contributions toward elucidating the mechanisms

Biological Nutrient Removal

4.14.28.3.2 Biological P removal microorganisms The basic requirement for BEPR is the presence in the AS system of microorganisms which can accumulate P in excess of normal metabolic requirements, in the form of polyP stored in granules called volutins. In the BEPR models, all organisms in the AS system accumulating polyP in this fashion and exhibiting the classical observed BEPR behavior – anaerobic P release, aerobic P uptake, and associated processes – are lumped together and represented by the generic PAO group. PolyPs can be accumulated by a wide range of bacteria. In general, they are accumulated as a phosphate reserve in relatively low amounts. Only very few types of bacteria seem to be able to harvest the energy that is stored in polyPs to take up VFAs and store them as PHAs under anaerobic conditions (in the absence of an external electron acceptor such as oxygen or nitrate). In the original research on BEPR microbiology conducted with cultivation studies, it was incorrectly considered that PAOs were of the genus Acinetobacter (Fuhs and Chen, 1975; Buchan, 1983; Wentzel et al., 1986), Microlunatus phosphovorus (Nakamura et al., 1995), Lampropedia (Stante et al., 1997), and Tetrasphaera (Maszenan et al., 2000). More recently, cultureindependent methods have shown that Accumulibacter phosphatis, a member of the genus Rhodocyclus (a beta proteobacterium), is a PAO which can be grown in enriched cultures (at up to 90% purity, as shown by fluorescence in situ hybridization (FISH) molecular probes) but not yet in axenic cultures (Wagner et al., 1994; Hesselmann et al., 1999; Crocetti et al., 2000; Martin et al., 2006; Meyer et al., 2006; Oehmen et al., 2007). From a modeling and design perspective, however, the identification of the exact organisms responsible for BEPR is of minor importance, although this may provide information that can be used to refine the models and design procedures; these are not based on the behavior of specific organisms, but rather on the observed behavior of groups of organisms identified by their function, in this case the PAOs.

4.14.28.3.3 Prerequisites To achieve BEPR in AS systems, the growth of organisms that accumulate polyP (PAOs) has to be stimulated. To accomplish this, two conditions are essential: (1) an anaerobic and aerobic (or anoxic) sequence of reactors/conditions and (2) the addition or formation of VFAs in the anaerobic reactor/period.

Glycogen

Concentration

of BEPR, including Shapiro et al. (1967), Fuhs and Chen (1975), Nicholls and Osborn (1979), Rensink et al. (1981), Marais et al. (1983), Lotter (1985), Comeau et al. (1986), Wentzel et al. (1986, 1991), Mino et al. (1987, 1994, 1998), Kuba et al (1993), Smolders et al. (1994a, 1994b, 1995), van Loosdrecht et al. (1997), Maurer et al. (1997), Seviour et al. (2003), Martin et al. (2006), and Oehmen et al. (2007). In this section, an explanation of the basic concepts underlying the more sophisticated mechanistic models for the biological P removal phenomenon is presented. For detailed description of the mechanisms, the reader is referred to the references mentioned earlier in this paragraph.

497

Poly P

PHA VFA Anaerobic

PO4 Aerobic

Figure 44 Schematic diagram showing the changes as a function of time in concentrations of volatile fatty acids (VFAs), P (PO4), polyphosphates (polyPs), polyhydroxyalkanoate (PHA), and glycogen through the anaerobic–aerobic sequence of reactors in a BEPR system.

4.14.28.3.4 Observations With the prerequisites for BEPR present, the following observations have been made at full, pilot, and laboratory scale (Figure 44). Under anaerobic conditions, bulk solution VFAs and intracellular polyP and glycogen decrease and soluble phosphate, Mg2þ, Kþ, and intracellular poly-b-hydroxyalcanoates (PHAs) increase (Randall et al., 1970; Rensink et al., 1981; Hart and Melmed, 1982; Fukase et al., 1982; Watanabe et al., 1984; Arvin, 1985; Hascoe¨t et al., 1985a; Wentzel et al., 1985; Comeau et al., 1986, 1987; Murphy and Lo¨tter, 1986; Gerber et al., 1987; Wentzel et al., 1988; Satoh et al., 1992; Smolders et al., 1994a; Maurer et al., 1997). Under aerobic conditions; intracellular polyP and glycogen increase; soluble phosphate, Mg2þ, Kþ, and intracellular PHA decrease (Fukase et al., 1982; Arvin, 1985; Hascoe¨t et al., 1985a; Comeau et al., 1986; Murphy and Lo¨tter, 1986; Gerber et al., 1987; Wentzel et al., 1988; Satoh et al., 1992; Smolders et al., 1994b; Maurer et al., 1997).

4.14.28.3.5 Biological P removal mechanism In describing the mechanisms of BEPR, a clear distinction is made between the PAOs and OHOs. In the anaerobic/aerobic sequence of reactors, it is considered that VFAs are present in the influent waste stream entering the anaerobic reactor or produced in the anaerobic reactor by fermenting organisms (accepted to be the OHOs in models). In the anaerobic reactor (zero nitrate and oxygen in or entering reactor), the OHOs cannot utilize the VFAs due to the absence of an external electron acceptor, oxygen or nitrate. The PAOs, however, can take up the VFAs from the bulk liquid and store them internally by linking the VFAs together to form complex long-chain carbon molecules of poly-b-hydroxyalkanoates (PHAs). The two common PHAs are poly-b-hydroxybutyrate (PHB: four-carbon compound synthesized from two acetate molecules) and polyhydroxyvalerate (PHV: five-carbon compound from one acetate and one propionate molecules) (Figure 45(a)). Forming PHAs from the VFAs requires energy for three functions: active transport of VFAs across the cell membrane, energization of VFAs into coenzyme A compounds (e.g.,

498

Biological Nutrient Removal Liquid Cell

PHA e−

Glycogen

VFA-CoA Energy

VFA

(PO4)n

VFA

(PO4)n−1 Pi

PAO

Pi

(a)

Liquid CO2

CO2

Catabolism PHA

Cell

e− Glycogen ETC

Anabolism

Energy H2O

New cells

O2 (PO4)n Pi

(b)

(PO4)n−1 PAO

Pi

Figure 45 (a) Simplified biochemical model for PAOs under anaerobic conditions. Anaerobic uptake of volatile fatty acids (VFAs), originating from the influent or from fermentation in the anaerobic reactor, and storage of polyhydroxyalkanoates (PHAs) by the PAOs with associated P release. (b) Simplified biochemical model for PAOs under aerobic conditions. Aerobic utilization of PHAs and growth of PAOs, with P uptake by existing and new PAOs.

acetyl-CoA) and reducing power (NADH) for PHA formation. PolyP degradation is associated with the formation of ADP from AMP, with the phosphokinase enzyme 2 ADP are converted to adenosine triphosphate (ATP) and adenosine monophosphate (AMP) (van Groenestijn et al., 1987). When ATP is used, orthophosphates are released and accumulate in the cell interior together with the counterions of polyP (potassium and magnesium). The efflux of these compounds might be related to building a proton motive force, which either can help in the uptake of acetate or in the generation of a small amount of extra ATP. It is observed (Smolders et al., 1994a) that the energy requirements for acetate uptake increase with increasing pH. This can be associated with the fact that the energy needed for acetate transport increases with pH. ATP is used, notably, for the energization of acetate and propionate into acetyl-CoA and propionyl-CoA. Glycogen degradation also results in ATP formation, NADH production, and intermediates that are

transformed into acetyl-CoA (or propionyl-CoA). Finally, acetyl-CoA and propionyl-CoA are stored as PHA (Comeau et al., 1986; Wentzel et al., 1986; Mino et al., 1998; Smolders et al., 1994b; Martin et al., 2006; Oehmen et al., 2007; Saunders et al., 2007). Thus, the PAOs in the anaerobic reactor have taken up for their exclusive use the VFAs under anaerobic conditions where the OHOs are unable to use these organics. To accomplish this, some of the stored polyP has been consumed and P released to the bulk solution. To stabilize the negative charges on the polyP, the cations Mg2þ, Kþ, and sometimes Ca2þare complexed, which add to the inorganic settleable solids (TSS) in the system (Ekama and Wentzel, 2004). When polyPs are consumed and P is released, mainly Mg2þ and Kþ cations are released in the approximate molar ratio P:Mg2þ:Kþ of 1:0.33:0.33 (Comeau et al., 1987; Brdjanovic et al., 1996; Pattarkine and Randall, 1999). In the subsequent aerobic reactor (presence of DO). In the presence of dissolved oxygen (or of nitrate under anoxic conditions) as an external electron acceptor, the PAOs utilize the stored PHA as a carbon and energy source for energy generation and growth of new cells as well as for regenerating the glycogen consumed in the anaerobic period. The stored PHA is also used as an energy source to take up P from the bulk solution to regenerate the polyP used in the anaerobic reactor, and to synthesize polyP in the new cells that are generated – P uptake (Figure 45(b)). The uptake of P to synthesize polyP in the new cells generated means that more P is taken up than is released in the anaerobic reactor, giving a net removal of P from the liquid phase in the AS system. Accompanying the P uptake, the cations Mg2þ and Kþ also are taken as countercharge for the negatively charged polyP polymer, in the approximate molar ratio P:Mg2þ:Kþ of 1:0.33:0.33. The PAOs, with stored polyP, are removed from the aerobic reactor of the system (where the internally stored polyP concentration in the PAOs is the highest in the system) via the waste sludge stream (wastage from the underflow recycle stream is possible, but not desirable for hydraulic control of sludge age; see Section 4.14.14). At steady state the mass of PAOs wasted per day (with stored polyP) equals the mass of new PAOs generated per day (with stored polyP). Thus, for a fixed sludge age, loading, and system operation, the mass of PAOs in the biological reactors remains constant, so that in the AS system at steady state there is neither a buildup nor a loss of PAOs, and the P/VSS ratio stays approximately constant. The mass of new PAOs formed depends on the mass of stored substrate (PHA) available to the PAOs. Accordingly, the enhanced P removal attained will depend on the mass of PHA stored in the anaerobic reactor.

4.14.28.3.6 Fermentable COD and slowly biodegradable COD As indicated above, under anaerobic conditions, PAOs can take up and store VFAs. However, some wastewaters contained very little VFAs, yet exhibited significant BEPR. This was ascribed to the influent RBOs, (Sbsi) which comprises both VFAs (Sbsai) and fermentable RBO (FBSO, Sbsfi) (Siebritz et al., 1983; Wentzel et al., 1985, 1990; Nicholls et al., 1985; Pitman et al., 1988; Randall et al., 1994). This influent FBSO is

Biological Nutrient Removal Liquid Cell F-RBCOD

F-RBCOD

Energy

VFA OHO

VFA

VFA

PAO

Figure 46 Simplified biochemical model for fermentation of RBSO to VFA by OHOs under anaerobic conditions – VFAs released by OHOs are taken up by PAOs.

fermented to VFAs by the OHOs in the anaerobic reactor, the VFAs becoming available for uptake and storage by the PAOs because the OHOs cannot utilize them due to the absence of an electron acceptor (NO3 or O) (Figure 46). Slowly biodegradable organics (SBO, XS), even though these can be hydrolyzed into RBO under anaerobic conditions, has been shown not to be linked to anaerobic phosphate release. This aspect is of crucial importance as it will influence both the design and operation of BNR systems, such as sizing and determining the number of anaerobic reactors, inclusion of primary sedimentation and maximum BEPR achievable. For the purpose of the BEPR models, the experimental evidence linking BEPR to the RBO is accepted, but a conversion of SBO to RBO is considered to be small enough to be negligible. Accordingly, where VFA production does occur, this will essentially be from the RBO. One exception to this consideration is when primary sludge is fermented in a separate fermentation reactor upstream of the anaerobic reactor – in these dedicated fermenters, some hydrolysis of SBO to RBO and VFAs takes place to augment the influent VFA and RBO concentrations (Lilley et al., 1992).

4.14.28.3.7 Functions of the anaerobic zone From the description of the mechanisms above, with normal domestic wastewater as influent, the anaerobic zone/reactor serves two functions: (1) it stimulates conversion of fermentable organics to VFAs by OHOs, that is, facultative acidogenic fermentation and (2) because it is not possible for the OHOs to metabolize the VFAs (no external electron acceptor), the PAOs take up the released VFAs and store them as PHA. Thereafter, the PAOs do not have to compete for substrate when an external electron acceptor becomes available in the aerobic (or anoxic) zone. Of the above two processes, the former is the slower and determines the size of the anaerobic reactor (Wentzel et al., 1985, 1990). Should primary sludge fermentation be implemented at the treatment plant, the first process would not be needed as much and the size of the anaerobic reactor could be decreased.

499

4.14.28.3.8 Influence of recycling oxygen and nitrate to the anaerobic reactor Numerous investigators (e.g., Barnard, 1976; Venter et al., 1978; Siebritz et al., 1980; Hascoe¨t and Florentz, 1985b) noted that the recycling of oxygen and/or nitrate to the anaerobic reactor causes a corresponding decrease in BEPR. In terms of the mechanisms described above, if oxygen and/or nitrate is recycled to the anaerobic reactor, the OHOs are able to utilize the fermentable COD for energy and growth using the oxygen or nitrate as external electron acceptor. For every 1 mgO recycled to the anaerobic reactor 3 mgCOD of fermentable RBO are consumed and for every 1 mgN of nitrate recycled 8.6 mgCOD of fermentable RBO are consumed by the OHOs. The ratio of 3 mgCOD/ mgO consumed comes from the catabolic oxygen requirement in organics utilization (i.e., 1/(1  fcvYH)E3) (Equation (46)). Similarly, considering that 1 mgNO3-N is equivalent to 2.86 mgO (Section 4.14.24.2), a ratio of 2.86/ (1  fcvYH)E8.6 mgCOD consumed by mgNO3-N reduced is obtained. The fermentable RBOs metabolized by the OHOs are not released to the bulk liquid as VFAs. Therefore, the amount of VFAs generated and released to the bulk liquid is reduced by the amount of RBO consumed by the OHOs. Consequently, the mass of VFAs available to the PAOs for storage is reduced, and correspondingly so is the P release, P uptake, and the net P removal. Should the influent RBO already consist of VFAs and oxygen and/or nitrate be recycled, the PAOs and OHOs will compete for the VFAs, the PAOs to take up the VFAs, and the OHOs to metabolize it. Accordingly, even in this situation recycling of oxygen and/or nitrate will reduce the BEPR. Thus, preventing the recycling of oxygen and nitrate to the anaerobic reactor is one of the primary considerations in the design and operation strategy for BEPR systems.

4.14.28.3.9 Denitrification by PAOs The extent of denitrification with associated anoxic P uptake by the PAOs appears to be highly variable (Ekama and Wentzel, 1999b), from near-zero anoxic P uptake (e.g., Wentzel et al., 1989a, Clayton et al., 1989, 1991) to anoxic P uptake dominant over aerobic P uptake (e.g. Sorm et al., 1996; Hu et al., 2000). Experimental evidence tends to suggest that magnitude of anoxic P uptake is influenced by the anoxic mass fraction and the mass of nitrate loaded on the anoxic reactor relative to its denitrification potential (Hu et al., 2002). For the purpose of design it will be considered that anoxic P uptake is not significant. Anoxic P uptake decreases the magnitude of P removal in the system (Ekama and Wentzel, 1999a, 1999b; Hu et al., 2002), and from a design point of view in which maximizing P removal is a priority, anoxic P uptake should be avoided in the system. Hence, in this chapter, anoxic P uptake will not be considered. It must be emphasized, however, that due to the anaerobic conversion of RBO to VFA which are taken up by PAOs, the kinetics of denitrification in the subsequent anoxic reactor change compared with that in the primary anoxic reactor of an MLE system.

500

Biological Nutrient Removal

4.14.29 Principles of Maximizing BEPR The principles of maximizing BEPR can be grouped into seven categories. A number of configurations or systems that are based on these principles are identified by specific names (Figure 47). 1. Oxygen entrainment in the anaerobic reactor should be minimized. For this purpose, mixing vortexes, upstream cascades, and screw pumps or air lift pumps should be avoided.

2. Nitrate (and nitrite) entering in the anaerobic reactor should be minimized. A number of named configurations were developed precisely for this purpose (Section 4.14.34). Based on observations a number of laboratory-, pilot- and full-scale systems (Barnard, 1974, 1975a, 1975b; Nicholls, 1975b), to achieve BEPR in the simplest configuration, Barnard (1976) proposed the Phoredox system (Figure 47(a) also known as the A/O process). This system comprises only an anaerobic and aerobic reactor and is intended not to nitrify to avoid nitrate entering the

Primary anoxic reactor

Anaerobic reactor Anaerobic reactor

Aerobic reactor

Mixed liquor recycle a Waste flow

Waste flow Settler Effluent

Influent

Settler Effl.

Infl.

Sludge recycle s

Sludge recycle s

(a)

(b)

Mixed liquor recycle a

Primary anoxic reactor

Anaerobic reactor

Primary Secondary Aerobic Reaeration anoxic anoxic reactor reactor reactor reactor

Anaerobic reactor

Aerobic reactor

r Recycle a Recycle Waste flow

Waste flow

Settler Effl.

Infl.

Settler Effl.

Infl.

Sludge recycle s

Sludge recycle s

(c)

(d) Primary anoxic reactor

Anaerobic reactor Anaerobic reactor

1st primary 2nd primary anoxic anoxic reactor reactor

r Recycle

Waste flow

a Recycle

Effl.

Sludge recycle s (e)

(f)

r Recycle

Settler Effl.

Infl.

Settler

Infl.

1st prim. Anaerobic anoxic reactor reactor

Aerobic reactor

Mixed liquor recycle a

Aerobic reactor Waste flow

2nd prim. anoxic reactor

Anoxic or aerobic Aerobic reactor reactor

a Recycle

PST

Aerobic reactor

SST Effluent

1° Solids

Waste

Phostrip II

Settler

Infl.

Sludge recycle s

Underflow anoxic reactor

Influent

Waste flow Effl.

Supernatant

Sludge recycle s (g)

Aerobic reactor

Mix Stripper tank Lime

Biospherics, USA

(h)

Figure 47 System configurations for BEPR: (a) phoredox or A/O; (b) three-stage Bardenpho or (A2/O); (c) five-stage Bardenpho; (d) University of Cape Town (UCT) or Virginia Initiative Process (VIP); (e) modified UCT; (f) Johannesburg (JHB); (g) biological chemical flexible system (BCFS); and (h) phostrip.

Biological Nutrient Removal

anaerobic reactor. Because nitrification can take place even at short sludge ages, particularly in warm climates, one or more anoxic reactors for denitrification are included in the two-reactor anaerobic–aerobic system to protect the anaerobic reactor from nitrate entering it. The position of the anoxic reactor(s) has led to a number of different configurations: (1) one between the anaerobic and aerobic reactors with the return sludge discharged to the anaerobic reactor (three-stage Bardenpho or A2/O systems, Figure 47(b)), (2) anoxic reactors before and after the aerobic reactor with the return sludge discharged to the anaerobic reactor (five-stage Bardenpho, Figure 47(c)), (3) one or two anoxic reactors between the anaerobic and aerobic reactors with the sludge return discharged to the first or only anoxic reactor (UCT, Siebritz et al., 1980 or VIP, Daigger et al., 1987; Figure 47(d) and modified UCT systems; Figure 47(e)), and (4) an anoxic reactor between the anaerobic and aerobic reactors and another in the sludge return flow (JHB system; Figure 47(f)). 3. VFA uptake by PAOs in the anaerobic reactor should be maximized. Primary sludge fermentation is an efficient way to increase the VFA content of the influent even though it also contributes to an increased loading in organic matter and ammonia to the AS system. Sodium acetate or fermentable industrial wastes can be added directly to the anaerobic reactor or industries that produce fermentable organics (e.g., breweries or food processing factories) should not be penalized for discharging their high RBO containing wastewater to the sewer. The sludge mass fraction of the anaerobic reactor can be increased to favor in situ fermentation of the influent or added fermentable organic matter. 4. Effluent particulate phosphorus should be minimized by removing TSSs efficiently. The particulate phosphorus content can reach as high as 18% gP/gTSS for enriched cultures. With a more typically 5–10% P content for municipal wastewater (Figure 43), every 10 mgTSS l1 in the effluent will contribute 0.5 to 1 mgP l1. Thus, efficient secondary clarification, avoiding floating sludge from denitrification in the settling tank, sand filtration, or even ultrafiltration (in a membrane bioreactor) are means of reducing the effluent TSS concentration. 5. Effluent soluble phosphorus should be minimized. Besides optimizing the BEPR process, chemical coagulants such as iron (e.g., FeCl3), aluminum (e.g., alum), or calcium (e.g., lime) salts can be added in the mainstream for pre-, co-, or post-precipitation (in the primary settling tank, in the AS process, downstream of the secondary settling tank, respectively, de Haas et al., 2001). Extracting the supernatant from the anaerobic tank or taking some sludge from the return AS and coagulating them can also lead to lower effluent soluble phosphorus (Sehayek and Marais, 1981; van Loosdrecht et al., 1998; e.g., BCFS process; Figure 47(g)). Sidestream lime precipitation of phosphate released anaerobically from the return sludge can also be done. More efficient phosphate release can be achieved in this sidestream tank by diverting some influent containing readily biodegradable COD (e.g., PhoStrip process, Figure 47(h)). These systems support the biological

501

process by phosphate stripping and potential P recovery in the main line, stabilizing the sludge settling properties and optimizing the control of nitrogen removal. In the BCFS system, a third recycle is added from the aerated reactor to the first anoxic reactor in order to maximize denitrification and to be able to aerate the second anoxic reactor during peak flows or cold temperatures. In this way both ammonium and nitrate can be better controlled to low effluent values (ammonium typical below 0.5 gN l1 and nitrate around 5–8 mg N l1). The recycle flows are controlled by a simple redox electrode-based controller (van Loosdrecht et al., 1998). Compartmentalizing the reactors and low effluent ammonia concentration contributes to a stable low SVI – around 120 ml g1 (Kruit et al., 2002; Tsai et al., 2003). Biological phosphorus removal can be supplemented by addition of precipitants to the anaerobic tank. Since phosphate concentrations are high in this tank, the precipitants are used effectively. Dosing chemicals, however, should be done carefully. Too much precipitation will make the phosphate unavailable for PAOs and deteriorate the BEPR efficiency (de Haas et al., 2001). A complicating factor is that the WWTP will respond rapidly to changes in addition of chemicals whereas the biological phosphorus removal process might have a response time of several days if not weeks. In the BCFS process, a small baffle is placed at the end of a plugflow anaerobic tank. The sludge will locally settle back into the anaerobic tank and a clear supernatant can be withdrawn for phosphate precipitation. The phosphorus can then be recovered (Barat and van Loosdrecht, 2006) or the chemical sludge produced can be prevented from accumulating in the AS which would limit the overall capacity of the plant by reducing the sludge age. Should anaerobic or aerobic digestion be performed with the wasted secondary sludge, essentially all of the polyPs will be hydrolyzed to ortho-P and the phosphate released in solution (Jardin and Po¨pel, 1994; Harding et al., 2009; Mebrahtu et al., 2010). Phosphorus recovery in the form of struvite (MgNH4PO4) or hydroxyapatite (Ca10(PO4)6OH2), which can be used as fertilizers, are also means of reducing the loading of soluble phosphate back to the AS process and, eventually, to the effluent. 6. Phosphorus uptake for cell synthesis should be maximized. Although more limited than the other maximization principles in its potential efficiency, maintaining the sludge age as short as possible will result in an increase in phosphorus removal by sludge production (cell synthesis). Although the endogenous respiration rate of the PAOs is low (0.04 d1), another small benefit of reducing the sludge age is that the PAOs degrade to a lower extent their polyP reserves for cell maintenance. 7. Because anoxic P uptake BEPR reduces the P content of the PAOs (Ekama and Wentzel, 1999a, b; Hu et al., 2002), growth of denitrifying PAOs should be avoided to maximize aerobic P uptake BEPR to maximize PAO P content – up to 0.38 mgP/mgPAOVSS (Wentzel et al., 1989b, 1990). For a review of how these developments took place, the reader is referred to Henze et al. (2008). In order to efficiently construct all the tanks in these complex BNR systems, it is possible

502

Biological Nutrient Removal

Figure 48 Zandvliet nitrification–denitrification (ND)BEPR WWTP Cape Town, South Africa. By arranging interconnecting recycle flows between the anaerobic (centre), anoxic (inner ring), and aerobic (outer ring) reactors, the system has the flexibility to be operated as a UCT, threestage Bardenpho or JHB system. Photo: GA Ekama.

to shift the construction from rectangular tanks to one round tank with a sloping outside wall divided in rings for the different aerobic/anoxic/anaerobic zones. In this way, the amount of concrete needed is minimized as all the walls require much less strength (Figure 48).

4.14.30 Model Development for BEPR 4.14.30.1 Early Developments When the first mainstream NDBEPR system was proposed (the five-stage Bardenpho, Figure 47(a); Barnard, 1976), initial conceptualization of the phenomenon extended little beyond recognition of (1) the necessity of an anaerobic/aerobic sequence of reactors and (2) the adverse influence of nitrate recycled to the anaerobic zone. With the inclusion of the secondary anoxic reactor, it was believed that nearly complete denitrification of nitrate would be achieved, thereby discharging very low nitrate concentrations to the anaerobic reactor. Design procedures were based on empirically based estimates for sizing denitrification and anaerobic reactors in terms of nominal hydraulic retention time, and sizing of the anaerobic reactor appeared to be linked to depression of the redox potential below some critical value. No rational method for predicting N and P removal was available and for design, removals were estimated largely from experience gained in operating experimental systems similar to the proposed systems (McLaren and Wood, 1976; Simpkins and McLaren, 1978; Osborn and Nicholls, 1978).

4.14.30.2 RBO and Anaerobic Mass Fraction In seeking an explanation for the different P release and enhanced P removal behavioral patterns in lab-scale modified UCT (Figure 47(e)) and MLE (Figure 34(b)) systems, Siebritz et al. (1980, 1983) applied the concept of RBO developed in denitrification and aerobic studies (Dold et al., 1980; van Haandel et al., 1981) to BEPR systems. They noted that the only evident difference between the modified UCT and MLE

systems lay in the concentration of RBO surrounding the organisms in the anaerobic reactor. (They also observed that the UCT and MLE systems with the same anoxic mass fractions yielded approximately the same effluent nitrate concentrations and the ND kinetic models (such as ASM1, Henze et al., 1987 or UCTOLD, Dold et al., 1991) predicted the NDBEPR system response reasonably well even at full scale (Nicholls, 1982). This implied that the anaerobic reactor did not appear to have a detrimental effect on the denitrification (the questions this raises regarding denitrification in NDBEPR systems are discussed in Section 4.14.34.) In the modified UCT system the RBO concentration in the anaerobic reactor is the maximum possible as no nitrate is recycled to the anaerobic reactor; in contrast, in the MLE system sufficient nitrate is recycled to the anoxic reactor to utilize all the RBO. Therefore, the different behavioral patterns of the systems would be consistently described if it is assumed that the concentration of RBO from the influent in the anaerobic reactor surrounding the organisms is a key parameter determining whether or not P release and BEPR take place. (Later it became clear that the parameter influent RBO concentration in the anaerobic reactor surrounding the organisms represented the influent and produced VFAs taken up by the PAOs in the anaerobic reactor.) The validity of this RBO hypothesis was established by Siebritz et al. (1983) at laboratory scale and Nicholls et al. (1985) at full scale, who found that the magnitude of the P release was proportional to the influent RBO concentration. This opened the way for enquiry into other factors affecting the P release and the BEPR and quantifying BEPR. It was concluded that the BEPR depended on two main parameters, viz. (1) influent RBO concentration and (2) the anaerobic sludge mass fraction. Testing the concepts of the parametric model did, in general, demonstrate the utility of the model. At laboratory scale, the concepts were tested in the modified UCT system at different sludge ages, temperatures, anaerobic mass fractions, and influent COD concentrations in which the RBSO fraction of the influent (unsettled municipal sewage) was augmented by the addition of glucose or acetate. Based on the influent RBO concentration and anaerobic mass fraction parameters, the predicted P removal compared quite consistently with the measured P removal. At full scale, evaluation of the Goudkoppies and Northern Works WWTPs with the parametric model provided a consistent explanation when good or poor P removal was obtained (Nicholls et al., 1985; 1986; 1987). Thus, the parametric model allowed some quantitative approach to design of N and P removal plants and provided a basis for evaluating the performance of existing plants (Ekama et al., 1983). This parametric BEPR model, as well the organics removal, ND models presented earlier in this chapter, were published in the NDBEPR system design guide (WRC, 1984). At the time of its publication (1984), the NDBEPR system design approach was criticized and rightly so, primarily because the influent RBO was used twice, once by the PAOs for P removal (uptake in the anaerobic reactor) and again by the OHOs for denitrification in the primary anoxic reactor. This would be possible only if in NDBEPR systems the PAOs utilize all the RBO in the primary anoxic reactor with nitrate as electron acceptor for growth and polyP accumulation in the same fashion as the RBO is completely utilized by the OHOs

Biological Nutrient Removal

in the primary anoxic reactor of the ND system. In this event the major portion of the P uptake and polyP storage by the PAOs should take place in the primary anoxic reactor of the NDBEPR systems. However, P uptake was observed taking place principally in the aerobic zone. This indicated that the denitrification behavior in NDBEPR systems is not the same as that observed ND systems so that the good predictions that had been obtained by the ND models for the NDBEPR systems were fortuitous. Denitrification behavior in NDBEPR systems is discussed in Section 4.14.34 after presenting the BEPR model based on PAO behavior. Essentially up to this time, models of NDBEPR system behavior did not recognize the presence of any specific organism mediating BEPR, only the OHOs for COD removal, denitrification, and RBO fermentation, and the ANOs for nitrification (Table 1). The parametric model in fact considered the active biomass as one group (OHOs) to represent a BEPR sludge with a propensity for P removal; variation in BEPR between different systems was modeled as changes in the propensity for P removal of OHO biomass caused by changes in influent RBSO concentration, anaerobic mass fraction, and/ or nitrate discharge to the anaerobic reactor. However, parallel research in the natural sciences had identified specific organism groups that have the propensity to store large quantities of P in the form of polyP (e.g., Buchan, 1983). This led to a shift in the approach to modeling BEPR in NDBEPR systems, from a representative OHO biomass to a specific organism group mediating BEPR, like the ANOs, the specific organism groups that mediate nitrification. The BEPR organism group became generically termed polyP organisms (Wentzel et al., 1986), bio-P organisms (Comeau et al., 1986), or PAOs (ASM2, Henze et al., 1995).

4.14.30.2.1 NDBEPR system kinetics Wentzel et al. (1988) set out to develop a general model that describes NDBEPR system behavior. They assumed that in an NDBEPR system treating municipal wastewaters, a mixed culture would develop which could be categorized into three groups of organisms: (1) heterotrophic organisms able to accumulate polyP, termed PAO; (2) heterotrophic organisms unable to accumulate polyP, termed OHOs; and (3) autotrophic organisms mediating nitrification, termed ANOs (Table 1). With regard to OHOs and ANOs, they accepted the ND models described in this chapter, viz., the steady-state (WRC, 1984) and general kinetic model (Dold et al., 1980, 1991; van Haandel et al., 1981). These models were extended to incorporate PAO behavior. To achieve this, the kinetic and stoichiometric characteristics of the PAOs in the AS environment needed to be established. From attempts to obtain information on the characteristics of the PAOs using mixed liquor from NDBEPR systems treating municipal wastewaters, Wentzel et al. (1988) noted that the OHO behavior masked the PAO behavior except in its P release, P uptake, and P removal. Accordingly, to isolate the PAO biomass characteristics, they developed enhanced cultures of PAOs in open (nonsterile) AS systems. (Serendipitously, because the UCT laboratory did not have the equipment to develop pure cultures, this was never attempted – in hindsight, this would have been the wrong approach because even today, a pure culture of PAOs has not

503

yet been established.) By enhanced culture is meant a culture in which (1) the growth of PAOs is selected into the system to the extent that they become the principal organism group so that their behavior dominates the system in all the measured parameters (OUR, VSS) and (2) growth of competing organisms is selected out but not positively excluded; neither are predation or other interaction effects.

4.14.30.3 Enhanced PAO Cultures 4.14.30.3.1 Enhanced culture development From the biochemical models, Wentzel et al. (1988) were able to identify conditions to be imposed in an NDBEPR AS system to produce an enhanced PAO culture – anaerobic/aerobic sequence with adequate anaerobic mass fraction; influent fed to the anaerobic reactor with acetate as substrate and with adequate macro- and micronutrients, in particular Mg2þ, Kþ, and to a lesser degree Ca2þ, and pH control in the aerobic reactor. Using the UCT and three-stage modified Bardenpho systems, with system sludge ages ranging from 7.5 to 20 days, they developed enhanced cultures of PAOs with greater than 90% of the organisms cultured aerobically being identified as Acinetobacter spp. using the analytical profile index (API) 20NE procedure. (The API 20NE procedure has subsequently been shown to overestimate Acinetobacter spp. numbers due to the testing technique (Lotter et al. 1986; Venter et al. 1989) and selection in culturing (e.g., Wagner et al. 1994). However, for the development of the design and simulation models exact identification of the PAOs in the enhanced cultures has been of minor consequence as the models are based on quantitative experimental observations.) The response of the enhanced culture systems indicated that significant concentrations of PAOs developed. For example, the UCT system (anaerobic mass fraction 15%, sludge age 10 days, and influent of acetate at 500 mgCOD l1) gave phosphate release of 253 mgP l1, phosphate uptake of 314 mgP l1, and phosphate removal of 61 mg l1, all as mgP l1 influent flow. This BEPR behavior was much higher than observed in a mixed culture NDBEPR systems with municipal wastewater influent of 500 mgCOD l1 giving a phosphate release of 45 mg l1, phosphate uptake of 57 mg l1, and phosphate removal of 12 mgP l1. In fact, the behavior of the enhanced culture systems corresponded closely to that of the mixed culture system in terms of the influent RBO/VFA fed – at 100% and 20% influent RBO/VFA respectively for 500 mgCOD l1 feed, the enhanced culture system removed 5 times more P (61 mgP l1) than the mixed culture system. The enhanced culture mixed liquor in the aerobic zone contained 0.25–0.20 mgP/mgVSS and had a VSS/ TSS ratio of 0.46–0.48 as sludge age increased from 7.5 to 20 days, much higher than for mixed culture systems at a P/VSS ratio of 0.1 and a VSS/TSS fraction of 0.78. The low VSS/TSS ratio for the enhanced culture systems is due to the high concentration of polyP with associated counterions in the PAOs, a phenomenon later included in the model by Ekama and Wentzel (2004).

4.14.30.3.2 Enhanced culture kinetic model From experimental observations on the enhanced culture steady-state systems and on a variety of batch tests (anaerobic, anoxic, and aerobic) on mixed liquor harvested from the

Biological Nutrient Removal

4.14.30.3.3 Simplified enhanced culture steady-state model Wentzel et al. (1990) simplified the enhanced culture kinetic model, to develop a steady-state model for the enhanced culture systems under constant flow and load conditions. From an examination of the kinetics of the processes under steady-state conditions, many of the processes were virtually complete so these kinetic relationships no longer serve an important function under steady-state conditions and could be replaced by stoichiometric relationships. The three examples are given as follows: (1) The anaerobic mass fractions provided in the enhanced culture systems were sufficient to ensure that all the acetate substrate was sequestered in the anaerobic zone, that is, the kinetics of acetate storage need not be incorporated. (2) Virtually, all the substrate taken up by the PAOs in the anaerobic zone was utilized in the subsequent aerobic zone, that is, the kinetics of PHA substrate utilization (and polyP storage) did not need to be incorporated. This implied that for the

10

OUR (mgO I−1 h−1)

9

Simulated Experimental

8 7 6 5 4 3 2 1 0

0

1

2

3

(a)

7

8

9

10

200

150

100

50

0

Simulated Experimental 0

1

2

3

4

5

6

7

8

9

10

Time (days)

(b)

100

20

90

PO4

80

15

70 60

NO3

50

10

40 30 20

Simulated

10

Experimental

0 0

(c)

4 5 6 Time (days)

1

2

3

4

5

6

7

8

9

5

Nitrate concentration (mgN I−1)

Soluble COD concentration (mgCOD I−1)

steady-state enhanced PAO systems, Wentzel et al. (1989a) elucidated the characteristics and kinetic response of the PAO biomass. Two characteristics of the PAOs in these enhanced cultures were of particular interest: (1) very little propensity to denitrify so that no provision for this process needed to be made in modeling PAO behavior – this has important implications in modeling denitrification in mixed culture NDBEPR systems (see Section 4.14.34) and (2) an extremely low endogenous mass loss rate, 0.04 mgPAOVSS/(mgPAOVSS d) which is much lower than that of OHOs in aerobic AS system at 0.24 mgOHOVSS/(mgOHOVSS d) (Marais and Ekama, 1976). A similar observation had been made by Wentzel et al. (1985) in studies on mixed culture NDBEPR systems treating municipal wastewaters; they noted from plots of phosphate uptake versus phosphate release for various sludge ages that, for a given phosphate release, the phosphate uptake was relatively insensitive to sludge age. In modeling PAO endogenous mass loss, Wentzel et al. (1989a) used the classical endogenous respiration approach (Equation (53)), as distinct from the death-regeneration approach used for the OHOs (Section 4.14.5.4.2), except that provision was made for the situations where no external electron acceptor is available. Taking note of the above, Wentzel et al. (1989a) developed a conceptual model for PAO behavior in the enhanced cultures incorporating the characteristics, processes, and compounds identified as important from the experimental investigation. Using the conceptual model as a basis, Wentzel et al. (1989b) formulated mathematically the process rates and their stoichiometric interactions with the compounds, to develop a kinetic model for the enhanced cultures of PAO. The kinetic and stoichiometric constants of the PAOs in the enhanced cultures were quantified by a variety of experimental procedures (Wentzel et al., 1989b). With these constants, application of the kinetic model to the various batch test responses observed with the enhanced cultures gave good correlation between observations and simulations (Figures 49 and 51). The model was then applied to simulate the steadystate behavior of the enhanced culture UCT and three-stage modified Bardenpho systems, for which good correlation was also obtained. (Wentzel et al., 1989b).

Soluble P concentration (mgP I−1)

504

0 10

Time (days)

Figure 49 Experimentally observed and simulated (a) oxygen utilization rate (OUR), (b) total soluble phosphorus (PO4) and nitrate (NO3) concentrations and (c) filtered COD concentrations with time in a batch aerobic digestion test of mixed liquor from an enhanced PAO culture system. Modified from Wentzel MC, Dold PL, Ekama GA, & Marais GR (1989b) Enhanced polyphosphate organism cultures in activatedsludge systems 3. Kinetic model. Water SA 15(2): 89–102.

PAOs, like for the OHOs, the growth process could be accepted as complete so that at steady state, for a given sludge age, a constant relationship exists between the flux of acetate fed to the system and the mass of PAOs formed with stored polyP. (3) P release for anaerobic maintenance energy requirements was small compared with P release for VFA uptake energy requirements, that is, the kinetics of phosphate release for anaerobic

60

60

50

50

PO4

40

40

30

30

20

20

10

10

Acetate

0 0

1

2

3

4

0

200

Simulated Experimental

150

150

PO4 100

100

50

50 Acetate

0

0

1

2 3 Time (h)

4

5

0

Figure 50 Experimentally observed and simulated total soluble phosphorus (PO4) and acetate concentration–time profiles with anaerobic addition of (a) 0.11 mgCODacetate/mgVSS (low) and (b) 0.265 mgCOD/mgVSS (moderate) to a mixed liquor drawn from a threestage Bardenpho enhanced PAO culture system. Modified from Wentzel MC, Dold PL, Ekama GA, and Marais GR (1989b) Enhanced polyphosphate organism cultures in activatedsludge systems 3. Kinetic model. Water SA 15(2): 89–102.

maintenance energy did not need to be incorporated. However, because the endogenous respiration process is never complete, it had to be retained in the steady-state model and, as for the OHOs, was accepted to take place in all the reactors of the system. Applying these simplifications and assumptions in the steady-state PAO model indicated that the P content of the PAOs was constant with sludge age at 0.38 gP/gPAOVSS, of which 0.03 was biomass P content and 0.35 was polyP content, to account for the observed P removal. What did vary was the relative proportion of PAOs (with stored polyP) in the VSS which accounted for the difference in P removal with sludge age. The resulting steady-state PAO model was identical to the OHO model (Section 4.14.31.1.5), including the value for the PAO yield coefficient (YG ¼ 0.45 mgPAOVSS/mgCOD), but the values for the PAO unbiodegradable residue fraction (fEG) and endogenous respiration rate (bG) were different to those of the OHOs (i.e., 0.25 and 0.04 d1, respectively). The PAO steady-state model provided the means for quantifying the PAO VSS mass and its endogenous residue in mixed culture NDBEPR systems receiving municipal wastewaters as influent.

OUR 100

0

5

PO4

0

5

10 15 Time (h)

20

500

0 25

20

450

Simulated Experimental

400 350

15

300 250

10

OUR

200 150

PO4

5

100 50 0

0

5

(b)

(c)

15

10

(a)

Soluble P concentration (mgP I−1)

200

Simulated Experimental

300

Time (h) 200

20

10 15 Time (h)

20

0 25

20

200 Simulated Experimental

15

150

100

5

50

0

10

OUR

PO4

0

5

10 15 Time (h)

20

0 25

Carbon oxygen utilization rate (mgO I−1h−1) Carbon oxygen utilization rate (mgO I−1h−1)

70

Soluble P concentration (mgP I−1)

Soluble P concentration (mgp I−1)

80

70

400

505

Carbon oxygen utilization rate (mgO I−1h−1)

80

(a)

(b)

90

Simulated Experimental

Soluble P concentration (mgP I−1)

100

90

Acetate concentration (mgHAc I−1)

100

Acetate concentration (mgHAc I−1)

Soluble P concentration (mgP I−1)

Biological Nutrient Removal

Figure 51 Experimentally observed and simulated total soluble phosphorus (PO4) concentration and carbonaceous oxygen utilization rate (OUR)–time profiles on aeration following anaerobic acetate addition of (a) 0.207 mgCOD/mgVSS (low),(b) 0.363 mgCOD/mgVSS (moderate), and (c) 0.220 mgCOD/mgVSS (high) to mixed liquor drawn from a three-stage Bardenpho enhanced PAO culture system. Modified from Wentzel MC, Dold PL, Ekama GA, and Marais GR (1989a) Enhanced polyphosphate organism cultures in activated sludge systems 2. Experimental behaviour. Water SA 15(2): 71–88.

4.14.30.3.4 Steady-state mixed culture NDBEPR systems Mixed culture steady-state model. Having developed the steadystate model for enhanced culture systems, Wentzel et al. (1990) extended this model to incorporate mixed cultures of PAOs and OHOs present in NDBEPR systems receiving

506

Biological Nutrient Removal

domestic wastewater as influent, to give a steady-state mixed culture model. This extension proved to be possible because (1) enhanced cultures rather than pure cultures were used to establish the kinetic and stoichiometric characteristics of the PAOs. In the enhanced cultures, PAOs present in mixed culture AS were enriched and no single species was artificially selected (as in pure cultures); (2) competing organisms and predators were not artificially excluded (as in pure cultures) so that the PAOs were subjected to the same selective pressures in enhanced as in mixed cultures; (3) the PAOs were also subjected to the same conditions present in mixed culture AS systems (e.g., anaerobic/aerobic sequencing, long SRT 45 days, etc.); and (4) per influent RBO/VFA, the PAOs exhibited the same behavioral patterns in the enhanced cultures as they did in mixed culture AS systems (i.e., P release/uptake, PHA/ polyP accumulation, etc.) – in fact, the similar, though magnified behavior of the PAO enhanced culture compared to the mixed culture systems was one criterion used to establish that the correct enhanced cultures had been established. In extending the model one aspect that emerged was the difference in the endogenous mass loss rate between PAO enhanced culture sludges and the normal aerobic OHO AS. As noted earlier, the high specific endogenous mass loss rate with OHO systems had been attributed to a high rate of predation and regrowth, formulated as death regeneration in the ND kinetic model by Dold et al. (1980). The low specific endogenous mass loss rate with PAOs in the enhanced cultures systems led Wentzel et al. (1989a) to conclude that the PAOs were not predated to the same degree as OHOs, and to adopt an endogenous respiration approach in modeling PAO endogenous mass loss. (From subsequent simulations with the steady-state mixed culture model, it was found that if the PAOs were subjected to a high predation rate, then significant BEPR in the mixed culture NDBEPR system would not be possible – the rate of death of the PAOs would be so high that no significant mass of these organisms could accumulate in the system, and BEPR would be near zero.) The low predation rate on the PAOs, and the fact that the PAOs and OHOs essentially do not compete for the same substrate, implied that PAO and OHO populations act virtually independently of each other in normal mixed culture NDBEPR systems. This allowed modeling the two population groups as essentially separate, except for the fermentation F-RBO to VFA conversion process in the anaerobic reactor, which could be used to quantify the proportion of the biodegradable organics (BO) obtained by the PAOs. This rate of conversion is much slower than the rate of VFA uptake, so that the rate of conversion controls the rate of VFA uptake. Hence, the flux of VFAs that becomes available in the anaerobic reactor to the PAOs is governed by the kinetics of conversion mediated by the OHOs. The work of Me´ganck et al. (1985) and Brodisch (1985) supported this conversion approach, which is also included in the NDBEPR kinetic models (UCTPHO, Wentzel et al., 1992; ASM2, Henze et al., 1995) – they showed that anaerobic/aerobic systems developed organisms which

convert sugars and similar compounds into VFAs in the anaerobic reactor. If nitrate (or oxygen) is recycled to the anaerobic reactor, RBO is utilized preferentially by the OHOs with nitrate (or oxygen) as external electron acceptor, thereby reducing the flux of VFAs available for uptake by the PAOs. A schematic diagram showing the proportion of the influent RBO obtained by the PAOs is shown in Figure 52. OHOs obtain BO that is not obtained by PAOs. From the above, the RBSO is subdivided into two fractions, VFAs (e.g., acetate) and fermentable RBSO (FBSO, e.g., glucose). Both these fractions are measured as RBO in the conventional bioassay (e.g., Ekama et al., 1986; Wentzel et al., 1995, 1999, 2000) and filtration (e.g., Dold et al. 1986; Mamais et al., 1993; Mbewe et al., 1994) tests (see Section 4.14.4.2.2). The rate of VFA uptake by PAOs is so rapid that all influent VFAs will be taken by the PAOs even in very small anaerobic reactors (Figure 50). The F-RBO is converted to VFAs by the OHOs in the anaerobic reactor and the resultant VFAs is available for uptake by the PAOs (Figure 46). The model for this conversion is given by Wentzel et al. (1985) and will be described below. The above model provided Wentzel et al. (1990) with the means for calculating the flux of BO (influent VFA and

Unbiodegradable soluble (effluent)

Influent wastewater COD Biodegradable COD

Unbiodegradable particulate COD

RBCOD F-RBCOD

SBCOD External acid fermentation

Internal acidification

Inert VSS accumulation

Volatile fatty acids (VFAs) P accumulating organisms (PAOs)

Ordinary heterotrophic organisms (OHOs)

Enhanced culture steady-state equations PAO activemass 0.03 mgP/ mg-VSS

O2,NO3−

Usual activated sludge steady-state equations

Usual OHO activemass 0.03 mgP/mg-VSS

PAD endogenous mass 0.03 mgP/mg-VSS

OHO Endogenous mass 0.03 mgP/mg-VSS

Inert mass 0.03 mgP/ mg-VSS

Mixed VSS in system has variable P content (mass P/mass VSS %) Depending on proportion of biodegradable COD obtained by PAOs

Figure 52 Schematic diagram showing the fate of various influent COD fractions in relation to the various OHO and PAO active, endogenous, and inert masses of the sludge.

Biological Nutrient Removal

Predicted P release (mgP I−1 influent)

100 90 80 70 60 50

R S (days) 3 4 5 6 8 10 15 20 21 25 28

40 30 20 10 0

0 (a)

10 20 30 40 50 60 70 80 90 100 Measured P release (mgP I−1 influent)

30

Predicted P removal (mgP I−1)

converted of F-RBO) taken up by the PAOs in the anaerobic reactor. The remainder of the BO flux is obtained by the OHOs. In effect, the conversion model splits the influent BO (COD) into two fractions, one eventually utilized by the PAOs and the other to be utilized by the OHOs. Because of the independent action of these two organism groups, the masses of PAOs (MXBG) and their endogenous residue (MXEG) in the system could be calculated from the enhanced PAO culture steady-state model and the masses of OHOs (MXBH) and their endogenous residue (MXEH) could be calculated from the steady-state OHO model. The mass UPO in the reactor from the influent (XI) could be calculated from the unbiodegradable particulate COD fraction (fS’up) as before (Section 4.14.9.3.2). The five VSS components, each with their P content – 0.38 mgP/mgPAOVSS for the PAOs and 0.025 mgP/mgVSS for the other four components – give the average P content of the VSS. The P removal achieved by the NDBEPR system is the P in sludge mass wasted per day from the system. Wentzel et al. (1990) evaluated the predictive power of the steady-state mixed culture BEPR model against observations made on 30 laboratory-scale NDBEPR systems over a 6-year period. The system configurations were Phoredox, three-stage modified Bardenpho, UCT, MUCT, and JHB with system sludge ages ranging from 3 to 28 days. For the evaluation, the measured nitrate in the recycle to the anaerobic zone was used to estimate the fermentable COD removal in the anaerobic zone by the OHOs with nitrate as external electron acceptor. The fermentable COD remaining was available for conversion in the anaerobic reactor to VFAs and uptake and storage as PHA by the PAOs. Plots of the predicted versus measured P release, P removal, and VSS concentration (Figures 53(a)– 53(c)) show good correlation.

507

25 20 R S (days)

15

3 4 5 6 8 10 15 20 21 25 28

10 5 0 0

(b)

4.14.31 Mixed Culture Steady-State Model 4.14.31.1 Division of Biodegradable Organics between PAOs and OHOs

Sbsi ¼ Sbsai þ Sbsfi

ð173Þ

The VFA in the influent (Sbsai) is directly available to the PAOs for uptake in the anaerobic reactor.

30

4000

Predicted VSS (mg VSS I−1)

4.14.31.1.1 Subdivision of influent RBO From the mechanism for BEPR, only VFAs can be taken up directly by the PAOs in the anaerobic reactor. Accordingly, the influent RBO (Sbsi) is subdivided into two fractions: (1) VFA (Sbsai) and (2) fermentable RBO (FBSO, Sbsfi). Hence,

5 10 15 20 25 Measured P removal (mgP I−1)

3000

R S (days)

2000

3 4 5 6 8 10 15 20 21 25 28

1000

0

4.14.31.1.2 Conversion of FBSO Wentzel et al. (1985) show that the FBSO component (Sbsfi) is converted to VFA in the anaerobic reactor by the OHOs, thereby making additional VFA available to the PAOs for uptake. The rate of conversion is much slower than the rate of VFA uptake, so that the rate of conversion controls the rate of uptake of generated VFA. Wentzel et al. (1985) proposed a

0 (c)

1000

2000

3000

4000

Measured VSS (mgVSS I−1)

Figure 53 Predicted vs. measured P release (a), P removal (b) and VSS concentration (c) in a variety of BEPR systems with various configurations. From Wentzel et al. (1990).

508

Biological Nutrient Removal

first-order conversion rate, viz.,

dSbsf ¼  KCT XBHn Sbsfn dt ðmgCOD l1 h1 Þ

equations for the conversion of FBSO to VFA can be developed. This yields equations for the concentration of FBSO in exiting the nth anaerobic compartment and the mass of OHOs in the entire NDBEPR reactor, MXBH viz.,

ð174Þ

where KCT is the first-order rate constant at temperature T ¼ 0.06 l/(mgOHOVSS d) at 20 1C, and XBHn and Sbsfn the concentrations of OHOs (mgOHOVSS l1) and FRBO (mgCOD l1) exiting the nth anaerobic compartment of the anaerobic reactor.

4.14.31.1.3 Effect of recycling nitrate or oxygen When nitrate or oxygen enter the anaerobic reactor via recycle and influent flows, the OHOs utilize FBSO with these electron acceptors. Hence, the OHOs do not release the VFA generated but completely metabolize the FBSO until the oxygen or nitrate is depleted. In the conversion model this is accommodated by reducing the concentration of FBSO available for conversion, that is,

S0bsfi

¼ Sbsfi  2:86=ð1  f cv YH ÞðrNnr þ Nni Þ  1=ð1  f cv YH ÞðrOr þ Oi Þ

ð175Þ

where S0bsfi is the FBSO available for conversion to VFA (mgCOD l1 influent), Sbsfi the influent FBSO concentration (mgCOD l1), r the recycle ratio to anaerobic reactor relative to the influent flow, Nnr, Or the nitrate and oxygen concentration in the recycle to anaerobic reactor (mgNO3-N l1 and mgO l1, respectively), Nni, Oi the nitrate and oxygen concentrations in the influent to anaerobic reactor (mgNO3-N l1 and mgO l1, respectively), 2.86/(1  fcvYH) ¼ 8.6 the mass of COD utilized per unit nitrate denitrified (mgCOD/mgNO3N), and 1/(1  fcvYH) ¼ 3.0 the mass of COD utilized per unit oxygen utilized (mgCOD/mgO). Kinetics of conversion of FBSO to VFA. The conversion model proposed by Wentzel et al. (1985) assumes that: 1. Only FBSO can be converted to a form suitable for uptake by the PAOs (i.e., VFA); within the timescale of the mixed liquor in the anaerobic reactor, conversion of SBO to VFA is assumed to be negligible. 2. The conversion is mediated by the OHOs in the absence of oxygen and nitrate only. 3. All VFA generated by conversion is immediately taken up by the PAOs. 4. All FBSO not converted to VFA in the anaerobic reactor is utilized subsequently by OHOs. 5. The rate of conversion of FBSO is first order with respect to the FBSO and OHO concentrations in the anaerobic reactor and given by Equation (174). 6. All VFA present in the influent to the anaerobic reactor is immediately taken up by the PAOs.

4.14.31.1.4 Steady-state FBSO conversion equation Applying Equations (174) and (175) within mass balances over the nth anaerobic compartment in a series of N equal volume anaerobic compartments in the anaerobic reactor receiving in a continuous flow NDBEPR system, the steady-state

S0bsfi =ð1 þ rÞ  n Sbsfn ¼  f xa MXBH 1 1 þ KCT N Qi ð1 þ RÞ ðmgCOD l1 Þ

ð176Þ

where fxa is the anaerobic mass fraction of the NDBEPR system, N the total number of compartments of equal volume in the anaerobic reactor, n the nth compartment of the series, n ¼ 1,2,yy,N, Sbsfn the concentration of FBSO exiting the nth compartment, MXBH the mass of OHOs in the system (mgOHOVSS), and Qi the influent flow rate (l d1). Equation (176) provides the means to calculate the flux of FBSO converted to VFA in a series of N anaerobic compartments, that is,

FSbCON ¼ Qi ½S0bsfi  ð1 þ rÞSbsfN 

ðmgCOD d1 Þ

ð177Þ

However, to calculate SbsfN, MXBH/Qi needs to be known. This is calculated from the flux of BO not obtained by the PAOs. All the VFA generated by conversion and all the VFA in the influent are taken up by the PAOs, so the flux of COD taken up by the PAOs, FSbPAO, is given by

FSbPAO ¼ FSbCON þ Qi Sbsai ¼ Qi ½S0bsfi  ð1 þ rÞSbsfN  þ Qi Sbsai

ðmgCOD d1 Þ

ð178Þ

and the flux of biodegradable COD taken up by the OHOs is given by

FSbOHO ¼ Qi Sbi  FSbPAO

ðmgCOD d1 Þ

ð179Þ

Hence, from Equation (103), the mass of OHOs in the NDBEPR system is given by

MXBH ¼

FSbOHO YH Rs ð1 þ bHT Rs Þ

ðmgOHOVSSÞ

ð180Þ

Substituting Equations (179) and (178) into Equation (180) and dividing by Qi yields the MXBH/Qi required in Equation (176), viz.,

MXBH ðSbi  ½S0bsfi  ð1 þ rÞSbsN þ Sbsai ÞYH Rs ¼ Qi ð1 þ bHT Rs Þ ðmgOHOVSSÞ

ð181Þ

Equations (176) and (181) need to be solved simultaneously to calculate the concentration of FBSO (SbsfN) exiting the last anaerobic compartment (N); the following procedure converges in three to four iterations: (1) Assume SbsfN ¼ 0 mgCOD l1, (2) calculate MXBH/Qi with Equation (181), (3) with MXBH/Qi known, calculate SbsfN with Equation (176), (4) recalculate MXBH/Qi using the new value for SbsfN, (5) repeat steps (3)–(5) until SbsfN and MXBH/Qi are constant. This procedure splits the influent BO between the OHOs and PAOs. Because the growth processes of two organism

Biological Nutrient Removal

groups after the anaerobic reactor are noncompetitive and VFA uptake process and the growth processes on the available organics are complete for both groups, the stoichiometric equations relating the flux of COD utilized and the biomass produced derived earlier (Equation (103)) can be applied to calculate the PAO and OHO masses and their endogenous residue masses.

509

(Supi, XIi) (Equation (67)), viz.,

MXI ¼ FSti f S0 up =f cv Rs

ðmgIVSSÞ

Total VSS in the NDBEPR system is the sum of the five VSS components:

MXv ¼ MXBH þ MXBG þ MXEH þ MXEG þ MXI ðmgVSSÞ

4.14.31.1.5 Mass of VSS in the NDBEPR system

ð188Þ

PAO mass

MXBG ¼ FSbPAO

YG Rs 1 þ bGT Rs

ðmgPAOVSSÞ

ð182Þ

where, YG is the PAO yield coefficient (mgPAOVSS/mgCOD utilized), FSbPAO the flux BO taken up by PAOs in the anaerobic reactor (mgCODd1), and bGT the PAO specific endogenous mass loss rate constant at temperature T (d1).

4.14.31.1.6 PAO P release From the mechanisms of BEPR (Wentzel et al., 1985, 1990), for every mole of VFA taken up, 1 mol of P is released to provide energy to synthesize and store the VFA as PHA. Accordingly, the P release in the anaerobic reactor is given by

FPrel ¼ f prel FSbPAO

ðmgP d1 Þ

ð189aÞ

PAO endogenous mass or

MXEG ¼ f EG bGT MXBG Rs

ðmgVSSÞ

ð183Þ

where fEG is the fraction of PAOs that is unbiodegradable particulate endogenous residue. PAO oxygen demand

FOGc ¼ FOGs ðsynthesisÞ þ FOGe ðendogenous respirationÞ ¼ ð1  f cv YG ÞFSbPAO þ f cv ð1  f EG ÞbGT MXBG   YG Rs ¼ FSbPAO ð1  f cv YG Þ þ f cv ð1  f EG ÞbGT 1 þ bGT Rs ðmgO d1 Þ

ð184Þ

OHO mass

MXBH

Prel ¼ f prel SbPAO

ðmgP l1 influentÞ

where fprel is the ratio P release/VFA uptake E1.0 molP/mol COD E0.5 mgP/mgCOD and SbPAO the concentration COD taken up by the PAOs per liter influent ¼ FSbPAO/Qi.

4.14.31.1.7 P removal The P removal via the waste sludge is calculated from the individual P content of the five VSS components, viz: By PAOs

  MXBG MXEG 1 DPG ¼ f XBGP þ f XEGP Rs Rs Qi ðmgP l1 influentÞ

YH Rs ¼ FSbOHO 1 þ bHT Rs

ðmg OHOVSSÞ

ð185Þ

where YH is the OHO yield coefficient (mgOHOVSS/mgCOD utilized), FSbOHO the flux BO taken up by OHOs in the anaerobic reactor (mgCOD d1), and bHT the OHO specific endogenous mass loss rate constant at temperature T (d1).

ðmgVSSÞ

By OHOs

ð186Þ

FOHc ¼ FOHs ðsynthesisÞ þ FOHe ðendogenous respirationÞ ¼ ð1  f cv YH ÞFSbOHO þ f cv ð1  f EH ÞbGT MXBH   YH Rs ¼ FSbOHO ð1  f cv YH Þ þ f cv ð1  f EH ÞbHT bHT Rs ðmgO d Þ

where PG is the P removal by the PAOs (mgP l1influent), fXBGP the P content of PAOs ¼ 0.38 mgP/mgPAOVSS, and fXEGP the P content PAO endogenous mass ¼ 0.03 mgP/ mgEVSS.

 MXBH MXEH 1 DPH ¼ f XBHP þ f XEHP Rs Rs Qi

where fEH is the fraction of OHOs that is unbiodegradable particulate endogenous residue. OHO oxygen demand

1

ð190Þ



OHO endogenous mass

MXEH ¼ f EH bHT MXBH Rs

ð189bÞ

ðmgP l1 influentÞ

ð191Þ

where PH is the P removal by the OHOs (mgP l1influent), fXBHP the P content of OHOs ¼ 0.03 mgP/mgOHOVSS, and fXEHP the P content OHO endogenous mass ¼ 0.03 mgP/ mgEVSS. By inert mass

  MXI 1 DPI ¼ f XIP Rs Qi

ðmgP l1 influentÞ

ð192Þ

ð187Þ

The same equations derived earlier in Section 4.14.7.1.1 apply for the UPO that accumulate in the reactor from the influent

where PI is the P removal due to inert mass (mgP l1 influent) and fXIP the P content inert VSS mass (mgP/mgIVSS) ¼ 0.025– 0.03 mgP/mgIVSS.

510

Biological Nutrient Removal

The total P removal is given by the sum of the individual P removals, i.e. Total P removal

DPT ¼ DPG þ DPH þ DPI

ðmgP l1 influentÞ

ð193Þ

The effluent P concentration is given by the difference between the influent P and the P removal, i.e. Effluent P concentration

Pte ¼ Pti  PT

ðmgP l1 Þ

mgPAOVSS), and the polyP ISS, which is 3.286 mgISS/mgP times the PAO polyP content, which is its total P content (fXBGP) minus its biomass P content (Ekama and Wentzel, 2004). Hence,

ð194Þ

If the P removal is greater than the influent P concentration, then the expectation is that the effluent P concentration will be below 0.5 mgP l1. How far below 0.5 mgP l1 is uncertain because currently this appears to be plant specific. Research is being conducted to investigate what the limits of BEPR technology are and what conditions in the NDBEPR system cause them (Neethling et al., 2009). Revised PAO P content (fXBGP). If the P removal is greater than the influent P concentration, then there is insufficient P in the influent for the PAOs to take up P up to their maximum P content of 0.38 mgP/mgPAOVSS. Their P content (fXBGP) will therefore be limited by the available P. Under these conditions, the PAO P content needs to be revised to match the available P. If this is not done, the reactor ISS concentration, which is strongly influenced by the PAO P content, will be overestimated. In the calculation for the revised PAO P content, it is assumed that the effluent P concentration (Pte) is equal to the P concentration of the unbiodegradable soluble organics (USO; see Section 4.14.4.4.3) and that the P content of the non-PAO VSS components remains unchanged. Unless data are available to indicate a nonzero USO P concentration (Pousi40), it is reasonable to accept it as zero. Clearly, if the wastewater contains USO P, then this will impact achieving the very low effluent P standards that are being set for NDBEPR systems these days. However, it would appear that USO P in municipal wastewaters is effectively zero, or at least masked by the scatter of the difference between membrane filtered effluent TP and OP concentrations. The revised PAO P content (fXBGP) is found by making fXBGP the subject of Equation (193) and the P removal equal to the difference between the influent P and USO P concentrations (Equation (39)), viz.,

f XBGP ¼ ½ðPti  Pousi ÞQi Rs  f XEGP MXEG  f XBHP MXBH  f XEHP MXEH  f XIP MXI =MXBG ðmgP=mgPAOVSSÞ ð195Þ

4.14.31.2 VSS and TSS Sludge Masses in the Reactor (System) The VSS mass in the NDBEPR reactor is the sum of the five VSS component masses (Equation (188)). The ISS concentration is the sum of the ISS that accumulates in the reactor from the influent (Equation (97)), the OHO ISS, and the PAO ISS. The OHO ISS is 15% of its VSS mass, that is, fiOHO ¼ 0.15 mgISS/ mgPHOVSS (Equation (98)). The PAO ISS is the sum of its biomass ISS, which is the same at the OHO ISS (0.15 mgISS/

XIO ¼ FXIOi Rs þ MXBH þ 3:286ðf XBGP  f XBGPBM ÞMXBG ðmgISSÞ ð196Þ where fXBGPBM is the PAO biomass P content ¼ OHO biomass P content ¼ 0.025–0.03 mgP/mgPAOVSS. The TSS mass in the NDBEPR system is the sum of the VSS and ISS masses, that is,

MXt ¼ MXv þ MXIO

ðmgTSSÞ

ð197Þ

This TSS mass is distributed in the various reactors of the NDBEPR system, not necessary at the same TSS concentration in each reactor. The reactor configuration (Figure 47) influences the TSS concentration in the different reactors of the system. Calculating the reactor concentrations from the various mass fraction of the reactors is discussed below.

4.14.31.3 BEPR System Design Considerations 4.14.31.3.1 Process volume requirements An approximate reactor volume, that is, a nonconfigurationspecific volume, can be estimated from a selected average reactor TSS concentration required for the system, that is,

Vp ¼ MXt =Xt

ðm3 Þ

ð198Þ

where Xt is the zone/reactor volume weighed average TSS concentration in the NDBEPR system (mgTSS l1). For all NDBEPR system configurations with SSTs, or with membranes (MBR), at steady-state and average dry weather flow (ADWF) conditions, the concentrations of TSS in the preanoxic (Figure 47(f)) and anaerobic (if present) and anoxic and aerobic zones (Xtpax, Xtana, Xtanx, Xtaer), as fractions of the average system TSS concentration Xt are equal to the ratio of the sludge mass fraction and volume fraction of the zones, that is,

Xtana f mana Xtanx f manx ¼ ; ¼ ; Xt f vana Xt f vanx Xtpax f mpax Xtaer f maer ¼ ; ¼ Xt f vaer Xt f vpax

ð199Þ

where fm, fv are the zone sludge mass and volume fractions respectively, and subscripts ana, anx, aer, and pax are the anaerobic, anoxic, aerobic, and pre-anoxic zones, respectively. For BNR systems with SSTs in which the sludge mass is uniformily distributed, that is, the TSS concentrations are the same in the anaerobic, anoxic, and aerobic zones of the reactor, the sludge mass and volume fractions are equal, such as in the three- and five-stage Bardenpho systems (Figures 47(b) and 47(c)) for N and P removal and the pre- (modified Ludzack–Ettinger, MLE) and post-(Wuhrmann) denitrification and four-stage Bardenpho systems for N removal. For example, if an MLE ND system (Figure 34(b)) requires anoxic and aerobic mass fractions (fmanx, fmaer) of 0.45 and 0.55,

Biological Nutrient Removal

respectively, or a three-stage Bardenpho (Figure 47(b)) system requires anaerobic, anoxic, and aerobic mass fractions (fmana, fmanx, fmaer) of 0.15, 0.35, and 0.50 respectively, the corresponding volume fractions of these zones (fvana, fvanx, fvaer) with respect to the reactor volume (VR) will also be 0.45 and 0.55 for the MLE system and 0.15, 0.35, and 0.50 for the three-stage Bardenpho system. This is because the influent flow dilutes the SST return sludge concentration in the first zone by the same amount as the SST concentrates it after the last zone. This equality of sludge mass and volume fractions does not apply to any multizone BNR system with membrane solid–liquid separation in the aerobic zone, because the aerobic zone concentration is in effect the equivalent of the return sludge concentration from the SST (if there were SSTs). For BNR systems with SSTs, in which the TSS concentrations are not the same in the pre-anoxic, anaerobic, anoxic, or aerobic zones (e.g., in the UCT (Figure 47(d)) or in the JHB (Figure 47(f)) systems), the volume and mass fractions are not equal. For the UCT system, the volume fractions (with respect to Vp) of the anaerobic, anoxic, and aerobic zones (fvana, fvanx, fvaer), and the anaerobic, anoxic, and aerobic TSS concentrations (Xtana, Xtanx, Xtaer) at steady-state and ADWF conditions are related to the anaerobic and aerobic mass fractions (fmana, fmaer), recycle ratio (r) from the anoxic to the anaerobic reactor and system average TSS concentration Xt , as follows:

f mana ðr þ 1Þ rB

ð200aÞ

ð1  f mana  f maer Þ B

ð200bÞ

f maer B

ð200cÞ

rB ðr þ 1Þ

ð200dÞ

f vana ¼ f vanx ¼

f vaer ¼

Xtana ¼ Xt

Xtanx ¼ Xtaer ¼ Xt B

ð200eÞ

  f mana 1þ r

ð200f Þ

where

B¼

For the JHB system with SSTs, assuming the influent flow to the pre-anoxic zone, which is sometimes included to increase pre-denitrification, is zero, the volume fractions (with respect to Vp) of the pre-anoxic, anaerobic, anoxic, and aerobic zones (fvpax, fvana, fvanx, fvaer), and the pre-anoxic, anaerobic, anoxic, and aerobic TSS concentrations (Xtpax, Xtana, Xtanx, Xtaer) at steady-state and ADWF conditions are related to the pre-anoxic, anaerobic, and aerobic mass fractions (fmpax, fmana, fmaer), underflow recycle ratio (s) from the SST to the pre-anoxic reactor and average TSS concentration Xt , as follows:

f mana C

ð201aÞ

ð1  f mana  f maer  f mpax Þ C

ð201bÞ

f vana ¼ f vanx ¼

f maer C

ð201cÞ

f mpax s Cðs þ 1Þ

ð201dÞ

f vaer ¼ f vpax ¼

Xtana ¼ Xtanx ¼ Xtaer ¼ Xt C

C¼

ð201eÞ

Cs ð1 þ sÞ

ð201f Þ

  f mpax 1 1þs

ð201gÞ

Xtpax ¼ Xt where

511

In BNR systems with membrane solid–liquid separation in the aerobic zone, the sludge mass distributes itself differently in the different zones of the system compared with systems with SSTs. This is because the effluent is withdrawn via the membranes from the aerobic zone which concentrates the sludge in this zone relative to that in the other zones. However, in recycling this concentrated aerobic zone sludge to an upstream zone, it is diluted by the less concentrated incoming sludge stream from the upstream zones. The higher the recycles from downstream zones to upstream zones, the more uniformily the sludge mass is distributed around the system and the closer the sludge concentrations in the different zones. For the UCT system with membranes, the volume fractions (with respect to Vp) of the anaerobic, anoxic, and aerobic zones (fvana, fvanx, fvaer), and the anaerobic, anoxic, and aerobic TSS concentrations (Xtana, Xtanx, Xtaer) at steady-state and ADWF conditions are related to the anaerobic and aerobic mass fractions (fmana, fmaer), recycle ratio (r) from the anoxic to the anaerobic zone, recycle ratio (a) from the aerobic to the anoxic zones, and system average TSS concentration Xt , as follows:

f mana ðr þ 1Þ Dr

ð202aÞ

ð1  f mana  f maer Þ D

ð202bÞ

af maer ða þ 1ÞD

ð202cÞ

rD ðr þ 1Þ

ð202dÞ

f vana ¼

f vanx ¼

f vaer ¼

Xtana ¼ Xt

Xtanx ¼ Xt D Xtaer ¼ Xt where

D¼

ða þ 1ÞD a

  f mana f maer 1þ  r ða þ 1Þ

ð202eÞ ð202f Þ

ð202gÞ

For the JHB system with membranes, the volume fractions (with respect to Vp) of the pre-anoxic, anaerobic, anoxic, and aerobic zones (fvpax, fvana, fvanx, fvaer), and the pre-anoxic, anaerobic, anoxic, and aerobic TSS concentrations (Xtpax, Xtana,

512

Biological Nutrient Removal

Xtanx, Xtaer) at steady-state and ADWF conditions are related to the pre-anoxic, anaerobic, and aerobic mass fractions (fmpax, fmana, fmaer), recycle ratio (s) from the aerobic to the pre-anoxic zones, recycle ratio (a) from the aerobic to the anoxic zones, and average TSS concentration Xt , as follows:

f vana ¼ f vanx ¼

f mana ð1 þ sÞ sE

ð1  f mana  f maer  f mpax Þða þ s þ 1Þ ða þ sÞE

ð203aÞ

ð203bÞ

f vaer ¼

f maer E

ð203cÞ

f vpax ¼

f mpax E

ð203dÞ

sE ðs þ 1Þ

ð203eÞ

Xtana ¼ Xt

Eða þ sÞ Xtanx ¼ Xt ða þ s þ 1Þ Xtaer ¼ Xtpax ¼ Xt E

ð203f Þ ð203gÞ

 ð1 þ sÞ ða þ s þ 1Þ E ¼ f mana þ f manx s ða þ sÞ  þf maer þ f mpax

ð203hÞ

Equation (203) applies also to the MLE ND system and the three-stage Bardenpho system with membranes. In Equation (203h) for E, for the MLE system, the anaerobic and pre-anoxic mass fractions are both set to zero, the anoxic mass fraction is 1 minus the aerobic mass fraction (i.e., fmanx ¼ 1  fmaer), and the mixed liquor recycle ratio (a) is also set to zero – only one recycle (s) is required to return nitrate and sludge to the anoxic reactor. For the three-stage Bardenpho system, only the pre-anoxic sludge mass fraction (fmpax) is set to zero. From Equations (200)–(203), the volumes of, and the TSS concentrations in, the various zones of common BNR systems with SST or membrane solid–liquid separation can be calculated for selected anaerobic, aerobic, and pre-anoxic mass fractions (fmaer, fmana, fmpax), and interzone recycle ratios (a, r, and s). In the derivation of these equations, steady-state conditions were assumed and the sludge waste flow rate was ignored – the effect of this is negligible (o2%), especially if the sludge age is long. Generally, a uniform distribution of sludge mass in BNR MBR systems will not occur, even in systems with a single recycle flow from the aerobic to the zone receiving the influent flow. For example, changing an MLE ND system, or a three-stage Bardenpho system with SSTs to membrane solid– liquid separation systems, will change these systems from uniformly distributed sludge mass systems in which the sludge mass and volume fractions are equal to nonuniformly distributed sludge mass systems in which the sludge mass and volume fractions are different, the magnitude of difference depending on the magnitude of the recycle ratios. In multizone BNR systems with membranes in the aerobic reactor and fixed volumes for the anaerobic, anoxic, and

aerobic zones (i.e., fixed volume fractions), the mass fractions can be varied (within a range) by varying the inter-reactor recycle ratios. For example, in a UCT system with anaerobic, anoxic, and aerobic zone volume fractions of 0.25, 0.35, and 0.40 and an r recycle ratio from the anoxic to the anaerobic zones of 1:1, the anaerobic, anoxic, and aerobic zone mass fractions can be varied from 0 to 0.131, 0 to 0.366, and 1 to 0.503, respectively, by changing the a recycle ratio from 0:1 to 5:1. Increasing the a recycle ratio concomitantly increases the nitrate load on the anoxic reactor, thereby increasing the denitrification and N removal as the anoxic mass fraction increases. Increasing the r recycle ratio increases the anaerobic mass fraction (at the expense of the other two zone mass fractions) and increases (not proportionally) the P removal. This zone mass fraction flexibility is a significant advantage of membrane BNR systems over conventional BNR system with SSTs because it allows changing the mass fractions to optimize biological N and P removal in conformity with influent wastewater characteristics and the effluent N and P concentrations required. If required, the performance of membrane BNR systems can be simulated with current BNR AS models such as UCTOLD (for ND, Dold et al., 1991), UCTPHO (for NDBEPR with 490 aerobic P uptake BEPR, Wentzel et al., 1992; Hu et al., 2003), and IWA ASM Nos 1, 2 (ND and BEPR, Henze et al., 1987) by returning the SST underflow into the aerobic zone from which the SST feed flow exits (Parco et al., 2009). However, such simulations require a priori information on the reactor and zone volumes and recycle flows, which would need to be determined with the steady-state procedures set out in this chapter.

4.14.31.3.2 Nitrogen requirements for sludge production The form of the equation for calculating the nitrogen requirement for sludge production (Ns, mgN l1 influent) is the same as set out in Section 4.14.21, Equation (142), that is,

Qi Ns ¼ f n MXv =Rs

ðmgN d1 Þ

However, for the BEPR system the term MXv needs to take account of the changes in VSS components, that is, it must be calculated using Equation (188). Effect of this is to increase Ns because MXv is greater in the NDBEPR system than in the same sludge age ND system receiving the same wastewater. The increase in Ns decreases the nitrification capacity (Nc) (Equation (152)), and hence also the nitrification oxygen demand (FOn, mgO d1). For the rest, the nitrification model calculations remain the same.

4.14.31.3.3 Total oxygen demand The carbonaceous oxygen demand (FOc) is the sum of oxygen demands of PAOs (Equation (184)) and OHOs (Equation (187)):

FOc ¼ FOGc þ FOHc ¼ ð1  f cv YG ÞFSbPAO þ f cv ð1  f EG ÞbGT MXBG þ ð1  f cv YH ÞFSbOHO þ f cv ð1  f EH ÞbHT MXBH ðmgO d1 Þ

ð204Þ

Biological Nutrient Removal

4.14.32 Influence of BEPR on the System 4.14.32.1 Influence on VSS, TSS, and Carbonaceous Oxygen Demand The model for BEPR systems presented above enables the VSS and TSS of the mixed liquor (Equations (188) and (197), respectively) and the carbonaceous oxygen demand (Equation (204)) to be calculated. A comparison of the masses of VSS and TSS in the reactor and the carbonaceous oxygen demand per kg COD load on the bioreactor versus sludge age with and without BEPR are shown in Figures 54(a) and 54(b) for the example raw and settled wastewaters respectively, with influent RBO fractions with respect to the biodegradable COD (fSb’s) of 0.25 and 0.38, respectively (Table 14) and an VFA fraction of 25% of the RBO, that is, influent RBO and VFA concentration of 146 and 36 mgCOD l1 for both wastewaters. The features of the BEPR system are a UCT configuration operated at 20 1C with two equal-sized in-series anaerobic compartments with a total anaerobic mass fraction (fxana) of 0.15, an anoxic to anaerobic (r) recycle of 1:1, and no nitrate recycled to the anaerobic reactor. From Figures 54(a) and 54(b), BEPR in the AS system increases the VSS slightly, by about 5–12% and 15–25% for

Raw wastewater

0.8

8 MLTSS Oxygen demand

0.6

0.4

4 MLVSS

0.2

2

0.0

0 0

5

Although there is only a small difference in VSS production between a BEPR and a non-BEPR system, the constituent

10 15 20 25 Sludge age (days)

30

Settled wastewater

10 Sludge mass − kgVSS/(kgCOD/d) reactor

With BEPR No BEPR 20 °C

6

4.14.32.2 VSS Composition

0.1

Oxygen demand − (kgO/d)/(kgCOD/d)

Sludge mass − kgVSS/(kgCOD/d) reactor

10

raw and settled wastewaters respectively depending on sludge age – the longer the sludge age the greater the difference. This increase in VSS is due to the lower endogenous mass loss/ death rate of the PAOs (0.04 d1 at 20 1C) compared with the OHOs (0.24 d1 at 20 1C). However, the TSS is increased substantially, by about 20–25% and 45–55% for raw and settled wastewaters, respectively, depending on the sludge age. This higher TSS production is due to the large quantities of stored inorganic polyP and the associated inorganic cations necessary to stabilize the polyP chains – principally Mg2þ and Kþ (Fukase et al., 1982; Arvin, 1985; Comeau et al., 1986; Wentzel et al., 1988; Ekama and Wentzel, 2004). The high inorganic content of the PAOs causes the VSS/TSS to be much lower than that of the OHOs, 0.46 mgVSS/mgTSS compared with 0.87 mgVSS/mgTSS (excluding the influent ISS). Thus, the higher the PAO fraction of the mixed liquor, the higher the BEPR, but the lower the VSS/TSS ratio of the mixed liquor. The increase in TSS with the inclusion of BEPR needs to be taken into account in the design of the bioreactor volume (Equation (198)) and daily sludge production. Also, since the inorganic cations that stabilize the polyP are derived from the influent wastewater, there must be sufficient concentrations of these cations in the influent; otherwise, the BEPR may be adversely affected (Wentzel et al., 1989a; Lindrea et al., 1994). Further, because the VSS mass generated per kg COD load is greater with BEPR than without, the oxygen demand with BEPR is correspondingly reduced, by about 5–6% and 8–9% for raw and settled wastewaters, respectively (depending on sludge age, Figures 54(a) and 54(b)).

1.0

With BEPR No BEPR 20 °C 8

0.8 Oxygen demand 0.6

6 MLTSS 4

0.4

MLVSS

2

0.2

0

Oxygen demand − (kgO/d)/(kgCOD/d)

The total oxygen demand (FOt, mgO d1) is the sum of the carbonaceous and nitrification oxygen demands, taking due account of the change in nitrogen requirements for sludge production (Ns) and nitrification capacity (Nc). For a nonnitrifying BEPR systems, the total oxygen demand FOt is given by FOc. Including nitrification in the BEPR system necessarily means that denitrification must also be included; the effect of nitrification and denitrification on the total oxygen demand will be considered in Section 4.14.34.

513

0.0 0

5

15 20 25 10 Sludge age (days)

30

Figure 54 Predicted masses of volatile solids (MXV, MLVSS) and total solids (MXt MLTSS) and daily carbonaceous oxygen demand (FOC) per kg COD load on the biological reactor in ND (thin line) and BEPR (bold line) systems treating raw (a) raw and settled (b) wastewater.

Biological Nutrient Removal

Settled wastewater with BEPR

200

I−1

750 mg COD 25% RBCOD fraction % Of VSS mass (settled WW)

% Composition of VSS mass

Inert 60 OHO endogenous 40 OHO active PAO endog 20

90 80

Additional VSS mass in system treating raw WW, i.e., raw WW produces about 100% more activated sludge VSS mass

140 120

70 60

100

50 Inert

80

40

OHO endogenous

60

30

OHO active

PAO endog

40

20 10

PAO active

0

0 0

5

10

15

20

25

30

Sludge age (days)

(a)

0 0

5

% Of VSS mass (settled WW)

Inert mass 60 OHO endogenous mass

40

20

OHO active mass

20

25

30

100

450 mg COD I−1 38% RBCOD fraction

180

80

15

Settled wastewater no BEPR

200

−1

750 mg COD I 25% RBCOD fraction

10

Sludge age (days)

(b)

Raw wastewater no BEPR

100

% Composition of VSS mass

160

20

PAO active

90 80

160 Additional VSS mass in system treating raw WW. i.e., raw WW produces about 100% more activated sludge VSS mass

140 120

70 60 50

100 Inert

40

80 OHO endogenous

60

30 20

40 OHO active

20

10 0

0

0 0 (c)

450 mg COD I 38% RBCOD fraction

180

80

100

−1

5

10

15

20

25

0

30

Sludge age (days)

(d)

% Of VSS mass (raw WW)

Raw wastewater with BEPR

100

% Of VSS mass (raw WW)

514

5

10

15

20

25

30

Sludge age (days)

Figure 55 Percentage composition of VSS mass for BEPR systems (a, c) and ND systems (No BEPR, b, d) treating raw (a, b) and settled (c, d) wastewater.

sludge fractions for the two systems differ markedly. This can be readily demonstrated by comparing the percentage composition of the VSS mass generated in systems exhibiting BEPR with the ND system that does not. To illustrate, percentage composition of the VSS mass is shown in Figures 55(a) to 55(d) for systems at 20 1C with no BEPR (Figures 55(b)– 55(d)) and with BEPR (Figures 55(a) and 55(c)) respectively treating the example raw (Figures 55(a) and 55(b)) and settled (Figures 55(c) and 55(d)) wastewaters. Note that the BEPR system has a smaller OHO active mass than the no BEPR ND system, but the BEPR system has additionally a significant concentration of PAO biomass.

4.14.32.3 P/VSS ratio A parameter often used to evaluate the BEPR performance of an AS system is the P/VSS (or P/TSS) ratio of the mixed liquor. In Figures 56(a) and 56(b), the calculated P/VSS ratio for a BEPR system with a two-compartment anaerobic reactor and the example raw and settled wastewater characteristics are plotted versus sludge age. A zero discharge of nitrate to the anaerobic reactor is assumed. From Figures 56(a) and 56(b), as the system sludge age increases, the P/VSS ratio increases up to a sludge age of about 10 days. Further increases in sludge age cause a decrease in P/VSS ratio. The initial increase in

Biological Nutrient Removal 15

20

10

0.05 fxana

5

0.0

0

15 0.10 0.05

10

fxana 5 0.0

0 5

0 (a)

10 15 20 Sludge age (days)

25

30

0

5

(b)

10 15 20 Sludge age (days)

25

30

20

15

Settled WW

10

%P in TSS (mgP/mgVSS*100)

Raw WW %P in TSS (mgP/mgVSS*100)

0.25 0.20 0.15

Settled WW

0.25 0.20 0.15 0.10

%P in VSS (mgP/mgVSS*100)

%P in VSS (mgP/mgVSS*100)

Raw WW

0.25 0.20 0.15 0.10

5 fxana

0.05 0.0

15

0.25 0.20 0.15 0.10

10

0.05 fxana

5

0.0

0

0 0 (c)

515

5

10 15 20 Sludge age (days)

25

0

30 (d)

5

10 15 20 Sludge age (days)

25

30

Figure 56 Predicted percentage phosphorus to VSS (P/VSS  100; a, b) and TSS (P/TSS  100; c, d) ratios vs. sludge age for mixed liquor in a BEPR system with various anaerobic mass fractions (fxana) treating the example raw (a, c) and settled (b, d) wastewater.

P/VSS with sludge age is due increasing OHO mass with sludge age, which increase the fermentable RBO to VFAs conversion efficiency in the anaerobic reactor and accordingly yields an increased PAO mass (with associated P content of 0.38 mgP/mgVSS). The decrease in P/VSS can be ascribed to the endogenous respiration effect on PAOs. The P/VSS ratio is therefore a consequence of the selection of the system design parameters sludge age and anaerobic mass fraction and wastewater characteristics. Accordingly, the P/VSS ratio can neither fulfill a function in BEPR plant design nor in BEPR performance assessment between different BEPR plants.

4.14.33 Factors Influencing Magnitude of BEPR The influence of the main design parameters on the magnitude of P removal is demonstrated with the mixed culture steady-state BEPR model. These main parameters are: raw settled wastewater (Sti ¼ 750 and 450 mgCOD l1 respectively, Tables 7, 11, 14), sludge age (SRT ¼ 20 days), anaerobic sludge

mass fraction (fxana ¼ 0.15), influent RBO COD fraction (fSb’s ¼ 0.25 for raw and 0.385 for settled), discharge of nitrate and oxygen to the anaerobic reactor (0 for both) and subdivision of anaerobic reactor into compartments (N ¼ 2). The numbers in brackets are the default wastewater characteristics and system design parameter values.

4.14.33.1 Sludge Age and Anaerobic Mass Fraction Using the characteristics of the example raw and settled wastewater with a total influent COD of 750 and 450 mgCOD l1 respectively, assuming (1) no nitrate and DO enters the anaerobic reactor, (2) a recycle ratio to the anaerobic (r) of 1:1 and the anaerobic reactor is subdivided into two compartments, the P removal (normalized with respect to influent COD, mgPmg1 influent COD) versus sludge age is shown in Figures 57(a) (raw) and 57(b) (settled) for anaerobic mass fractions of 0.00 (no BEPR), 0.05, 0.10, 0.15, 0.20, and 0.25. In the same figures, the actual P removal in mgP l1 is shown on the right-hand axis.

Biological Nutrient Removal

0.02

0.25 0.20 0.15 0.10

22.5

15.0

Influent P 0.05

fxana

0.01

7.5

0.0

5

10 15 20 25 Sludge age (days)

0.04 0.03

Influent P

0.02

0.25 0.20 0.15 0.10

18.0 13.5 9.0

0.05

fxana 4.5

0.01 0.0

0.0

0.00

0.0 0

22.5 Settled WW

0

30 (b)

P removal (mgP I−1)

0.03

0.00 (a)

0.05

30.0 Raw WW

P removal (mgP/mg Infl COD)

P removal (mgP/mg Infl COD)

0.04

P removal (mgP I−1)

516

5

10 15 20 25 Sludge age (days)

30

Figure 57 Predicted P removal vs. sludge age for various anaerobic mass fractions (fxana) for a two-compartment anaerobic reactor BEPR system at 20 days sludge age, treating the example raw (a) and settled (b) wastewater.

The effect of sludge age on P removal is complex. For sludge age o3 days, the P removal increases with increase in sludge age and for sludge age 43 days, P removal decreases with increase in sludge age. The reason for this is that an increase in sludge age causes an increase in the system OHO mass, which in turn causes an increase in fermentable RBO conversion and, therefore, an increase in P release, P uptake, and P removal. However, the increased sludge age also causes a decrease in PAO biomass, its associated P content, due to endogenous respiration which decreases the P removal. At sludge age o3 days, the former effect dominates the P removal, while at sludge age 43 days the latter dominates. The decrease in both PAO biomass with increase in sludge age is crucially affected by the specific endogenous mass loss rate of the PAOs – should the endogenous mass loss rate of the PAOs (0.04 d1) have been the same as that of the OHOs (0.24 d1), virtually no BEPR would have been obtained. The effect of anaerobic mass fraction (fxana) on P removal also is shown in Figures 57(a) and 57(b). For a selected sludge age, an increase in fxana always gives rise to an increase in P removal. This is due to the increased conversion of fermentable RBO with larger anaerobic mass fractions. The improvement in P removal, however, diminishes with each step increase in fxana, due to the first-order nature of the RBO conversion kinetics. From Figures 57(a) and 57(b), it can be seen that for fxana 40.15 only small additional increases in P removal are obtained, which usually are not justified due to the decrease in unaerated sludge mass fraction this causes and the consequent impact on the minimum sludge age for nitrification.

4.14.33.2 Raw and Settled Influent The effect of primary settled wastewater on P removal can be seen by comparing Figures 57(a) and 57(b), which show the P removal for the raw wastewater of original COD of 750 mgCOD l1 and the settled wastewater produced from the raw wastewater with a 450 mgCOD l1. It can be seen that although the P removal per mg influent COD is higher for the settled WW, the P removal in mgP l1 is lower. This decrease is due to the decrease in the flux of biodegradable COD entering

the AS system which causes a reduction in OHO biomass and hence in the fermentable RBO converted and in the mass of PAOs generated. The P removal per influent COD entering the biological reactor is higher for the settled because the fraction of the biodegradable organics that is RBO (Sbsi/Sbi) is higher for settled than for unsettled wastewater because primary settling removes only the settleable organics (although not strictly true, RBO loss or gain in primary settling appears to be small; it is assumed that the RBO is not changed during primary settling).

4.14.33.3 Influence of Influent RBO Fraction Assuming zero discharge of nitrate to the two-compartment anaerobic reactor of mass fraction (fxana) of 0.15, the effect of the influent RBO fraction (fSb’s ¼ Sbsi/Sbi) is illustrated in Figures 58(a) and 58(b) for raw and settled wastewaters, respectively. At any anaerobic mass fraction, the higher the influent RBO fraction, the higher the P removal. In design, one option to improve the P removal is supplementation of influent RBO by, for example, acid fermentation of primary sludge (Pitman et al., 1983; Barnard, 1984; Osborn et al., 1989) or dosing other RBO or VFA into the anaerobic reactor.

4.14.33.4 Influence of Recycling Nitrate and Oxygen to the Anaerobic Reactor The influence of nitrate recycled to the anaerobic reactor is illustrated in Figures 59(a) and 59(b) which show P removal versus nitrate concentration recycled to the anaerobic reactor in a recycle ratio 1:1. Clearly, in agreement with numerous experimental and full-scale NDBEPR systems, recycling nitrate has a markedly deleterious influence on the magnitude of P removal. As the nitrate concentration recycled to the anaerobic reactor increases, the P removal decreases. The same applies to oxygen entering the anaerobic reactor, except that its effect is 1/2.86 times that of nitrate because the oxygen equivalent of nitrate is 2.86 mgO/mgNO3-N. If oxygen and/or nitrate are recycled to the anaerobic reactor, the OHOs no longer convert fermentable RBO to VFAs but instead themselves utilize it for energy and growth with the oxygen or nitrate as external electron acceptor. For every 1

Biological Nutrient Removal

0.02

13.5

9.0

0.05

fxana

0.01

4.5

0.0

0.00 0.1

0.2

0.3

0.4

0.5

22.5

Influent P

0.02

15.0 0.05

fxana 0.01

0.6

7.5

0.0

0.0 0

Influent RBO fraction (fSb’s)

(a)

0.03

0.00

0.0 0

30.0

0.25 0.20 0.15 0.10

Raw WW

P removal (mgP I−1)

Influent P

0.03

0.04

18.0

0.25 0.20 0.15 0.10

P removal (mgP/mg Infl COD)

Settled WW

P removal (mgP I−1)

P removal (mgP/mg Infl COD)

0.04

517

0.1

0.2

0.3

0.4

Influent RBO fraction (fSb’s)

(b)

Figure 58 Predicted P removal vs. readily biodegradable COD (RBCOD, Sbsi) as a fraction of the biodegradable COD (Sbi) fSb’s ¼ Sbsi/Sbi) for various anaerobic mass fractions (fxana) for a two-compartment anaerobic reactor BEPR system at 20 days sludge age, treating the example raw (a) and settled (b) wastewater.

0.04

0.25 0.20 0.15 0.10

0.02

22.5 Influent P

15.0

0.05

fxana

0.01

7.5

0.0

0.00 (a)

8 2 4 6 Nitrate in recycle (NO3−N I−1)

0.25 0.20 0.15 0.10

0.03

Influent P

13.5

0.02

9.0 0.05

fxana 0.01

4.5 0.0

0.00

0.0 0

18.0 Settled WW

10

0.0 0

(b)

P removal (mgP I−1)

0.03

P removal (mgP/mg Infl COD)

30.0 Raw WW P removal (mgP I−1)

P removal (mgP/mg Infl COD)

0.04

2 4 6 8 Nitrate in recycle (NO3−N I−1)

10

Figure 59 Predicted P removal vs. nitrate concentration in recycle to anaerobic (recycle 1:1) for various anaerobic mass fractions (fxana) for a twocompartment anaerobic reactor BEPR system at 20 days sludge age treating the example raw (a) and settled (b) wastewater.

mgO2 and 1 mgNO3-N recycled to the anaerobic reactor, 3.0 and 8.6 mgCOD, respectively, are utilized (Equation (175)). Consequently, allowing oxygen and/or nitrate to enter the anaerobic reactor reduces the flux of VFAs available to the PAOs for storage, and correspondingly reduces the P release, P uptake, and P removal. From Figures 59(a) and 59(b), when the nitrate concentration in the recycle exceeds about 12 mgN l1, the P removal decreases to 4 (raw) and 2.2 mgP l1 (settled) which is the same as that of an ND system (fxana ¼ 0) with zero BEPR. In this situation, all the influent RBO is denitrified by the OHOs with the result that no VFAs are released and no VFAs are available to the PAOs, and BEPR no longer takes place – the P removal obtained is due to wastage of sludge with normal metabolic P content (0.03 mgP/mgVSS). If the influent RBO concentration increases or decreases, the concentration of recycled nitrate that completely consumes the RBO will increase or decrease correspondingly below about 12 mgN l1 (provided the recycle ratio remains unchanged). Clearly, one of the principal orientations in any design for BEPR is to minimize oxygen entrainment and nitrate recycling to the anaerobic reactor. To achieve this in situations where nitrification is obligatory or unaviodable, a number of different system configurations have been developed (Figure 47).

4.14.33.5 Subdivision of the Anaerobic Reactor into Compartments The effect of subdividing the anaerobic reactor into compartments is shown in Figures 60(a) (raw) and 60(b) (settled). Increasing the anaerobic reactor from a single completely mixed one to two compartments in series significantly improves the P removal, but increasing the number of compartments to greater than 3 yields little additional increase. This increase is due to the increased fermentable RBO conversion with in-series anaerobic reactor operation as a result of the first-order nature of the conversion kinetics. For design, at least two equal-sized in-series anaerobic reactors should be used.

4.14.34 Denitrification in NDBEPR Systems 4.14.34.1 Introduction Because usually N removal is also a requirement for BNR systems, nitrification is included and hence also denitrification to benefit from its advantages (Section 4.14.24). In the steady mixed culture BEPR model, the nitrate recycled to the anaerobic reactor needs to be known considering the adverse

Biological Nutrient Removal

0.02

0.10 0.05

22.5

15.0

Influent P

fxana

0.01

7.5

0.0

2

4 3 No anaerobic compartments

18.0

0.25 0.20

0.03

13.5 0.15 0.10

0.02

Influent P

9.0

0.05

fxana

0.01

4.5

0.0

0.00

0.0 1

Settled WW

5

0.0 1

(b)

P removal (mgP I−1)

0.25 0.20 0.15

0.03

0.00 (a)

0.04

30.0 Raw WW

P removal (mgP/mg Infl COD)

P removal (mgP/mg Infl COD)

0.04

P removal (mgP I−1)

518

2

4 3 No anaerobic compartments

5

Figure 60 Predicted P removal vs. the number of compartments in the anaerobic reactor for various anaerobic mass fractions (fxana) in a BEPR system at 20 days sludge age treating the example raw (a) and seattled (b) wastewater.

influence of recycling nitrate to the anaerobic reactor on P removal. Indeed, one of the principal orientations in the design procedure for P removal is to prevent nitrate recycling. Where nitrification is not required, this can be suppressed in a simple configuration such as the Phoredox (or the A/O) system (Figure 47(a)) but this option may not viable in some countries because nitrification is either obligatory or unavoidable due to high wastewater temperature. Accordingly, reliable and accurate quantification of denitrification in NDBEPR systems is essential not only for security in P removal but also for estimating the N removal by the NDBEPR system. The early approach (1976–85) to quantify denitrification in NDBEPR systems was to use the theory and procedures for ND systems, as set out in Sections 4.14.17–4.14.27 (Nicholls, 1982; Ekama et al., 1983; WRC, 1984). Experimental data indicated that this approach appeared to predict the observed denitrification quite closely. However, from the mechanisms for BEPR, which emerged later, this approach was theoretically inconsistent. The influent RBO was apparently used twice – first in the anaerobic reactor where it is converted to VFAs which are taken up and stored as PHA by the PAOs, and again in the primary anoxic reactor for denitrification via the K1 rate (Section 4.14.25.1). This situation is possible only if the PAOs denitrified significantly using most of their internally stored PHA with nitrate as electron acceptor in the anoxic reactor. This implies that the most of the P uptake should take place in the primary anoxic reactor. However, this was not usually observed. Practically, all (490%) the P uptake took place in the aerobic reactor of the UCT NDBEPR systems operated during the 1980s (Wentzel et al., 1985, 1990).

4.14.34.2 Experimental Basis for Denitrification Kinetics in NDBEPR Systems Clayton et al. (1991) undertook an experimental investigation into the denitrification kinetic behavior in mixed culture NDBEPR systems. A laboratory-scale modified UCT system (Figure 47(e)) was set up and operated for a year and a half. For the first 6 months, the first primary anoxic reactor was a plugflow reactor, thereafter a completely mixed one. The response of the system was monitored daily and profiles on the

plugflow primary anoxic reactor measured periodically. In addition, a variety of anoxic batch tests that reproduced the conditions in primary and secondary anoxic reactors were conducted on mixed liquor harvested from the MUCT system. In the plugflow reactor and batch tests, all the important parameters were measured to delineate the behavior of the OHOs and PAOs. No differences in the concentration time profiles from the plugflow reactor and batch tests were noted. From these tests: (1) Under the steady-state conditions of the MUCT system, the general denitrification formulation for ND systems dNO3/ dt ¼  KXBH applied also to NDBEPR systems. (2) In the primary anoxic reactor, (1) the rapid rate of denitrification associated with RBO in ND systems (K0 1, Figure 35; the K0 rate here for NDBEPR systems is used to distinguish it from the K rate in ND systems) was usually absent or of very short duration, (2) the slower rate of denitrification associated with BPO (K0 2) continued over the entire duration of the plugflow retention time or batch test (as in ND systems) but its rate was approximately 2 1/2 times faster than in the primary anoxic reactor of ND systems, i.e., 0.224 mgNO3-N/(mgOHOVSS d), where the OHOVSS concentration was calculated from the experimental system data with the ND model, not the BEPR model, that is, ignoring the reduction in OHOVSS due to the presence of the PAOs (Clayton et al., 1991). Based on the BEPR model, which yields lower OHOVSS due to the presence of PAOs, the K0 2 rate would be even higher compared with ND systems (Table 17). (3) In the secondary anoxic reactor, the denitrification rate (K0 3) was approximately 1 1/2 times the rate measured in secondary anoxic reactors of ND systems (K3, Figure 35(b)), also based on the ND model (Table 17). Clayton et al. (1991) proposed three possible explanations for the increased denitrification rate constant K0 2 observed in the primary anoxic zone of NDBEPR systems: 1. PAOs can denitrify, thereby contributing to the denitrification rate by utilizing the intracellular PHB acquired in the anaerobic zone. 2. PAOs cannot denitrify and the influent BPO is modified in the anaerobic zone to a more readily hydrolyzable form, thereby inducing a faster denitrification rate by the OHO in the primary anoxic reactor.

Biological Nutrient Removal

519

Table 17 Specific denitrification rates (K) at 20 1C observed by Clayton et al. (1991) in the MUCT NDBEPR system based on the WRC (1984) and Wentzel et al. (1990) models compared with those in ND systems (WRC, 1984) System

NDBEPR systems based on ND model

NDBEPR systems based on BEPR model

ND systems based on ND model

Units

mgNO3–N/(mgVSS d)

mgNO3–N/(mgAHVSS d)

mgNO3-N/(mgAHVSS d)

Primary anoxic

K10 ¼ 0.61a K20 ¼ 0.224 K30 ¼ 0.100

K10 ¼ 0.70a K20 ¼ 0.255 K30 ¼ 0.114

K1 ¼ 0.720 K2 ¼ 0.101b K3 ¼ 0.072

Secondary anoxic a

Denitrification by this rate contributes negligibly to the N removal of the NDBEPR system. In single-reactor intermittent aeration ND systems Warburton et al. (1991) obtained K2 ¼ 0.128 at 20 1C.

b

Denitrification in N removal systems Primary anoxic reactor

Aerobic reactor

Denitrification in BioP removal systems

Secondary anoxic Reaeration reactor reactor

Mixed liquor recycle a Influent

Waste flow Settler Effluent Sludge recycle s

Influent RBCOD + SBCOD

K1 + K2 K2

Influent SBCOD only

K3

Endogenous SBCOD only

Anaerobic Anoxic reactor reactors

Aerobic reactor

Mixed liquor recycle Influent

Secondary anoxic reactor Waste flow Settler

a

r

Effluent RBCOD acidified and taken up by PAOs No initial rapid K1 rate

Sludge recycle s

K 2′

Influent SBCOD only

K 3′

Endogenous SBCOD only

Figure 61 Comparison of steady-state specific denitrification rates (K, mgNO3-N/(mgOHOVSS d)) in the primary and secondary anoxic reactors of ND (a) and NDBEPR (b) systems. The rates are compared in Table 15.

3. PAOs cannot denitrify and the BPO is not modified in the anaerobic zone but a higher rate of BPO hydrolysis/utilization is stimulated in the OHOs in NDBEPR systems by the anaerobic–anoxic–aerobic sequencing. The PHB concentrations measured in the (1) anaerobic, anoxic, and aerobic zones of the MUCT parent system; (2) anoxic batch tests on mixed liquor harvested from the MUCT system; and (3) anoxic batch tests on mixed liquor harvested from the enhanced PAO cultures of Wentzel et al. (1988, 1989a) demonstrated that PHB did not serve as a substrate source for denitrification (negligible decrease). Therefore, the PAOs did not contribute significantly to the K’2 denitrification rate in the primary anoxic reactor and so cause (1) had to be rejected. This conclusion was supported from the observation that in the mixed and enhanced culture systems and the batch tests, the P uptake was predominantly (490%) aerobic – negligible anoxic P uptake was observed. If the anaerobic reactor pretreats the influent BPO to a more readily utilizable form, then the K denitrification rates should be lower when the NDBEPR sludge is mixed with influent wastewater. Batch tests on sludge from the MUCT system fed the same wastewater as the parent system, yielded the same high K0 2 denitrification rates. Therefore, the anaerobic reactor did not modify the BPO to a more utilizable form. So cause (2) was rejected and default cause (3) had to be accepted. No experimental means was devised to test this third

cause. However, it did at least provide a consistent explanation also for the higher K0 3 in the secondary anoxic reactor – causes (1) and (2) explain only a higher K0 2 rate. A comparison between the denitrification rates in the ND and NDBEPR systems is shown in Figure 61. Because the PAOs did not significantly contribute to the denitrification, the K0 rates had to be recalculated so that the denitrification process in NDBEPR systems is correctly ascribed to the OHO group performing it. The proportion of OHOs in the VSS of NDBEPR systems is smaller than in ND systems (Figure 55) because the PAOs obtain most of the influent RBO. Ekama and Wentzel (1999a) calculated the OHO fraction (favOHO) for NDBEPR systems iteratively with the aid of the steady-state BEPR model (Section 4.14.31) using the measured value for the influent RBO fraction and varying the influent UPOs fraction (UPOCOD, fS’up) until the calculated system VSS mass, now comprising active and endogenous PAO and OHO components and unbiodegradable particulate VSS from the influent, matched that measured. When the correct fS’up had been found, the calculated P removal was matched to that measured by changing the PAO P content (fXBGP) from the enhanced PAO culture value of 0.38 mgP/mgPAOVSS. For MUCT system of Clayton et al. (1991), the recalculated average VSS fractionation results are fS’up ¼ 0.15, favOHO ¼ 0.21, and fXBGP ¼ 0.388 mgP/mgPAOVSS. Because the favOHO based on ND model (Section 4.14.9.4) was 0.24, on average the K0 rates were 0.24/0.21 ¼1.14 or 14%

520

Biological Nutrient Removal

higher (Table 17). Following this calculation procedure, Ekama and Wentzel (1999a) also calculated the fS’up, favOHO, fXBGP, and K0 2 for four other UCT investigations, viz., Musvoto et al. (1994), Pilson et al. (1995), Sneyders et al. (1997), and Mellin et al. (1997). For the same wastewater source, reasonably consistent fS’up values are expected. For fully aerobic and ND systems, this has been the case in the UCT laboratory. For the Mitchells Plain unsettled wastewater usually fed to the experimental systems, this value was found to be around 0.12 for widely differing aerobic and ND systems, e.g., 0.10870.052 for aerobic systems (Mbewe et al., 1994), and 0.13570.060 (Warburton et al., 1991) and 0.1270.04 (Ubisi et al., 1997a, b) for anoxicaerobic systems. However, for NDBEPR systems, this was not the case. Not only was fS’up higher for NDBEPR systems fed the Mitchells Plain unsettled wastewater, it also varied widely in the different NDBEPR systems, from 0.0470.055 (Sneyders et al., 1997) to 0.29370.063 (Musvoto et al., 1994). Because of the method calculating the fS’up fraction, by reconciling the calculated VSS mass with the measured VSS mass, the variation in fS’up changes the favOHO. This, in turn, affects the K0 2 and K0 3, rates, which are higher for higher fS’up and lower for lower fS’up (Ekama and Wentzel, 1999b). Clearly, there are factors that affect the sludge production per unit COD load in the NDBEPR system that the models do not recognize. Two such factors appear to be the unaerated sludge mass fraction (fxt) and sludge settleability (measured as diluted sludge volume index, DSVI). The higher the fxt, the higher the fS’up, which could be due to an accumulation of undegraded BPO in the system. If this were the only factor, then the method of calculating fS’up and favOHO would be acceptable because undegraded BPO in effect is unbiodegradable particulate organics. However, this is not the only factor because systems with the same fxt yielded different fS’up and favOHO values depending on the DSVI (Musvoto et al., 1994; Casey et al., 1994a, 1994b). As the DSVI and hence AA (low F/M) filament abundance increased, so the system VSS mass decreased and vice versa. The calculated K0 rates varied accordingly, decreasing as the system VSS mass increased and vice versa. No explanation for this variation with DSVI can be advanced. The NDBEPR models, both steady-state (e.g., Wentzel et al., 1990) and dynamic simulation (e.g., ASM2, Henze et al., 1995), are extensions of their predecessors (WRC, 1984, ASM1; Henze et al., 1987) by including the kinetics of BEPR. Relatively few interactions between the ND and BEPR processes take place in these models, the main ones being that (1) the VFAs for the PAOs are generated by the OHOs in the anaerobic zone from the influent RBO, and more importantly, (2) the reduction factor for the BPO hydrolysis/utilization rate, Z, is increased from 0.33 in ASM1 (and UCTOLD, Dold et al., 1991) to 0.60 in ASM2 (and UCTPHO, Wentzel et al., 1992) to account for the increased K0 2 and K0 3 rates observed by Clayton et al. (1991). Insofar as the BEPR kinetics in the ASM2 and UCTPHO models are concerned, P release and uptake occur exclusively in the anaerobic and aerobic reactors respectively, in conformity with the observations of Siebritz et al. (1983), Wentzel et al. (1985, 1989b, 1990), Clayton et al. (1991), and Sneyders et al. (1997). Therefore, for the last two mentioned investigations, given the correct input fS’up and Z values, the NDBEPR models will satisfactorily predict the

performance of the M/UCT systems. However, in three other investigations, viz., Musvoto et al. (1994), Pilson et al. (1995), and Mellin et al. (1997), the P release, P uptake, and P removal behavior were significantly different to that observed on which the models are based. Not only was the excess P removal depressed at about 60% of that expected from the model of Wentzel et al. (1990), but also the P release to removal ratio was decreased. With the depressed P removal, significant P uptake took place in the (second) anoxic reactors of the MUCT systems. This was confirmed in the anoxic batch tests; whereas in the tests of Clayton et al. (1991) and Sneyders et al. (1997) negligible anoxic P uptake took place, in those of Mellin et al. (1997) significant (B40%) P uptake took place. The significant decrease in BEPR with anoxic P uptake BEPR was subsequently confirmed by Vermande et al. (2002) in parallel aerobic P uptake BEPR and anoxic P uptake BEPR systems. It is possible that different species of PAOs find a niche in the systems that can accomplish anoxic P uptake, but which have lower RBCOD to P release, P release to P removal, and fXBGP ratios. Biochemical assays have indicated that some PAOs can denitrify (Lo¨tter, 1985; Lo¨tter et al., 1986) and even anaerobic–anoxic (no aerobic) BEPR systems have been operated successfully (Kuba et al., 1993). Also, in several other studies significant anoxic P uptake has been observed (e.g., Vlekke et al., 1988; Kerrn-Jespersen and Henze, 1993; Bortone et al., 1996; Kuba et al., 1996; Kuba and van Loosdrecht, 1996; Hu et al., 2000, 2007a, 2007b). Denitrification by PAOs is included in the biochemical model of Wentzel et al. (1986, 1991) but is not included in ASM2 (Henze et al., 1995) and UCTPHO (Wentzel et al., 1992) kinetic models. Anoxic P uptake behavior of PAOs has been included in ASM2d (Henze et al., 1999) but it merely allows the P uptake to commence in the anoxic reactor without changing the P uptake, that is, anoxic P uptake is modeled with the same stoichiometry and kinetics as aerobic P uptake, which clearly is not observed experimentally. Realistic anoxic P uptake behavior of PAOs therefore cannot be simulated with current suite of IWA ASM models. Proposals to include PAO denitrification have been made (e.g., Mino et al., 1995; Barker and Dold, 1997; Hu et al., 2007a, 2007b), with varying success. One of the main problems with modeling anoxic P-uptake BEPR is that the triggers that stimulate it are not well understood. Hu et al. (2002) conclude that anoxic P-uptake BEPR is undesirable in NDBEPR systems due to the reduction in P removal per influent RBO it causes. For maximum BEPR with (usually) limited influent RBO, aerobic P-uptake BEPR is required. Large aerobic mass fractions (fxto0.5) and underloaded primary anoxic reactors with nitrate appear to favor aerobic P-uptake BEPR.

4.14.34.3 Denitrification Potential in NDBEPR Systems The denitrification potential is the maximum amount of nitrate per liter influent flow that can be removed by biological means in the anoxic reactors. As the experimental investigation into denitrification kinetics in NDBEPR systems indicated that the formulation

dNO3 =dt ¼ KXBH

Biological Nutrient Removal

developed for ND systems can also be applied to NDBEPR systems, the techniques set out in Section 4.14.25.2 to develop equations for denitrification potential in ND systems can be followed for NDBEPR systems also. For development of these equations, the experimental observation that the PAOs do not denitrify is accepted. Denitrification in the primary anoxic reactor is via utilization of any RBO leaking through the anaerobic reactor, and BPO. Procedures to determine the amount of RBO leaking through the anaerobic reactor to the primary anoxic reactor were set out in Section 4.14.31.1.4, where SbsfN is the concentration of FRBO exiting the last anaerobic compartment. Hence, SbsfN (1 þ recycle ratio) is the concentration FRBO per liter influent flow exiting the anaerobic reactor and available for denitrification in the primary anoxic reactor by OHOs. Accordingly, the denitrification potential in the primary anoxic reactor (Dp1) can be expressed as

Dp1 ¼ SbsfN ð1 þ rÞð1  f cv YH Þ=2:86 þ K02 XBH Rnp

ð205Þ

Following the procedures set out in Section 4.14.26.3, Equation (205) can be modified and simplified to give

Dp1 ¼ SbsfN ð1 þ rÞð1  f cv YH Þ=2:86 þ f x1 K02T ðSbOHO ÞYH Rs =ð1 þ bHT Rs Þ ðmg NO3 -N l1 influentÞ ¼ a0 þ f x1 K02T b0

ð206Þ

where fx1 is the primary anoxic sludge mass fraction and a0 ¼ SbsfN ð1 þ rÞð1  f cv YH Þ=2:86 and b0 ¼ ðSbOHO ÞY H Rs = ð1 þ bHT Rs Þ. In Equation (206), it is assumed that the initial rapid rate of denitrification (K0 1T) on FRBO leaking through the anaerobic reactor, SbsfN(1 þ r) is always complete, that is, the actual retention time in the primary anoxic reactor is longer than the time required to utilize this usually low concentration FRBO. As with ND systems, an equation can be developed to determine the minimum primary anoxic mass fraction f’x1min to deplete this RBO. This minimum will be a very low value (o0.05) which is much smaller than the primary anoxic reactors in NDBEPR systems, so generally Equation (206) is valid. However, Equation (206) is not without complication. To calculate the primary anoxic denitrification potential (Dp1), the concentration of RBO in the outflow from the anaerobic reactor (SbsfN) is required. To calculate SbsfN, the concentration of nitrate recycled to the anaerobic reactor is required which in turn requires Dp1 to be known. This problem is overcome by assuming initially that the nitrate concentration exiting the primary anoxic reactor is zero, as was done for the primary anoxic reactor of the MLE system in Section 4.14.25.2, which for the UCT systems means zero nitrate discharge to the anaerobic reactor. For brevity, other NDBEPR configurations are not considered in this chapter. However, if required, the denitrification potential of the secondary anoxic reactor is found using the principles set out in Section 4.14.25.2, viz.,

Dp3 ¼ f x3 K03T ðSbOHO Þ YH Rs =ð1 þ bHT Rs Þ ¼ f x3 K03T b0

ðmgNO3 -N l1 influentÞ

where fx3 is the secondary anoxic sludge mass fraction.

ð207Þ

521

Equation (207) applies to secondary anoxic reactors situated both in the mainstream (e.g., five-stage Bardenpho) and in the underflow recycle (e.g., JHB system). However, in applying Equation (207) to secondary anoxic reactors situated in the underflow recycle, care must be taken in evaluating fx3, because the mixed liquor concentration is increased by a factor (1 þ s)/s in the underflow anoxic reactor compared with the mainstream reactors.

4.14.34.4 Principles of Denitrification Design Procedures for NDBEPR Systems In NDBEPR systems design is oriented to achieve in a single sludge system COD removal, N removal (ND), and P removal (BEPR). Conflict between the last two objectives may arise, for example, the proportion of the total unaerated sludge mass assigned to the anoxic reactor(s) (N removal) and the anaerobic reactor (P removal). For each design, the priorities for treatment need to be assessed and a compromise reached to optimize the system for the particular effluent quality required. Because P is the element that is the main driver for eutrophication, for most designs of NDBEPR systems the focus is on BEPR with denitrification as a secondary design priority. Accordingly, the principle in denitrification design for NDBEPR systems is to ensure that the anaerobic reactor is protected from recycling of nitrate, which causes a disproportionate decrease in the magnitude of P removal (Figure 59). This principle guides selection of the system configuration (five-stage modified Bardenpho, JHB, and M/ UCT; Figure 47) and provides a starting point for sizing the anoxic reactors. When selecting a system configuration for BEPR, it is necessary to establish whether complete denitrification can be achieved. For the wastewater characteristics (i.e., influent TKN and COD concentrations (Nti and Sti)), maximum specific growth rate of the nitrifiers at 20 1C (mAm20), and the average minimum water temperature, the maximum unaerated sludge mass fraction (fxm) and the nitrification capacity (Nc) can be calculated for a selected sludge age (Rs) (Section 4.14.20.3). This fxm needs to be divided between anaerobic (for BEPR) and anoxic (for denitrification) mass fractions. Consequently, the maximum anoxic sludge mass fraction (fxdm) is the difference between the maximum unaerated mass fraction (fxm) and the selected anaerobic sludge mass fraction (fxa), that is,

f xdm ¼ f xm  f xa

ð208Þ

where fxm is given by Equation (136) for a selected Rs, mAm20, Sf, and Tmin. The fxdm then can be subdivided between primary and secondary anoxic sludge mass fractions (fx1 and fx3) and this division fixes the denitrification potential of these two reactors (Dp1 and Dp3) and hence also of the system. If the denitrification potential of the system exceeds the nitrification capacity (i.e., Dp1 þ Dp34Nc), then complete denitrification is possible and the secondary anoxic reactor can be situated in the mainstream, that is, a five-stage Bardenpho system can be selected. If complete denitrification is not possible, then depending on the magnitude of the effluent nitrate

522

Biological Nutrient Removal

concentration, the underflow (s) recycle cannot be discharged directly to the anaerobic reactor. If the nitrate concentration is low (o3 mgN l1), the secondary anoxic sludge mass fraction (fx3) can be combined with the primary anoxic sludge mass fraction to form a three-stage Bardenpho, which, with a higher a recycle ratio, may produce a lower nitrate concentration in the sludge underflow (and effluent) than the fivestage Bardenpho because the K0 2 rate is higher than the K0 3 (Table 17). If the nitrate concentration is high (43 mgN l1), the secondary anoxic reactor can be moved into underflow recycle to form the JHB system, in which event the denitrification potential of the secondary anoxic reactor (Dp3) must exceed the nitrate and oxygen loads via the underflow s recycle. If this requirement is not met, nitrate will leak through the underflow secondary anoxic reactor to the anaerobic reactor. In this event, since the denitrification potential of the primary anoxic reactor (Dp1) is greater than that of the secondary anoxic reactor (Dp3) for equal anoxic mass fractions, incorporation of a secondary anoxic reactor becomes an inefficient utilization of anoxic mass fraction, and the secondary anoxic mass fraction is added to the primary anoxic reactor, the underflow s recycle needs to be denitrified in the primary anoxic reactor to form the M/UCT system. With each change of configuration, more nitrogen removal is sacrificed (i.e., the effluent nitrate concentration increases) to protect maximum BEPR, that is, zero or very low nitrate discharge to the anaerobic reactor.

4.14.34.5 Analysis of Denitrification in NDBEPR Systems Analysis of the denitrification behavior in the NDBEPR system is essentially the same as for the ND system (Section 4.14.26.3) except: (1) The maximum anoxic mass fraction for denitrification (fxdm) for the NDBEPR system is given by Equation (208), whereas fxdm for the ND system is given by Equation (166). Hence, for the same maximum unaerated sludge mass fraction (fxm), the NDBEPR system has a lower fxdm than the ND system, by an amount equal to fxa. (2) The specific denitrification rates for ND systems (K2 and K3) are substituted with the rates for NDBEPR systems (K0 2 and K0 3, Table 17). (3) The denitrification potentials for the primary and secondary anoxic reactors are modified from Equations (163) and (164) for ND systems to Equations (206) and (207) for the NDBEPR system to take account of the uptake of COD by the PAOs in the anaerobic reactor, the zero denitrification by the PAOs, and the faster OHO denitrification rates in NDBEPR systems. Taking account of the above, denitrification equations can be developed for all the NDBEPR configurations (Figure 47). However, in the interests of brevity, only the UCT configuration will be considered.

4.14.35 Denitrification in the UCT System In the UCT system the denitrification behavior is very similar to that in the MLE system, because the a and s recycles discharge into the primary anoxic reactor, so that taking due account of the effect of incorporating the anaerobic reactor, the design equations and procedures developed for the MLE system can be applied to the UCT system. Since complete

denitrification is not possible in the UCT system (high effluent nitrate concentration), the entire anoxic mass fraction (fxdm) available is used as a primary anoxic reactor (fx1). As in the MLE system, the a and s recycle ratios determine the split of the nitrate generated in the aerobic reactor (nitrification capacity, Nc) between the primary anoxic reactor and the effluent. The a recycle ratio is selected so that the equivalent nitrate load on the primary anoxic reactor via the a and s recycles is equal to its denitrification potential (Dp1). For a selected s recycle ratio, the a recycle ratio that loads the primary anoxic reactor via to its denitrification potential is the optimum a recycle ratio (aopt). The denitrification potential of the (primary) anoxic reactor (Dp1) is found from Equation (206) with fx1 ¼ fxdm. Following the same reasoning as in Section 4.14.26.3, the optimum a recycle ratio (aopt) is given by Equation (169), with the proviso that the Nc and Dp1 are applicable to NDBEPR systems, that is, Nc is lower due to the higher sludge production (Section 4.14.31.5.2) and Dp1 is based on Equation (206) with K0 2. As for the MLE system, at a ¼ aopt, Equation (170) gives the minimum effluent nitrate concentration (Nne) achievable. Equation (170) is valid for all aoaopt because for all aoaopt the assumption on which Equation (169) is based is valid, that is, zero nitrate concentration exiting the primary anoxic reactor. If the system is operated with a4aopt, the equivalent nitrate load on the primary anoxic reactor via the a and s recycles exceeds the denitrification potential and nitrate will also be recycled via the r recycle to the anaerobic reactor, to the detriment of BEPR. Furthermore, if nitrate does leak through the primary anoxic reactor, then the nitrate concentration in the outflow from the primary anoxic reactor no longer is zero, and consequently, Equation (170) for the effluent nitrate concentration (Nne) is no longer valid. Equations for the effluent nitrate concentration for a4aopt can be derived by following the principles applied above for aoaopt, but are not considered because aoaopt is required for zero discharge of nitrate to the anaerobic reactor and maximum BEPR. If Equation (169) yields aopt ¼ 0, then the equivalent nitrate load via the s recycle is sufficient to match the denitrification potential of the primary anoxic reactor; if aopto0, the equivalent nitrate load via the s recycle exceeds Dp1 and nitrate will be recycled via the r recycle to the anaerobic reactor. The implication of this is that the Nc that gives aopt ¼ 0 represents the upper limit (equivalently the maximum influent TKN/ COD concentration ratio) that the UCT system is able to treat and still protect the anaerobic reactor against nitrate entry. All Nc (equivalently influent TKN/COD ratios) above this limit will result in nitrate recycle to the anaerobic reactor, which cannot be controlled in the UCT system (a ¼ 0) except by reducing the s recycle ratio. From the above, the minimum a recycle ratio is a ¼ 0. The maximum a recycle ratio (amax) is determined by some practical upper limit (aprac), usually in the range 5–6, beyond which the higher pumping costs outweigh the small gain in lower effluent nitrate concentration (see Section 4.14.26.3). However, for oxidation ditch type systems, or for systems with ‘‘through the wall’’ a recycles via low head high volume pumps, the a recycle ratio ( ¼ aopt) may be significantly higher than the aprac of 5–6. If aopt 4 aprac and aprac is selected, then the

Biological Nutrient Removal

primary anoxic reactor is oversized. This unused denitrification potential (Figure 38) can be kept (i.e., fx1 not decreased) as a factor of safety (for uncertainty if K0 rate) or the size of the fx1 of primary anoxic reactor reduced to match the its denitrification potential (Dp1) to the equivalent nitrate load, as was done for the balanced MLE system (Section 4.14.26.3.2), which will allow a reduction in sludge age. The procedure for the balance MLE sytem can be followed to determine the new sludge age. All the aspects discussed in Sections 4.14.11 and 4.14.14– 4.14.16 regarding reactor concentration selection, system design and control, selection of sludge age, and treatment of the primary and/or secondary sludge produced also apply to NDBEPR systems and should be referred to there.

4.14.36 Conclusion The ND and NDBEPR models such as IWA ASM1 and 2 (Henze et al., 1987, 1995), UCTOLD (Dold et al., 1991), and UCTPHO (Wentzel et al., 1992) are very helpful for biological nutrient removal process description and simulation. However, models always need to be used with great circumspection and experience of real systems. It would appear that the ND models such as ASM1 and UCTOLD give an acceptably reliable description of the ND AS systems – the model predictions compare favorably with observed results and the wastewater characteristic, stoichiometric, and kinetic constants in the models to achieve this are reasonably consistent. For these models some scientific maturity is apparent, where the default kinetic and stoichiometric constants predict the performance of an ND system with acceptable risk of deviation. For the NDBEPR models, this is not the case. The experiments described in the literature point to three important observations in real NDBEPR systems not recognized in NDBEPR models that model users need to be aware of for prudent and proper application: that is, (1) the large variation in the unbiodegradable particulate COD fraction (fS’up) and hence the OHO active fraction (favOHO) and denitrification rate (K0 2); (2) the large variation in biological P removal behavior and P content of PAOs (fXBGP) with anoxic P-uptake BEPR stimulated in some systems for reasons not well defined yet; and (3) the unaccounted loss of influent COD in NDBEPR systems, in that even in carefully controlled laboratory systems, only 75–85% of the influent COD can be recovered in a COD mass balance (Ekama et al., 1999a,b).

References Alexander WV, Ekama GA, and Marais GvR (1980) The activated suldge process Part 2 – Application of the general kinetic model to the contact stabilization process. Water Research 14: 1737--1747. Antoniou P, Hamilton J, Koopman B, et al. (1990) Effect of temperature and pH on the effective maximum specific growth rate of nitrifying bacteria. Water Research 24(1): 97--101. Arvin E (1985) Biological removal of phosphorus from wastewater. Critical Reviews in Environmental Control 15: 25--64. Barat R and van Loosdrecht MCM (2006) Potential phosphorus recovery in a WWTP with the BCFSs process: Interactions with the biological process. Water Research 40(19): 3507--3516.

523

Barker PS and Dold PL (1997) General model for biological nutrient removal activated – sludge systems: Model presentation. Water Environment Research 69(5): 969--984. Barnard JL (1973) Biological denitrification. Water Pollution Control 72(6): 705--720. Barnard JL (1974) Cut P and N without chemicals. Water and Wastes Engineering 11: 33--36 (41-44??). Barnard JL (1975a) Biological nutrient removal without the addition of chemicals. Water Research 9: 485--490. Barnard JL (1975b) Nutrient removal in biological systems. Water Pollution Control 74(2): 143--154. Barnard JL (1976) A review of biological phosphorus removal in the activated sludge process. Water SA 2: 136--144. Barnard JL (1984) Activated primary tanks for phosphate removal. Water SA 10(3): 121--126. Batstone DJ, Keller J, Angelidaki I, et al. (2002) Anaerobic Digestion Model No 1 (ADM1), Scientific and Technical Report No 9. London: International Water Association (IWA). Blackall LL, Crocetti G, Saunders AM, and Bond PL (2002) A review and update of the microbiology of enhanced biological phosphorus removal in wastewater treatment plants. Antonie Van Leeuwenhoek International Journal of General and Molecular Microbiology 81(1–4): 681--691. Bortone G, Saltarelli R, Alonso V, Sorm R, Wanner J, and Tilche A (1996) Biological anoxic phosphorus removal – the DEPHANOX process. Water Science and Technology 34(1/2): 119--128. Bratby J (1978) Aspects of sludge thickening by dissolved-air flotation. Water Pollution Control 77(3): 421--432. Bratby J, Walker S, Barnes C, and Jones G (2008) Float on and on – taking flotation thickeners to the limits of performance. Water Environment and Technology 30(7): 56--59. Brdjanovic D, Hooijmans CM, van Loosdrecht MCM, Alaerts GJ, and Heijnen JJ (1996) The dynamic effects of potassium limitation on biological phosphorus removal. Water Research 30(10): 2323--2328. Brodisch KEU (1985) Interaction of different groups of microorganisms in biological phosphate removal. Water Science and Technology 17(11/12): 89--97. Brouckaert CJ, Ikumi DS and Ekama GA (2010) A 3 phase anaerobic digestion model. Proceedings of the 12th IWA AD conference, Guadalajara, Mexico, 1–4 November. Buchan L (1983) Possible biological mechanism of phosphorus removal. Water Science and Technology 15(3/4): 87--103. Burke RA, Wentzel MC, Dold PL, Ekama GA, and Marais GvR (1989) Biological excess phosphorus removal in short sludge age activated sludge systems. In: Proceedings of the 1st Biennial WISA Conference. Cape Town, March Water Institute of Southern Africa, PO Box 6011, Halfway House, 1685, South Africa. Casey TG, Wentzel MC, Ekama GA, Loewenthal RE, and Marais GvR (1994a) An hypothesis for the cause and control of anoxic-aerobic (AA) filament bulking in nutrient removal activated sludge systems. Water Science and Technology 29(7): 203--212. Casey TG, Wentzel MC, Ekama GA, and Marais GvR (1994b) Causes and control of anoxic-aerobic (AA) (or low F/M) filament bulking in long sludge age nutrient removal activated sludge systems. WRC Report No 286/2/94. Pretoria: Water Research Commission. Chao AC and Keinath TM (1979) Influence of process loading intensity on sludge clarification and settling characteristics. Water Research 13(12): 1213--1224. Clayton JA, Ekama GA, Wentzel MC, and Marais GvR (1989) Denitrification kinetics in biological nitrogen and phosphorus removal activated-sludge systems. Research Report W63, Dept. of Civil Eng., Univ of Cape Town, Rondebosch, 7701, South Africa. Clayton JA, Ekama GA, Wentzel MC, and Marais GvR (1991) Denitrification kinetics in biological nitrogen and phosphorus removal activated-sludge systems treating municipal wastewaters. Water Science and Technology 23(4–6): 1025--1035. Comeau Y, Hall KJ, Hancock REW, and Oldham WK (1986) Biochemical model for biological enhanced phosphorus removal. Water Research 20: 1511--1521. Comeau Y, Rabinowitz B, Hall KJ, and Oldham WK (1987) Phosphorus release and uptake in enhanced biological phosphorus removal from wastewater. Journal of the Water Pollution Control Federation 59: 707--715. Crocetti GR, Hugenholtz P, Bond PL, Schuler A, Keller J, and Jenkins D (2000) Identification of polyphosphate-accumulating organisms and design of 16S rRNAdirected probes for their detection and quantitation. Applied and Environmental Microbiology 66(3): 1175--1182. Daigger GT, Randall CW, Waltrip GD, Romm ED, and Morales LM (1987) Factors affecting biological phosphorus removal for the VIP process, a high-rate University of Cape Town process. In: Proceedings of the IAWPRC Biological Phosphate

524

Biological Nutrient Removal

Removal from Wastewater Seminar, Rome. London: International Water Association (IWA) Publishing. de Haas DW, Wentzel MC, and Ekama GA (2000) The use of simultaneous chemical precipitation in modified activated sludge systems exhibiting biological excess phosphate removal Part 1: Literature review. Water SA 26(4): 439--452. de Haas DW, Wentzel MC, and Ekama GA (2001) The use of simultaneous precipitation in activated sludge systems exhibiting biological excess phosphate removal. Part 5: Experimental periods using ferrous-ferric chloride blend. Water SA 27(2): 117--134. Dick RI (1976) Folklore in the design of final settling tanks. Journal of the Water Pollution Control Federation 48(4): 633--644. Dold PL, Bagg WK, and Marais GvR (1986) Measurement of the readily biodegradable COD fraction (Sbs) in municipal wastewater by ultrafiltration. Research Report No. W57, Rondebosch: University of Cape Town. Dold PL, Buhr HO, and Marais GvR (1982) Design and control of equalization tanks. Research Report No. W 42, Rondebosch: Department of Civil Engineering, University of Cape Town. Dold PL, Buhr HO, and Marais GvR (1984) An equalization control strategy for activated sludge process control. Water Science and Technology 17(Amsterdam): 221--234. Dold PL, Ekama GA, and Marais GvR (1980) A general model for the activated sludge process. Progress in Water Technology 12(Toronto): 47--77. Dold PL, Ekama GA, and Marais GvR (1985) pH control and cost savings in aerobic digestion. In: Drake RAR (ed.) Proceedings of the IAWPRC Specialized Seminar on Wastewater Systems Control. Houston, April 1985. Advances in Water Pollution Control: Instrumentation and Control of Water and Wastewater and Transport Systems, pp. 375–382. Oxford: Pergamon. Dold PL and Marais GvR (1986) Evaluation of the general activated sludge model proposed by the IAWPRC Task Group. Presented at the IAWPRC Specialized Seminar on Modelling of Biological Wastewater Treatment, Copenhagen, August 1985. Water Sciemce Technology, Vol. 18, pp. 63–89. Dold PL, Wentzel MC, Billing AE, and Marais GvR (1991) Activated sludge system simulation programs: Nitrification and nitrification/denitrification systems, Version 1. WRC Report No. TT 52/91. Pretoria, S.A. Downing AL, Painter HA, and Knowles G (1964) Nitrification in the activated sludge process. Journal and Proceedings of the Institute of Sewage Purification 64(2): 130--153. Ekama GA (2009) Using bioprocess stoiciometry to build a steady state plant wide wastewater treatment plant model. Water Research 43(8): 2101--2120. Ekama GA, Barnard JL, Gu¨nthert FW, et al. (1997) Secondary Settling Tanks: Theory, Design, Modelling and Operation. IAWQ STR No 6, pp. 216. London: International Association on Water Quality. Ekama GA, Casey TG, Wentzel MC, and Marais GVR (1996b) Filamentous organism bulking in nutrient removal activated sludge systems. Paper 6: Review, evaluation and consolidation of results. Water SA 22(2): 147--152. Ekama GA, Dold PL, and Marais GvR (1986) Procedures for determining influent COD fractions and the maximum specific growth rate of heterotrophs in activated sludge systems. Water Science and Technology 18(6): 91--114. Ekama GA and Marais GvR (1986) Sludge settleability and secondary settling tank design. Water Pollution Control 85(2): 100--113. Ekama GA and Marais P (2004) Assessing the applicability of the 1D flux theory to full scale secondary settling tank design with a 2D hydrodynamic model. Water Research 38(3): 495--506. Ekama GA, Siebritz IP, and Marais GvR (1983) Considerations in the process design of nutrient removal activated sludge processes. Water Science and Technology 15(3/4): 283--318. Ekama GA, So¨temann SW, and Wentel MC (2006b) Mass balances based plant wide wastewater treatment plant models – Part 3: Biodegradability of activated sludge organics under anaerobic conditions. Water SA 32(3): 287--296. Ekama GA and Wentzel MC (1999a) Denitrification kinetics in biological N & P removal activated sludge systems treating municipal wastewaters. Water Science and Technology 39(6): 69--77. Ekama GA and Wentzel MC (1999b) Difficulties and developments in biological nutrient removal technology and modelling. Water Science and Technology 39(6): 1--11. Ekama GA and Wentzel MC (2004) A predictive model for the reactor inorganic suspended solids concentration in activated sludge systems. Water Research 38(19): 4093--4106. Ekama GA, Wentzel MC, and So¨temann SW (2006a) Mass balance-based plant-wide wastewater treatment plant models – Part 2: Tracking the influent inorganic suspended solids. Water SA 32(3): 277--286. Ekama GA, Wentzel MC, Casey TG, and Marais GvR (1996a) Filamentous organism bulking in nutrient removal activated sludge systems. Paper 3: Stimulation of the selector effect under anoxic conditions. Water SA 22(2): 119--126.

Fuhs GW and Chen M (1975) Microbiological basis of phosphate removal in the activated sludge process for the treatment of wastewater. Microbial Ecology 2: 119--138. Fukase T, Shibata M, and Miyaji Y (1982) Studies on the mechanism of biological removal of phosphorus. Japanese Journal of Water Pollution and Research 5: 309--317. Gabb DMD, Still DA, Ekama GA, Jenkins D, and Marais GvR (1991) The selector effect on filamentous bulking in long sludge age activated sludge systems. Water Science and Technology 23(4/6): 867--877. Garrett MT (1958) Hydraulic control of activated sludge growth rate. Sewage and Industrial Wastes 30: 253. Gerber A, Devilliers RH, Mostert ES, and Winter CT (1987) Interactions between phosphate, nitrate and organic substrate in biological nutrient removal processes. Water Science and Technology 19(1–2): 183--194. Gleisberg D (1993) Reduced P-Input from detergents. In: 9th EWPCA-ISWA Symposium, Munich, pp. 23–45. European Water Pollution Control Association, now European Water Association, Hennep, D53773, Germany. Gujer W and Jenkins D (1975a) The contact stabilization activated sludge process – oxygen utilization, sludge production and efficiency. Water Research 9(5/6): 553--560. Gujer W and Jenkins D (1975b) A nitrification model for the contact stabilization activated sludge process. Water Research 9(5/6): 561--566. Harding TH, Ikumi DS, and Ekama GA (2009) A steady state stoichiometric model describing the anaerobic digestion of biological excess phosphorus removal waste activated sludge. MSc Thesis, Research Report No. W132. Rondebosch: Department of Civil Engineering, University of Cape Town. Hart MA and Melmed LN (1982) Microbiology of nutrient removing activated sludge. Poster pres. Special in Seminar of the IAWPRC, Pretoria, SA. Hascoe¨t MC and Florentz M (1985b) Influence of nitrates on biological phosphorus removal from wastewater. Water SA 11: 1--8. Hascoe¨t MC, Florentz M, and Granger P (1985a) Biochemical aspects of enhanced biological phosphorus removal from wastewater. Water Science and Technology 17(11/12): 23--41. Henze M (1992) Characterization of wastewater for modelling of activated sludge processes. Water Science and Technology 25(6): 1--15. Henze M, Dupont R, Grau P, and De La Sola A (1993) Rising sludge in secondary settlers due to denitrification. Water Research 27(2): 231--236. Henze M, Grady CPL, Gujer W, Marais GvR, and Matsuo T (1987) Activated Sludge Model No.1. IAWPRC Scientific and Technical Reports No. 1, ISSN 1010-707X. 33pp. London: International Association on Water Pollution Research and Control, IAWPRC. Henze M, Gujer W, Mino T, Matsuo T, Wentzel MC, and Marais GvR (1995) Activated Sludge Model No. 2, IAWQ Scientific and Technical Report No. 3, ISSN 1025-0913, 32pp. London: IAWQ. Henze M, Gujer W, Mino T, et al. (1999) Activated sludge Model No. 2d, ASM2d. Water Science and Technology 39(1): 165--182. Henze M, Kristensen GH, and Strube R (1994) Rate-capacity characterization of wastewater for nurient removal precesses. Water Science and Technology 29(7): 101--107. Henze M, van Loosdrecht MCM, Ekama GA, and Brdjanovic D (2008) Biological Wastewater Treatment–Principles, Modelling and Design. London: IWA Publishing. ISBN13 9781843391883. Hesselmann RPX, Werlen C, Hahn D, van der Meer JR, and Zehnder AJB (1999) Enrichment, phylogenetic analysis and detection of a bacterium that performs enhanced biological phosphate removal in activated sludge. Systematic and Applied Microbiology 22(3): 454--465. Ho¨rler A (1969) Discussion of ‘‘Performance of (activated sludge) secondary settling tanks’’ by Pflanz P (1969) In: Jenkins SH (ed.) Proceedings of the 4th IAWPR Conference, pp. 569–593. Prague. Oxford: Pergamon. Hu Z-R, Wentzel MC, and Ekama GA (2000) External nitrification in biological nutrient removal activated sludge systems. Water SA 26(2): 225--238. Hu ZR, Wentzel MC, and Ekama GA (2001) External nitrification in biological nutrient removal activated sludge systems. Water Science and Technology 43(1): 251--260. Hu Z-R, Wentzel MC, and Ekama GA (2002) The significance of denitrifying polyphosphate accumulating organisms in biological nutrient removal activated sludge systems. Water Science and Technology 46(1/2): 129--138. Hu Z-R, Wentzel MC, and Ekama GA (2003) Modelling biological nutrient removal activated sludge systems–A review. Water Research 37(14): 3430--3444. Hu ZR, Wentzel MC, and Ekama GA (2007a) A general kinetic model for biological nutrient removal activated sludge systems: Model development. Biotechnology and Bioengineering 98(6): 1242--1258.

Biological Nutrient Removal

Hu Z, Wentzel MC, and Ekama GA (2007b) A general model for biological nutrient removal activated sludge systems–model evaluation. Biotechnology and Bioengineering 98(6): 1259--1275. Jardin N and Po¨pel HJ (1994) Phosphate release of sludge from enhanced biological P-removal during digestion. Water Science and Technology 30(6): 281--292. Jenkins D, Daigger GT, and Richard MG (1993) Manual on the Causes and Control of Activated Sludge Bulking and Foaming, 2nd ed. Chelsea, MA: Lewis Publishers. Jenkins D and Tandoi V (1991) The applied microbiology of enhanced biological phosphate removal – Accomplishments and needs. Water Research 25(12): 1471--1478. Kappeler J and Gujer W (1992) Estimation of kinetic parameters of heterotrophic biomass under aerobic conditions and characterization of wastewater for activated sludge modelling. Water Science and Technology 25(6): 125--139. Kerrn-Jespersen JP and Henze M (1993) Biological phosphorus uptake under anoxic and oxic condition. Water Research 27(4): 617--624. Korentajer L (1991) A review of the agricultural use of sewage sludge: Benefits and potential hazards. Water SA 17(3): 189--196. Kruit J, Hulsbeek J, and Visser A (2002) Bulking sludge solved? Water Science and Technology 46(1–2): 457--464. Kuba T, Murnleitner E, van Loosdrecht MCM, and Heijnen JJ (1996) A metabolic model for biological phosphorus removal by denitrifying organisms. Biotechnology and Bioengineering 52: 685--695. Kuba T, Smolders G, van Loosdrecht MCM, and Heijnen JJ (1993) Biological phosphorus removal from wastewater by anaerobic-anoxic sequencing batch reactor. Water Science and Technology 27(5–6): 241--252. Kuba T and van Loosdrecht MCM (1996) Phosphorus and nitrogen removal with minimal COD requirement by integration of denitrifying dephosphatation and nitrification in a two-sludge system. Water Research 30(7): 1702--1710. Lawrence and McCarty (1972) Unified basis for biological treatment design and operation. Journal of the Sanitary Engineering Division – ASCE 96(SA3): 757--778. Lee BJ, Wentzel MC, and Ekama GA (2006) Measurement and modelling of ordinary heterotrophic organism active biomass concentration in anoxic/aerobic activated sludge mixed liquor. Water Science and Technology 54(1): 1--10. Lilley ID, Wentzel MC, Loewenthal RE, Ekama GA and Marais GvR (1991) Acid fermentation of primary sludge at 201C. Proceedings of the 2nd Biennial Water Institute of Southern Africa (WISA) Conference, Kempton Park, May. Lindrea KC, Pigdon SP, Boyd B, and Lockwood GA (1994) Biomass characterization in a nitrification-denitrification biological enhanced phosphorus removal (NDBEPR) plant during start-up and subsequent periods of good and poor phosphorus removal. Water Science and Technology 29(7): 91--100. Loewenthal RE and Marais GvR (1977) Carbonate Chemistry of Aquatic Systems: Theory and Application. Michigan, MI: Ann Arbor Science. Lo¨tter LH (1985) The role of bacterial phosphate metabolism in enhanced phosphorus removal from the activated sludge process. Water Science and Technology 17(11/12): 127--139. Lo¨tter LH, Wentzel MC, Ekama GA, and Marais GvR (1986) A study of selected characteristics of Acinetobacter spp. isolated from activated sludge in anaerobic/ anoxic/aerobic and aerobic systems. Water SA 12(4): 203--208. Loveless JE and Painter HA (1968) The influence of metal ion concentration and pH value in the growth of Nitrosomonas strain isolated from activated sludge. Journal of General Microbiology 52: 1--14. Ludzack FJ and Ettinger MB (1962) Controlling operation to minimize activated sludge effluent nitrogen. Journal of Water Pollution Control Federation 34: 920--931. Malan WN and Gouws EP (1966) Geaktiveerde slyk vir rioolwatersuiwering op Bellville. Research Report. Council for Scientific and Industrial Research, November 1966. Mamais D and Jenkins D (1992) The effects of MCRT and temperature on enhanced biological phosphorus removal. Water Science and Technology 26(5/6): 955--965. Mamais D, Jenkins D, and Pitt P (1993) A rapid physical – chemical method for the determination of readily biodegradable soluble COD in municipal wastewater. Water Research 27(1): 195--197. Marais GvR (1974) Faecal bacterial kinetics in stabilization ponds. Journal of the Environmental Engineering Division – ASCE 100(EE1): 119--139. Marais GvR and Ekama GA (1976) The activated sludge process part 1–Steady state behaviour. Water SA 2(4): 163--200. Marais GvR, Loewenthal RE, and Siebritz IP (1983) Review: Observations supporting phosphate removal by biological excess uptake. Water Science and Technology 15(3/4): 15--41.

525

Martin HG, Ivanova N, Kunin V, et al. (2006) Metagenomic analysis of two enhanced biological phosphorus removal (EBPR) sludge communities. Nature Biotech 24(10): 1263--1269. Maszenan AM, Seviour RJ, Patel BKC, et al. (2000) Three isolates of novel polyphosphate-accumulating Gram-positive cocci, obtained from activated sludge, belong to a new genus, Tetrasphaera gen. nov., and description of two new species, Tetrasphaera japonica sp, nov and Tetrasphaera australiensis sp nov. International Journal of Systematic and Evolutionary Microbiology 50: 593--603. Maurer M, Abramovich D, Siegrist H, and Gujer W (1999) Kinetics of biologically induced phosphorus precipitation in waste-water treatment. Water Research 33(2): 484--493. Maurer M, Gujer W, Hany R, and Bachmann S (1997) Intracellular carbon flow in phosphorus accumulating organisms from activated sludge systems. Water Research 31(4): 907--917. Mbaya A, Germanis J, Wilsenach JA, and Ekama GA (2010) The impact of source separation of urine on biological nutrient removal activated sludge plants. In: Proceedings of the WISA Biennial Conference, Durban, 18–22 April. Water Institute of Southern Africa, PO Box 6011, Halfway House, 1685, South Africa. Mbewe A, Wentzel MC, and Ekama GA (1994) Characterization of municipal wastewaters. Research Report W84, Department of Civil Engineering, Rondebosch: University of Cape Town. Mbewe A, Wentzel MC, Lakay MT, and Ekama GA (1998) Characterization of the carbonaceous materials in municipal wastewaters. In: Proceedings of the WISA Biennial Conference, Cape Town, 3–5 May, 2-E7. Water Institute of Southern Africa, PO Box 6011, Halfway House, 1685, South Africa. McKinney RE and Symons JM (1964) Growth and endogenous metabolism – a discussion. Proceedings of the 1st International Conference on Water Pollution Control. London, 1962. Oxford: Pergamon. McLaren AR and Wood RJ (1976) Effective phosphorus removal from sewage by biological means. Water SA 2(1): 47--50. Mebrahtu MK, Vogts M, and Ekama GA (2010) Anoxic-aerobic digestion of waste activated sludge from biological nitrogen and phosphorus removal systems. In: Proceedings of the WISA Biennial Conference. Durban, 18–22 April. Water Institute of Southern Africa, PO Box 6011, Halfway House, 1685, South Africa. Me´ganck M, Malnou D, Le Flohic P, Faup GM, and Rovel JM (1985) The importance of the acidogenic microflora in biological phosphorus removal. Water Science and Technology 17(11/12): 199--212. Mellin HKO, Lakay MT, Wentzel MC, and Ekama GA (1997) The effect of high temperatures (301C) on biological nutrient removal performance and AA (low F/M) filament bulking in ND and NDBEPR systems. Research Report No. W91. Rondebosch: Department of Civil Engineering, University of Cape Town. Metcalf and Eddy Inc (1991) Wastewater Engineering: Collection, Treatment, Disposal And Reuse, 3rd edn. New York: McGraw-Hill. ISBN 0-07-100824-1 Meyer RL, Saunders AM, and Blackall LL (2006) Putative glycogen accumulating organisms belonging to alphaproteobacteria identified through rRNA-based stable isotope probing. Microbiology–SGM 152: 419--429. Mino T, Arun V, Tsuzuki Y, and Matsuo T (1987) Effect of phosphorus accumulation on acetate metabolism in the biological phosphorus removal process. Paper presented at the IAWPRC Biological Phosphate Removal from Wastewater, Rome. Mino T, Liu WT, Kurisu F, and Matsuo T (1995) Modeling glycogen storage and denitrification capability of microorganisms in enhanced biological phosphate removal processes. Water Science and Technology 31(2): 25--34. Mino T, Van Loosdrecht MCM, and Heijnen JJ (1998) Microbiology and biochemistry of the enhanced biological phosphate removal process. Water Research 32(11): 3193--3207. Mino T, Wen-Tso L, Kurisu F, and Matsuo T (1994) Modelling glycogen storage and denitrification capability of microorganisms in enhanced biological phosphate removal processes. Paper presented at the Specialized Conference on Modelling of Wastewater Treatment. Kollekolle, Denmark. Monod J (1950) Technique of continuous culture – Theory and application (Translation from French). Annales de l’Institut Pasteur 79: 167. Monteith HD, Bridle TR, and Sutton PM (1980) Industrial waste carbon sources for biological denitrification. Progress in Water Technology 12(Tor): 127--141. Moosbrugger RE, Wentzel MC, Ekama GA, and Marais GvR (1992). Simple titration procedures to determine H2CO3  alkalinity and short-chain fatty acids in aqueous solutions containing known concentrations of ammonium, phosphate and sulphide weak acid/bases. Water Research Commission, Gezina, RSA. Muller AW, Wentzel MC, and Ekama GA (2006) Estimation of nitrification capacity of rock media trickling filters in external nitrification BNR. Water SA 32(5): 611--618. Murphy M and Lo¨tter LH (1986) The effect of acetate on polyphosphate formation and degradation in activated sludge with particular reference to Acinetobacter calcoaceticus: A microscopic study. Water SA 12: 63--66.

526

Biological Nutrient Removal

Musvoto EV, Casey TG, Ekama GA, Wentzel MC, and Marais GvR (1994) The effect of incomplete denitrification on anoxic-aerobic (low F/M) filament bulking in nutrient removal activated sludge systems. Water Science and Technology 29(7): 295--299. Musvoto EV, Samson K, Taljard M, Fawcett K, and Alexander WV (2002) Calculation of peak oxygen demand in the design of full scale nitrogen removal activated sludge plants. Water SA, Special Edn Wisa 2002: 56–60. Nakamura K, Ishikawa S, and Kawaharasaki M (1995) Phosphate-uptake and release activity in immobilized polyphosphate-accumulating bacterium Microlunatus phosphovorus strain NM-1. Journal of Fermentation and Bioengineering 80(4): 377--382. Neethling JB, Clark D, Pramanik A, Stensel HD, Sandino J, and Tsuchihashi R (2009) WERF nutrient challenge investigates limits of nutrient removal technologies. In: Proceedings of the 2nd IWA Specialized Conference Nutrient Management in Wastewater Treatment Processes. Krakow 6–9 September, 353–362. London: International Water Association (IWA), IWA Publishing. Nicholls HA (1975a) Biological Nitrogen and Phosphorus Removal in Large Plants. Internal Thesis Report, Department of Civil Engineering, University of Cape Town, Rondebosch, RSA. Nicholls HA (1975b) Full scale experimentation on the new Johannesburg extended aeration plant. Water SA 1(3): 121--132. Nicholls HA (1982) Application of the Marais–Ekama activated sludge model to large plants. Water Science and Technology 14(Cape Town): 581--598. Nicholls HA and Osborn DW (1979) Bacterial stress: Prerequisite for biological removal of phosphorus. Journal of the Water Pollution Control Federation 51: 557--569. Nicholls HA, Pitman AR, and Osborn DW (1985) The readily biodegradable fraction of sewage: Its influence on phosphorus removal and measurement. Water Science and Technology 17(11/12): 73--87. Nicholls HA, Osborn DW, and Pitman AR (1986) Biological phosphorus removal at the Johannesburg Northern and Goudkoppies wastewater purification plants. Water SA 12(1): 13--18. Oehmen A, Lemos PC, Carvalho G, Yuan Z, Keller J, and Blackall LL (2007) Advances in enhanced biological phosphorus removal: From micro to macro scale. Water Research 41(11): 2271--2300. Osborn DW and Pitman AR (1987) Improvement to the stability of the biological phosphate removal process at the Johannesburg Northern Works. In: Biological Phosphate Removal From Wastewaters, 261–272. London: International Water Association (IWA), IWA Publishing. Osborn DW, Lo¨tter LH, Pitman AR, and Nicholls HA (1989) Two-year study on the enhancement of biological phosphate removal by altering process feed composition (plant and laboratory studies). WRC Report No. 137/2/89. Pretoria: Water Research Commission. Osborn DW and Nicholls HA (1978) Optimisation of the activated sludge process for the biological removal of phosphorus. Progress in Water Technology 10(1/2): 261--277. Pagilla K, Sattayatewa C, Urgen-Demirtas M, and Baek S (2009) Effect of influent N speciation on organic nitrogen occurrence in activated sludge process effluents. In: Proceedings of the 2nd IWA Specialized Conference Nutrient Management in Wastewater Treatment Processes. Krakow 6–9 September, pp. 221–228. London: International Water Association (IWA), IWA Publishing. Parco V, du Toit GJG and Ekama GA (2009) The impact of high mixed liquor TSS concentration (3-13 gVSS/l) on the kinetics of N and P removal in biological nutrient removal activated sludge systems. Procs. 82nd Annual Water Environment Federation Technical Exhibition and conference - WEFTEC 2009. WEF, Washington DC. Parker DS, Kaufman WJ, and Jenkins D (1971) Physical conditioning of activated sludge floc. Journal of the Water Pollution Control Federation 43(9): 1817--1833. Pattarkine VM and Randall CW (1999) The requirement of metal cations for enhanced biological phosphorus removal by activated sludge. Water Science and Technology 40(2): 159--165. Pilson RA, Ekama GA, Wentzel MC, and Casey TG (1995) The effect of temperature on denitrification kinetics and biological excess phosphorus removal in biological nutrient removal systems in temperate climates (121 C–201 C). Research Report W86, Rondebosch: Department of Civil Engineering, University of Cape Town. Pincince AB, Braley BG, Sangrey KH, and Reardon RD (1995) Minimizing costs of activated sludge systems: Application to the Deer Island treatment plant. In: Proceedings of the 68th Annual WEF Conference and Exposition, Miami, vol. 1, pp. 693–699. Water Environment Federation, Washington, DC. Pitman AR (1991) Design considerations for nutrient removal activated sludge plants. Water Science and Technology 23(4–6): 781--790. Pitman AR (1999) Management of biological nutrient removal plant sludges–change the paradigms? Water Research 33(5): 1141--1146.

Pitman AR, Vandalsen L, and Trim BC (1988) Operating experience with biological nutrient removal at the Johannesburg Bushkoppie works. Water Science and Technology 20(4–5): 51--62. Pitman AR, Venter SLV, and Nicholls HA (1983) Practical experience with biological phosphorus removal plants in Johannesburg. Water Science and Technology 15(3/4): 233--259. Poinapen J and Ekama GA (2010) Biological sulphate reduction using primary sewage sludge in a upflow anaerobic sludge bed reactor–Part 5: Development of a steady state model. Water SA 36(2): 193--202. Randall AA, Benefield LD, and Hill WE (1994) The effect of fermentation products on enhanced biological phosphorus removal, polyphosphate storage, and microbial population dynamics. Water Science and Technology 30(6): 213--219. Randall CW, Marshall DW, and King PH (1970) Phosphate release in activated sludge process. Journal of the Sanitary Engineering Division, ASCE 96(SA2): 395--408. Randall EW, Wilkinson A, and Ekama GA (1991) An instrument for the direct determination of oxygen utilization rate. Water SA 17(1): 11--18. Rensink JH, Donker HJGW, and Vries HPD (1981) Biological P-removal in domestic wastewater by the activated sludge process. In: Proceedings of the 5th Europe Sewage and Refuse Symposium, Munich. European Water Association, Hennep, D53773, Germany or International Solid Waste Association (ISWA), Vienna, A-1080, Austria. Richard MG, Jenkins D, Hao O, and Shimizu G (1982) The isolation and characterization of filamentous micro-organisms from activated sludge bulking. Progress Report No. 81-2. Berkeley: SERL, University of California. Riddell MDR, Lee JS, and Wilson TE (1983) Method for estimating the capacity of an activated sludge plant. Journal of the Water Pollution Control Federation 55(4): 360--368. Samson KA and Ekama GA (2000) An assessment of sewage sludge stability with a specific oxygen utilization rate (SOUR) test method. Water Science and Technology 42(9): 37--40. Satoh H, Mino T, and Matsuo T (1992) Uptake of organic substrates and accumulation of polyhydroxyalkanoates linked with glycolysis of intracellular carbohydrates under anaerobic conditions in the biological excess phosphate removal processes. Water Science and Technology 26(5–6): 933--942. Saunders AM, Mabbett AN, McEwan AG, and Blackall LL (2007) Proton motive force generation from stored polymers for the uptake of acetate under anaerobic conditions. FEMS Microbiology Letters 274(2): 245--251. Sehayek L and Marais GVR (1981) Supplementary phosphorus removal by side-line addition of lime in the activated sludge process. Research Report W40. Rondebosch: Department of Civil Engineering, University of Cape Town. Sen D, Mitta P, and Randall CW (1994) Performance of fixed film media integrated in activated sludge reactors to enhanced nitrogen removal. Water Science and Technology 30(11): 13--24. Setter LR, Carpenter WT, and Winslow GC (1945) Practical application of modified sewage aeration. Sewage Works Journal 17(4): 669. Seviour RJ, Mino T, and Onuki M (2003) The microbiology of biological phosphorus removal in activated sludge systems. FEMS Microbiology Reviews 27(1): 99--127. Shapiro J, Levin GV, and Humberto HZ (1967) Anoxically induced release of phosphate in wastewater treatment. Journal of the Water Pollution Control Federation 39: 1810--1818. Siebritz IP, Ekama GA, and Marais GVR (1980) Excess biological phosphorus removal in the activated sludge process at warm temperature climate. In: Proceedings of the Waste Treatment and Utilization 2, pp. 233–251, Toronto: Pergamon. Siebritz IP, Ekama GA, and Marais GVR (1983) A parametric model for biological excess phosphorus removal. Water Science and Technology 15(3/4): 127--152. Simpkins MJ and McLaren AR (1978) Consistent biological phosphate and nitrate removal in an activated sludge plant. Progress in Water Technology 10(5/6): 433--442. Smolders GJF, van der Meij J, van Loosdrecht MCM, and Heijnen JJ (1994a) Stoichiometric model of the aerobic metabolism of the biological phosphorus removal process. Biotechnology and Bioengineering 44(7): 837--848. Smolders GJF, van der Meij J, van Loosdrecht MCM, and Heijnen JJ (1994b) Model of the anaerobic metabolism of the biological phosphorus removal process: Stoichiometry and pH influence. Biotechnology and Bioengineering 43: 461--470. Smolders GJF, Meij Jvd, Loosdrecht MCMv, and Heijnen JJ (1995) A structured metabolic model for anaerobic and aerobic stoichiometry and kinetics of the biological phosphorus removal process. Biotechnology and Bioengineering 47: 277--287. Sneyders MJ, Wentzel MC, and Ekama GA (1997) The effect of unstabilized landfill leachate addition on biological nutrient removal performance in activated sludge

4.15 Biofilms in Water and Wastewater Treatment Z Lewandowski, Montana State University, Bozeman, MT, USA JP Boltz, CH2M HILL, Inc., Tampa, FL, USA & 2011 Elsevier B.V. All rights reserved.

4.15.1 4.15.2 4.15.2.1 4.15.2.2 4.15.2.2.1 4.15.2.2.2 4.15.2.2.3 4.15.2.2.4 4.15.2.2.5 4.15.2.2.6 4.15.2.2.7 4.15.2.2.8 4.15.3 4.15.3.1 4.15.3.1.1 4.15.3.1.2 4.15.3.2 4.15.3.3 4.15.3.4 4.15.3.4.1 4.15.3.4.2 4.15.3.4.3 4.15.3.4.4 4.15.3.4.5 4.15.3.5 4.15.3.5.1 4.15.3.5.2 4.15.3.5.3 4.15.3.5.4 4.15.3.5.5 4.15.3.5.6 4.15.4 4.15.4.1 4.15.4.1.1 4.15.4.1.2 4.15.4.2 4.15.4.3 4.15.4.4 References

Introduction Part I: Biofilm Fundamentals Biofilm Formation and Propagation The Concepts of Biofilms and Biofilm Processes Quantifying microbial activity, hydrodynamics, and mass transport in biofilms Biofilm heterogeneity and its effects Biofilm activity Quantifying local biofilm activity and mass transport in biofilms from microscale measurements Horizontal variability in diffusivity and microbial activity in biofilms Mechanism of mass transfer near biofilm surfaces Biofilm processes at the macroscale and at the microscale Biofilms in conduits Part II: Biofilm Reactors Application of Biofilm Reactors Techniques for evaluating biofilm reactors Graphical procedure Empirical and Semi-Empirical Models Mathematical Biofilm Models for Practice and Research Biofilm Model Features Attachment and detachment process kinetics and rate coefficients Concentration gradients external to the biofilm surface and the mass transfer boundary layer Diffusivity coefficient for the rate-limiting substrate inside the biofilm Parameters: estimation and variable coefficients Calibration protocol Biofilm Reactors in Wastewater Treatment Biofilm reactor compartments Moving bed biofilm reactors Biologically active filters Expanded and fluidized bed biofilm reactors Rotating biological contactors Trickling filters Part III. Undesirable Biofilms: Examples of Biofilm-Related Problems in the Water and Wastewater Industries Biofilms on Metal Surfaces and MIC Differential aeration cells on iron surfaces SRB corrosion Biofilms on Concrete Surfaces: Crown Corrosion of Sewers Biofilms on Filtration Membranes in Drinking Water Treatment Biofilms on Filtration Membranes in Wastewater Treatment

4.15.1 Introduction Fundamental principles describing biofilms exist as a result of focused research. The use of reactors for the treatment of municipal wastewater is a common application of biofilms. Applied research exists that provides a basis for the mechanistic understanding of biofilm reactors. The empirical information derived from such applied research has been used to develop design criteria for biofilm reactors and remains the basis for biofilm reactor design despite the emergence of

529 529 529 530 531 532 532 533 535 535 536 538 541 542 542 543 543 544 546 546 547 548 548 548 549 549 549 550 556 556 558 562 563 564 564 565 565 566 567

mathematical models as reliable tools for research and practice. Unfortunately, little information exists to bridge the gap between our current understanding of biofilm fundamentals and reactor-scale empirical information. Therefore, there is a clear dichotomy in literature: micro- (biofilm) and macro(reactor) scales. This chapter highlights the division. Part I is dedicated to basic research and communicating the state of the art with respect to understanding biofilms. Part II is practice oriented and describes the use of biofilms for the sanitation of municipal wastewater. A basis for addressing this

529

530

Biofilms in Water and Wastewater Treatment

disconnection is presented by (1) describing the fundamental biofilm principles that can be uniformly applied to biofilms in several disciplines extending from medicine to environmental biotechnology and (2) describing a fundamentalbased approach in order to understand and apply biofilms in reactors. The use of mathematical biofilm models is common in both research and practice, but only a cursory presentation of their mathematical description is presented here. Finally, Part III gives examples of undesirable biofilms in water and wastewater industries and describes the attempts to mitigate their effects. Metabolic reactions mediated by microorganisms residing in biofilms promote the biodeterioration of materials, including metals, concrete, and plastics. It is estimated that microbially influenced corrosion (MIC) alone costs the US economy billions of dollars every year.

4.15.2 Part I: Biofilm Fundamentals 4.15.2.1 Biofilm Formation and Propagation Biofilm formation is a process that consists of a sequence of steps. It begins with the adsorption of macromolecules (e.g., proteins, polysaccharides, nucleic acids, and humic acids) and smaller molecules (e.g., fatty acids, lipids, and pollutants such as polyaromatic hydrocarbons and polychlorinated biphenyls) onto surfaces. These adsorbed molecules form conditioning films which may have multiple effects, such as altering the physicochemical characteristics of the surface, acting as a concentrated nutrient source for microorganisms, suppressing or enhancing the release of toxic metal ions from the surface, detoxifying the bulk solution through the adsorption of inhibitory substances, supplying the nutrients and trace elements required for a biofilm, and triggering biofilm sloughing. Once the surface is prepared, cells begin to attach. The initial stages of biofilm formation are well documented, mostly because acquiring images of microorganisms at this stage of biofilm formation is relatively easy. The adherence of bacteria to a surface is followed by the production of slimy adhesive substances, extracellular polymeric substances (EPS). These are predominantly made of polysaccharides and proteins. Although the association of EPS with attached bacteria has been well documented in the literature, there is little evidence to suggest that EPS participates in the initial stages of adhesion. However, EPS definitely assists the formation of mature biofilms by forming a slimy substance called the biofilm matrix. Figure 1 shows the steps in the formation of mature biofilms. The existence of these three phases of biofilm development, as depicted in Figure 1, is generally acknowledged, although the terminology may vary among authors. For example, Notermans et al. (1991) called these phases: (1) adsorption, (2) consolidation, and (3) colonization. Once a mature biofilm has been established on a surface, it actively propagates and eventually covers the entire surface. The mechanisms of propagation in mature biofilms are more complex than those of initial attachment, and several of these mechanisms of biofilm propagation are depicted in Figure 2. Although biofilms can be seen with an unaided eye, imaging their structure, microbial community structure, and

Biofilm formation Attachment

Colonization

Growth

Bulk fluid

Surface Figure 1 Steps in biofilm formation. & 1995 Center for Biofilm Engineering, MSU-BOZEMAN.

distribution of EPS requires the use of several types of microscopy combined with various probes, such as fluorescent in situ hybridization (FISH) probes and fluorescent proteins (FPs) used as reporter genes. The favorite types of microscopy among biofilm researchers are those that allow the examination of living and fully hydrated biofilms. In addition, sophisticated image acquisition devices are often needed that can selectively stimulate and image various probes when more than one type of multicolored probe is used simultaneously. Using these techniques in conjunction with a suitable microscopy, biofilm researchers can detect the presence of the selected physiological groups of microorganisms in the biofilm, their position in the biofilm with respect to other microorganisms and surface, and even their physiological state – dead, injured, or alive. The in vitro FISH techniques, popular in medical diagnostics, require that DNA or RNA be isolated from the sample and separated on a gel, and that the fluorescent probes then be added to the sample. The in situ variety of the hybridization technique, which is extensively used in biofilm research, does not require isolating DNA or RNA prior to the use of the probes; instead, the probes are hybridized to the respective nucleotide sequences inside the cells (Biesterfeld et al., 2001; Delong et al., 1999; Ito et al., 2002; Jang et al., 2005; Manz et al., 1999). In situ hybridization uses fluorescence-labeled complementary DNA or RNA probes, often derived from fragments of DNA that have been isolated, purified, and amplified. In microbial ecology, ribosomal RNA in bacterial cells is targeted by fluorescencelabeled oligonucleotide probes. Figure 3 shows an image of manganese-oxidizing bacteria (MOB) Leptotrix discophora stained with a FISH probe (green) and counterstained with propidium iodide (red). Propidium iodide is a general stain which is quite popular with biofilm researchers (GrayMerod et al., 2005; McNamara et al., 2003; Nancharaiah et al., 2005). In mature biofilms, microorganisms are imbedded in the layer of EPS. Figure 4 shows an image of a mature biofilm acquired using scanning electron microscopy (SEM). It shows microbes embedded in a matrix of EPS attached to a surface, although the EPS in this image were reduced to an entangled network of dry strands because the sample had to be dehydrated before the biofilm was imaged using electron microscopy.

Biofilms in Water and Wastewater Treatment

531

Streaming Detaching

Seeding dispersal

Rippling

Rolling

Figure 2 Mechanisms of biofilm propagation (MSU-CBE, P.Dirx).

Figure 4 SEM image of a biofilm of Desulfovibrio desulfuricans G20 embedded in EPS (Beyenal et al., 2004).

Figure 3 L. discophora stained with FISH probes and counterstained with propidium iodide. Red indicates cells that were stained with propidium iodide, and green indicates cells that react positively to the fluorescent FISH probe. Yellow indicates green and red overlay. The scale bar is 20 mm (Campbell, 2003).

4.15.2.2 The Concepts of Biofilms and Biofilm Processes It is difficult to offer precise definitions of biofilms and biofilm processes that will satisfy everyone who is interested in studying biofilms and biofilm-based technologies. Several currently used definitions have roots in historical approaches to biofilm studies. These approaches initially referred to biofilms as physical objects – microbial deposits on surfaces – but later expanded the concept to consider biofilms as a mode of

microbial growth, an alternative to microbial growth in suspension. Life scientists often emphasize the definitions that refer to biofilms as a mode of microbial growth. Engineers often find that the definitions that refer to aggregates of microorganisms which are embedded in a matrix composed of microbially excreted EPS and attached to a surface are useful for their applications. Here, we will refer to biofilms as microorganisms and microbial deposits attached to surfaces. We will use the term biofilm processes in reference to all physical, chemical, and biological processes in biofilm systems that affect, or are affected by, the rate of biofilm deposition or the microbial activity in biofilms. Biofilm processes are carried out in biofilm reactors. Colloquially, the terms biofilm reactors and biofilm systems are used interchangeably. However, biofilm systems exist with or without human intervention, while biofilm reactors are produced by our actions.

532

Biofilms in Water and Wastewater Treatment

When we promote or suppress a biofilm process in a biofilm system, or even when we quantify a biofilm process in a biofilm system without affecting its rate, the biofilm system becomes a biofilm reactor. For example, wetlands can be natural or constructed. However, even natural wetlands become biofilm reactors once we start monitoring biofilm processes in them. We will use the term biofilm system to refer to a group of compartments and their components determining biofilm structure and activity. Biofilm systems are composed of four compartments:

• • • •

the the the the

surface to which the microorganisms are attached; biofilm (the microorganisms and the matrix); solution of nutrients; and gas phase (if present).

Each compartment of a biofilm system can have a number of components. The exact number of components in each compartment may vary, depending on the needs of a particular description. For example, for some analyses it may be convenient to identify two components of the biofilm: (1) the EPS (matrix) and (2) the microorganisms. In another study, it may be convenient to identify three components of the biofilm: (1) the EPS, (2) the microorganisms, and (3) the particular matter trapped in the matrix. Similarly, in some studies it may be convenient to single out two components of the surface – (1) the bulk material and (2) the biomineralized deposits – or, if MIC is studied, it may be convenient to describe the surface by identifying three components: (1) the metal substratum, (2) the corrosion products, and (3) the biomineralized deposits on the surface. The needs of the specific study or analysis dictate the number of components identified in each compartment of the biofilm system. Biofilm studies can be characterized as studies of the relations among the compartments, the properties of one or more compartments, or one or more components of a compartment. Among many factors that are used to quantify biofilm processes, biofilm activity is most often used. Biofilm reactors are often designed and operated to optimize biofilm activity, as are the biofilm reactors used for wastewater treatment discussed later in the text. Typically, biofilm activity is identified with the rate of utilization of the growth-limiting nutrient. In some instances, however, rates other than the rate of substrate utilization or biofilm accumulation are better descriptors of the system dynamics. For example, in studies of MIC, the rate of anodic dissolution of the metal affected by the process may be a more useful descriptor of biofilm activity than the rate at which the growth-limiting substrate is utilized. The choice of the process for evaluating biofilm activity is dictated by the nature of the study, and sometimes by analytical convenience. Monitoring the rate of biofilm accumulation is important in many applications, whether we want to enhance or inhibit the growth of biofilms. The methods employed include optical microscopy (Bakke and Olsson, 1986; Bakke et al., 2001), measuring light intensity reflected from microbially colonized surfaces (Bremer and Geesey, 1991; Cloete and Maluleke, 2005), collecting and analyzing images of biofilm depositions (Milferstedt et al., 2006; Pons et al., 2009), surface sensors based on piezoelectric devices (Nivens et al., 1993; Pereira

et al., 2008), and electrochemical sensors in which stainless steel electrodes change their electrochemical behavior as a result of biofilm deposition (Licina et al., 1992; Borenstein and Licina, 1994).

4.15.2.2.1 Quantifying microbial activity, hydrodynamics, and mass transport in biofilms Microbial activity (biofilm activity), hydrodynamics, and mass transport in biofilms are difficult to discuss separately as they affect each other in many ways. Biofilm activity at the microscale is quantified as the flux, from the bulk solution to the biofilm surface, of the substance selected for evaluating biofilm activity. Since fluxes at the microscale are quantified locally, rather than averaged over the entire surface area as is done when biofilm activity is evaluated at the macroscale, the concentration profiles of the selected substance must be measured with microsensors to assure adequate spatial resolution. The idealized model of hydrodynamics and mass transfer in biofilms shown in Figure 5 is a good starting point for a discussion of biofilm activity at the microscale. In this model the overall flow velocity in the main stream is considered to be the average flow velocity, Cb. This decreases toward the surface of the biofilm, as required by hydrodynamics, and reaches concentration Cs at the biofilm surface. The layer of liquid just above the biofilm surface, where the flow velocity decreases as a result of proximity to the surface, is the hydrodynamic boundary layer, and it is denoted by j. As the flow velocity decreases toward the biofilm surface, the mechanism of mass transport changes from being dominated by convection at locations away from the biofilm, where the flow velocity is high, to being dominated by diffusion at locations near the biofilm surface, where the flow velocity is low. As the microorganisms in the biofilm consume nutrients at the rate at which they are delivered and the mass transport becomes less efficient near the biofilm surface, the nutrient concentration decreases near the surface, forming a nutrient concentration profile within the hydrodynamic boundary layer. The layer of liquid above the biofilm surface where the nutrient concentration decreases is the mass transport boundary layer, and it is denoted by LL and RL is the mass transfer resistance external to the biofilm.

Substrate concentration profile

Flow velocity profile

N = k(Cb − Cs) k =

vb

DW 1 = RL LL Cb C LF

LL

ϕ Biofilm Substratum

Figure 5 Profiles of flow velocity and growth-limiting nutrient concentration near the surface of an idealized biofilm.

Biofilms in Water and Wastewater Treatment 7 6 CO2 (mg l−1)

4.15.2.2.2 Biofilm heterogeneity and its effects The term biofilm heterogeneity refers to the extent of the nonuniform distribution of any selected constituent in any of the compartments of the biofilm system, such as the distribution of the biomass, selected nutrients, selected products of microbial metabolism, or selected groups of microorganisms. Since there are many choices for the constituents selected to evaluate biofilm heterogeneity, the term biofilm heterogeneity is usually combined with an adjective referring to the selected constituent, such as structural heterogeneity, chemical heterogeneity, or physiological heterogeneity. The term biofilm heterogeneity was initially used exclusively to refer to the nonuniform distribution of the biomass in a biofilm. As time has passed, more types of heterogeneity have been described, and the term biofilm heterogeneity is not self-explanatory anymore: the specific feature of the biofilm with respect to which the heterogeneity is quantified needs to be specified. Quantifying biofilm heterogeneity is equivalent to quantifying the extent of nonuniform distributions, such as the distribution of biomass in the biofilm. Several tools from the statistical toolbox are available for evaluating the extent of nonuniform distribution; the most popular is the standard deviation. The procedure for estimating the heterogeneity of a selected constituent of a biofilm is identical with the procedure for evaluating the standard deviation of a set of experimental data with one important difference: the deviations from the average are not due to errors in measurement but reflect a feature of the biofilm – heterogeneity. One of the most profound effects of biofilm heterogeneity is that microscale measurements in biofilms deliver different results at different locations. This is an obvious concern as most models referring to microbial growth and activity have been developed for well-mixed reactors, in which the result of a measurement does not depend on the location. Figure 6 shows this effect: three very different profiles of carbon dioxide concentration were measured at three locations in a biofilm. Because of the biofilm heterogeneity, it is impossible to determine a representative location to make the local measurements of biofilm activity that are used to validate models of biofilm processes. To include the effects of biofilm heterogeneity in mathematical models of biofilm processes, the extent of these effects – the spatial variability of the features measured in biofilms – needs to be evaluated experimentally using tools that can take measurements in biofilms to a high spatial resolution. Such tools are routinely used in biofilm research in the form of microelectrodes and various types of microscopy, often enhanced with fluorescent probes. These types of measurements deliver information about selected locations in the biofilm, and their results are referred to as local properties. The most common such measurements are local biofilm activity, local mass transfer coefficient, local diffusivity, and local flow velocity. The definition of the local mass transport coefficient is derived from the measurement procedure: the coefficient of the mass transport of an electroactive species to the tip of an electrically polarized microelectrode. The local mass transport coefficient is measured using an amperometric microelectrode without a membrane operated at the limiting current condition (masstransfer-limited). Local diffusivity is computed from these

533

A B C

5 4 3 2 1 0 0

100 200 300 400 Distance from the bottom (µm)

500

Figure 6 Carbon dioxide concentration profiles measured perpendicularly to the bottom (substratum) at three locations in a biofilm microcolony.

measurements by calibrating local mass transport microelectrodes in gels of known diffusivities (Beyenal et al., 1998).

4.15.2.2.3 Biofilm activity Biofilm activity in a biofilm reactor can be evaluated from the mass balance on the growth-limiting nutrient in the reactor:

Biofilm activity ¼

ðCInfluent  CEffluent Þ  Q A

ð1Þ

where C is the concentration of the growth-limiting nutrient (kg m3), Q the volumetric flow rate in the reactor (m3 s1), and A the surface area covered by the biofilm (m2). Therefore, biofilm activity at the scale of the reactor is the average flux of nutrients across the biofilm surface, which corresponds to the approach delineated in Equations (12) and (13) used in graphical procedure to evaluate pilot-plant observations. Average biofilm activity in a reactor is a useful descriptor of reactor performance. However, when the underlying biofilm processes are to be studied, an image of local biofilm activity is often required. This information can be extracted from growth-limiting substrate concentration profiles measured at

534

Biofilms in Water and Wastewater Treatment 7

surface on the oxygen profile coincides with the inflection point of the nutrient concentration profile. It is not easy to determine the exact position of the surface, though. We use a simplified procedure, explained later in Figure 16, to find the approximate position of biofilm surface on concentration profiles measured with microelectrodes. One use of such data is to estimate the local biofilm activity in terms of the flux of the growth-limiting nutrient at the location where the profile was measured. The flux of the nutrient across the biofilm surface, JLF at the location of the measurement is computed as the product of the slope of the concentration profile at the biofilm bulk solution interface by the diffusivity coefficient in water of the substance whose concentration was measured:

Oxygen concentration (mg l−1)

6 5 4 3 2 1

JLF ¼ Dw

0 0

300

600

900

1200

1500

Distance from the bottom (µm) Figure 7 Oxygen concentration profile. The vertical line marks the approximate position of the biofilm surface (Rasmussen and Lewandowski, 1998).

selected locations in the biofilm, as shown in Figure 7. The results from the two scales of observation – (1) the local biofilm activity evaluated from the concentration profiles and (2) the average biofilm activity evaluated from the mass balances around the reactor – provide different types of information. The measurements at the microscale deliver information that cannot be extracted from the measurements at the macroscale. For some biofilm processes, it is important to quantify the extreme values of biofilm activity because the locations in the biofilm where these extreme values occur exhibit extreme properties. For example, in studying MIC, which causes highly localized damage to metal surfaces, it is important to evaluate the extreme values of biofilm activity because the extreme, and highly localized, microbial activity in biofilms determines the extent of microbial corrosion. The average biofilm activity estimated from measurements at the macroscale cannot deliver this information.

4.15.2.2.4 Quantifying local biofilm activity and mass transport in biofilms from microscale measurements The profiles of flow velocity and growth-limiting substrate concentration shown in the conceptual image depicted in Figure 5 can be measured experimentally. Their interpretation leads to a better understanding of the processes occurring in biofilms. Figure 7 shows an oxygen concentration profile measured in a biofilm using an oxygen microelectrode. Nutrient concentration profiles, such as the one shown in Figure 7, are composed of two parts, the part above and the part below the biofilm surface. Different factors shape these parts of the profile: the shape of the profile above the surface is dominated by bulk liquid hydrodynamics, whereas the shape of the profile below the surface is dominated by microbial respiration in the biofilm. These two parts are described by different equations but are connected at the biofilm surface by the requirement of oxygen flux continuity. The position of the

  dC dx ðxxs Þ¼0

ð2Þ

where Dw is the diffusivity in water of the substance selected for the evaluation of biofilm activity, usually the growthlimiting nutrient (m2 s1). Diffusivity of this substance in the biofilm is not constant, but instead it varies with distance, as explained below. Early mathematical descriptions of biofilm activity and the shape of the concentration profile within the biofilm were based on the conceptual model of so-called uniform biofilms, depicting biomass uniformly distributed in the space occupied by the biofilm (Atkinson and Davies, 1974; Williamson and McCarty, 1976). Formally, these early mathematical models of microbial activity in biofilms imitated the models of microbial activity in suspension, with the addition of mass transport resistance. They quantified the equilibrium between the rate of utilization of the growth-limiting nutrient and the rate of mass transport in one dimension, toward the surface:

 2      qC q C mmax Xf C ¼ Df  ; qt f qx2 f Yx=s Ks þ C

0 r x r xs

ð3Þ

At steady state, this equation delivers

Df

d 2C mmax CXf ¼ 2 Yx=s ðKs þ CÞ dx

ð4Þ

Two boundary conditions were generally used to specify the concentrations of oxygen at the bottom and surface of the biofilm:



dC dx

 ¼ 0;

Cðx¼xs Þ ¼ Cs;

t0

ð5Þ

ðx¼0Þ

where Df is the averaged effective diffusivity of growth-limiting nutrient in the biofilm (m2 s1); x the distance from the bottom (m); xs the distance from the biofilm surface in the new system of coordinates (m); Xf the averaged biofilm density (kg m3); Yx/s the yield coefficient (kg microorganisms/kg nutrient); mmax the maximum specific growth rate (s1); Ks the Monod half-rate constant (kg m3); C the growth-limiting substrate concentration (kg m3); and Cs the growth-limiting substrate concentration at the biofilm surface (kg m3). These early models were subsequently refined by adding additional factors affecting biofilm processes, such as bacterial

Biofilms in Water and Wastewater Treatment

  d dC d 2 C dDfl dC mmax CXfl Dfl ¼ Dfx 2 þ ¼ dx dx Yx=s ðKs þ CÞ dx dx dx

ð6Þ

where Dfl is the local effective diffusivity of the growth-limiting nutrient (m2 s1) and Xfl the local biofilm density (kg m3). Accepting that diffusivity and biofilm density are variable introduces two new variables into the equation, and functions describing changes in effective diffusivity and biofilm density need to be quantified before the equations can be solved. Experimental data show that density changes are surprisingly regular in biofilms and can be described as a linear function of biofilm depth. Relative surface-averaged effective diffusivity

0.70 0.65

D*fz = 0.001z + 0.2968

0.60 0.55 D fz

growth and decay in a steady-state biofilm (Rittmann and McCarty, 1980a, 1980b) and then the model was extended to include unsteady states and dual nutrient limitations (Rittmann and Brunner, 1984; Rittmann and Dovantzis, 1983). One of the most popular biofilm models, initially marketed as a software called BIOSIM (Wanner and Gujer, 1986), was later improved to include irregular biofilm structure and renamed AQUASIM (Wanner et al., 1995; Wanner and Reichert, 1996). The growing popularity of the conceptual model of heterogeneous biofilms coincided with the growing popularity of cellular automata (CA) (Wolfram, 1986), and it is not surprising that the heterogeneous biofilm structures were modeled using CA procedures (Wimpenny and Colasanti, 1997a, 1997b). Soon after, Picioreanu et al. (1998a, 1998b) improved this model using more realistic assumptions and used differential equations to describe mass transport with the discrete model describing the structure (Picioreanu et al., 1998a, 1998b). Since its early applications, CA remains the most popular model used to generate biofilm structure. Further improvement of the biofilm model came from Kreft et al. (2001), who developed a two-dimensional (2-D) multinutrient, multi-species model of nitrifying biofilms to predict biofilm structures, that is, surface enlargement, roughness, and diffusion distance. These authors compared the predicted structure of the biofilm with the predictions of the biomass (cells and EPS)-based model developed by Picioreanu et al. (1998a, 1998b), and concluded that the two models had similar solutions. Meanwhile, biofilm researchers urgently needed mathematical description of the biofilm processes that could be used to describe recent progress in understanding biofilm processes. The main problems that needed to be addressed were horizontal and vertical profiles in mass transport and activity in biofilms. These were experimentally verified and the assumption that the effective diffusivity and biofilm density were constant across the biofilm had become difficult to defend. Biofilm diffusivity decreases toward the bottom of the biofilm and biofilm density increases. There have been attempts to include these results in the modeling of biofilm processes but they lead to more complicated mathematical expressions in which diffusivity and biofilm density are functions of distance. To simplify these expressions it is possible to model a biofilm as a stack of layers with constant diffusivity and density, which change from layer to layer rather than continuously. At steady state, this approach delivers the mass transport and activity related to the local properties of the biofilm:

535

*

0.50 0.45 0.40 0.35 0.30 0.25

50

100

150

200

250

300

Distance from the bottom, z (µm) Figure 8 The surface-averaged relative effective diffusivity (Dfz*) is multiplied by the diffusivity of the growth-limiting nutrient in the water to calculate the surface-averaged effective diffusivity (Dfz). Since, in the example, the growth-limiting nutrient is oxygen, to calculate the effective diffusivity of oxygen at various distances from the bottom, we must multiply the relative effective diffusivity at various distances from the bottom by the diffusivity of oxygen in water (2.1  105 cm2 s1) (Beyenal and Lewandowski, 2005).

profile, reproduced from Beyenal and Lewandowski (2005), is shown in Figure 8. Assuming that biofilm density varies with depth in a linear fashion, as shown in Figure 8, the diffusivity gradient (x) is constant:

dDfx ¼z dx

ð7Þ

At steady state, this simplifies Equation (5) to the form

Dfl

d 2C dC mmax CXfl ¼ þz 2 dx dx Yx=s ðKs þ CÞ

ð8Þ

Further, it has been demonstrated that in biological aggregates, including biofilms, density is related to effective diffusivity (Fan et al., 1990):

Dfl ¼ 1 

0:43X0:92 fl 11:19 þ 0:27X0:99 fl

ð9Þ

Using this equation, we can estimate biofilm density from the variation in local effective diffusivity (Figure 9).

4.15.2.2.5 Horizontal variability in diffusivity and microbial activity in biofilms Concentration profiles of growth-limiting nutrients, such as the one shown in Figure 7, are taken at a specific location in a biofilm. Based on the results, the biofilm activity at that location can be computed. However, when the next profile is taken at another location, even as close as several micrometers from the first location, the two profiles can be significantly different. This is not surprising, considering that biofilms are heterogeneous. However, it brings into question the practice of

536

Biofilms in Water and Wastewater Treatment

evaluating biofilm activity based on a single measurement at an arbitrarily selected location. For microscale measurements in stratified biofilms, the selected variable, such as local effective diffusivity or local dissolved oxygen concentration, is measured at locations on a grid (Figure 10). Grids are positioned at various distances from the bottom. The results are then presented as maps of the distributions of the selected parameter at the specified distances from the bottom, as shown in Figure 11. One of the main advantages of this approach is that it allows us to average the concentrations of oxygen at the selected distances from the bottom and arrive at a representative profile of oxygen that illustrates its distribution across the biofilm and also shows the deviations from the average due to biofilm heterogeneity. The maps of oxygen distributions shown in Figure 11 served to construct the representative profile of oxygen across this biofilm shown in Figure 12.

100 Pseudomonas aeruginosa (v = 3.2 cm s−1) Mixed culture (v = 1.6 cm s−1) Mixed culture (v = 3.2 cm s−1)

Biofilm density (g l−1)

80

60

40

20

0

0

100 200 300 400 Distance from the bottom, z (μm)

500

Figure 9 Variation in biofilm density with distance from the bottom (Beyenal et al., 1998).

4.15.2.2.6 Mechanism of mass transfer near biofilm surfaces When the local nutrient concentrations measured across a biofilm are plotted versus distance, they form a nutrient concentration profile. It would be expected that the shape of the nutrient concentration profile will follow the shape of the local mass transport coefficient profile when they are measured at the same location. It would also be expected that, at locations where the local mass transport coefficient is high, the local nutrient concentration will be high as well, at least higher than at a location where the local mass transport coefficient is low. Figure 13 shows profiles of oxygen concentration and local mass transport coefficient measured at the same location in a biofilm (Rasmussen and Lewandowski, 1998). As can be seen in Figure 13, the mass transport coefficient profile does not correlate well with the oxygen concentration profile. Approaching the biofilm surface, for example, the oxygen concentration decreases rapidly and reaches quite low levels at the biofilm surface, while the local mass transport coefficient remains quite high at that location. This observation seems difficult to explain: since there is no oxygen consumption in the bulk, the oxygen concentration profile would be expected to follow the shape of the mass transport coefficient profile much closer than it does in Figure 13. However, although these two profiles do not match, each of them is consistent with our knowledge of the system’s behavior. We expect to measure a low concentration of oxygen at the biofilm surface: this result fits the concept of a mass transfer boundary layer of high mass transport resistance above the biofilm surface. Measuring a high mass transport coefficient near the biofilm surface is also not surprising because, as we have estimated, convection is the predominant mass transport mechanism in that zone. The two features cannot coexist: high mass transport resistance and convection. To explain this apparent discrepancy, we need to examine the procedure for measuring flow velocity in biofilms. All available flow velocity measurements in biofilms report only one component of the

Figure 10 Microscale measurements in stratified biofilms. The selected variable, such as the local effective diffusivity or local dissolved oxygen concentration, is measured at the locations where the gridlines intersect. Such grids are positioned at various distances from the bottom (MSU-CBE, P.Dirx).

Biofilms in Water and Wastewater Treatment

400

100

200

500

300 m)

y (c

CL (mg l−1)

CL (mg l−1)

400

100

200

500

500

200 x ( 300 cm 400 )

300 m)

7 6 5 4 3 2 1 0 200 x ( 300 cm 400 ) 500

400 500

500

300 200 m) 100 ( y c

7 6 5 4 3 2 1 0 200 x ( 300 cm 400 )

160 µm

400 300 200 cm) 100 y(

7 6 5 4 3 2 1 0 200 x ( 300 cm 400 ) 500

500

500 100

400 300 m) 200 c ( y

500

CL (mg l−1)

7 6 5 4 3 2 1 0 200 500 400 300 300 x( 400 cm 500 100 200 y (cm) )

400 300 200 m) y (c 100

500

180 µm

CL (mg l−1)

280 µm

500

0 µm (boottom)

CL (mg l−1)

CL (mg l−1)

7 6 5 4 3 2 1 0 200 x ( 300 cm 400 ) 500

400 300 200 (cm) y 100

40 µm

7 6 5 4 3 2 1 0

y (c

320 µm

CL (mg l−1)

400 300 200 cm) 100 y(

200 µm

7 6 5 4 3 2 1 0 200 x ( 300 cm 400 ) 500

7 6 5 4 3 2 1 0 200 x ( 300 cm 400 )

7 6 5 4 3 2 1 0 200 x ( 300 cm 400 ) 500

CL (mg l−1)

360 µm

80 µm

CL (mg l−1)

7 6 5 4 3 2 1 0 200 x ( 300 cm 400 ) 500

240 µm

CL (mg l−1)

CL (mg l−1)

400 µm (top)

537

500

100

400 300 200 m) y (c

500

7 6 5 4 3 2 1 0 500 200 400 300 x( 300 400 cm 200 m) ) 500 100 y (c

Figure 11 Distribution of oxygen measured in a biofilm at the specified distances from the bottom (Veluchamy, 2006).

flow velocity vector, parallel to the bottom. Based on these results, we estimated that mass transport is controlled by convection near biofilms. However, the convective mass transport rate equals the nutrient concentration times the flow velocity component normal to the reactive surface. The component of the flow velocity parallel to the surface has nothing to do with the convective mass transport toward that surface. Consequently, the estimate of the mass

transport mechanism based on flow velocity holds only in the direction in which the flow velocity was measured. Indeed, when the flow near a surface is laminar, the laminas of liquid slide parallel to the surface, and there is little or no convection across these layers: the mass transport parallel to the surface is convective, while the mass transport perpendicular to the surface remains diffusive. This mechanism is visualized in Figure 14.

538

Biofilms in Water and Wastewater Treatment 7

Bulk

Biofilm 6

CSA (mg l−1)

5 4 3 2 1 0 0

100

200

300

400

Distance from the bottom, x (µm)

1.0

7

0.9

6

0.8 5

0.7 0.6

4

0.5 3

0.4

2

0.3 Oxygen

1

k/kmax

Dissolved oxygen concentration (mg l−1)

Figure 12 Surface averaged oxygen concentrations (CSA) and standard deviations computed for each data set in Figure 11. The average oxygen concentrations form a representative profile of oxygen concentration, characterizing the area covered with the biofilm, and the envelope of the standard deviation is a measure of the heterogeneity of the measured variable, oxygen concentration in this case (Veluchamy, 2006).

0.2 0.1

k/kmax

0 0

0.0

200 400 600 800 1000 1200 1400 Distance from substratum (µm)

Figure 13 Profiles of oxygen and local mass transfer coefficient through a thin biofilm cluster (’, dissolved oxygen; , local mass transfer coefficient). The vertical line marks the observed thickness of the biofilm. At distances of less than 30 mm, the wall effect caused the local mass transport coefficient to decrease. The biofilm thickness was 70 mm in this location. The value of k/kmax was only slightly affected by the presence of the biofilm up to a distance of less than 30 mm from the substratum (Rasmussen and Lewandowski, 1998).

4.15.2.2.7 Biofilm processes at the macroscale and at the microscale Accurate mathematical models are necessary for advances in biofilm research. Biofilm researchers use mathematical models of biofilm processes not only to predict the outcome of these processes, but also to interpret the results of biofilm studies. In the absence of suitable models, the interpretation of biofilm studies is impaired. Biofilm science and technology are relatively young, and mathematical descriptions of biofilm

processes often lag behind the rapidly expanding knowledge of biofilm processes. On the other hand, most of the experience that was accumulated in modeling biofilm processes in water and wastewater treatment was based on the operating reactors with suspended biomass. Biofilm reactors are different, and some effects common in biofilm reactors are much less usual in reactors with suspended biomass. One effect that is particularly difficult to accommodate in biofilm models is the influence of biofilm heterogeneity on biofilm processes. Biofilm models that describe biofilm processes on the scale of the entire reactor assume that the biofilm is uniformly distributed and its effects do not depend on the location in the reactor. This assumption, which is justified in the case of well-mixed reactors, may or may not be justified in biofilm reactors. With the current sophistication in exploring biofilm processes at the microscale, it is not surprising to observe that the local conditions quantified in biofilms deviate widely from the average conditions described by the biofilm models. One hopes that these deviations from the idealized models cancel each other and that overall, at the macroscale, they do not matter much. One particularly troubling problem is the definition of and the existence of a steady state in biofilm reactors. Defining a steady state in a biofilm reactor may well be the most important question facing biofilm researchers, both those who focus on experiment and those who focus on modeling. The existence of a steady state is obvious in flow reactors, where microbial growth occurs in suspension. In such reactors, the interplay among the microbial growth rates, biomass concentration, and hydraulic and biomass retention times leads to a steady state in which process variables do not change for a long time. In contrast, the reasons for the existence of a steady state in a biofilm reactor are much less clear because an important condition for a steady state is not satisfied in a biofilm reactor: the concentration of biomass in a

Biofilms in Water and Wastewater Treatment

Convection Diffusion Convection and diffusion

539

Direction of mass transport Convection Diffusion Direction of measured flow velocity

Figure 14 Alternating zones of convective and diffusive mass transport in heterogeneous biofilms. This hypothetical model of mass transport is consistent with the results in Figure 13. Mass transport in the space occupied by the biofilm is convective, but the amount of nutrient delivered to this space is limited by the diffusive mass transport just above the biofilm surface (MSU-CBE, P.Dirx).

biofilm reactor is not a simple function of retention time and growth rate. Some biofilm technologies actually take advantage of this fact and grow biofilm microorganisms using retention times at which the microorganisms would be washed out from reactors operated with suspended microorganisms. Practically, this problem corresponds to the fact that we are uncertain what function describes detachment in biofilms, and what mechanisms are involved in biofilm detachment, except perhaps for shear stress. The mechanism of biofilm sloughing remains unknown. A steady state for the biomass concentration assumes that the same amount of biomass is generated as is removed by various processes, particularly biofilm detachment. One can argue that if the biofilm reactor is large enough, the microscale biofilm processes will average out on the scale of the reactor, and that this average may be stable even if the components of the average vary over time. This argument, even if it is true, however, does not settle the issue. A question follows: how large does the reactor have to be to ensure that the variations in the microscale biofilm processes average out and the reactor reaches a steady state at the macroscale? There are also difficulties at the microscale. Experimentally measured concentration profiles and flow velocity profiles corroborate the conceptual model shown in Figure 5. However, when it comes to interpreting experimental data, the idealized image of biofilms in Figure 5 is not adequate for many reasons. One reason is shown in Figure 15: the difficulty with locating the position of the biofilm surface. The position of the biofilm surface is important: one of the boundary conditions in the equation describing biofilm activity and mass transport specifies the conditions at the biofilm surface. As can be seen in Figure 15, however, locating it is not trivial. This problem has been addressed experimentally by judiciously locating the surface on a nutrient concentration profile at the location where the profile ends its curvature near the bottom. The rule of locating the biofilm surface at that location has been developed based on the results of studies in which an oxygen electrode and an optical sensor were used to measure the oxygen concentration profile and detect the biofilm surface, where optical density changed (Figure 16). The position of the biofilm surface coincides with the location where the oxygen profile becomes linear. The biofilm surface in Figure 7 was positioned using this principle.

Figure 15 Surface of a biofilm grown at a flow velocity of 0.81 m s1 (Groenenboom, 2000).

4.15.2.2.8 Biofilms in conduits Among the many possible effects that biofilms may have in water conduits, we will discuss two effects in more detail: (1) the effect on flow characteristics – pressure drop in conduits and (2) the effect on material performance – MIC. Flow velocity near the biofilm surface. It is well known that flow velocity affects biofilm processes. Figure 5 shows an example of the effect of flow velocity on mass transport dynamics near the biofilm surface. However, biofilm also affects flow velocity: flow velocity near a wall covered with biofilm is different from that near a wall with no biofilm. Figure 17 shows this effect. The effect of biofilm on flow velocity distribution most certainly influences the dynamics of mass transfer. However, this is not the only effect that biofilm has on hydrodynamics. For example, it is well known that biofilms increase the pressure drop in conduits, but it is not clear what the mechanism of this process is or how to quantify it. To predict pressure drop in pipes the Moody diagram is used, which correlates the Reynolds number and the relative roughness to provide the friction factor, f. This friction factor is then

Biofilms in Water and Wastewater Treatment 9.0

0.30

Oxygen concentration (mg l−1)

8.0

0.27

Oxygen conc. Log (lo /l)

7.0

0.24

6.0

0.21

5.0

0.18

4.0

0.15

3.0

0.12

2.0

0.09

1.0

0.06

0.0

0.03

0.00

0.30

0.60 0.90 1.20 1.50 Distance from the bottom (mm)

1.80

2.10

Log (lo /l )

540

0.00 2.40

Figure 16 Profiles of oxygen concentration and optical density in a biofilm. A combined microsensor – an oxygen microelectrode and an optical density microprobe – permitted locating the biofilm surface at 0.60 mm from the bottom. This distance, when marked on the oxygen concentration profile, indicates that the biofilm surface is at the beginning of the linear part of the oxygen profile within the mass transfer boundary layer; I is the local light intensity, and Io is the maximum light density (Lewandowski et al., 1991).

1.2 Biofilm 1.0

0.4 Biofilm

0.6

0.3 V /Vmax

V /Vmax

0.8

0.4

0.2 0.1

0.2 0 0

1000

2000

250 Depth (µm) 3000

4000

500 5000

Depth (µm) Figure 17 The flow velocity profile near a wall covered with a biofilm is different from the flow velocity profile near the same wall without the biofilm (DeBeer et al., 1994).

plugged into the Darcy–Weisbach equation to calculate the pressure drop:

HL ¼ f

l V2 D 2g

ð10Þ

where HL is the head loss due to friction, l the pipe length, V the average fluid velocity, g the gravitation constant, D the pipe diameter, and f the friction factor provided by the Moody diagram. When the flow velocity increases, the thickness of the boundary layer decreases, and the roughness elements protrude through the boundary layer, further affecting the drag and the pressure drop.

Unfortunately, the Moody diagram is of little help in predicting the pressure drop in conduits covered with biofilms. The pressure drop in such conduits is caused by different factors than the pressure drop in conduits without biofilms because different mechanisms are responsible for the shape of the pressure drop in each of these conduits. These differences sometimes demonstrate themselves in the form of puzzling experimental results, such as decreasing pressure drop resulting from increasing flow velocity, which is a consequence of the elastic and viscoelastic properties of biofilms. Microcolonies are made of bacterial cells embedded in gelatinous EPS that can change shape under stress. At high flow velocities the hydrodynamic boundary layer separates from the

Biofilms in Water and Wastewater Treatment

3 2 1 m

2.5 m

8 4

m

2.5 m

m

2.5 m

2.0 cm

1 2.0 cm

8 4

m

2.5 m

V (cm s−1)

4

2

2.5 m

2.0 cm

8

3

m

2.0 cm

V (cm s−1)

V (cm s−1)

Without biofilm

V (cm s−1)

V (cm s−1)

With biofilm

V (cm s−1)

microcolonies, causing pressure drag downstream of the microcolony and pulling the material in this direction. The microcolonies slowly flow under the strain, forming elongated shapes that we call streamers. Streamers are often seen when biofilms grow at high flow velocities. The streamers contribute to pressure drop by moving rapidly and dissipating the kinetic energy of the flowing water. Another important consequence of a streamer’s oscillations is that they are transmitted to the underlying microcolonies, which also oscillate rhythmically. This system reacts with turbulent boundary layers much differently than the rigid surface roughness elements of clean pipes do. One way to gain experimental access to the interactions between flowing water and biofilm is to monitor flow velocity profiles. Imaging flow velocity profiles makes it possible to evaluate the effect of biofilm formation on the flow in conduits by quantifying its effect on the entry length in the conduit. The hydrodynamic entry length is defined as the distance needed to develop a steady flow, after the water has passed through the entrance to the reactor. If the presence of biofilm makes the entry length longer, then the biofilm contributes to flow instability, and vice versa. There is a simple relation between the Reynolds number and the entry length: the higher the Reynolds number, the longer the entry length. This effect was used as a base for quantifying the effects of biofilm on the flow in conduits. Flow velocity distribution was measured in a rectangular reactor when the flow velocity was increasing from one measurement to another. As the flow velocity and the Reynolds number increased, the flow stability was monitored in a rectangular conduit using nuclear magnetic resonance (NMR) imaging. The results, shown in Figure 18, demonstrate that the presence of biofilm actually made the flow more stable. The entry length was shorter and the flow reached stability closer to the entrance in the presence of biofilm than in its absence. It is difficult to interpret this result immediately because it is well known that the presence of biofilm increases pressure drop in conduits: traditionally, pressure drop in pipes is related to friction. As pressure drop is larger in biofilm-covered pipes, a natural conclusion was that biofilms must increase friction and therefore the presence of the biofilm should introduce flow instability rather than reduce it. The relation between flowing water and biofilms is determined by two facts: (1) biofilms are made of viscoelastic polymers which actively interact with the oscillations generated by the flow of water and (2) the flow of water affects the biofilm structure. Based on what we now understand, at low flow velocities biofilms can effectively smooth surfaces and stabilize the flow because the oscillating layer of elastic polymeric matrix can effectively damp the vibrations coming from the flowing water. This effect delays the onset of turbulence in conduits covered with biofilm and explains the results shown in Figure 18. However, as the flow velocity increases further, the elastic polymeric matrix must oscillate faster and faster and, eventually, the frequency of its oscillation cannot follow the frequency of the incoming eddies. At that point the biofilm oscillation is out of phase and the biofilm not only fails to damp the flow instabilities but also actively introduces instability by randomly oscillating at a different frequency than the incoming eddies. The pressure drop in the conduit

541

2.0 cm

8 4

m

2.5 m

2.0 cm

Figure 18 Flow velocity profiles in a rectangular conduit whose walls were colonized with a biofilm. The increasing flow velocity did not affect the character of the velocity profiles in the reactor with biofilm. On the other hand, the same increase in velocity had a pronounced effect on the reactor without biofilm.

increases rapidly. This effect was, in early biofilm works, mistaken for a similar effect caused by rough surface elements. For example, Picologlou et al. (1980) observed a considerable increase in frictional resistance after the film thickness reached a value approximately equal to the calculated thickness of the hydrodynamic boundary layer for a clean surface. In clean pipes covered with surface roughness elements, when flow velocity increases the boundary layer becomes thinner and at some flow velocity the boundary layer thickness is smaller than the height of the roughness elements. When this happens, the roughness elements protrude through the boundary layer and cause an additional drag, which exhibits itself in a sudden increase of the pressure drop for flow velocities exceeding this critical flow velocity. This model was commonly accepted and was used to explain the pressure drop in conduits covered with biofilms, although even at that time some authors warned that this might not be the true mechanism of the process (Characklis, 1981). Currently, there are no models that can account accurately for pressure drop in conduits covered with biofilm.

4.15.3 Part II: Biofilm Reactors Biological systems treating municipal wastewater require (1) the accumulation of active microorganisms in a bioreactor and (2) the separation of the microorganisms from treated effluent. In suspended growth reactors, such as the activated

542

Biofilms in Water and Wastewater Treatment

sludge process, microorganisms grow and bioflocculate; the resultant flocs are suspended freely in the bulk phase. Flocculated bacteria are then separated from the bulk liquid by sedimentation or membranes. Clarifier-coupled suspended growth reactors rely on return activated sludge, or underflow, from the coupled clarifier to provide the desired active biomass concentration in the bioreactor. Consequently, clarification unit processes may be limited by the hydraulic loading rate (HLR) or solids loading rate (SLR). Biofilm reactors retain bacterial cells in a biofilm that is attached to the fixed or free moving carriers. The biofilm matrix consists of water and a variety of soluble (C) and particulate (X) components that include soluble microbial products, inert material, and EPS. Without suspended biomass, the bioreactor is decoupled from the liquid–solids separation unit. Active biomass concentrations inside the biofilm are large at 10–60 g of volatile suspended solids (VSS) l1 of biofilm. This biomass range can be compared with the range of concentrations expected for suspended growth reactors, which is typically 3–8 g VSS l1 of reactor volume. The lower value in this range is associated with clarifier-coupled activated sludge processes, and the upper range with membrane bioreactors. In biofilm reactors, bacteria attached to carriers periodically detach from the biofilm matrix and exit the system in the effluent stream. Figure 19 provides a conceptual illustration of different biofilm reactor types. Biofilm reactors can be classified based on the number of phases involved – gas, liquid, solid – according to the biofilm being attached to a fixed or moving carrier within the reactor. They are also classified based on how electron donors or acceptors are applied to seven basic types as listed below (adapted from Harremo¨es and Wilderer (1993)): 1. Three-phase system – fixed biofilm-laden carrier, bulk water, and air. Water trickles over the biofilm surface and

2.

3.

4.

5.

6.

7.

air moves upward or downward in the third phase (e.g., trickling filter (TF)) (Figure 19(a)). Three-phase system – fixed (or semifixed) biofilm-laden carrier, bulk water, and air. Water flows through the biofilm reactor with gas bubbles (e.g., aerobic biologically active filter (BAF)). Gravel is a fixed media and polystyrene beads are semifixed (Figures 19(b) and 19(c)). Three-phase system – moving biofilm-laden carrier, bulk water, and air. Water flows through the biofilm reactor. Air is introduced with gas bubbles (e.g., aerobic moving bed biofilm reactor (MBBR)) (Figure 19(g)). Two-phase system – moving biofilm-laden carrier and bulk water. Water flows through the biofilm reactor with the electron donor and electron acceptor (e.g., denitrification fluidized bed biofilm reactor (FBBR)) (Figure 19(g)). Two-phase system – fixed biofilm-laden carrier material and bulk water. Water flows through the biofilm reactor with the electron donor and electron acceptor (e.g., denitrification filter) (Figures 19(b) and 19(c)). Three-phase membrane system – a microporous hollowfiber membrane with biofilm and water on one side and gas on the other, diffusing through the membrane to the biofilm (e.g., membrane biofilm reactor (MBfR)) (Figure 19(h)). Two-phase membrane system – a proton exchange membrane separating a compartmentalized biofilm-laden anode from a compartmentalized cathode with water on both sides, but with the electron donor on one side and electron acceptor on the other (e.g., biofilm-based microbial fuel cell (MFC)).

Biofilms are ubiquitous in nature and in engineered systems and can be used beneficially in municipal water and wastewater treatment. Biofilm and suspended growth reactors can meet similar treatment objectives for carbon oxidation,

Air

Air (a)

(b)

(c)

(f)

(g)

Air

Air (e)

(d)

(h)

Figure 19 Types of biofilm reactors: (a) trickling filter; (b) submerged fixed bed biofilm reactor operated as up flow or (c) down flow mode; (d) rotating biological contactor; (e) suspended biofilm reactor including airlift reactor; (f) fluidized bed reactor; (g) moving bed biofilm reactor; and (h) membrane attached biofilm reactors. From Morgenroth (2008) Modelling biofilm systems. In: Henze M, van Loosdrecht MCM, Ekama G, and Brdjanovic D (eds.) Biological Wastewater Treatment – Principles, Modelling, and Design, pp. 457–492. London: IWA Publishing.

Biofilms in Water and Wastewater Treatment

nitrification, denitrification, and desulfurization. Biofilm reactors have also been used for the treatment of a variety of oxidized contaminants including perchlorate and bromate. The same microorganisms are responsible for biochemical reactions in both activated sludge and biofilm systems, and respond in the same way to local environmental conditions (i.e., pH, temperature, electron donor, electron acceptor, and macronutrient availability) (Morgenroth, 2008). A key component to be considered by anyone who is evaluating a biofilm reactor is the effect of multiple substrates and biomass fractions and the manner in which the reactor is affected by mass-transport limitations. Substrates typically considered are: 1. soluble compounds, including electron donors (e.g., readily biodegradable chemical oxygen demand (rbCOD),   NHþ 4 , NO2 , and H2), electron acceptors (e.g., O2, NO3 ,  2 3 NO2 , and SO4 ), and nutrients and buffers (e.g., PO4 ,  NHþ 4 , and HCO3 ) and 2. particulate compounds, including electron donors (e.g., slowly biodegradable COD (sbCOD)), active biomass fractions (e.g., heterotrophic and autotrophic bacteria), inert biomass, and EPS.

4.15.3.1 Application of Biofilm Reactors This section exists to provide the reader with a general overview of biofilm reactor applications. While general biofilm reactor applicability is described here, several treatment scenarios exist that are not conveniently generalized yet warrant the use of biofilm reactor technology. Water-quality regulations exist to protect human health and the water environment. Organic matter and the nutrients such as nitrogen and phosphorus are major contributors to water-quality impairment. In municipal wastewater-treatment scenarios, biofilm reactors are generally applied for the removal of carbon-based organic matter and/or nitrogenous compounds. Specifically, these biofilm reactors may achieve carbon oxidation, combined carbon oxidation and nitrification, tertiary nitrification, or tertiary denitrification. Biofilm reactors are not commonly used for biological phosphorus removal. Biofilm reactors treating industrial wastewaters have been applied to meet treatment objectives similar to those in municipal wastewater treatment and industrial pretreatment. The objective of pretreatment is to process industrial waste streams until their characteristics are similar to raw sewage (see Metcalf and Eddy (2003) for a description). As a result the industry can then discharge their treated wastewater into municipal sewers where further processing is accomplished at a municipal wastewater-treatment plant. Biofilm reactors are common for industrial applications because the processes are reliable, robust, easy to operate, and resilient to toxic or shock loading.

4.15.3.1.1 Techniques for evaluating biofilm reactors Several approaches exist to evaluate biofilm reactors. The primary objective of a biofilm or biofilm reactor model is to predict soluble substrate flux (J) through the biofilm surface. This flux (M L2 T1) can be used to obtain an estimate of the (1) overall biofilm reactor performance, (2) required biofilm surface area, (3) electron acceptor (e.g., dissolved oxygen), (4)

543

external electron donor (e.g., methanol or hydrogen), and (5) biosolids management requirements. This section discusses the relative benefits and limitations to some general methods of evaluating biofilm reactors. The use of mathematical biofilm models is common in both research and practice, but only a cursory presentation of their mathematical description is presented. Excellent resources exist describing aspects of mathematical modeling of biofilms and biofilm reactors (for additional information, see Wanner et al. (2006) and Morgenroth (2008)). The approaches discussed here include a graphical procedure, empirical models, semiempirical models, and mechanistic mathematical models.

4.15.3.1.2 Graphical procedure A graphical procedure can be used to determine the total hydraulic load (THL) required to decrease a substrate concentration, and by definition the biofilm surface area required to provide a desired substrate concentration remaining in the effluent stream. These items can be determined directly. The graphical procedure can be used to determine effluent substrate concentration from any series of continuous flow stirred tank reactors (CFSTRs). A stepwise procedure must be used when a series of CFSTRs will be used. Antoine (1976) and Grady et al. (1999) developed the graphical procedure described here and the approach is valid for any biofilm-based CFSTR. If multiple stages are expected to have different characteristics, then the graphical method requires different flux curves to describe system response in each of the CFSTRs. The procedure requires a graphical representation of substrate flux (J) as a function of bulk-liquid substrate concentration (CB). This relationship between flux and bulk-liquid substrate concentration can be obtained from numerical simulations, full-scale or pilot-plant observations. In practice, this graphical procedure is typically used to extend pilot-plant observations to full-scale biofilm reactor design criteria. The process designer should recognize that the relationship between flux and bulk-liquid substrate concentration is based on the system and location. Therefore, the flux curve required to implement the graphical procedure may not be obtained from or correlate well with values reported in the literature or from different systems. As a result, the process designer should consider carefully the conditions under which the flux curve was developed before applying results. A flux curve representing mass transfer and environmental conditions characteristic of a specific system and operating mode may not be the representative of different biofilm reactor types designed to meet the same treatment objectives. A flux curve generated for the same biofilm reactor type under similar operating conditions, however, may offer some direction in the absence of system-specific numerical simulation or pilot/full-scale observations. When using the graphical procedure to evaluate pilot-plant observations, fluxes should be compared to rates in full-scale systems. Any flux that deviates significantly from those reported for biofilm reactors in published studies should be used only after careful consideration. Pilot or experimental systems may promote a greater flux than expected. The basis for the graphical procedure is a material balance on a

544

Biofilms in Water and Wastewater Treatment

biofilm-based CFSTR:

0 = Q ⋅ Cin ,i − Q ⋅ C B ,i − J LF ,i ⋅ A − rB ,i ⋅ VB mass per time input

mass per time output

biofilm transformation rate

suspended growth transformation rate

ð11Þ

where Q is the flow rate through the system (m3 d1); Cin,i the influent concentration of soluble substrate i (g m3); CB,i the effluent, or bulk-liquid, concentration of soluble substrate i (g m3); JLF,i the flux of soluble substrate i across the biofilm surface equal to the average biofilm activity in the reactor, as shown in Equation (1) (g m2 d1); A the biofilm surface area (m2); rB,i the rate of substrate i conversion because of suspended biomass (g m2 d1); and VB the bulk-liquid volume (m3). Assuming that transformation occurring in the bulk liquid is negligible, the suspended growth transformation rate (Equation (11)) can be neglected. Rearranging Equation (11) provides the rationale for the graphical procedure:

JLF;i ¼

Q Q  Cin;i   CB;i A A |fflfflfflffl{zfflfflfflffl} |{z} const:

ð12Þ

slope

The slope, or (  (Q/A)), is referred to as the operating line and represents the total hydraulic load on each stage. Figure 20 illustrates the graphical method. The flux curves have been created based on observations in the first and second stage of a post-denitrification biofilm reactor. The ordinate represents nitrate–nitrogen flux and the abscissa nitrate–nitrogen concentration remaining in the effluent stream. The graphical solution indicates that the

first-stage denitrification biofilm reactor effluent nitrate– nitrogen concentration is approximately 3.9 mg l1. The secondstage effluent nitrate–nitrogen concentration is approximately 1.1 mg l1 with fluxes of approximately 1.6 and 1.1 g m2 d1 in the first and second stage, respectively. The graphical procedure depends on the substrate flux curve(s). The method requires development of multiple flux curves if the performance characteristics of respective stages vary significantly. When using pilot-plant data to generate a flux curve, appropriate scale considerations must be given when designing the pilot unit and experiments.

4.15.3.2 Empirical and Semi-Empirical Models Empirical models can be implemented easily by hand or using a spreadsheet, but they have limited applicability because of their black-box consideration of system parameters. Because environmental conditions and bioreactor configuration affect biofilm reactor performance, a system can respond differently from the description provided by an empirical model. The limited descriptive capacity of empirical models typically results from parameter values and model features based on data that were obtained from few system installations or operating conditions. Therefore, the engineer or scientist should be aware of conditions under which system-specific model parameters have been defined. Significant sources of variability in values include differences in biofilm carrier type and configuration, the extent of concentration gradients external to the biofilm surface, and biofilm composition. Despite their ease of implementation, empirical models can produce results that vary 50–100% of actual system performance.

3.5 Denitrification rate (g m−2 d−1 as NO3−N)

Stage 1 operating line 3.0 Stage 2 operating line

Stage 1 flux response curve

2.5 Stage 2 flux response curve

J LF1

2.0

1.5 −Q/A J LF2

1.0

0.5 CB -stage

0.0 0

1

2

3

CB -stage 1 4

C in 5

6

7

8

9

10

CNO3−N (mg-N l−1) Figure 20 Graphical procedure for describing the response of a denitrification moving bed biofilm reactor to defined conditions, including (1) firstand second-stage operating lines and (2) flux curves based on observations at a pilot-scale denitrification moving bed biofilm reactor (Boltz et al., 2010b).

Biofilms in Water and Wastewater Treatment

Coefficient values, and sometimes the empirical models, are typically created to describe system response for the removal of a specific material. The models can be used as an indicator of system viability to meet treatment objectives with respect to the specific process governing transformation. Empirical models are, however, inadequate for describing complex processes such as the explicit evaluation of two-step ammonium oxidation first to nitrite by ammonia-oxidizing bacteria and then to nitrate by nitrite-oxidizing bacteria. Therefore, empirical models have limited application in defining the conditions that either promote or deter complex processes in biological systems. Historically, biofilm reactors have been designed using empirical criteria and models, but this trend is changing. One should recognize that the coefficients in empirical models describing biofilm reactors include system, and many times, location-specific mass-transfer resistances (Grady et al., 1999). For this reason, the values typically differ from apparent or intrinsic values reported in the literature. Once a flux has been determined, Equation (11) can be rearranged, neglecting bulkphase conversion processes, to calculate the material concentration remaining in the effluent:

CB;i ¼ Cin 

JLF;i  A Q

ð13Þ

If sufficient data exist to allow for the development of parameter values and mathematical relationships capable of describing a complete range of conditions expected when treating municipal wastewater, then empirical models can be used. The addition of model components to account for specific phenomenon encroaches on the premise of mechanistic mathematical model development. For this reason, a distinction is made between empirical and semi-empirical models. Gujer and Boller (1986) and Sen and Randall (2008) provided an example of the latter describing nitrifying TFs, and MBBRs and IFAS systems, respectively.

C

LF

LL

4.15.3.3 Mathematical Biofilm Models for Practice and Research Mathematical modeling can be used to describe certain features of a biofilm or biofilm system (such as a bioreactor) by selecting and solving mathematical expressions. Biofilm reactor research and design commonly involve the use of mathematical biofilm models. These mathematical models are tools that allow the user to efficiently evaluate a variety of complex scenarios. Empirical models fail to provide information that is a concern for biofilm researchers and environmental protection such as biofilm composition and competition among bacteria for multiple substrates and space inside the biofilm, and the influence of individual processes on the interaction between several bacterial types. Mathematical biofilm models have been used as a research tool, but only recently modern biofilm reactors have encouraged the use of biofilm models in engineering practice. Submerged and completely mixed biofilm reactors allow for the application of modern biofilm knowledge, and are conducive to simulation with existing biofilm models (Boltz and Daigger, 2010). As a result, a majority of existing wastewater-treatment plant simulators have been improved to include a biofilm reactor module(s) that is based on the mathematical description of a 1-D biofilm. A user should understand the mathematical biofilm model basis, underlying assumptions, and limitations before applying the model to research or design. A biofilm schematic is shown in Figure 21. The schematic illustrates diffusion and reaction occurring inside a 1-D biofilm. In addition, concentration gradients external to the biofilm surface are illustrated in the manner that they are modeled, namely an external mass transfer resistance represented by a mass transfer boundary layer. The partial differential equation describing molecular diffusion, substrate utilization inside a biofilm, and dynamic accumulation has been presented as Equation (3). It should be emphasized that the basis for a mathematical description of the 1-D biofilm, as described by Equation (3), is simultaneously

C

LF

LL

CB

CB

C LF

C LF

Distance from growth medium

Distance from growth medium

Z Distance from surface, X

545

Z Distance from surface, X

Figure 21 Schematic of a 1-D biofilm of thickness LF having an assumed homogeneous (a) and heterogeneous, or layered, (b) biomass distribution. Soluble substrate concentration profile is illustrated with a bulk-liquid concentration (CB) decreasing through a mass transfer boundary layer of thickness LL until reaching the liquid–biofilm interfacial concentration (CLF), and then decreasing through the biofilm.

546

Biofilms in Water and Wastewater Treatment

occurring molecular diffusion and biochemical reaction. Molecular diffusion is based on Fick’s law. Monod-type kinetics is typically applied to describe the biochemical transformation rate. Analytical solutions to Equation (3) are available only for first- and zero-order rate expressions and assuming steady state. Zero-order kinetics are valid if the bulkliquid substrate concentration is well above the half-saturation concentration (i.e., CB,i 4 Ki), and first-order kinetics is applicable for low substrate concentrations (i.e., CB,ioKi). Solving the second-order differential equations requires constants that can be derived from two boundary conditions described by Equation (5). From the concentration profile (Cf,i(x)) the flux through the biofilm surface (JLF) is calculated as Equation (2). This substrate flux, JLF, is used in biofilm reactor material balances (see Equation (11)). The concentration gradient external to the biofilm surface is not explicitly modeled. Rather, it is modeled as a mass transfer resistance:

JMTBL ¼

1 ðCB;i  CLF;i Þ RL

ð14Þ

•

•

Here, JMTBL is the substrate flux in the stagnant liquid layer and RL the mass transfer resistance external to the biofilm. It is helpful to visualize RL by introducing the concept of a mass transfer boundary layer. Defining the thickness of this mass transfer boundary layer provides a more intuitive understanding compared to the mass transfer resistance. Resistance to mass transfer and the mass transfer boundary layer thickness are related according to Equation (15):

RL ¼

LL Dw

ð15Þ

Here, LL is the mass transfer boundary layer thickness and Dw the solute diffusion coefficient in the water phase. The substrate flux through the mass transfer boundary layer (Equation (15)) is linked to the substrate flux across the biofilm surface (Equation (2)). This provides an additional Equation (16) (boundary condition) that is required to calculate the additional unknown value of the substrate concentration at the liquid–biofilm interface (JLF):

JMTBL ¼ JLF

ð16Þ

One of the most difficult aspects of choosing an approach to model biofilms and biofilm reactors is to choose the appropriate level of complexity. An overview of the different model approaches is provided below (after Taka´cs et al., 2010):

•

•

0-D biofilm. One aspect of modeling biofilms is that bacteria are retained in the system and are not washed out with effluent water. The simplest approach for biofilm modeling would be to assume that all biomass in the reactor is exposed to bulk phase concentrations neglecting the effect of mass transport limitations (i.e., 0-D). In wastewater treatment biofilms are relatively thick and are usually masstransfer-limited. Thus, the 0-D modeling approach that neglects mass transfer limitations is not useful except in special cases. 1-D homogeneous biofilm (single limiting substrate). This approach takes into account mass transfer limitations into

•

the biofilm and the corresponding effects on concentration profiles and substrate flux into the biofilm. It is assumed that active bacteria are homogeneously distributed over the thickness of the biofilm. The approach is valid only if calculations are performed for the limiting substrate which has to be determined a priori by the user as described in Morgenroth (2008). The flux of the nonlimiting substrates can be calculated based on reaction stoichiometry. 1-D homogeneous biofilm (multiple substrates and multiple biomass components). One key aspect of modeling biofilms is to evaluate the competition and coexistence of different groups of bacteria and local environmental conditions. Local process conditions can be accurately determined by calculating penetration depths for different soluble substrates. Based on the fluxes the growth of individual groups of bacteria can be determined. To simplify calculations it can be assumed that all bacterial groups are homogeneously distributed over the thickness of the biofilm (Rauch et al., 1999; Boltz et al., 2009a). 1-D heterogeneous biofilm. Different groups of bacteria are competing in a biofilm not only for substrate but also for space where bacteria toward the surface are less influenced by mass transport limitations. Bacteria growing toward the base of the biofilm are often rate limited by substrate availability resulting from mass transfer limitations. On the other hand, these bacteria are better protected from detachment. These 1-D heterogeneous biofilm models must keep track of local growth and decay of the different bacterial groups and of detachment to calculate biomass distributions over the biofilm thickness. 2-D and 3-D biofilm models. Practically, biofilms are not as smooth and flat as is assumed in 1-D biofilm models. Mathematical models have been developed that predict the development of biofilms in two or three dimensions, the influence of the heterogeneous structure on fluid flow, and ultimately the combination of fluid flow and biofilm structure on substrate availability and removal inside the biofilm. For most questions related to practical biofilm reactor studies, such multi-dimensional models are not necessary. However, it is important for model users to recognize that biofilm structure influences local fluid dynamics and external mass transport, which are simultaneously affected by biofilm reactor appurtenances and mode of operation. Such interactions are not accounted for in existing 1-D biofilm models due to a rigid segregation of the bulk phase, mass transfer boundary layer, and biofilm (which is assumed to have a uniform thickness and smooth surface). Multi-dimensional biofilm models have been used to quantify the influence of biofilm structure on local fluid dynamics and external mass transport (Eberl et al., 2000).

Different scales of heterogeneity are relevant for biofilm reactors. The length scale of the biofilm thickness, which is on the order of 100–1000 mm, is taken into account in 1-D and multi-dimensional biofilm models. Substrate fluxes from these simulations can then be integrated into models describing overall reactor performance where the length scale is on the order of 1 m. However, heterogeneities can also be observed in biofilm reactors in between these scales where, in some cases, patchy biofilms are observed and where certain

Biofilms in Water and Wastewater Treatment

parts of the biofilm support medium is bare while at other areas dense biofilms develop (B1–10 cm). These heterogeneities in between the small and the large scale are typically not considered in biofilm models and it is not clear to what extent they are relevant (Taka´cs et al., 2010). No simple and general recommendations can be given on what approach is the most appropriate for describing biofilm reactors. Wanner et al. (2006) provided a detailed description of different modeling approaches and a discussion on how the modeling approaches compare for different modeling scenarios. Many commercially available wastewater-treatment plant simulators used for biofilm reactor design and evaluation takes into account multiple substrates and biomass fractions in either a heterogeneous or a homogeneous 1-D biofilm. Examples of software, and references to the biofilm model that constitutes the biofilm reactor module, that is applied to design, optimize, and evaluate, typically pilot- or full-scale biofilm reactors are summarized in Table 1.

4.15.3.4 Biofilm Model Features Excellent guides exist that describe the mathematical modeling of biofilms (see Wanner et al., 2006; Morgenroth, 2008). However, the state of biofilm modeling is subject to several uncertainties. In the context of this chapter, Boltz et al. (2010a) summarized the following items which cause uncertainty when using 1-D biofilm models to describe biofilm reactors: (1) the fate of particulate substrates, (2) biofilm distribution in the reactor and the effect biofilms have on reactor components, (3) dynamics and fate of biofilm detachment, (4) quantifying concentration gradients external to the biofilm surface, and (5) a lack of generally accepted biofilm reactor Table 1

model calibration protocol. Parameter estimation and model calibration are serious concerns for process engineers who apply biofilm models in engineering practice. Therefore, parameters that are critical components of biofilm reactor models (that use a 1-D mathematical biofilm model) are introduced, including: attachment (kat) and detachment (kdet) coefficients, the mass transfer boundary layer, rate-limiting substrate diffusivity coefficient inside the biofilm (Df,ratelimiting), and the biokinetic parameters maximum growth rate (m) and the ratelimiting substrate half-saturation coefficient (Ki,ratelimiting) (Boltz et al., 2010b).

4.15.3.4.1 Attachment and detachment process kinetics and rate coefficients An accurate mathematical description of particle attachment and detachment processes is a critical component of biofilm (reactor) models. Unfortunately, attachment/detachment process mechanics are poorly understood. Conceptually, particles suspended in the bulk liquid are hydrodynamically transported to the vicinity of the biofilm. From the bulk phase, particles are subjected to concentration gradients external to the biofilm surface. Particles enter the biofilm matrix through channels, crevasses, and other structural irregularities where they attach to the biofilm surface (see Reichert and Wanner (1997) for a description of particle transport within the biofilm matrix). Once entrapped, the particles can be hydrolyzed by extracellular polymeric enzymes resulting in soluble substrate that diffuses into the biofilm. Then, the soluble substrate is subject to well-known biochemical transformation processes that yield biomass. Alternatively, particles that have attached to the biofilm surface from the bulk phase remain unaltered and exit the system after detaching from the biofilm

Biofilm models used in practice (Boltz et al. 2010b)

Software

Company

Biofilm model type and biomass distribution

Reference

AQUASIMTM

EAWAG, Swiss Federal Institute of Aquatic Science and Technology, Du¨bendorf, Switzerland (www.eawag.ch/index_EN) Aquaregen, Mountain View, California (www.aquifas.com) EnviroSim Associates Ltd., Flamborough, Canada (www.envirosim.com)

1-D, DY, N; heterogeneous

Wanner and Reichert (1996) (modified)

1-D, DY, SE and N, heterogeneous 1-D, DY, N, heterogeneous

Sen and Randall (2008)

Hydromantis Inc., Hamilton, Canada (www.hydromantis.com) CH2M HILL Inc., Englewood, Colorado (www.ch2m.com/corporate) ifak GmbH, Magdeburg, Germany (www.ifak-system.com) WRc, Wiltshire, England (www.wateronline.com/ storefronts/wrcgroup.html) MOSTforWATER, Kortrijk, Belgium (www.mostforwater.com)

1-D, DY, N, heterogeneous

AQUIFASTM BioWinTM

GPS-XTM Pro2DTM SimbaTM STOATTM WESTTM

547

1-D, SS, N(A), homogeneous (constant Lf) 1-D, DY, N, heterogeneous 1-D, DY, N, heterogeneous 1-D, DY, N(A)a, Nb, homogeneousa, heterogeneousb

Wanner and Reichert (1996) (modified), Taka´cs et al. (2007) Hydromantis (2006) Boltz et al. (2009a; 2009b) Wanner and Reichert (1996) (modified) Wanner and Reichert (1996) (modified) Rauch et al. (1999)a, Wanner and Reichert (1996) (modified)b

a

Rauch et al. (1999) is linked with the definition ’N(A)’ and ’homogeneous’. Wanner and Reichert (1996) (modified) is linked with the definition ’N’ and ’heterogeneous’.

b

1-D, one dimensional; DY, dynamic; N, numerical; N(A), numerical solution using analytical flux expressions; SE, semi-empirical; SS, steady-state. Hydromantis, Inc. (2006) Attached growth models. In: GPS-X Technical Reference, pp. 157–185 (unpublished).

548

Biofilms in Water and Wastewater Treatment

matrix. Most of the heterogeneous 1-D biofilm models listed in Table 1 describe the rate of particle attachment (rat) as a first-order process ðrat ¼ kat  XTSS;bulk Þ depending on an attachment rate coefficient (kat) and the bulk-liquid particle concentration. Boltz and La Motta (2007) presented a model describing variability in this parameter with influent particle concentrations. The researchers postulated that increasing particle concentrations ultimately reduced the biofilm surface area available for particle attachment; thereby, the particle attachment coefficient decreases until reaching a plateau. The plateau was considered commensurate with a condition in which a minimum biofilm area was consistently available as a result of continuously detaching biofilm fragments (during steady operating conditions – variable hydrodynamics can influence biofilm structure). Given the current state of the science, experimental data are required to develop/validate or evaluate existing approaches for simulating the fate of particles in biofilm reactors. Steady-state biofilm models have assumed a constant biofilm thickness in which case biofilm growth is balanced by internal loss (e.g., decay and hydrolysis, or endogenous respiration) and/or detachment. This approach has been successfully applied to simulate biofilm reactors at steady state, but their dynamic simulation requires that a detachment model is included despite rather limited mechanistic understanding. The rate (Morgenroth and Wilderer, 2000; Boltz et al., 2010a) and category (i.e., abrasion, erosion, sloughing, and predator grazing) of detachment can have a significant influence on biofilm reactor simulation and performance (Morgenroth, 2003). Kissel et al. (1984) stated that problems inherent to biofilm detachment modeling include a poor understanding of fundamental (biofilm detachment) process mechanics and the inability to predict exactly at what location inside the biofilm that detachment will occur. Detachment location is important when taking into account a heterogeneous biofilm distribution throughout the reactor either by combining multiple 1-D simulations or by 2- or 3-D modeling (Morgenroth et al., 2000). Unlike attachment, Boltz et al. (2010a) described eight different biofilm detachment rate expressions (rdet) for the heterogeneous 1-D biofilm models listed in Table 1. Detachment rate equations can be categorized based on the aspect controlling detachment: biofilm thickness (LF), shear, or growth/activity. Mixed-culture biofilms, such as those growing in a combined carbon oxidation and nitrification MBBR, are subject to competition for substrate between fast-growing heterotrophic and slow-growing autotrophic organisms (primarily for dissolved oxygen). Morgenroth and Wilderer (2000) performed a modeling study that demonstrated ammonium flux was significantly influenced by the mode of simulated detachment. Essentially, biofilm (thickness) dynamics influenced competition for substrate between heterotrophic and autotrophic organisms; high variations in biofilm thickness dynamics favored the faster growing heterotrophic organisms.

4.15.3.4.2 Concentration gradients external to the biofilm surface and the mass transfer boundary layer Biofilms growing virtually in all full-scale biofilm reactors are subject to some degree of substrate concentration gradients

external to the biofilm surface. Concentration gradients external to the biofilm surface are not explicitly simulated in 1-D biofilm models. Rather, the reduction in concentration of any substrate is modeled as a mass-transfer resistance, RL ( ¼ LL/Dw). Based on the observation that the external masstransfer resistance, RL, is more dependent on biofilm reactor bulk-liquid hydrodynamics than biofilm thickness or surface heterogeneity, the impact of RL can be accounted for by empirical correlations (Boltz et al., 2010a). However, a realistic description of hydrodynamic effects ultimately depends on an accurate estimate of the mass-transfer boundary layer thickness LL. Therefore, the mass-transfer boundary layer thickness is an important facet of biofilm-reactor models that use a 1-D biofilm model. Despite the potential significant impact the mass-transfer boundary layer thickness may have on biofilmreactor model results and process design, factors influencing the interface between the biofilm model and reactor scale is one important feature of biofilm-reactor models that is not well understood.

4.15.3.4.3 Diffusivity coefficient for the rate-limiting substrate inside the biofilm Soluble substrates are primarily transported into biofilms by a combination of advection and molecular diffusion. Generally, the most important mechanism is molecular diffusion (Zhang and Bishop, 1994). The largest component of biofilm is water, but the diffusivity of a solute inside the biofilm is generally less than that in water because of the tortuosity of the pores and minimal biofilm permeability. Consequently, an effective diffusivity must be applied. Many biofilm reactor models treat this value as 80% of the diffusivity in water (i.e., Dw ¼ Df/0.80) (Stewart, 2003). However, it has been demonstrated that the effective diffusion coefficient (Df,i) for any soluble substrate i can vary with depth inside the biofilm (Beyenal and Lewandowski, 2000). The effective diffusivity decreases with depth because of increasing density and decreasing porosity and permeability of the biofilm with depth. Flow velocity past the biofilm is a major influencing factor determining biofilm density. Varying liquid velocity in the vicinity of the biofilm surface can influence a soluble substrate effective diffusivity inside a biofilm. Consequently, the varying flow rate can affect the rate of internal mass transfer and transformation rates (Bishop, 2003). Turbulent, high-sheer stress environments result in planar and denser biofilms while quiescent, low-sheer stress environments will result in rough and less dense biofilms (van Loosdrecht et al., 1995). Picioreanu (1999) defined a growth number ðG ¼ L2f  mmax  Xf =ðDf  CB ÞÞ that can be related to biofilm roughness. According to Picioreanu (1999), the biofilm may have a dense solid matrix and a flat surface when Go5. However, if G 4 10 the biofilm may develop complex structures such as mushroom clusters and streamers.

4.15.3.4.4 Parameters: estimation and variable coefficients A parameter is an arbitrary constant whose value characterizes a system member. Biokinetic parameter estimation is a serious concern for those who seek to use biofilm models for biofilm reactor process design and research because most parameter values cannot be measured directly in full-scale municipal

Biofilms in Water and Wastewater Treatment

wastewater-treatment plants (Brockmann et al., 2008). Parameters exist for every aspect of biofilm models, including stoichiometry, kinetics, mass transfer, and the biofilm itself. A majority of parameter values in modern process models (e.g., those described by Henze et al. (2000)) have a substantial database that serves to define a relatively narrow range of values that are applicable to a majority of municipal wastewater-treatment systems. Existing biofilm models are relatively insensitive to changes in a majority of the biokinetic parameter values, most of which are described by Henze et al. (2000), within a range of values reported in the literature except for, as an example, the autotrophic nitrifier maximum growth rate (m). However, the mathematical description of some processes includes variable, or lumped, parameters. These parameter values are often system specific and subject to significant uncertainty. The lumped parameters account for an incomplete mechanistic description of the simulated process. Lumped parameters in a majority of biofilm models, including those described in this chapter, are the oxygen affinity constant for autotrophic nitrifiers (KO2,A), endogenous respiration rate constants (bres), attachment rate coefficient (kat), detachment rate coefficient (kdet), mass-transfer boundary layer thickness (LL), ratio of diffusion in biofilm to diffusion in water (Df/Dw). Indentifying parameter subsets that require definition for biofilm model calibration has been the subject of several investigations by Smets et al. (1999), Van Hulle et al. (2004), and Brockmann et al. (2008).

549

model for another period. Similarly, Bilyk et al. (2008) reported the calibration of a denitrification filter model by adjusting assumed biofilm thickness and incorporating the assimilative denitrification reaction. Both of these biofilm reactor model calibration efforts were based on bulk-phase measurements, but only Sin et al. (2008) utilized measured characteristics of the biofilm. Such adjustments to systemspecific biofilm and biokinetic parameters in order to match observed data may not produce a properly calibrated model that is capable of describing a variety of design conditions for a wastewater-treatment plant. As previously discussed, the attachment coefficient, for example, has been experimentally demonstrated (and described mathematically) to change as a function of particle (total suspended solids) load (Boltz and La Motta, 2007). Then, it may be argued that adjusting the attachment coefficient (during calibration) to match an observed dataset would naturally render the calibrated model incapable of describing another scenario with a different particle load. Suffice it to say that a reliable and transparent description of recommended approaches for the application and calibration of biofilm models are required for the models to gain general acceptance, understanding, and become effectively used in engineering practice. Protocol defining methodology for sampling, testing, evaluating and applying data to mathematical biofilm reactor models is required. It is likely that existing biofilm reactor models will require improvement for reliable dynamic simulation in practice.

4.15.3.5 Biofilm Reactors in Wastewater Treatment 4.15.3.4.5 Calibration protocol Application of a dynamic biofilm model to describe full-scale municipal wastewater-treatment processes requires a calibration of the selected model. Ad hoc expert-based trial and error and standardized systematic approaches have been used to calibrate process models. Sin et al. (2005) presented a critical comparison of systematic calibration protocols for activated sludge models. These protocols have many similarities that are applicable to biofilm reactor models including goal definition, data collection/testing/reconciliation, and validation. The major differences between protocols reported by Sin et al. (2005) are related to the measurement campaign, experimental methods for influent wastewater characterization, and parameter subset selection and calibration. The major differences speak to areas of systematic calibration protocols for activated sludge models that will almost certainly be exasperated when creating systematic protocol for calibrating a biofilm reactor model. Certainly, additional tests will be required to characterize the physical attributes of both suspended growth and biofilm compartments, and mathematical biofilm models have more parameters than activated sludge models. Furthermore, the biofilm compartment parameters must be estimated from bulk-phase measurements in order to have a timely and costeffective approach to calibrating biofilm reactor models. Sin et al. (2008) reported the calibration of a dynamic biologically active (continuously backwashing) filter model using traditional expert-based manual trial and error. The researchers manipulated system-specific parameters related to attachment, detachment, and biofilm thickness. After calibration, Sin et al. (2008) successfully tested the calibrated

Biofilm reactors play an important role in environmental biotechnology, but many aspects of their design and operation remain poorly understood. Biofilm reactors can be traced to origination of modern water sanitation. Corbett (1903) reported the use of continuously distributed sewage flow over a fixed bed, and Stoddart (1911) reported the use of a coarse biofilm-covered medium dosed with a continuous trickling flow. These accounts are acknowledged as the creation of the TF process. Approximately 100 years following these reports significant advances in the design, academic understanding, and mathematical modeling of biofilms have led to the development of new and emerging biofilm reactors conducive to fundamentally based design approaches and the application of fundamentally based design and operation procedures for traditional biofilm reactors. Two processes – mass transfer and biochemical conversion – are characteristics of all biofilm reactors and influence biofilm structure and function. Compartments that are common to every biofilm reactor exist to optimize mass-transfer and biochemical conversion.

4.15.3.5.1 Biofilm reactor compartments Biofilm reactors have five primary compartments: (1) influent wastewater (distribution) system; (2) containment structure; (3) biofilm carrier; (4) effluent water collection system; and (5) an aeration system (for aerobic processes and scour) or mixing system (for anoxic processes that require bulk-liquid agitation and biofilm carrier distribution). Five components influence local conditions inside the biofilm: (1) biofilm carrier surface (i.e., substratum); (2) biofilm (including both particulate and liquid fractions); (3) mass-transfer boundary

550

Biofilms in Water and Wastewater Treatment

layer; (4) bulk liquid; and (5) gas phase (when significant). The components typical of biofilm reactors are described in context of some commercially available biofilm reactors. The five biofilm reactors described include the MBBR, BAF, FBBR, rotating biological contactor (RBC), and TF.

4.15.3.5.2 Moving bed biofilm reactors The MBBR is a two- (anoxic) or three- (aerobic) phase system with a buoyant free-moving plastic biofilm carrier that requires mechanical mixing or aeration to distribute carriers throughout the tank. The process includes a submerged, completely mixed biofilm reactor and liquid–solids separation unit (Ødegaard, 2006). A range of pollutant loading and bulkphase external carbon sources in denitrification MBBRs and dissolved oxygen concentrations in carbon-oxidation and/or nitrification MBBRs have been applied, and system response evaluated (Lazarova and Manem, 1994). It has been demonstrated that MBBRs are capable of processing wastewater to meet a variety of effluent water-quality standards ranging, for example, from the US Environmental Protection Agency definition of secondary treatment (30 mg TSS l1 and 30 mg BOD5/l monthly average) to more stringent enhanced nitrogen removal limits (e.g., total nitrogen less than 3–5 mg l1) under a variety of loading conditions. The MBBR process is capable of meeting similar treatment objectives as the activated sludge process for carbon oxidation, nitrification, and denitrification, but the MBBR makes use of a smaller tank volume than a clarifier-coupled activated sludge system. Biomass retention is clarifier independent; therefore, solids loading in liquid–solids separation unit are significantly reduced when compared with the activated sludge process. Because it is a continuously flowing process, the MBBR does not require a special operational cycle for biofilm thickness control (e.g., backwashing in a BAF or flushing in a TF). Hydraulic head loss and operational complexity is minimal. The MBBR offers much of the same flexibility to manipulate the process flow sheet (to meet specific treatment objectives) as the activated sludge process. Multiple reactors can be configured in series without the need for intermediate pumping or return activated sludge pumping (to accumulate mixed liquor). Liquid–solids separation may be achieved with a variety of processes including sedimentation basins, dissolved air flotation, cloth-disk and membrane filters. The MBBR is well suited for retrofit installation into existing municipal wastewater-treatment plants. An MBBR may be a single reactor or several reactors in a series. Typically, each MBBR has a length-to-width ratio (L:W) in the range of 0.5:1–1.5:1. Plans with an L:W greater than 1.5:1 can result in nonuniform distribution of the biofilm carriers. MBBRs contain a plastic biofilm carrier volume up to 67% of the liquid volume. Screens are typically installed with one MBBR wall and allow treated effluent to flow to the next treatment step while retaining the free-moving plastic biofilm carriers. Aerobic MBBRs use a diffused aeration system to evenly distribute the plastic biofilm carriers and meet process oxygen requirements. On the other hand, anoxic MBBRs use mechanical mixers to evenly distribute the plastic biofilm carriers because there is no process oxygen requirements. Each process mechanical component is submerged. Figure 22

depicts the Williams-Monaco WWTP, Commerce City, Colorado, a two-train bioreactor that consists of four MBBRs in series. The biofilm carriers are extruded or molded from either virgin or recycled high-density polyethylene (HDPE). Table 2 summarizes characteristics of several commercially available plastic biofilm carriers. The carriers are slightly buoyant and have a specific gravity between 0.94 and 0.96 g cm3. Both native and biofilm-covered plastic biofilm carriers have a propensity to float in quiescent water. Biofilms primarily develop on the protected surface inside the plastic biofilm carrier. For this reason, the specific surface areas of plastic biofilm carriers listed in the table exclude areas not inside the plastic carrier. The listed bulk-specific surface area, which is based on 100% carrier fill, is characteristic of a plastic biofilm carrier. The net specific surface area is characteristic of plastic biofilm carrier and fill percentage. For example, if a plastic biofilm carrier has a 500 m2 m3 bulk-specific surface area, then the net specific surface area at 50% carrier fill is 250 m2 m3. Similarly, the net liquid volume displacement at 50% carrier fill is 0.0725 for a plastic biofilm carrier having a characteristic 0.15-bulk-liquid volume displacement (at 100% carrier fill). Plastic biofilm carriers are retained in an MBBR by horizontally configured cylindrical screens or vertically configured flat screens as shown in Figure 23. Aerobic zones typically contain cylindrical screens; anoxic zones contain the flat wall screens. Cylindrical screens are desired. They extend horizontally into the upward-flowing air bubbles imparted by the diffuser grid which aids in scouring any accumulated debris. Energy imparted by the mechanical mixers is insufficient to dislodge debris accumulated on the flat wall screen. Therefore, scouring of flat screens is accomplished with a sparging air header in a denitrification MBBR. Removing the debris retained on a screen aids in maintaining hydraulic throughput. Hydraulically, an MBBR is commonly designed to process a maximum approach velocity (based on the tank cross-sectional area perpendicular to forward flow) in the range 30– 35 m h1. Screen area is defined by the maximum allowable head loss through the screens, which is typically in the range of 5–10 cm. The screen superficial hydraulic load is typically in the range of 50–55 m h1 for average design conditions. The screens and their supporting structural assemblies, if required, are typically constructed from stainless steel and may be from wedge-wire mesh or perforated plates. Low-pressure diffused air is applied to aerobic MBBRs. The airflow enters the reactor through a network of air piping and diffusers that are attached to the tank bottom. Airflow has the dual purpose of meeting process oxygen requirements and uniformly distributing plastic biofilm carriers. To promote uniform distribution of the plastic biofilm carriers, the diffuser grid layout and drop pipe arrangement provide a rolling water circulation pattern. Coarse-bubble diffusers are typically used in moving bed reactors (Figure 25). Coarse-bubble diffusers typically used in MBBRs are stainless steel pipes with circular orifices along the underside. These coarse-bubble diffusers are less affected by scaling and fouling because of the large dimension and turbulent airflow through the discharge orifice (Stenstrom and Rosso, 2008). As a result, coarse-bubble diffusers require less maintenance than fine-bubble diffusers. The coarse-bubble diffusers are designed with a structural end

Biofilms in Water and Wastewater Treatment

551

Aerated reactor #2

Aerated reactor #1

RECIR

Mixer Mixed bed reactor #2

Screen

Effluent overflow Effluent

Airflow distribution area

RECIR pump Effluent basin

RECIR

RECIR

Mixed bed reactor #1

Effluent

Influent Influent splitter box

Aerated reactor #3

RECIR

Aerated reactor #4

(a)

Mixed bed reactor #4

Mixed bed reactor #3

RECIR

Effluent

RECIR

(b)

Figure 22 (a) Moving bed biofilm reactor at the Williams-Monaco Wastewater Treatment Plant, Colorado, USA. (b) Schematic representation of the photographed system which illustrates the system consisting of two parallel trains each with four reactors in series.

support that enables them to withstand the weight of plastic biofilm carriers when the MBBR is out of service and drained. Denitrification MBBRs use mechanical mixers to agitate the bulk of the liquid and to distribute plastic biofilm carriers uniformly throughout the tank. The mechanical mixers are typically rail-mounted submersible (wet motor) units. Stateof-the-art submersible mechanical mixers typically have a maximum 120-rpm impeller speed and a minimum of three blades per impeller. The mixer uses a stainless steel backwardcurve propeller with a round bar welded along its leading edge to avoid damage to the plastic biofilm carriers and impeller wear. The mixer has a large diameter impeller with a fairly low rotational speed (90 rpm at 50 Hz and 105 rpm at 60 Hz). The plastic biofilm carriers float in quiescent water. As a result, the mixers need to be located near the water surface but not so close as to create an air-entraining vortex. A slight negative

inclination of mixer orientation helps maintain the rollingwater circulation pattern and uniformly distribute plastic biofilm carriers (see Figure 24). Rail-mounted units facilitate access to the mixer when maintenance is required. The mixers are typically sized to input 25 W m3 of reactor volume. Carbon-oxidizing MBBRs are classified as low-rate, normalrate, or high-rate bioreactors. Low-rate carbon-oxidizing MBBRs promote conditions for nitrification in downstream reactors. High- and normal-rate MBBRs are strictly carbon-oxidizing bioreactors. In the absence of site-specific pilot-scale observations or a calibrated mathematical model, high-rate MBBRs are typically designed to receive a filtered BOD5 load in the range of 15–20 g m2 d1 at 15 1C. This corresponds to total BOD5 loads as high as 45–60 g m2 d1 at 15 1C (Ødegaard, 2006). To reach secondary treatment effluent standards, a hydraulic residence time less than 30 min is not

552

Biofilms in Water and Wastewater Treatment

Table 2

Moving bed biofilm reactor plastic biofilm carrier characteristicsa

Manufacturer

Name

Bulk specific surface area, weight, gravity

Nominal carrier dimensions (depth; diameter)

Veolia Inc.

AnoxKaldnesTM K1

500 m2 m3 145 kg m3 0.96–0.98

7.2 mm; 9.1 mm

AnoxKaldnesTM K3

500 m2 m3 95 kg m3 0.96–0.98

10 mm; 25 mm

AnoxKaldnesTM Biofilm Chip (M)

1,200 m2 m3 234 kg m3 0.96–1.02

2.2 mm; 45 mm

AnoxKaldnesTM Biofilm Chip (P)

900 m2 m3 173 kg m3 0.96–1.02

3 mm; 45 mm

ActiveCellTM 450

450 m2 m3 134 kg m3 0.96

15 mm; 22 mm

ActiveCellTM 515

515 m2 m3 144 kg m3 0.96

15 mm; 22 mm

ABC4TM

600 m2 m3 150 kg m3 0.94–0.96

14 mm; 14 mm

ABC5TM

660 m2 m3 150 kg m3 0.94–0.96

12 mm; 12 mm

BioPortzTM

589 m2 m3

14 mm, 18 mm

Infilco Degremont Inc.

Aquise

Entex Technologies Inc.

Carrier photo

a

As reported by manufacturer. Modified from Boltz JP, Morgenroth E, deBarbadillo C, et al. (2010b) Biofilm reactor technology and design. In: Design of Municipal Wastewater Treatment Plants, WEF Manual of Practice No. 8, ASCE Manuals and Reports on Engineering Practice No. 76, 5th edn, vol. 2, ch. 13, (ISBN P/N 978-0-07-166360-1 of set 978-0-07-166358-8; MHID P/N 0-07166360-6 of set 0-07-166358-4). New York: McGraw-Hill.

recommended. Medium-rate MBBRs designed for meeting basic secondary treatment standards are typically designed for a loading of 5–10 g BOD5 m2 d1 at 10 1C, depending on the choice of liquid–solids separation process. Values in the higher range are used when coagulation occurs before the separation unit; values in the lower range are used without coagulation. Studying a pilot-scale combined carbon oxidation and nitrification MBBR receiving primary effluent and a (tertiary) nitrification MBBR receiving secondary effluent while maintaining a 4–6 g m3 bulk-liquid dissolved-oxygen concentration in both units, Hem et al. (1994) observed that a total BOD5 load of 1–2 g m2 d1 resulted in nitrification rates

from 0.7 to 1.2 g m2 d1, a total BOD5 load of 2–3 g m2 d1 resulted in nitrification rates from 0.3 to 0.8 g m2 d1, and a total BOD5 load greater than 5 g m2 d1 resulted in virtually no nitrification.

4.15.3.5.3 Biologically active filters BAFs have natural mineral, structured or random plastic media that supports biofilm growth and serves as a filtration medium. Solids accumulated from filtration and biochemical transformation are removed by backwashing. Media density influences BAF configuration and backwash regimes. BAF

Biofilms in Water and Wastewater Treatment

553

(a)

(b)

Figure 23 (a) Horizontal cylindrical screens constructed of wedge wire. Stainless steel coarse-bubble diffusers typically used in aerobic MBBRs are also pictured on the tank floor. (b) Flat wall screen constructed of wedge wire. A single air-header is pictured. Air is periodically introduced to scour debris accumulated on the screen.

A

B

30°

D (a)

(b)

Figure 24 (a) Schematic and (b) picture of mechanical mixers that are specially designed for anoxic moving bed biofilm reactors.

influent requires preliminary and primary treatment. Historically, the acronym BAF has meant biological aerated filters which have been used to refer to aerated biofilters used for secondary treatment. However, Boltz et al. (2010b) revised the acronym BAF to cover all BAFs, including those that operate under anoxic conditions for denitrification. BAFs are characterized by their media configurations and flow regime.

Downflow BAFs with media heavier than water include the Biocarbones process, which was marketed during the 1980s for secondary and tertiary treatment, and packed-bed tertiary denitrification filters such as the Tetra Denites process. These BAFs are backwashed using an intermittent counter-current flow. Upflow BAFs with media heavier than water such as the Infilco Degremont Biofors process have been used for

554

Biofilms in Water and Wastewater Treatment

secondary and tertiary treatment. The systems make use of expanded clay or another mineral media. These BAFs are backwashed using an intermittent concurrent flow. BAFs with floating media such as the Veolia Biostyrs process have also been used for secondary and tertiary treatment, and uses polystyrene, polypropylene, or polyethylene media. These BAFs operate with an intermittent backwash counter-current flow. Continuous backwashing filters operate in an upflow mode and contains media that is heavier than water. The media continuously moves counter-current to the wastewater flow (i.e., downward), and is continuously channeled to a center air lift where it is scoured, rinsed, and returned to the top of the media bed. Nonbackwashing submerged filters consist of a submerged static media bed, and have been called submerged aerated filters (SAFs). Solids are not retained in these filters. Therefore, nonbackwashing submerged filters require a dedicated liquid–solids separation process. A downflow BAF with media heavier than water, such as the Tetra Denites filter, is illustrated in Figure 25. The

Denites process has been used since the late 1970s for meeting stringent total nitrogen limits while providing a filtered effluent. Methanol or another external carbon source is added to the influent wastewater stream to promote biological denitrification. A typical installation includes 1.8 m of 2–3 mm diameter sand media over 457 mm of graded support gravel. In a downflow denitrification BAF, the backwash cycle typically consists of a brief air scour followed by an air–water backwash and water rinse cycle. Backwash water and air scour flow rates are typically 15 and 90 m3 m2 h1, respectively. Backwash water usage is typically 2–3% of the average flow being treated. Nitrogen gas accumulates in the media. A releasing mechanism is pumping backwash water up through the media bed for a short duration. The denitrification capacity between nitrogen release cycles typically ranges from 0.25 to 0.5 kg NOX-N m2. An upflow BAF with media heavier than water, such as the Infilco Degremont Biofors, is illustrated in Figure 26. The Degremont Biofors operates such that solids are trapped Proces air

Raw water

Air

Backwash water extraction

Water Biofilter media Support layer

Air scour Backwash water Treated water Figure 25 Downflow BAF with media heavier than water (e.g., Biocarbones and Tetra Denites). From ATV (1997) Biologische und weitergehende Abwasserreinigung (German), 4th edn. Berlin: Ernst and Sohn as presented by Tschui (1994).

Water

Process air

Biofilter media

Backwash water extraction

Air

Support layer

Air scour Treated water Backwash water

Raw water Figure 26 Upflow BAF with media heavier than water (e.g., Infilco Degremont Biofors). From ATV (1997) Biologische und weitergehende Abwasserreinigung (German), 4th edn. Berlin: Ernst and Sohn as presented by Tschui (1994).

Biofilms in Water and Wastewater Treatment

mostly in the lower part of the filter medium during normal operation and are removed through backwashing and applying scour air. As the backwash consists of concurrent scour air and backwash water, accumulated solids travel up through the media bed before being released at the top. Three types of media can be used in the Biofors depending on the application; the media types include expanded clay, expanded shale (both in the form of spherical grains with an effective size of 3.5 or 4.5 mm), and angular grains (with an effective size of 2.7 mm). The media form a submerged, fixed bed in the bottom of the reactor. The media bed typically has a height of 3–4 m with approximately 1-m freeboard. The grains-specific surface area is approximately 1640 m2 m3. Influent water to the bed flows through a plenum and nozzle air/water distribution system. The nozzles are installed in a false floor located approximately 1 m above the filter floor. Nozzles in the false floor are subject to clogging. Therefore, backwash water and scour air flow through the same plenum/nozzle system. Process air is introduced through separate air diffusers located in the media bed above the inlet nozzles. A key issue with the backwash of sunken media systems is the potential for boils during backwashing. The flow will short-circuit through the line of least resistance. This will result in a boil, or violent eruption of the flow through the point of least resistance. Similar short circuits and boils can also occur if the nozzles are blocked. These boils can result in excessive media loss during backwashing. Therefore, to achieve even backwashing the water must be well distributed across the BAF plan area. Therefore, the headloss across the distribution system must be greater than the headloss through the bed. An upflow BAF with floating media, such as the Veolia Biostyrs, is illustrated in Figure 27. These processes use a floating bed of media to provide area for biofilm development and filtration. Coarse-bubble aeration diffusers exist at the bottom of the media to enhance the contact of air, water, and biomass (Rogalla and Bourbigot, 1990). The Biostyrs process uses light weight expanded polystyrene (specific gravity of 0.05). Alternatively, the Biobeads process uses

555

recycled polypropylene with a specific gravity slightly lower than 1. The Biostyrs reactor is partially filled with (2–6 mm) polystyrene beads. Process objectives determine selection of the bead size; larger beads can be more heavily loaded. The beads, which are lighter than water, form a floating bed in the upper portion of the reactor, typically a height of 3–4 m with approximately 1.5 m of freeboard. The top of the bed is restrained by a slab fitted with filtration nozzles to evenly collect treated wastewater. The clean specific surface area of spherical beads is 1000–1400 m2 m3. In the bottom of the reactor, influent is distributed by troughs formed in the base of the cells. Process air is distributed through diffusers located along the bottom of the reactor or within an aeration grid in the media bed. The latter is used if an anoxic zone is required for denitrification. Backwashing consists of counter-current air scour and backwash water flow. The Biobeads BAF process is similar to Biostyrs, except that the media is larger and heavier, using polypropylene or polyethylene with a density of approximately 0.95. To prevent media attrition, a metal grid is fixed near the top of the reactor. Upflow floating media BAFs may also require a certain number of mini-backwashes (typically 4–8 and, in extreme cases, more than 10) to bump the filter, remove some solids, and lower headloss to achieve a complete filtration cycle of 24 or 48 h (which is the time between normal backwashes). The requirement for minibackwashes plus normal backwashes can generate a significant backwash water volume. During demonstration testing in San Diego, California, USA, a single-stage carbon-oxidation BAF with floating media generated a backwash water volume in the range of 10.3–13.9% of influent flow, compared to a sunken media BAF which produced a backwash water volume in the range of 7.4–7.9% (Newman et al., 2005). An upflow continuous backwash BAF, such as the Parkson Dynsands, is illustrated in Figure 28. Moving bed, continuous backwash filters operate in an upflow mode and consist of media heavier than water. The media continuously moves downward, counter-current to the wastewater flow. These filters are used widely for tertiary suspended solids and turbidity Backwash water

Air scour Process air

Air Aerobic filter zone

Treated water

Anoxic filter zone Water

Recirculation pump Raw water

Backwash water extraction Figure 27 Upflow BAF with floating media (e.g., Veolia Biostyrs). Adapted from ATV (1997) Biologische und weitergehende Abwasserreinigung (German), 4th edn. Berlin: Ernst and Sohn as presented by Tschui (1994).

556

Biofilms in Water and Wastewater Treatment Central reject compartment (H)

Feed (influent) (A)

Rejects (L) Top of airlift pump (G) Filtrate weir (J)

Reject weir (K) Sand washer (L) Effluent (E)

Downward moving sand bed (D)

Downward feed (B) Feed radials (C)

Bottom of airlift pump (F) Figure 28 Parkson Dynasands process schematic, continuous backwash BAF.

removal but have also been applied to separate stage nitrification and denitrification. Two commercially available systems using this technology are the Parkson DynaSands and Paques Astrasands filters. The filter cells are supplied as 4.65-m2 modules with center airlift assembly. The effective media depth is typically 2 m, and sand media size typically ranges from approximately 1 to 1.6 mm. Influent wastewater enters the filter bed through radials located at the bottom of the filter. The flow moves up through the downward-moving sand bed and effluent flows over a weir at the top of the filter. The media, with the accumulated solids, is drawn downward to the bottom cone of the filter. Compressed air is introduced through an airlift tube extending to the conical bottom of the filter and rises upward with a velocity exceeding 3 m s1 creating an air pump that lifts the sand at the bottom of the filter through the center column. The turbulent upward flow in the airlift provides scrubbing action that effectively separates solids from the media before discharge to a wash box. There is a constant upward flow of liquid into the wash box (backwash water) controlled by the wash box discharge weir. Moving bed filter manufacturers typically set the reject weir to provide a wash water flow rate equivalent to approximately 10% of the forward flow at an average filter loading rate of 4.9 m h1. The backwash frequency is quantified by the bed turnover rate. To maintain sufficient biomass for denitrification, the bed turnover rate must be reduced to approximately 100–250 mm h1.

Several media types are available for use in BAFs. Media selection is integral to meeting treatment objectives, flow and backwashing regimes. Typically, media can be categorized as mineral media and plastic media. In most cases, mineral media is denser than water and plastic media is buoyant. The media needs to resist breakdown from abrasion during backwashing and chemical degradation by constituents in municipal wastewater. Commercially available BAF systems and their media are listed in Table 3. Backwashing BAFs maximizes solids capture and filter run time. Proper backwashing requires filter bed expansion and rigorous scouring followed by efficient rinsing. Accumulation of solids and media (mud balling) results in wastewater short-circuiting and can result in excessive media loss. Feed characteristics and type of treatment provided by the BAF affect solids production and frequency requirements for backwashing. Biomass yield in tertiary BAF systems is typically low, so backwashing is relatively infrequent (i.e., one backwash per 36–48 h). Reactor characteristics and media type influence backwash frequency. More openly structured media capture fewer solids which reduces backwash frequency. During backwashing the media bed is typically expanded or fluidized (depending on the system) to allow for grain separation and free movement in order to remove as much accumulated solids as possible. Table 4 compares typical BAF backwashing requirements. BAFs designed for carbon oxidation and suspended solids removal in secondary treatment typically have volumetric BOD loading rates in the range of 1.5–6 kg m3 d1. Average and peak HLRs for secondary and tertiary treatment systems are typically in the range of 4–8 and 10–20 m h1, respectively. As BAFs for secondary treatment are typically placed immediately downstream of primary clarifiers, the applied volumetric mass loading rate is almost always the limiting design parameter. Combined carbon oxidation and nitrification will proceed when the organic loading at lower temperatures is limited to 2.5 kg BOD m3 d1 (Rogalla et al., 1990). Under these conditions a total Kjeldahl nitrogen removal rate of 0.4 kg N m3 d1 may be achieved. Inversely, Rogalla et al. (1990) found that nitrification decreases when soluble COD loadings approach 4 kg m3 d1. Ammonium removal of 80– 90% can be achieved for ammonium loads in the range of 2.5–2.9 kg m3 d1 (Peladan et al., 1996).

4.15.3.5.4 Expanded and fluidized bed biofilm reactors Expanded bed biofilm reactors (EBBRs) and FBBRs use small media particles that are suspended in vertically flowing wastewater, so that the media becomes fluidized and the bed expands. Individual particles become suspended once the drag force of the relatively fast flowing wastewater (30–50 m h1) overcomes gravity and they are separated. In municipal applications, fluidized beds are typically used for tertiary denitrification. Design criteria for denitrifying FBBRs are listed in Table 5. When treating groundwater or industrial wastewater, FBBRs are used for the removal of oxidized contaminants such as nitrate and perchlorate. Suspension of the media maximizes the contact surface between microorganisms and wastewater. It also increases treatment efficiency by improving mass transfer because there

Biofilms in Water and Wastewater Treatment Table 3

557

Biologically active filter systems and commercially available media

Process

Supplier

Flow regime

Media

Specific gravity

Size (mm)

Astrasands Biobeads Biocarbones Biofors Biolest Biopur

Paques/Siemens Brightwater F.L.I. OTV/Veolia Degremont Stereau Sulzer/Aker Kvaerner Kruger/Veolia Severn Trent Severn Trent Parkson FB Leopold Severn Trent

Upflowa Upflow Downflow Upflow Upflow Downflow

Sand Polyethylene Expanded shale Expanded clay Pumice/pouzzolane Polyethylene

42.5 0.95 1.6 1.5–1.6 1.2

1–1.6

Upflow Upflow Downflow Upflowa Downflow Up/down

Polystyrene Sand Sand Sand Sand Slag

0.04–0.05 2.6 2.6 2.6 2.6 2–2.5

3.3–5 2–3 2–3 1–1.6 2 28–40

Washed gravel

2.6

19–38

Biostyrs ColoxTM Denites Dynasands Eliminites Submerged activated filter

2–6 2.7, 3.5, and 4.5

Specific surface area (m2 m3)

1400–1600

Structured 1000 656 656

240

a

Moving bed. From Boltz JP, Morgenroth E, deBarbadillo C, et al. (2010b) Biofilm reactor technology and design. In: Design of Municipal Wastewater Treatment Plants, WEF Manual of Practice No. 8, ASCE Manuals and Reports on Engineering Practice No. 76, 5th edn, vol. 2, ch. 13 (ISBN P/N 978-0-07-166360-1 of set 978-0-07-166358-8; MHID P/N 0-07-166360-6 of set 0-07-166358-4). New York: McGraw-Hill.

Table 4

Summary of biologically active filter (BAF) backwashing (BW) requirements

Upflow, sunken media Normal BW Energetic BWa Upflow, floating media Normal BW Mini-BWb Downflow, sunken media Upflow, moving bedf

Backwash water rate, m h1

Air scour rate, m h1

Total duration minc

Total backwash water volume per cellc

Total backwash water volume per celld

20 (8.2) 30 (12.3)

97 (5.3) 97 (5.3)

50 25

9.2 m3 m2 9.2 m3 m2

12 m3 m2 10 m3 m2

55 (22.5) 55 (22.5) 15 (6) 0.5–0.6

12 (0.65) 12 (0.65) 90 (5) Continuous through air lift

16 5 20–25 Continuous

2.5 m3 m3 mediae 1.5 m3 m3 mediae 3.75–5 m3 m2 55–67 m3 d1

2.5 m3 m3 mediae 1.5 m3 m3 mediae 3.75–5 m3 m2 55–67 m3 d1

(0.2–0.24) a

Energetic backwash once every 1–2 months depending on trend in ‘‘clean bed’’ headloss following normal backwash. Mini-backwash applied as interim measure when pollutant load exceeds design load. c Backwash duration reflects total duration of the typical backwash cycle, which includes valve cycle time and pumping and nonpumping steps. The duration of each step is adjustable via programmable logic controller and supervisory control and data acquisition control systems. d The total backwash water volume includes drain and filter to waste steps, where applicable. e Backwash volume requirements for upflow floating media BAF typically are based on media volume rather than cell area because depths vary. f Continuous backwash filter BW is based on a standard 4.65 m2 cell and a typical weir setting for reject flow of approximately 2.3–2.8 m3 h1 cell1. From Boltz JP, Morgenroth E, deBarbadillo C, et al. (2010b) Biofilm reactor technology and design. In: Design of Municipal Wastewater Treatment Plants, WEF Manual of Practice No. 8, ASCE Manuals and Reports on Engineering Practice No. 76, 5th edn, vol. 2, ch. 13 (ISBN P/N 978-0-07-166360-1 of set 978-0-07-166358-8; MHID P/N 0-07-166360-6 of set 0-07-166358-4). New York: McGraw-Hill. b

is significant relative motion between the biofilm and flowing wastewater. Because of the balance of forces involved in particle fluidization and bed expansion, the smallest particles are found at the top and the largest at the bottom of the fluid bed. Therefore, the media particles should be graded to a relatively tight size range. The degree of bed expansion determines whether a bed is deemed expanded or fluidized. The transition lies between 50% and 100% expansion over the static bed height. This discussion assumes the upper limit: beds less than double static bed height

(o100% expanded) are considered expanded; those more than double the static bed height (4100% expanded) are fluidized. A lower degree of bed expansion is advantageous, because it requires a lower flow velocity, less energy, and increases effective biomass concentration, which reduces the reactor footprint. In aerobic processes, however, it increases volumetric oxygen demand because of increased biomass concentration. The FBBR/EBBR is illustrated in Figure 29. The system consists of a column in which the particles are fluidized and a

558

Biofilms in Water and Wastewater Treatment

Table 5

Design criteria for denitrifying fluidized bed biofilm reactors

Parameter

Value

Packing Type Effective size Sphericity Uniformity coefficient Specific gravity Initial depth Bed expansion Empty-bed upflow velocity Hydraulic loading rate Recirculation ratio NO 3  N loading: 13 1C 20 1C Empty-bed contact time C:N (methanol) Specific surface areaa Biomass concentrationa

Unit

Range

Typical

mm Unitless Unitless Unitless m % m h1 m3effluent m2 Bioreactor Unitless

Sand 0.3–0.5 0.8–0.9 1.25–1.50 2.4–2.6 1.5–2.0 75–150 36–42 400–600 2:1–5:1

Sand 0.4 0.8–0.85 r1.4 2.6 2.0 100 36 500 3:1

2.0–4.0 3.0–6.0 10–20 3.0–3.5 1000–3000 15 000–40 000

3.0 5.0 15 3.2 2000 30 000

area

kg m3 d1 kg m3 d1 min Unitless m2 m3 mg l1

d1

a

Specific surface area range based on sand particles; alternate media used in fluidized bed reactors such as carbon or glassy coke may have a different specific surface area range. From Boltz JP, Morgenroth E, deBarbadillo C, et al. (2010b) Biofilm reactor technology and design. In: Design of Municipal Wastewater Treatment Plants, WEF Manual of Practice No. 8, ASCE Manuals and Reports on Engineering Practice No. 76, 5th edn, vol. 2, ch. 13 (ISBN P/N 978-0-07-166360-1 of set 978-0-07-166358-8; MHID P/N 0-07-166360-6 of set 0-07-166358-4). New York: McGraw-Hill.

Figure 29 Fluidized bed biofilm reactor process flow diagram (Shieh and Keenan, 1986).

Process flow enters at the bottom of the reactor and flows through a distribution system to ensure even dispersion and uniform fluidization. Silica sand (0.3–0.7 mm diameter) and granular activated carbon (GAC; 0.6–1.4 mm) are typically used. Other materials, however, have been used at pilot scale, such as 0.7–1.0 mm glassy coke (McQuarrie et al., 2007). Small carrier particles (1 mm) provide a large specific surface area for biofilm growth (up to 2400 m2 m3 when expanded 50%), which is one of the key advantages of this process technology. In a study of tertiary nitrification of activated sludge-settled effluent using a pilot-scale EBBR, Dempsey et al. (2006) found that the process also removed up to 56% CBOD and 62% TSS from the influent stream. Removal of these materials was attributed to the activities of protozoa (free-living and stalked) and metazoa (rotifers, nematodes, and oligochaetes) as shown in Figure 30.

recycle line that is used to maintain a fixed, vertical hydraulic flow. In this way, bed expansion is kept constant and biofilm covered particles are retained independent of influent flow. Aeration typically is achieved during recycle, during which influent wastewater mixes with effluent recycled from the top of the bed. If aeration is conducted within the fluidized bed, then a significant volume of gas disturbs the fluidized state by causing turbulence and increased force of interparticle collisions. This can cleave biofilm from the substratum. Nevertheless, this approach has been used. The advantage of adding air to the recycle stream is that biomass is not stripped from the media by turbulence of rising gas bubbles; therefore, the treated effluent typically has a lower suspended concentration (Jeris et al., 1981).

The RBC process has been applied where average effluent water-quality standards are less than or equal to 30 mg l1 BOD5 and TSS. The RBC employs a cylindrical, synthetic media bundle that is mounted on a horizontal shaft. Figure 31 illustrates the shaft-mounted media. The bundled media is partially submerged (typically 40%) and slowly (1–1.6 rpm) rotates to expose the biofilm to substrate in the bulk of the liquid (when submerged), and to air (when not submerged). Detached biofilm fragments suspended in the RBC effluent stream are removed by liquid– solids separation units. The RBC process is typically configured with several stages operating in series. Each reactorin-series may have one or more shafts. Parallel trains are

Effluent

Excess biomass

Recycle Separator

Influent

Media Reactor Bioparticle

O2 Chemicals (optional) 1 2

1 Medium 2 Biofilm

4.15.3.5.5 Rotating biological contactors

Biofilms in Water and Wastewater Treatment

(a)

(b)

(c)

(d)

(e)

(f)

559

Figure 30 Particulate biofilms with associated protozoa and metazoan from expanded bed: (a) bioparticles in expanded bed; (b) bioparticles with surface attached; (c) closeup of rotifer attached to bioparticle; (d) stalked protozoa on surface of particulate biofilms; (e) testate amoeba grazing on biofilm; and (f) oligochaete worm grazing on bioparticles (Dempsey et al., 2006).

implemented to provide additional surface area for biofilm development. Media-supporting shafts typically are rotated by mechanical drives. Diffused air-drive systems and an array of airentraining cups that are fixed to the periphery of the media (to capture diffused air) have been used to rotate the shafts. RBCs have failed as a result of shaft, media, or media support system structural failure; poor treatment performance; accumulation of nuisance macrofauna; poor biofilm thickness control; and inadequate performance of air-drive systems for shaft rotation. Typically, the RBC tank is sized at 4.9 103 m3 m2 of media for low-density units. Disks typically have a 3.5-m diameter and are situated on a 7.5-m-long rotating shaft. The RBCs may contain low- or high-density media. Low-density media has a 118-m2 m3 biofilm active specific surface; high-density units have 180 m2 m3. Low-density media typically are used in the first stages of RBC systems which are designed for BOD5 removal to reduce potential media clogging and weight problems resulting from substantial biofilm accumulation. High-density media typically is used for nitrification. Mechanical shaft drives consist of an electric motor, speed reducer, and belt or chain drive. Typically, 3.7-kW mechanical drives have been provided for full-scale RBCs. Air-driven shafts require a remote blower for air delivery. Air headers are equipped with coarse-bubble diffusers. The air flow rate is typically in the range of 4.2–11.3 m3 min1 per shaft. Air quantity required by systems using air-driven shaft rotation, however, must be evaluated on a site-specific basis. Mechanical drive units have been designed for operation from 1.2 to 1.6 rpm. Air-drive units have been designed for 1.0–1.4 rpm. Ideally, shaft rotational speed is consistent. The development of an evenly distributed biofilm is desirable to avoid an uneven weight distribution, which may cause cyclical loadings in mechanical-drive systems and loping (uneven rotation) in air-driven shaft rotating systems. A loping condition often

accelerates rotational speed and, if not corrected, may lead to inadequate treatment and the inability to maintain shaft rotation. Air-drive systems should provide ample reserve air supply to maintain rotational speeds, restart stalled shafts, and provide short-term increased speeds (2–4 times normal operation) to control excessive or unbalanced biofilm thicknesses. Available data indicate that in excess of an 11.3-m3 min1 airflow rate per shaft may be required to maintain a 1.2-rpm shaft rotational speed during peak organic loading conditions (Brenner et al., 1984). Large-capacity air cups (150 mm diameter) typically are provided in the first stage of the process to exert a greater torque on the shaft and reduce loping. The RBC process is typically covered to avoid ultraviolet (UV) light-induced media deterioration and algae growth, to prevent excessive cooling, and to provide odor control. RBCs have been installed in buildings or under prefabricated fiberglass-reinforced plastic (FRP) covers (as pictured in Figure 31).

4.15.3.5.6 Trickling filters The TF is a three-phase biofilm reactor with fixed carriers. Wastewater enters the bioreactor through a distribution system, trickles downward over the biofilm surface, and air moves upward or downward in the third phase where it diffuses through the flowing liquid and into the biofilm. TF components generally include an influent water distribution system, containment structure, rock or plastic media, and underdrain and ventilation system. Wastewater treatment using the TF results in a net production of total suspended solids. Therefore, liquid–solids separation is required, and is typically achieved with circular or rectangular secondary clarifiers. The TF process generally includes an influent/ recirculation pump station, the TF(s), and liquid–solids separation unit(s).

560

Biofilms in Water and Wastewater Treatment

Figure 31 Photograph of the Envirexs rotating biological contactor cylindrical synthetic media bundle mounted on a horizontal shaft (a) and rotating biological contactor covers (b). Photographs courtesy Siemens Water Technologies.

Figure 32 (a) Hydraulically driven rotary distributors use variable frequency drive controlled gates that either open or close distributor orifices which adjust with varying pumped flow rates to maintain a constant preset rotational speed. (b) Electrically driven rotary distributor. Photographs courtesy WesTech, Inc.

Primary effluent or screened and degritted wastewater is either pumped or flows by gravity to the TF distribution system. Essentially, there are two types of TF distribution systems: fixed-nozzle and rotary distributors. Because their efficiency is poor, distribution with fixed nozzles should not be used (Harrison and Timpany, 1988). Rotary distributors may be hydraulically or electrically driven. A properly designed rotary distribution system allows for effective media wetting and the intermittent application of wastewater to biofilm carriers. The intermittent application of influent wastewater allows the biofilm to have periods of resting which primarily serves as a process aeration mechanism. Poor media wetting may lead to dry pockets, ineffective treatment zones, and odor. An electrically or modern hydraulically driven rotary distributor

(Figure 32) controls rotational speed independent of the influent wastewater flow rate, and may be used to maintain the desired hydraulic dosing rate. Ideal TF media provides a high specific surface area, low cost, high durability, and high enough porosity to avoid clogging and promote ventilation (Metcalf and Eddy, 2003). TF media types include rock (RO), random (RA) (synthetic), vertical flow (synthetic) (VF), and cross-flow (synthetic) (XF). Both VF and XF media are constructed with smooth and/or corrugated plastic sheets. Another commercially available synthetic media, although not commonly used, is vertically hanging plastic strips. Horizontal redwood or treated wooden slats have also been used, but are generally no longer considered viable because of high cost or limited supply. Modules

Biofilms in Water and Wastewater Treatment

of plastic sheets (i.e., self-supporting VF or XF modules) are used almost exclusively for new and improved TFs, but several TFs with rock media exist, and have proven capable of meeting treatment objectives when properly designed and operated. Table 6 compares the characteristics of some TF media. The higher specific surface area and void space in modular synthetic media allow for higher hydraulic loading, enhanced oxygen transfer, and biofilm thickness control in comparison to rock media. Rock media has, ideally, a 50-mm diameter, but may range in size. Due to structural requirements associated with the large unit weight of rock, rock-media TFs are shallow in comparison to synthetic-media TFs. Their large surface area makes them more susceptible to excessive cooling. Generally, rock media is considered to have a low specific surface area, void space, and high unit weight. Although recirculation is common, the low void ratio in rock-media TFs limits hydraulic application rates. Excessive hydraulic application can result in ponding, limited oxygen transfer, and poor bioreactor performance. Performance of existing rock-media TFs may sometimes be improved by providing mechanical ventilation, solids contact channels, and/or deepened secondary clarifiers that include energy dissipating inlets and flocculator-type feed wells. Grady

Table 6

et al. (1999) suggested that under low organic loading (i.e., o1 kg BOD5 d1 m3) rock- and synthetic-media TFs are capable of equivalent performance. However, as organic loading increases, synthetic-media TFs are less susceptible to operational problems and have reduced potential for plugging. Synthetic TF media has a higher specific surface area and void space, and lower unit weight than rock media. Modular synthetic media is generally manufactured with the following specific surface areas: 223 m2 m3 as high density, 138 m2 m3 as medium density, and 100 m2 m3 as low density. Both VF and XF media are reported to remove BOD5 and NH3–N (Harrison and Daigger, 1987), but sufficient scientific evidence exists to surmise that there is a difference in the treatment efficiency of TFs constructed with XF and VF media even when manufactured with virtually identical specific surface areas. Plastic modules with a specific surface area in the range of 89–102 m2 m3 are well suited for carbon oxidation and combined carbon oxidation and nitrification. Parker et al. (1989) recommended medium-density XF media against the use of high-density XF media in nitrifying TFs. This is supported by observations from a pilot-scale nitrifying TF application data and conclusions of Gujer and Boller (1983, 1984)

Properties of some trickling filter media Nominal size (m)

Bulk density (kg m3)

Specific surface area (m2 m3)

Void space (%)

0.024–0.076

1442

62

50

0.076–0.128

1600

46

60

0.61  0.61  1.22

24–45

100, 138, and 223

95

Vertical flow

0.61  0.61  1.22

24–45

102 and 131

95

Randomb

0.185 ø  0.051 H

27

98

95

Media type Rock River

Slag

Plastica Cross flow

a

561

Material

Manufacturers of modular plastic media: (formerly) BF Goodrich, American Surf-Pac, NSW, Munters, (currently) Brentwood Industries, Jaeger Environmental, and SPX Cooling. Manufacturers of random plastic media: (formerly) NSW Corp. and (currently) Jaeger Environmental.

b

562

Biofilms in Water and Wastewater Treatment

which show lower nitrification (flux) rates for lower-density modular synthetic media. The researchers claim that lower rates occur with high-density media due to the development of dry spots below the flow interruption points (i.e., higherdensity media has more flow interruptions and, therefore, less effective wetting). Using medium-density media also reduces plugging potential. Vertically oriented modular synthetic (VF) media is generally accepted as being ideally suited for highstrength wastewater (perhaps industrial) and high organic loadings such as with a roughing TF. In some cases, XF media has been placed in the top layer to enhance wastewater distribution and VF media comprises the remainder of the TF media. Typically, the top layer of a TF’s modular plastic media is covered with FRP or HDPE grating. The grating protects modular plastic media from deterioration by UV light and potential structural damage that may result from waterinduced load exerted during periods of high-intensity dosing. Figure 33 illustrates a typical TF column and a picture of the grating. Rock and random synthetic media are not self-supporting and require structural support to contain the media within the bioreactor. These containment structures are typically precast or panel-type concrete tanks. When self-supporting media such as plastic modules are used, other materials such as wood, fiberglass, and coated steel have been used as containment structures. The containment structure serves to avoid wastewater splashing, and to provide media support, wind protection, and flood containment. In some cases TF containment structures have been designed to allow flooding of the media, which increases operator flexibility in controlling macrofauna accumulation. The TF underdrain system is designed to meet two objectives: collect treated wastewater for conveyance to downstream unit processes and create a plenum that allows for the transfer of air throughout the TF media (Grady et al., 1999). Clay or concrete underdrain blocks are commonly used for rock-media TFs because of the required structural support. A variety of support systems including concrete piers and reinforced fiberglass plastic grating are used for other media types. The volume created between concrete and media bottom creates the underdrain.

TFs require oxygen to sustain aerobic biochemical transformation processes. The VF of air through the media can be induced mechanically or by natural draft. Natural air ventilation results from a difference in ambient air temperature outside and inside the TF. The temperature causes air to expand when warmed or contract when cooled. The net result is an air-density gradient throughout the TF, and an air front either rises or sinks depending on the differential condition. This rising or sinking action results in a continuous air flow through the bioreactor. Natural ventilation may become unreliable or inadequate in meeting process air requirements when neutral temperature gradients do not produce air movement. Currently, the provision of adequate underdrain and effluent channel sizing to permit free air flow is standard. Passive devices for ventilation include vent stacks on the TF periphery, extensions of underdrains through TFs side walls, ventilating manholes, louvers on the sidewall of the tower near the underdrain, and discharge of TF effluent to the subsequent settling basin in an open channel or partially filled pipes. Drains, channels, and pipes should be sufficiently sized to prevent submergence greater than 50% of their crosssectional area under design hydraulic loading. Ventilating access ports with open grating covers should be installed at both ends of the central collection channel. Large diameter TFs typically have branch channels (to collect the treated wastewater). These branches should also include ventilating manholes or vent stacks installed at the TF periphery. The open area of the slots in the top of the underdrain blocks should not be less than 15% of the TF area. One square meter gross area of open grating in ventilating manholes and vent stacks should be provided for each 23 m2 of TF area. Typically, 0.1 m2 of ventilating area is provided for every 3–4.6 m of TF periphery, and 1–2 m2 of ventilation area in the underdrain area per 1000 m3 of TF media. Another criterion for rockmedia TFs is the provision of a vent area at least equal to 15% of the TF cross-sectional area. Mechanical ventilation enhances and controls air flow with low-pressure fans that continuously circulate air throughout the TF. Therefore, a majority of new and improved TFs use low-pressure fans to mechanically promote air flow. The air flow resulting from natural draft will distribute itself. This will

Figure 33 Skid-resistant (polyethylene or fiberglass-reinforced plastic) grating placed on top of a typical modular plastic media trickling filter column.

Biofilms in Water and Wastewater Treatment

not occur with mechanical ventilation. Pressure loss through synthetic TF media is typically low, often less than 1-mm H2O/ m of TF depth (Grady et al., 1999). The low pressure drop typically results in low fan power requirements (B3–5 kW). The head on the fan is typically less than 1500-mm H2O. Unfortunately, the low pressure drop allows air to rise upward through the TF media without distributing itself across the bioreactor section. Therefore, fans are typically connected to distribution pipes. The air flow distribution piping has openings sized such that air flow through each is equal and air flow distribution is uniform. The pipes typically have a velocity in the range of 1100–2200 m h1 in order to further promote uniform air flow distribution. Air flow requirements are calculated based on process oxygen requirements and characteristic oxygen-transfer efficiency which is typically in the range of 2–10%. The mechanical air stream may flow upward to downward. Down-flow systems can be designed without covers. However, covers are required for systems that do not have air distribution through a network of pipes under the media. Covering TFs offers a wintertime benefit of limiting cold airflow and minimizing wastewater cooling. Mechanical ventilation and covered TFs may be used to destroy odorous compounds. A critical unit in the TF process is the pump station that lifts primary effluent (or screened raw sewage), and recirculates unsettled trickling effluent (here, referred to as underflow) to the influent stream. In general, TF underflow is recirculated to the distribution system to achieve the hydraulic load (influent þ recirculation) required for proper media wetting and biofilm thickness control, and decouple hydraulic and organic loading. TF influent is generally pumped to allow TF underflow to flow by gravity to the suspended growth reactor (or solids contact basin), secondary clarifier, or other downstream of the TF. When fit with weirs, a single pump station can be used to convey both influent and recirculation streams.

Table 7

563

TFs can be classified as roughing, carbon oxidation, carbon oxidation and nitrification, and nitrification. Table 7 summarizes characteristics of each TF. The performance ranges are associated with average design conditions. Single day or average week observations may significantly be greater.

4.15.4 Part III. Undesirable Biofilms: Examples of Biofilm-Related Problems in the Water and Wastewater Industries Biofilms are unavoidably associated with water environments, so biofilm control, a component of many industrial processes, is especially important in water and wastewater treatment. Depending on the particular setting, biofilms may cause process performance problems, material performance problems, health problems, and esthetic problems. The specific problems that biofilms cause in industrial settings are as diverse as the technological processes affected by the biofilms. In this section, we discuss four biofilm-related problems that have been reported in the water and wastewater industries: 1. biofilms on metal surfaces and MIC; 2. biofilms on concrete surfaces and crown corrosion of sewers; 3. biofilms on filtration membranes in drinking water treatment; and 4. biofilms on filtration membranes in wastewater treatment.

4.15.4.1 Biofilms on Metal Surfaces and MIC In the manufacturing of metals and metal alloys, raw materials – the ores – are chemically reduced and their internal chemical energy increases. These materials are used by microorganisms as sources of energy in a sequence of processes in which the chemical energy of the affected material decreases, bringing the energy levels of the products closer to

Trickling filter classification

Design parameter

Roughinga

Carbon oxidizing (cBOD5 removal)a

Carbon oxidation and nitrificationa

Nitrificationa

Media typically used

VF

RO, XF, or VF

RO, XF, or VF

XF

Wastewater source

Primary effluent

Primary effluent

Primary effluent

Secondary effluent

Hydraulic loading m3 d1 m2 BOD5 and NH3  N Load kg m3 d1 g m2 d1

52.8–178.2

14.7–88.0

14.7–88.0

35.2–88.0

1.6–3.52 NA

0.32–0.96 NA

0.08–0.24 0.2–1.0

NA 0.5–2.4

Conversion (%) or effluent concentration (mg l1) Macro fauna

50–75% filtered cBOD5 conversion No appreciable growth

20–30 mg l1 cBOD5 and TSSb Beneficial

0.5–3 mg l1 as NH3  Nb

Depth, m (ft)

0.91–6.10

r12.2

o10 mg l1 as cBOD5; o3 mg l1 as NH3  Nb Detrimental (nitrifying biofilm) r12.2

a

Detrimental r12.2

Applicable to shallow trickling filters. gpm ft2, gallons per minute per square foot of trickling filter plan area. Concentration remaining in the clarifier effluent stream. From Boltz JP, Morgenroth E, deBarbadillo C, et al. (2010b) Biofilm reactor technology and design. In: Design of Municipal Wastewater Treatment Plants, WEF Manual of Practice No. 8, ASCE Manuals and Reports on Engineering Practice No. 76, 5th edn, vol. 2, ch. 13, p. 238 (ISBN P/N 978-0-07-166360-1 of set 978-0-07-166358-8; MHID P/N 0-07166360-6 of set 0-07-166358-4). New York: McGraw-Hill. b

564

Biofilms in Water and Wastewater Treatment

the energy levels of the materials from which they were made. MIC can affect a variety of materials, both metallic and nonmetallic. If nonmetallic materials are affected, the term biodeterioration of materials is more often used than MIC, although this terminology is not very consistent and, for example, the term crown corrosion of sewers, which in fact refers to the biodeterioration of concrete, is quite popular among water professionals. Accelerated corrosion of metals in the presence of microorganisms stems from microbial modifications to the chemical environment near metal surfaces (Beech et al., 2005; Geiser et al., 2002; Lee and Newman, 2003; Lewandowski et al., 1997). Such modifications depend, of course, on the properties of the corroding metal and on the microbial community structure of the biofilm deposited on the metal surface (Beech and Sunner, 2004; Dickinson et al., 1996; Flemming, 1995; Olesen et al., 2000, 2001). Beech et al. (2005) described MIC as a consequence of coupled biological and abiotic electron transfer reactions, that is, redox reactions of metals enabled by microbial ecology (Beech et al., 2005). Hamilton (2003) attempted to generate a unified concept of MIC but found common features in only some of the possible mechanisms (Hamilton, 2003). It is unlikely that a unified concept of MIC can be generated at all. Rather, there are many mechanisms by which microorganisms may affect metal surfaces, and we demonstrate some of them here. These do not exhaust the possibilities, of course, but are rather used to exemplify the possible mechanisms. As we have restricted the discussion of MIC to metal surfaces only, it is convenient to define corrosion as anodic dissolution of a metal. In this way we can easily separate the corrosion reaction, the anodic dissolution of the metal, from many other anodic reactions that can occur at a metal surface covered with a biofilm. These other anodic reactions deliver electrons originating from substances metabolized near the metal surface, but only the reaction in which the metal itself is oxidized is defined as corrosion. The presence of the other anodic reactions causes confusion in MIC studies, as the current between the anode and the cathode is made up of electrons originating from many anodic reactions occurring at the surface, not only from the corrosion reaction. Microorganisms generate chemical environments that are conducive to corrosion reactions even if they do not take part in the process themselves. As in most industrial processes microorganisms are always present on metal surfaces, it is not immediately obvious whether the microorganisms attached to the surface accelerate the corrosion process or are just innocent bystanders. The only way to resolve this is by demonstrating that a specific mechanism of MIC is present because a product of microbial metabolism consistent with this mechanism can be detected. Many mechanisms of MIC have been proposed. Accelerated corrosion may result from the action of acid-producing bacteria, such as Thiobacillus thiooxidans and Clostridium aceticum; iron-oxidizing bacteria, such as Gallionella, Sphaerotilus, and Leptothrix; MOB, such as L. discophora; or hydrogenproducing bacteria. These mechanisms have been studied and the results described in numerous publications. We describe here representative examples of such mechanisms: the effects of differential aeration cells, sulfate-reducing bacteria (SRB corrosion), and MOB corrosion.

4.15.4.1.1 Differential aeration cells on iron surfaces MIC caused by differential aeration cells is an example of a nonspecific mechanism of MIC, because it depends on the presence of biofilm, and not on the type of microorganisms that reside in the biofilm. If the oxygen concentrations at two adjacent locations on an iron surface are different, then the half-cell potentials at these locations are different as well. The location where the oxygen concentration is higher will have a higher potential (more cathodic) than the location where the oxygen concentration is lower (more anodic). The difference in potential will give rise to a current flow from the anodic locations to the cathodic locations and to the establishment of a corrosion cell. This is the mechanism of differential aeration cells, and the prerequisite to this mechanism is that the concentration of oxygen varies among locations (Acuna et al., 2006; Dickinson and Lewandowski, 1996; Hossain and Das, 2005). Indeed, many measurements using oxygen microsensors have demonstrated that oxygen concentrations in biofilms can vary from one location to another (Lewandowski and Beyenal, 2007). If the anodic reaction is the oxidation of iron,

Fe-Fe 2þ þ 2e 

ð17Þ

and the cathodic reaction is the reduction of oxygen,

O2 þ 2H2 O þ 4e -4OH

ð18Þ

then the overall reaction describing the process is

2Fe þ O2 þ 2H2 O-2Fe2þ þ 4OH

ð19Þ

The Nernst equation quantifying the potential for this reaction is

E ¼ Eo 

0:059 ½Fe 2þ  2 ½OH   4 log 4 pðO2 Þ

ð20Þ

Figure 34 visualizes this mechanism.

4.15.4.1.2 SRB corrosion SRB causes corrosion of cast iron, carbon, and low alloy steels and stainless steels. SRB corrosion of potable water mains is a common (US EPA, 1984) and well-recognized problem (Seth and Edyvean, 2006; Tuovinen et al., 1980). MIC caused by SRB is an example of a mechanism that depends on the activity of a specific group of microorganisms in a biofilm. The corrosion of mild steel caused by SRB is the most notorious case of MIC, and it provides a direct and easy-to-understand link between microbial reactions and electrochemistry (Javaherdashti, 1999). According to the mechanism that was originally proposed by Von Wohlzogen Kuhr in 1934, SRB oxidizes cathodically generated hydrogen to reduce sulfate ions to H2S, thereby removing the product of the cathodic reaction and stimulating the progress of the reaction (Al Darbi et al., 2005). This mechanism was later found to be inadequate to explain the field observations. More involved mechanisms were implicated in this type of microbial corrosion, including the puzzling effect of oxygen, which can stimulate what is apparently an anaerobic process. It is now certain that the

Biofilms in Water and Wastewater Treatment

565

Aerated water Cathodic site; corrosion products

Biofilm

OH−

OH−

Cathode − e

Biof ilm

O2

e−

Anodic site

Biofilm

O2

O O2 Aerobic 2 O2 O2 O2 Anaerobic O2 O O2 2 O2 O2 Anaerobic O 2 O2 O2 O2 M+ M+ M+ Anode

O2 O2 OH−OH− e−

Cathode e−

1 mm

Metal (b)

(a)

Figure 34 Biofilm heterogeneity results in differential aeration cells. (a) This schematic shows pit initiation due to oxygen depletion under a biofilm (Borenstein, 1994). (b) An anodic site and pit under the biofilm and corrosion products deposited on mild steel.

possible pathways for cathodic reactions include sulfides and bisulfides as cathodic reactants (Videla, 2001; Videla and Herrera, 2005). The currently accepted mechanism of SRB corrosion is composed of a network of reactions that reflects the complexity of the environment near corroding metal surfaces covered with biofilms; the following paragraphs illustrate some of this complexity. The process starts with the microbial metabolism of SRB producing hydrogen sulfide by reducing sulfate ions. Hydrogen sulfide can serve as a cathodic reactant, thus affecting the rate of corrosion (Antony et al., 2007; Costello, 1974):

2H2 S þ 2e -H2 þ 2HS

ð21Þ

Ferrous iron generated from anodic corrosion sites precipitates with the metabolic product of microbial metabolism, hydrogen sulfide, forming iron sulfides, FeSx:

Fe 2þ þ HS  ¼ FeS þ H þ

ð22Þ

This reaction may provide protons for the cathodic reaction (Crolet, 1992). The precipitated iron sulfides form a galvanic couple with the base metal. For corrosion to occur, the iron sulfides must have electrical contact with the bare steel surface. Once contact is established, the mild steel behaves as an anode and electrons are conducted from the metal through the iron sulfide to the interface between the sulfide deposits and water, where they are used in a cathodic reaction. Surprisingly, the most notorious cases of SRB corrosion often occur in the presence of oxygen. As SRB is anaerobic microorganisms, this fact has been difficult to explain. This effect of oxygen can be explained based on a mechanism in which iron sulfides (resulting from the reaction between iron ions and sulfide and bisulfide ions) are oxidized by oxygen to elemental sulfur, which is known to be a strong corrosion agent (Lee et al., 1995). Biofilm heterogeneity plays an important role in this process, because the central parts of microcolonies are anaerobic while the outside edges remain aerobic

(Lewandowski and Beyenal, 2007). This arrangement makes this mechanism of microbial corrosion possible, because the oxidation of iron sulfides produces highly corrosive elemental sulfur, as illustrated by the following reaction:

2H2 O þ 4FeS þ 3O2 -4So þ 4FeOðOHÞ

ð23Þ

Hydrogen sulfide can also react with the oxidized iron to form ferrous sulfide and elemental sulfur (Schmitt, 1991), thereby aggravating the situation by producing even more elemental sulfur, and closing the loop through production of the reactant used in the first reaction, FeS:

3H2 S þ 2FeOðOHÞ-2FeS þ So þ 4H2 O

ð24Þ

The product of these reactions – elemental sulfur – increases the corrosion rate. Schmitt (1991) has shown that the corrosion rate caused by elemental sulfur can reach several hundred mpy (Schmitt, 1991). We have demonstrated experimentally that elemental sulfur is deposited in the biofilm during SRB corrosion (Nielsen et al., 1993), thereby detecting the component vital for this mechanism to occur. It is also well known that the sulfur disproportionation reaction that produces sulfuric acid and hydrogen sulfide is carried out by sulfur-disproportionating microorganisms (Finster et al., 1998). Also, several microbial species, such as T. thiooxidans, can oxidize elemental sulfur and sulfur compounds and produce sulfuric acid:

4S o þ 4H2 O-3H2 S þ H2 SO4

ð25Þ

In summary, the SRB corrosion of mild steel in the presence of oxygen is an acid corrosion: Anodic reaction:

Fe-Fe 2þ þ 2e

ð26Þ

2H þ þ 2e-H2

ð27Þ

Cathodic reaction:

566

Biofilms in Water and Wastewater Treatment 4.15.4.3 Biofilms on Filtration Membranes in Drinking Water Treatment

O2

2−

FeO(OH)

SO4

SO42−

SO42− H2S 0

S

FeS2 FeS

H+

O2

Fe2+ Metal

S0

HS− H2

O2

e

Figure 35 The SRB corrosion of mild steel in the presence of oxygen is an acid corrosion (Lewandowski et al., 1997).

The mechanism of SRB corrosion involves several loops, cycles in which reactants are consumed in one reaction and recycled in other reaction; the process is spontaneous at the expense of the energy released by the oxidation of the metal. This mechanism also demonstrates how the reactants and products of corrosion processes are included in the metabolic reactions of the microorganisms. For example, hydrogen, the product of the cathodic reaction above, is oxidized by some species of SRB to reduce sulfate and generate hydrogen sulfide, H2S (Cord-Ruwisch and Widdel, 1986), which is the reactant in the first reaction we referred to in this section. Hydrogen sulfide then dissociates to bisulfides:

H2 S ¼ Hþ þ HS

ð28Þ

which are then used in the reactions described above. Figure 35 shows the network of reactions described above.

4.15.4.2 Biofilms on Concrete Surfaces: Crown Corrosion of Sewers The mechanism of crown corrosion of sewers is very similar to the mechanism of MIC corrosion of metals caused by SRB. In sewers, SRB reduces sulfate ions to sulfides, which are oxidized by oxygen to elemental sulfur. Then the elemental sulfur is further oxidized, mainly by T. thiooxidans, but also by other Thiobacillus species, such as T. novellus/intermedius and T. neapolitanus, in a complex ecosystem on the sewer pipe (Vincke et al., 2001). As a result, sulfuric acid is produced, which dissolves the concrete and damages the sewers (Padival et al., 1995; Islander et al., 1991; Sand and Bock, 1984). The following reactions illustrate this action:

H2 SO4 þ CaCO3 -CaSO4 þ H2 CO3

ð29Þ

H2 SO4 þ CaðOHÞ2 -CaSO4 þ 2H2 O

ð30Þ

Crown corrosion of sewers depends on the presence of biofilm on the concrete surface and on the generation of sulfuric acid in immediate proximity to the concrete surface.

The common use of membranes in various technologies of water and wastewater treatment is probably the most visible mark of the changes that occurred in these applications in the last decade, and it is expected that filtration membranes will be even more popular in the future than they are now (Shannon et al., 2008). The traditional use of membranes in water treatment has been in the desalination of sea and brackish waters using the reverse osmosis (RO) process, and there is a large body of knowledge accumulated on this application. RO membrane filtration is becoming even more popular as the cost of desalination decreases because of various improvements in the technology that reduce the energy consumption and because of the use of new materials that produce less expensive and more robust membranes (Veerapaneni et al., 2007). Membrane processes have been introduced into other types of water treatment, besides desalination, such as water softening (Conlon et al., 1990). The main advantages of using membrane filtration in water treatment are that the process does not require using chemicals and that the membrane modules have a smaller footprint than the conventional treatment facilities. Membrane filtration can be used instead of other traditional processes in water treatment, such as coagulation, sand and activated carbon filtration, or ion exchange, without the necessity of adding chemicals to the water, which helps prevent the formation of disinfection byproducts, for example. Membrane filtration can be used alone in water treatment or in combination with other processes, in hybrid arrangements. For example, it can be used in combination with powdered activated carbon (PAC) to remove disinfection byproducts that exist in the raw water (Khan et al., 2009). Excessive biofouling of membranes is a problem in all membrane applications, but RO and nanofiltration (NF) processes are the most sensitive to biofouling (Vrouwenveldera et al., 2009). Much research has been done toward understanding the process of biofilm formation on these membranes and developing methods for cleaning the membranes. The removal of biofilm from RO membranes can be accomplished by mechanical or by chemical methods, or by a combination of mechanical and chemical methods. Mechanical methods include flushing with water or with water and air. Mechanical cleaning can be used alone or it can be followed by chemical cleaning. The simplest method of mechanical cleaning is the forward flush, in which the water flow rate above the membrane is increased to increase the shearing force and remove the deposits from the membrane. To increase the shearing force even further, air can be introduced into the conduit delivering the cleaning water. The air bubbles introduce additional instability into the flow field and increase the shearing force exerted on the surface. The backward flush is based on reversing the direction of filtration: cleaning water is filtered in the opposite direction and the particles trapped in membrane pores are removed. Depending on the contaminants deposited on the membranes, the surface can be cleaned chemically using various type of chemicals. If the deposits are predominantly inorganic scale, then the chemical cleaning can include agents that act mostly on scale, such as hydrochloric acid (HCl) or nitric acid (HNO3). If the

Biofilms in Water and Wastewater Treatment

biofilm is the main problem, then the cleaning substance may include antimicrobial agents to remove the biofilms. Two types of antimicrobial agents are in common use for this purpose: oxidizing and nonoxidizing biocides. The oxidizing biocides popular in membrane cleaning processes include chlorine, bromine, chloramine, chlorine dioxide, hydrogen peroxide, peroxyacetic acid, and ozone. Nonoxidizing biocides include formaldehyde, glutaraldehyde, and quaternary ammonium compounds. One recent study targeted cell–cell communications in biofilms to develop a novel approach in controlling membrane fouling (Yeon et al., 2009). Much effort has been directed toward the development of membranes with new or modified materials that can resist biofouling and toward modifying the surfaces of ultrafiltration (UF) and NF membranes by the graft polymerization of hydrophilic monomers that resist biofouling or allow more aggressive chemical treatment of the membranes (Hester et al., 2002; Wang et al., 2005; Asatekin et al., 2006, 2007). According to recent studies, in spiral-wound membrane modules, biofilm accumulation has a major impact on the spacer channel but the actual fouling of the membrane contributes to the overall pressure drop to a much smaller extent than previously assumed (Vrouwenveldera et al., 2009).

4.15.4.4 Biofilms on Filtration Membranes in Wastewater Treatment Membrane filtration is used in two types of wastewater technologies: (1) membrane bioreactors (MBRs) and (2) membrane biofilm reactors (MBfRs). This terminology is somewhat confusing: the names sound similar, and the fact that the obvious acronyms for the two technologies are the same does not help. It is therefore customary to call the MBRs and the MBfRs. From the biofouling point of view, microbial growth on membranes is undesirable (Le-Clech et al., 2006) while in MBfRs biofilm growth on the membrane is necessary for process performance. MBfRs are used to deliver dissolved gases, such as oxygen, hydrogen, and methane, to the microorganisms attached to the membrane (Brindle and Stephenson, 1996; Brindle et al., 1998; Suzuki et al., 2000; Lee and Rittmann, 2000; Pankhania et al., 1999; Modin et al., 2008). MBRs are used to replace gravity settling in the secondary sedimentation tanks used in traditional biological wastewater treatment; for example, the activated sludge process where membrane processes can be used to separate the biomass of suspended microorganisms from the effluent. The membranes used in MBRs are typically UF membranes. MBR technology is well established in wastewater treatment: it has been implemented on large scales (Melin et al., 2006), and textbooks have been published describing its application (Stephenson et al., 2000; Judd, 2006). Using membrane filtration to replace gravity settling has many advantages, and one of them is avoidance of the notorious problems with sludge bulking that plague many activated sludge treatment plants. Membranes in MBRs suffer from biofouling, which decreases the permeate flow (Howell et al., 2003; Young et al., 2006; Kimura et al., 2005) Large-scale operations suffer from this problem, particularly the irreversible fouling that cleaning does not remove (Wang et al., 2005). The most common solution to the excessive accumulation of biomass is bubbling air near the

567

membrane’s surface, which creates high shear and removes the biomass (Hong et al., 2002). Basic studies on biofilm formation (Davies et al., 1998) indicate that bacteria regulate their group behaviors, such as biofilm formation, in response to population density using small signal molecules called autoinducers, or quorumsensing molecules. It is expected that interference with microbial communication systems in biofilms may lead to novel approaches to preventing biofouling in many areas. Three strategies for interfering with autoinducer molecules have been proposed: blockage of autoinducer production, interference with signal receptors, and inactivation of autoinducer molecules (Rassmusen and Givskov, 2006). In a recent study, Yeon et al. (2009) demonstrated that inactivating the autoinducer molecules in a batch-type MBR reactor decreased the amount of EPS deposited on the membrane and that interfering with cell–cell communication in biofilms can alleviate the fouling of filtration membranes.

References Acuna N, Ortega-Morales BO, and Valadez-Gonzalez A (2006) Biofilm colonization dynamics and its influence on the corrosion resistance of austenitic UNSS31603 stainless steel exposed to Gulf of Mexico seawater. Marine Biotechnology 8: 62--70. Al Darbi MM, Agha K, and Islam MR (2005) Comprehensive modelling of the pitting biocorrosion of steel. Canadian Journal of Chemical Engineering 83: 872--881. Angathevar-Veluchamy RR (2006) Structure and Activity of Pseudomonas aeruginosa PAO1 Biofilms. Thesis Defense, Montana State University, Library, Bozeman, MT. Antoine RL (1976) Fixed Biological Surfaces – Wastewater Treatment. Cleveland, OH: CRC Press. Antony PJ, Chongdar S, Kumar P, and Raman R (2007) Corrosion of 2205 duplex stainless steel in chloride medium containing sulfate-reducing bacteria. Electrochimica Acta 52: 3985--3994. Asatekin A, Menniti A, Kang S, Elimelech M, Morgenroth E, and Mayes AM (2006) Antifouling nanofiltration membranes for membrane bioreactors from selfassembling graft copolymers. Journal of Membrane Science 285: 81--89. Atkinson B and Davies IJ (1974) The overall rate of substrate uptake (reaction) by microbial films. Part I: A biological rate equation. Transactions of the Institution of Chemical Engineers 52: 248--259. Bakke R, Kommedal R, and Kalvenes S (2001) Quantification of biofilm accumulation by an optical approach. Journal of Microbiological Methods 44: 13--26. Bakke R and Olsson PQ (1986) Biofilm thickness measurement by light microscopy. Journal of Microbiological Methods 5: 93--98. Beech IB, Sunner JA, and Hiraoka (2005) Microbe-surface interactions in biofouling and biocorrosion processes. International Microbiology 8: 157--168. Beech WB and Sunner J (2004) Biocorrosion: Towards understanding interactions between biofilms and metals. Current Opinion in Biotechnology 15: 181--186. Beyenal H, Tanyolac A, and Lewandowski Z (1998) Measurement of local effective diffusivity in heterogeneous biofilms. Water Science and Technology 38: 171--178. Beyenal H and Lewandowski Z (2005) Modeling mass transport and microbial activity in stratified biofilms. Chemical Engineering Science 60: 4337--4348. Beyenal H and Lewandowski Z (2000) Combined effect of substrate concentration and flow velocity on effective diffusivity in biofilms. Water Research 34(2): 528--538. Beyenal H, Sani RK, Peyton B, Dohnalkova AC, Amonette J, and Lewandowski Z (2004) Uranium immobilization by sulfate reducing biofilms. Environmental Science and Technology 38: 2067--2074. Biesterfeld S, Figueroa L, Hernandez M, and Russell P (2001) Quantification of nitrifying bacterial populations in a full-scale nitrifying trickling filter using fluorescent in situ hybridization. Water Environment Research 73: 329--338. Bilyk K, Taka´cs I, Rohrbacher J, Pitt P, Latimer R, and Dold P (2008) Full-scale testing advances fundamental understanding of denitrification filters. Proceedings of the 81st Annual Water Environment Federation Technical Exhibition and Conference (WEFTEC). Chicago, IL. Bishop PL (2003) The effect of biofilm heterogeneity on metabolic processes. In: Wuertz S, Wilderer PA, and Bishop PL (eds.) Biofilms in Wastewater Treatment. London: IWA Publishing.

568

Biofilms in Water and Wastewater Treatment

Boltz JP and Daigger GT (2010) Uncertainty in bulk-liquid hydrodynamics creates uncertainties in biofilm reactor design. Water Science and Technology 61(2): 307. Boltz JP, Johnson BR, Daigger GT, and Sandino J (2009a) Modeling integrated fixed film activated sludge (IFAS) and moving bed biofilm reactor (MBBR) systems I: Mathematical treatment and model development. Water Environment Research 81: 576--586. Boltz JP, Johnson BR, Daigger GT, Sandino J, and Elenter D (2009b) Modeling integrated fixed film activated sludge (IFAS) and moving bed biofilm reactor (MBBR) systems II: Evaluation. Water Environment Research 81: 555--575. Boltz JP and La Motta EJ (2007) The kinetics of particulate organic matter removal as a response to bioflocculation in aerobic biofilm reactors. Water Environment Research 79: 725--735. Boltz JP, Morgenroth E, and Sen D (2010a) Mathematical modelling of biofilms and biofilm reactors for engineering design. Water Science and Technology (in press). Boltz JP, Morgenroth E, deBarbadillo C, et al. (2010b) Biofilm reactor technology and design. In: Design of Municipal Wastewater Treatment Plants, WEF Manual of Practice No. 8, ASCE Manuals and Reports on Engineering Practice No. 76, 5th edn, vol. 2, ch. 13 (ISBN P/N 978-0-07-166360-1 of set 978-0-07-166358-8; MHID P/N 0-07-166360-6 of set 0-07-166358-4). New York: McGraw-Hill. Borenstein SB (1994) Microbiologically Influenced Corrosion Handbook. New York: Industrial Press. Borenstein W and Licina GJ (1994) An overview on monitoring techniques for the study of microbiologically influenced corrosion. In: CORROSION’94, Paper 611. Houston, TX, 1994. Houston, TX: NACE International. Bremer PJ and Geesey GG (1991) An evaluation of biofilm development utilizing nondestructive attenuated total reflectance Fourier transform infrared spectroscopy. Biofouling 3: 89--100. Brenner RC, Heidman JA, Opatken EJ, Petrasek AC (1984) Design Information on Rotating Biological Contactors, EPA–600/2–84–106. Washington, DC: US Environmental Protection Agency. Brindle K, Stephenson T, and Semmens MJ (1998) Nitrification and oxygen utilization in a membrane aeration bioreactor. Journal of Membrane Science 144: 197--209. Brockmann D, Rosenwinkel K-H, and Morgenroth E (2008) Practical identifiably of biokinetic parameters of a model describing two-step nitrification in biofilms. Biotechnology and Bioengineering 101(3): 497--514. Campbell SC (2003) Influence of the Distribution of the Manganese Oxidizing Bacterium Leptothrix discophora, on Ennoblement of 316L Stainless Steel. MS Thesis, Montana State University, Library, Bozeman, MT. Characklis WG (1981) Fouling biofilm development: A process analysis. Biotechnology and Bioengineering 23: 1923--1960. Cloete TE and Maluleke MR (2005) The use of the rotoscope as an online, real-time, non-destructive biofilm monitor. Water Science and Technology 52: 211--216. Conlon WJ, Hornburg CD, Watson BM, and Kiefer CA (1990) Membrane softening: The concept and its application to municipal water supply. Desalination 78: 157--175. Corbett J (1903) A Dozen years of sewage purification experiments on a large scale at Salford, England. Engineering News Record 49: 191--192. Cord-Ruwisch R and Widdel F (1986) Corroding iron as a hydrogen source for sulphate reduction in growing cultures of sulphate-reducing bacteria. Applied Microbiology and Biotechnology 25: 169--174. Costello JA (1974) Cathodic depolarization by sulfate-reducing bacteria. South African Journal of Science 70: 202--204. Crolet JL (1992) From biology and corrosion to biocorrosion. Oceanologica Acta 15: 87--94. Davies DG, Parsek MR, Pearson JP, Iglewski BH, Costerton JW, and Greenberg EP (1998) The involvement of cell-to-cell signals in the development of a bacterial biofilm. Science 280: 295--298. DeBeer D, Stoodley P, and Lewandowski Z (1994) Liquid flow in heterogeneous biofilms. Biotechnology and Bioengineering 44: 636--641. Delong EF, Taylor LT, Marsh TL, and Preston CM (1999) Visualization and enumeration of marine planktonic archaea and bacteria by using polyribonucleotide probes and fluorescent in situ hybridization. Applied and Environmental Microbiology 65: 5554--5563. Dempsey MJ, Porto I, Mustafa M, Rowan AK, Brown A, and Head IM (2006) The expanded bed biofilter: Combined nitrification, solids destruction, and removal of bacteria. Water Science and Technology 54(8): 37--46. Dickinson WH and Lewandowski Z (1996) Manganese biofouling and the corrosion behavior of stainless steel. Biofouling 10: 79--93. Dickinson WH, Lewandowski Z, and Geer RD (1996) Evidence for surface changes during ennoblement of type 316L stainless steel: Dissolved oxidant and capacitance measurements. Corrosion 52: 910--920.

Eberl HJ, Picioreanu C, Heijnen JJ, and van Loosdrecht MCM (2000) A threedimensional numerical study on the correlation of spatial structure, hydrodynamic conditions, and mass transfer and conversion on biofilms. Chemical Engineering Science 55: 6209--6222. Fan LS, Leyvaramos R, Wisecarver KD, and Zehner BJ (1990) Diffusion of phenol through a biofilm grown on activated carbon particles in a draft-tube 3-phase fluidized-bed bioreactor. Biotechnology and Bioengineering 35: 279–286. Finster K, Liesack W, and Thamdrup B (1998) Elemental sulfur and thiosulfate disproportionation by Desulfocapsa sulfoexigens sp. nov., a new anaerobic bacterium isolated from marine surface sediment. Applied and Environmental Microbiology 64: 119--125. Flemming HC (1995) Biofouling and biocorrosion-effects of undesired biofilms. Chemie Ingenieur Technik 67: 1425--1430. Geiser M, Avci R, and Lewandowski Z (2002) Microbially initiated pitting on 316 L stainless steel. International Biodeterioration and Biodegradation 49: 235--243. Grady LE, Daigger GT, and Lim H (1999) Biological Wastewater Treatment, 2nd edn. New York: Dekker. GrayMerod R, Hendrickx L, Mueller LN, Xavier JB, and Wuertz S (2005) Effect of nucleic acid stain Syto9 on nascent biofilm architecture of Acinetobacter sp BD413. Water Science and Technology 52: 195--202. Groenenboom M (2000) Increase of frictional resistance in closed conduit systems fouled with biofilms. MS Dissertation. Montana State University. Bozeman Gujer W and Boller M (1983) Operating experience with plastic media tertiary trickling filters for nitrification. In: de Emde V and Tench HB (eds.) Design and Operation of Large Treatment Plants. Oxford: Pergamon. Gujer W and Boller M (1984) Operating experience with plastic media tertiary trickling filters for nitrification. Water Science and Technology 16: 201--213. Gujer W and Boller M (1986) Design of a nitrifying trickling filter based on theoretical concepts. Water Research 20: 1353. Hamilton WA (2003) Microbially influenced corrosion as a model system for the study of metal microbe interactions: A unifying electron transfer hypothesis. Biofouling 19: 65--76. Harremo¨es P and Wilderer PA (1993) Fundamentals of nutrient removal in biofilters. In: Proceedings from the 9th Annual EWPCA–ISWA Symposium. Mu¨nchen, Germany, 11–13 May. Abwassertechnische Vereinigung e.V. Harrison JR and Daigger GT (1987) A comparison of trickling filter media. Journal WPCF 59: 679. Harrison JR and Timpany PL (1988) Design considerations with the trickling filter solids contact process. In: Proceedings of the Joint Canadian Society of Civil Engineers, ASCE National Conference on Environmental Engineering, pp. 753– 762. Vancouver, British Columbia. Hem L, Rusten B, and Ødegaard H (1994) Nitrification in a moving bed reactor. Water Research 28(6): 1425--1433. Henze M, Gujer W, Mino T, et al. (2000) Activated Sludge Model No. 2d. In: Henze M, Gujer W, Mino T, van Loosdrecht MCM (eds.) Activated Sludge Models ASM1, ASM2, ASM2d, and ASM3. Scientific and Technical Report No. 9. London: IWA Publishing. Hong SP, Bae T, Tak T, Hong S, and Randall A (2002) Fouling control in activated sludge submerged hollow fiber membrane bioreactors. Desalination 143: 219--228. Hossain MA and Das CR (2005) Kinetic and thermodynamic studies of microbial corrosion of mild steel specimen in marine environment. Journal of the Indian Chemical Society 82: 376--378. Howell JA, Arnot TC, and Liu W (2003) Membrane bioreactors for treating waste streams. Advanced Membrane Technology 984: 411--419. Islander RL, Devinny JS, Mansfeld F, Postyn A, and Shih H (1991) Microbial ecology of crown corrosion in sewers. Journal of Environmental Engineering 117: 751--770. Ito T, Nielsen JL, Okabe S, Watanabe Y, and Nielsen PH (2002) Phylogenetic identification and substrate uptake patterns of sulfate-reducing bacteria inhabiting an oxic-anoxic sewer biofilm determined by combining microautoradiography and fluorescent in situ hybridization. Applied and Environmental Microbiology 68: 356--364. Jang A, Okabe S, Watanabe Y, Kim IS, and Bishop PL (2005) Measurement of growth rate of ammonia oxidizing bacteria in partially submerged rotating biological contactor by fluorescent in situ hybridization (FISH). Journal of Environmental Engineering and Science 4: 413--420. Javaherdashti R (1999) A review of some characteristics of MIC caused by sulfatereducing bacteria: Past, present and future. Anti-Corrosion Methods and Materials 46: 173--180. Jeris JS, Owens RW, and Flood F (1981) Secondary treatment of municipal wastewater with fluidized bed technology. In: Cooper PF, Atkinson B, Cooper PF, and Atkinson

Biofilms in Water and Wastewater Treatment

B (eds.) Biological Fluidized Bed Treatment of Water and Wastewater, pp. 112--120. Chichester: Ellis Horwood for Water Research Laboratory, Stevenage Laboratory. Judd S (2006) Principles and Applications of Membrane Bioreactors for Water and Wastewater Treatment, 1st edn. Oxford: Elsevier Science. Khan MT, Lewandowski Z, Takizawa S, et al. (2009) Continuous and efficient removal of THMs from river water using MF membrane combined with high dose of PAC. Desalination 249: 713--720. Kimura K, Yamato N, Yamamura H, and Watanabe Y (2005) Membrane fouling in pilotscale membrane bioreactors (MBRs) treating municipal wastewater. Environmental Science and Technology 39: 6293--6299. Kissel JC, McCarty PL, and Street RL (1984) Numerical simulations of mixed-culture biofilms. Journal of Environmental Engineering 110: 393--411. Kreft JU, Picioreanu C, Wimpenny JWT, and van Loosdrecht MCM (2001) Individualbased modelling of biofilms. Microbiology – Society for General Microbiology 147: 2897--2912. Lazarova V and Manem J (1994) Advances in biofilm aerobic reactors ensuring effective biofilm activity control. Water Science and Technology 29(10–11): 319–327. Le-Clech P, Chen V, and Fane AG (2006) Fouling in membrane bioreactors used for wastewater treatment – a review. Journal of Membrane Science 284: 17--53. Lee AK and Newman DK (2003) Microbial iron respiration: Impacts on corrosion processes. Applied Microbiology and Biotechnology 62: 134--139. Lee KC and Rittmann BE (2000) A novel hollow-fiber membrane biofilm reactor for autohydrogenotrophic denitrificaion of drinking water. Water Science and Technology 41: 219--226. Lee W, Lewandowski Z, Nielsen PH, and Hamilton WA (1995) Role of sulfate-reducing bacteria in corrosion of mild-steel – a review. Biofouling 8: 165--194. Lewandowski Z and Beyenal H (2007) Fundamentals of Biofilm Research. Boca Raton, FL: CRC Press. Lewandowski Z, Dickinson W, and Lee W (1997) Electrochemical interactions of biofilms with metal surfaces. Water Science and Technology 36: 295--302. Lewandowski Z, Walser G, and Characklis WG (1991) Reaction kinetics in biofilms. Biotechnology and Bioengineering 38: 877--882. Licina GJ, Nekosku C, Howard RL (1992) An electrochemical method for on-line monitoring of biofilm activity in cooling water. In: CORROSION’92, Paper 177. Houston, TX: NACE. Manz W, Wendt-Potthoff K, Neu TR, Szewzyk U, and Lawrence JR (1999) Phylogenetic composition, spatial structure, and dynamics of lotic bacterial biofilms investigated by fluorescent in situ hybridization and confocal laser scanning microscopy. Microbial Ecology 37: 225--237. McQuarrie J, Dempsey MJ, Boltz JP, Johnson B (2007) The expanded bed biofilm reactor (EBBR) – an innovative biofilm approach for tertiary nitrification. In: Proceedings of the 80th Annual Water Environment Federation Technical Exposition and Conference (CD-ROM). San Diego, CA, 13–17 October. Alexandria, VA: Water Environment Federation. McNamara CJ, Lemke MJ, and Leff LG (2003) Underestimation of bacterial numbers in starvation-survival mode using the nucleic acid stain DAPI. Archiv fu¨r Hydrobiologie 157: 309--319. Melin T, Jefferson B, Bixio D, et al. (2006) Membrane bioreactor technology for wastewater treatment and reuse. Desalination 187: 271--282. Metcalf and Eddy (2003) Wastewater Engineering: Treatment and Reuse, 4th edn. New York: McGraw-Hill. Milferstedt K, Pons M-N, and Morgenroth E (2006) Optical method for long-term and large-scale monitoring of spatial biofilm development. Biotechnology and Bioengineering 94: 773--782. Modin O, Fukushi K, Nakajima F, and Yamamoto K (2008) A membrane biofilm reactor achieves aerobic methane oxidation coupled to denitrification (AME-D) with high efficiency. Water Science and Technology 58: 83--87. Morgenroth E (2003) Detachment-an often overlooked phenomenon in biofilm research and modelling. In: Wuertz S, Wilderer PA, and Bishop PL (eds.) Biofilms in Wastewater Treatment, pp. 264--290. London: IWA Publishing. Morgenroth E (2008) Modelling biofilm systems. In: Henze M, van Loosdrecht MCM, Ekama G, and Brdjanovic D (eds.) Biological Wastewater Treatment – Principles, Modelling, and Design, pp. 457--492. London: IWA Publishing. Morgenroth E, Eberl HJ, and van Loosdrecht MCM (2000) Evaluating 3-D and 1-D mathematical models for mass transport in heterogeneous biofilms. Water Science and Technology 41(4–5): 347--356. Morgenroth ET and Wilderer PA (2000) Influence of detachment mechanisms on competition in biofilms. Water Research 34(2): 417--426. Nancharaiah YV, Venuopalan VP, Wuertz S, Wilderer PA, and Hausner M (2005) Compatibility of the green fluorescent protein and a general nucleic acid stain for quantitative description of a Pseudomonas putida biofilm. Journal of Microbiological Methods 60: 179--187.

569

Newman J, Occiano V, Appleton R, et al. (2005) Confirming BAF performance for treatment of CEPT effluent on a space constrained site. In: Proceedings of the 78th Annual Water Environment Federation Technical Conference and Exposition. Washington, DC, 12–15 October. Alexandria, VA: Water Environment Federation. Nielsen PR, Lee WC, Lewandowski Z, Morrison M, and Characklis W (1993) Corrosion of mild steel in an alternating oxic and anoxic biofilm system. Biofouling 7: 267--284. Nivens DE, Chambers JQ, Anderson TR, and White DC (1993) Long-term, online monitoring of microbial biofilms using a quartz crystal microbalance. Analytical Chemistry 65: 65--69. Notermans S, Dormans JAMA, and Mead GC (1991) Contribution of surface attachment to the establishment of micro-organisms in food processing plants: A review. Biofouling 5: 21--36. Ødegaard H (2006) Innovations in wastewater treatment: The moving bed biofilm process. Water Science and Technology 53(9): 17--33. Olesen BH, Nielsen PH, and Lewandowski Z (2000) Effect of biomineralized manganese on the corrosion behavior of C1008 mild steel. Corrosion 56: 80--89. Olesen BH, Yurt N, and Lewandowski Z (2001) Effect of biomineralized manganese on pitting corrosion of type 304L stainless steel. Materials and Corrosion – Werkstoffe und Korrosion 52: 827--832. Padival NA, Weiss JS, and Arnold RG (1995) Control of Thiobacillus by means of microbial competition: Implications for corrosion of concrete sewers. Water Environment Research 67: 201--205. Pankhania M, Brindle K, and Stephenson T (1999) Membrane aeration bioreactors for wastewater treatment: Completely mixed and plug-flow operation. Chemical Engineering Journal 73: 131--136. Parker DS, Lutz M, Dahl R, and Berkkopf S (1989) Enhancing reaction rates in nitrifying trickling filters through biofilm control. Journal WPCF 61(5): 618--631. Peladan J-G, Lemmel H, and Pujol R (1996) High nitrification rate with upflow biofiltration. Water Science and Technology 34(1–2): 347--353. Pereira A, Mendes J, and Melo LF (2008) Using nanovibrations to monitor biofouling. Biotechnology and Bioengineering 99: 1407--1415. Picioreanu C (1999) Multidimensional Modeling of Biofilm Structure. PhD Thesis, Department of Biotechnology, Delft University of Technology, Delft, The Netherlands. Picioreanu C, van Loosdrecht MCM, and Heijnen JJ (1998a) A new combined differential-discrete cellular automaton approach for biofilm modeling: Application for growth in gel beads. Biotechnology and Bioengineering 57: 718--731. Picioreanu C, van Loosdrecht MCM, and Heijnen JJ (1998b) Mathematical modeling of biofilm structure with a hybrid differential-discrete cellular automaton approach. Biotechnology and Bioengineering 58: 101--116. Picologlou BF, Zelver N, and Characklis WG (1980) Biofilm growth and hydraulic performance. Journal of the Hydraulic Division, ASCE 106(HY5): 733--746. Pons M-N, Milferstedt K, and Morgenroth E (2009) Biofilm monitoring on rotating discs by image analysis. Biotechnology and Bioengineering 103: 105--116. Rasmussen K and Lewandowski Z (1998) Microelectrode measurements of local mass transport rates in heterogeneous biofilms. Biotechnology and Bioengineering 59: 302--309. Rassmusen TB and Givskov M (2006) Quorum sensing inhibitors: A bargain of effects. Microbiology 152: 895--904. Rauch W, Vanhooren H, and Vanrolleghem PA (1999) A simplified mixed-culture biofilm model. Water Research 33(9): 2148--2162. Reichert P and Wanner O (1997) Movement of solids in biofilms: Significance of liquid phase transport. Water Science and Technology 95(2–3): 289--299. Rittmann BE and Brunner CW (1984) The non-steady-state-biofilm process for advanced organics removal. Journal (Water Pollution Control Federation) 56: 874--880. Rittmann BE and Dovantzis K (1983) Dual limitation of biofilm kinetics. Water Research 17: 1727--1734. Rittmann BE and McCarty PL (1980a) Evaluation of steady-state-biofilm kinetics. Biotechnology and Bioengineering 22: 2359--2373. Rittmann BE and McCarty PL (1980b) Model of steady-state-biofilm kinetics. Biotechnology and Bioengineering 22: 2343--2357. Rogalla F and Bourbigot M-M (1990) New developments in complete nitrogen removal with innovative biological reactors. Water Science and Technology 22(1–2): 273--280. Rogalla F, Ravarini P, DeLarminat G, and Courtelle J (1990) Large scale biological nitrate and ammonia removal. Water and Environment Journal 4(4): 319--329. Sand W and Bock E (1984) Concrete corrosion in the Hamburg sewer system. Environmental Technology Letters 5: 517--528. Schmitt G (1991) Effect of elemental sulfur on corrosion in sour gas systems. Corrosion 47: 285--308. Sen D and Randall CW (2008) Improved computational model (AQUIFAS) for activated sludge, integrated fixed–film activated sludge, and moving-bed biofilm reactor

570

Biofilms in Water and Wastewater Treatment

systems. Part I: Semi-empirical model development. Water Environment Research 80: 439--453. Seth AD and Edyvean RGJ (2006) The function of sulfate-reducing bacteria in corrosion of potable water mains. International Biodeterioration and Biodegradation 58: 108--111. Shannon MA, Bohn PW, Elimelech M, Georgiadis JG, Marinas BJ, and Mayes AM (2008) Science and technology for water purification in the coming decades. Nature 452: 301--310. Shieh WK and Keenan JD (1986) Fluidized bed biofilm reactor for wastewater treatment. Advances in Biochemical Engineering Biotechnology 33: 133--169. Sin G, Van Hulle SWH, De Pauw DJW, van Griensven A, and Van rolleghem PA (2005) A critical comparison of systematic calibration protocols for activated sludge models: A SWOT analysis. Water Research 39: 2459--2474. Sin G, Weijma J, Spanjers H, and Nopens I (2008) Dynamic model development and validation for a nitrifying moving bed biofilter: Effect of temperature and influent load on the performance. Process Biochemistry 43: 384--397. Smets BF, Riefler RG, Lendenmann U, and Spain JC (1999) Kinetic analysis of simultaneous 2,4-dinitrotoluene (DNT) and 2,6-DNT biodegradation in an aerobic fluidized-bed biofilm reactor. Biotechnology and Bioengineering 63(6): 642--653. Stenstrom MK and Rosso D (2008) Aeration and mixing. In: Henze M, van Loosdrecht MCM, Ekama G, and Brdjanovic D (eds.) Biological Wastewater Treatment – Principles, Modelling, and Design. London: IWA Publishing. Stephenson T, Judd S, Jefferson B, and Brindle K (2000) Membrane Bioreactors for Wastewater Treatment. London: IWA Publishing (Alliance House). Stewart PS (2003) Diffusion in biofilms. Guest commentaires. Journal of Bacteriology 185(5): 1485--1491. Stoddart FW (1911) Nitrification and the absorption theory. Journal of the Society of Chemical Industry 30: 236--237. Suzuki Y, Hatano N, Ito S, and Ikeda H (2000) Performance of nitrogen removal and biofilm structure of porous gas permeable membrane reactor. Water Science and Technology 41: 211--217. Taka´cs I, Bye CM, Chapman K, Dold PL, Fairlamb PM, and Jones RM (2007) A biofilm model for engineering design. Water Science and Technology 55(8–9): 329--336. Takacs I, Dold P, Belia L., et al. (2010) Structured process models for nutrient removal. In: Johnson BR (ed.) Nutrient Removal, WEF Manual of Practice No. 34, ch. 11. New York: McGraw-Hill. Tschui M (1994) Submerse Festbettreaktoren. Skriptum der EAWAG (Du¨bendorf) fu¨r den Studiengang Umwelt- und Kulturingenieure an der ETH Zu¨rich, Vertieferblock, WS 94/95 [unvero¨ffentlicht]. Tuovinen OH, Button KS, Vuorinen A, Carlson L, Mair DM, and Yut LY (1980) Bacterial, chemical, and mineralogical characteristics of tubercles in distribution pipelines. Journal of American Water Works Association 72: 626--635. US EPA (1984) Corrosion manual for internal corrosion of water distribution systems. EPA/570/9-84/001. Washington, DC: Office of Drinking Water. van Loosdrecht MCM, Eikelboom D, Gjaltema A, Mulder A, Tijhuis L, and Heijnen JJ (1995) Biofilm structures. Water Science and Technology 32(8): 35--43.

Van Hulle SWH, Maertens J, De Pauw DJW, and Vanrolleghem P (2004) Using parameter sensitivity analysis of the CANON biofilm process: What to measure, where to measure and under which conditions? Water and Environmental Management Series Young Researchers 59--66. Veerapaneni S, Long B, Freeman S, and Bond R (2007) Reducing energy consumption for seawater desalination. Journal of the American Water Works Association 99: 95--106. Veluchamy RRA (2006) Structure and activity of Pseudomonas aeruginosa PAO1 biofilms. MS Dissertation. Montana State University. Bozeman. Videla HA (2001) Microbially induced corrosion: An updated overview. International Biodeterioration and Biodegradation 48: 176–201. Videla HA and Herrera LK (2005) Microbiologically influenced corrosion: Looking to the future. International Microbiology 8: 169--180. Vincke E, Boon N, and Verstraete W (2001) Analysis of the microbial communities on corroded concrete sewer pipes – a case study. Applied Microbiology and Biotechnology 57: 776--785. Vrouwenveldera JS, Graf von der Schulenburgc DA, Kruithofa JC, Johnsc ML, and van Loosdrechtb MCM (2009) Biofouling of spiral-wound nanofiltration and reverse osmosis membranes: A feed spacer problem. Water Research 43: 583--594. Wang Y, Wang T, Su Y, Peng F, Wu H, and Jiang Z (2005) Remarkable reduction of irreversible fouling and improvement of the permeation properties of poly (ether sulfone) ultrafiltration membranes by blending with pluronic F127. Langmuir 21: 11856--11862. Wanner O, Cunningham AB, and Lundman R (1995) Modeling biofilm accumulation and mass transport in a porous-medium under high substrate loading. Biotechnology and Bioengineering 47: 703–712. Wanner O, Eberl H, Morgenroth E, et al. (2006) Mathematical Modeling of Biofilms, Scientific and Technical Report No. 18. London: IWA Publishing. Wanner O and Gujer WA (1986) Multispecies biofilm model. Biotechnology and Bioengineering 28: 314--328. Wanner O and Reichert P (1996) Mathematical modeling of mixed-culture biofilms. Biotechnology and Bioengineering 49: 172--184. Williamson K and McCarty PL (1976) Verification studies of biofilm model for bacterial substrate utilization. Journal Water Pollution Control Federation 48: 281--296. Wimpenny JWT and Colasanti R (1997a) A more unifying hypothesis for biofilm structures – a reply. FEMS Microbiology Ecology 24: 185--186. Wimpenny JWT and Colasanti R (1997b) A unifying hypothesis for the structure of microbial biofilms based on cellular automaton models. FEMS Microbiology Ecology 22: 1--16. Wolfram S (1986) Theory and Application of Cellular Automata. Addison-Wesley, MA: World Scientific. Yeon K-M, Cheong W-S, Oh H-S, et al. (2009) Quorum sensing: A new biofouling control paradigm in a membrane bioreactor for advanced wastewater treatment. Environmental Science and Technology 43: 380--385. Zhang X and Bishop PL (1994) Evaluation of tortuosity factors and effective diffusivities in biofilm. Water Research 28: 2279--2287.

4.16 Membrane Biological Reactors FI Hai, University of Wollongong, Wollongong, NSW, Australia K Yamamoto, University of Tokyo, Tokyo, Japan & 2011 Elsevier B.V. All rights reserved.

4.16.1 4.16.2 4.16.2.1 4.16.2.2 4.16.3 4.16.3.1 4.16.3.2 4.16.3.2.1 4.16.3.2.2 4.16.3.2.3 4.16.3.2.4 4.16.3.2.5 4.16.3.2.6 4.16.3.3 4.16.3.4 4.16.3.4.1 4.16.3.4.2 4.16.3.4.3 4.16.3.4.4 4.16.3.4.5 4.16.4 4.16.4.1 4.16.4.2 4.16.4.3 4.16.4.4 4.16.4.4.1 4.16.4.4.2 4.16.4.4.3 4.16.4.4.4 4.16.4.4.5 4.16.4.4.6 4.16.4.4.7 4.16.4.5 4.16.4.6 4.16.4.7 4.16.4.7.1 4.16.4.7.2 4.16.4.7.3 4.16.4.7.4 4.16.5 4.16.5.1 4.16.5.1.1 4.16.5.1.2 4.16.5.1.3 4.16.5.1.4 4.16.5.2 4.16.5.3 4.16.5.4 4.16.5.4.1 4.16.5.4.2 4.16.5.4.3 4.16.5.5 4.16.5.5.1

Introduction Aeration and Extractive Membrane Biological Reactors Aeration Membrane Biological Reactor Extractive Membrane Biological Reactor History and Fundamentals of Biosolid Separation MBR Historical Development Process Comparison with Conventional Activated Sludge Process Treatment efficiency/removal capacity Sludge properties and composition Sludge production and treatment Space requirements Wastewater treatment cost Comparative energy usage Relative Advantages of MBR Factors Influencing Performance/Design Considerations Pretreatment Membrane selection and applied flux Sludge retention time Mixed liquor suspended solids concentration Oxygen transfer Worldwide Research and Development Challenges Importance of Water Reuse and the Role of MBR Worldwide Research Trend Modeling Studies on MBR Innovative Modifications to MBR Design Inclined plate MBR Integrated anoxic–aerobic MBR Jet-loop-type MBR Biofilm MBR Nanofiltration MBR Forward osmosis MBR Membrane distillation bioreactor Technology Benefits: Operators’ Perspective Technology Bottlenecks Membrane Fouling – the Achilles’ Heel of MBR Technology Fouling development Types of membrane fouling Parameters influencing MBR fouling Fouling mitigation Worldwide Commercial Application Installations Worldwide Location-specific drivers for MBR applications Plant size Development trend and the current status in different regions Decentralized MBR plants: Where and why? Commercialized MBR Formats Case-Specific Suitability of Different Formats MBR Providers Market share of the providers Design considerations Performance comparison of different providers Standardization of Design and Performance-Evaluation Method Standardization of MBR filtration systems

571 572 572 574 574 574 576 576 576 577 577 578 579 580 581 581 581 581 581 581 582 582 583 583 584 585 585 585 585 585 586 586 586 587 588 588 588 589 593 596 596 596 596 596 598 600 600 601 601 601 602 604 604

571

572

Membrane Biological Reactors

4.16.5.5.2 4.16.6 4.16.7 References

Standardization of MBR characterization methods Future Vision Conclusion

4.16.1 Introduction

Size

Membrane biological reactors combine the use of biological processes and membrane technology to treat wastewater. The use of biological treatment can be traced back to the late nineteenth century. It became a standard method of wastewater treatment by the 1930s (Rittmann, 1987). Both aerobic and anaerobic biological treatment methods have been extensively used to treat domestic and industrial wastewater (Visvanathan et al., 2000). After removal of the soluble biodegradable matter in the biological process, any biomass formed needs to be separated from the liquid stream to produce the required effluent quality. In the conventional process, a secondary settling tank is used for such solid/liquid Apprx. molecular weight 200 µm 0.001

20k

200k

0.01

0.1

Metal ions

605 605 605 605

separation and this clarification is often the limiting factor in effluent quality (Benefield and Randall, 1980). Membrane filtration, on the other hand, denotes the separation process in which a membrane acts as a barrier between two phases. In water treatment, the membrane consists of a finely porous medium facilitating the transport of water and solutes through it (Ho and Sirkar, 1992). The separation spectrum for membranes, illustrated in Figure 1, ranges from reverse osmosis (RO) and nanofiltration (NF) for the removal of solutes, to ultrafiltration (UF) and microfiltration (MF) for the removal of fine particulates. MF and UF membranes are predominantly used in conjunction with biological reactors (Pearce, 2007). UF can remove the finest particles found in water supply, with the removal rating dependent upon the

500k 1.0

10

100

Protein/enzymes Colloidal silica

Target of separation

Acid

Viruses Pesticide

Endotoxin

Herbicide

Yeast and fungi Bacteria Algae

Humic acids

Endocrine disruptors

Giardia cyst

Natural organic matter

Human hair

Crypt ospor idium

Starch

Membrane

Reverse osmosis Nano filtration Ultrafiltration Microfiltration • Reverse osmosis (RO): salts, pesticides, herbicides, metal ions, endocrine disruptors, disinfection by-products • Nanofiltration (NF): divalents salts, pesticides, herbicides, divalent metal ions • Ultrafiltration (UF): virus, bacteria, endotoxin • Microfiltration (MF): bacteria, (virus) Figure 1 The separation spectrum for membranes.

1000

Membrane Biological Reactors

• •

Easily meets regulatory levels Suitable for discharge to pristine environment

Reuse • • •

Membrane Bioreactor

Aeration (a) Oxygen phase Transfer of organics and nutrients

Membrane

Biofilm Wastewater side

(b)

Wastewater side (Biodegradable + inhibitory organics) Selective membrane

Meets standards for potable applications Increased value for industrial applications May be useful in obtaining development permit

Figure 2 Market drivers for membranes in wastewater. Information from Howell JA (2004) Future of membranes and membrane reactors in green technologies and for water reuse. Desalination 162: 1–11; and Pearce G (2007) Introduction to membranes: Filtration for water and wastewater treatment. Filtration and Separation 44(2): 24–27.

Membrane permeate

Suspended biomass

Advantages

Discharge

Wastewater

Selective transfer of degradable organics

Criteria

et al., 2000); for bubble-less aeration of the bioreactor (Brindle and Stephenson, 1996); and for extraction of priority organic pollutants from hostile industrial wastewaters (Stephenson et al., 2000). There are other forms of membrane biological reactors such as enzymatic membrane bioreactor (Charcosset, 2006) for production of drugs, vitamins, etc., or membrane biological reactors for waste-gas treatment (Reij et al., 1998), a discussion about which is beyond the scope of this chapter. Solid–liquid membrane-separation bioreactors employ UF or MF modules for the retention of biomass to be recycled into the bioreactor. Gas-permeable membranes are used to provide bubble-less oxygen mass transfer to degradative bacteria

Oxygen transfer

pore size of the active layer of the membrane. The complete pore-size range for UF is approximately 0.001–0.02 mm, with a typical removal capability of UF for water and wastewater treatment of 0.01–0.02 mm. MF typically operates at a particle size that is up to an order of magnitude coarser than this. In water treatment, the modern trend is to use a relatively tight MF with a pore size of approximately 0.04–0.1 mm, whereas wastewater normally uses a slightly more open MF with a pore size of 0.1–0.4 mm (though wastewater can be treated using UF membranes, or MF membranes used for water applications). The market drivers for membranes in wastewater are illustrated in Figure 2. However, as in any separation process, in membrane technology too, the management and disposal of concentrate is a significant issue. Environment-friendly management and disposal of the resulting concentrates at an affordable cost is a significant challenge to water and wastewater utilities and industry. To eradicate the respective disadvantages of the individual technologies, the biological process can be integrated with membrane technology. Although some recent studies have demonstrated case-specific feasibility of direct UF of raw sewage (Janssen et al., 2008), membranes by themselves are seldom used to filter untreated wastewater, since fouling prevents the establishment of steady-state conditions and because water recovery is very low (Schrader et al., 2005; Fuchs et al., 2005; Judd and Jefferson, 2003). However, membrane filtration can be efficiently used in combination with a biological process. The biological process converts dissolved organic matter into suspended biomass, reducing membrane fouling and allowing increase in recovery. On the other hand, in the membrane filtration process, the membranes introduced into the bioreactors not only replace the settling unit for solid–liquid separation but also form an absolute barrier to solids and bacteria and retain them in the process tank. As our understanding of membrane technology grows, we learn that membrane technology is now being applied to a wider range of industrial applications and is used in many new forms for wastewater treatment. Combining membrane technology with biological reactors for the treatment of municipal and industrial wastewaters has led to the development of three generic membrane processes within bioreactors (Figure 3): for separation and recycle of solids (Visvanathan

573

Biofilm Nutrient biomedium (c) Figure 3 Simplified representation of membrane biological reactors: (a) biosolid separation, (b) aeration, and (c) extractive membrane biological reactors.

574

Membrane Biological Reactors

present in the bioreactor. Additionally, the membrane can act as support for biofilm development, with direct oxygen transfer through the membrane wall in one direction and nutrient diffusion from the bulk liquid phase into the biofilm in the other direction. An extractive membrane process has been devised for the transfer of degradable organic pollutants from hostile industrial wastewaters, via a nonporous silicone membrane, to a nutrient medium for subsequent biodegradation. Biosolid separation is, however, the most widely studied process and has found full-scale applications in many countries. In a comprehensive review published in 2006, Yang et al. (2006) pointed out that the vast majority of research on membrane biological reactors since 1990 focused on biosolidseparation-type applications. There was no significant increase in the number of studies on gas diffusion and extractive membrane biological reactors over time. Publications on extractive and diffusive membrane biological reactors became available during 1994–95, after which a steady output of less than five publications a year was observed. This indicates that current research is predominantly in the water and wastewater-filtration area, in parallel with the commercial success in this field. In line with the current trend of research and commercial application, this chapter focuses on the biosolidseparation membrane biological reactors, which is more commonly known as membrane bioreactor (MBR). However, a brief outline of the other two types of membrane biological reactors is furnished in Section 4.16.2. The remainder of this chapter elaborates on the history, fundamentals, research and development challenges, as well as the commercial application of the biosolid-separation membrane biological reactors, which are henceforth referred to as MBRs.

4.16.2 Aeration and Extractive Membrane Biological Reactors 4.16.2.1 Aeration Membrane Biological Reactor Wastewater-treatment processes using high-purity oxygen have a greater volumetric degradation capacity compared to the conventional air-aeration process. However, conventional oxygenation devices have high power requirements associated with the need for high mixing rate, and cannot be used in conjunction with biofilm processes. In the membraneaeration biological reactors (MABRs), the capability of biofilm to retain high concentrations of active bacteria is coupled with the high oxygen transfer rate to the biofilm. The key characteristic advantages of MABRs are summarized as follows:

• •

High oxygen transfer rate, especially suitable for highoxygen-demanding wastewaters. In conventional aerobic biological wastewater treatment, volatile organic compounds (VOCs) can escape to the atmosphere without being biodegraded as a result of air bubbles stripping out the compounds from the bulk liquid. Since no oxygen bubbles are formed in MABRs, gas stripping of VOCs and foaming due to the presence of surfactants can be prevented (Rothemund et al., 1994; Semmens 1991; Wilderer et al., 1985) to a greater extent.

•

Membrane-attached biofilms are in intimate contact with the oxygen source, with direct interfacial transfer and utilization of oxygen within the biofilm. In contrast to conventional biofilm processes, in MABR biofilms, oxygen from the membrane and pollutant substrate(s) from the bulk liquid are transferred across the biofilm in countercurrent directions (Figure 4). Biofilm stratification in MABRs results from this distribution of the maximum oxygen and pollutant-substrate concentrations at different locations within the biofilm and the relative thickness of MABR biofilms; this enables the removal of more than one pollutant type. The high oxygen concentrations coupled with the low organic carbon concentrations near the membrane/biofilm interface encourage nitrification, an aerobic heterotrophic layer above this facilitates organic carbon oxidation, and an anoxic layer near the biofilm/ liquid interface supports denitrification (Stephenson et al., 2000).

MABRs have been used to treat a variety of wastewater types at the laboratory scale (Brindle and Stephenson, 1996). However, in line with the characteristics of MABRs discussed above, most investigations show that the process is particularly suitable for the treatment of high-oxygen-demanding wastewaters, biodegradation of VOCs, combined nitrification, denitrification, and/or organic carbon oxidation in a single biofilm. Bubble-less oxygen mass transfer can be accomplished using gas-permeable dense membranes or hydrophobic microporous membranes (Cote et al., 1988). Both plate and frame and hollow-fiber membrane configurations have been used to supply the oxygen. Oxygen diffusion through dense membrane material can be achieved at high gas pressures without bubble formation. In hydrophobic microporous membranes, the pores remain gas filled; and oxygen is transported to the shell side of the membrane through the pores by gaseous diffusion or Knudsen flow-transport mechanisms. The partial pressure of the gas is kept below the bubble point to ensure the bubble-less supply of oxygen (Ahmed and Semmens, 1992; Rothemund et al., 1994; Semmens, 1991; Semmens and Gantzer, 1993). Pressurized hollow fibers have been investigated in the dead-end and flow-through modes of operation. The evacuation of carbon dioxide from the bioreactor is a benefit of flow-through operation, though no quantitative work to determine removal rates has been undertaken (Cote et al., 1997; Kniebusch et al., 1990). Deadend operation has usually been avoided, due to significantly decreased performance and condensate formation in the lumen (Cote et al., 1997). The nonbiological fouling and loss of performance of dead-end porous hollow fibers due to iron oxidation, absorption of free oils and greases into pores, surfactants, and suspended solids, and fiber tangling have been reported (Semmens and Gantzer, 1993). Chemical treatment of the dead ends of these hollow fibers may provide a means for the condensate to escape. The liquid boundary layer normally has a greater impact upon the overall oxygen mass transfer than the membrane, with mixing of the liquid a key operational parameter (Cote et al., 1997; Kniebusch et al., 1990; Wilderer et al., 1985). However, wall thickness significantly affects the transport of

Membrane Biological Reactors

Anaerobic zone

Aerobic zone Microbial activity

Biofilm/liquid interface

Membrane/biofilm interface

575

Oxygen Carbon substrate

Biofilm

Oxygen Microbial activity

Carbon substrate

Biofilm/liquid interface

Aerobic zone

biofilm interface

Nonpermeable support/

Anaerobic zone

Biofilm Figure 4 Simplified representation of the steady-state concentration profiles of oxygen, carbon substrate, and microbial activity in case of MABR biofilm and conventional biofilm.

oxygen through dense polymer membranes (Wilderer et al., 1985). Oxygen transport is also controlled by the presence of membrane-attached biofilm and its thickness; the partial pressure of oxygen and flow-velocity characteristics on the lumen side; and the wastewater flow-velocity characteristics on the shell side of the membrane (Kniebusch et al., 1990; Pankania et al., 1994). Oxygen partial pressure provides the means for controlling the depth of oxygen penetration into the wastewater, with an increase in partial pressure resulting in an increase in the metabolic activity of the membraneattached biofilm (Rothemund et al., 1994). In bioreactors, most membranes used for oxygen mass transfer operate with the biofilm attached to the membrane surface. These biofilms are in intimate contact with the oxygen source and are protected against abrasion and grazing (Kniebusch et al., 1990; Rothemund et al., 1994). Scanning electron micrographs show that some attached bacteria inhabit the membrane pores, with the location of the oxygen and wastewater interphase very close to the bacteria (Rothemund et al., 1994). Thus, oxygen-transfer resistance due to the thickness of the porous membrane and the liquid boundary layer are not necessarily decisive limiting factors (Kniebusch et al., 1990; Rothemund et al., 1994; Wilderer et al., 1985).

Excessive biofilm accumulation can result in the transport limitation of oxygen and nutrients, plugging of membrane fibers, a decline in biomass activity, metabolite accumulation deep within the biofilm, and the channeling of flow in the bioreactor such that steady-state conditions may not be maintained (Debus and Wanner, 1992; Pankania et al., 1994; Yeh and Jenkins, 1978). To operate at maximum efficiency, occasional membrane washing, air scouring, backwashes, and high recirculation rate of wastewater to achieve high shear velocities have all been employed to control biomass accumulation. In the MABR process, oxygen is transferred without forming bubbles and therefore cannot be utilized to mix the bulk liquid. In laboratory scale MABRs, liquid-phase mixing has been achieved using recirculation pumps, impellers, magnetic stirrers, nitrogen, or air sparging.

4.16.2.2 Extractive Membrane Biological Reactor The extractive membrane biological reactor (EMBR) process enables the transfer of degradable organic pollutants from hostile industrial wastewaters, via a dense silicone membrane,

576

Membrane Biological Reactors

to a nutrient medium for subsequent degradation (Brindle and Stephenson, 1996). Membranes used for the extraction of pollutants into a bioreactor have been developed using pervaporation by exchanging the vacuum phase with a nutrient biomedium phase where biodegradation mechanisms maintain the concentration gradient needed to transfer organic pollutants present in hostile industrial wastewaters (Lipski and Cote, 1990; Nguyen and Nobe, 1987; Yun et al., 1992). The inorganic composition of the nutrient medium is unaffected by the industrial wastewater within the hydrophobic hollow-fiber membrane. Hence, the conditions within the bioreactor can be optimized to ensure high biodegradation rate (Brookes and Livingston, 1993; Livingston, 1993, 1994). The extraction and biodegradation of toxic volatile organic pollutants, such as chloroethanes, chlorobenzenes, chloroanilines, and toluene from hostile industrial wastewaters, with high salinity and extremes of pH, using EMBRs have been demonstrated at the laboratory scale (Stephenson et al., 2000). Further information on these two generic types of MBRs can be derived from the review papers by Brindle and Stephenson (1996) and McAdam and Judd (2006), and the book by Stephenson et al. (2000). Yang et al. (2006) argued that extractive or aeration MBRs present a significant opportunity for researchers as niche areas of application as these novel processes remain unexplored. Hazardous waste treatment and toxic waste cleanup present two potential areas for the EMBR (Brookes and Livingston, 1994; Dossantos and Livingston, 1995; Livingston et al., 1998), whereas hydrogenotrophic denitrification of groundwater (Clapp et al., 1999; Mo et al., 2005; Modin et al., 2008; Nuhoglu et al., 2002; Rezania et al., 2005) and gas-extractionassisted fermentation (Daubert et al., 2003; Lu et al., 1999) are potential research areas for the AMBR. It is also important to recognize the fact that these three membrane processes are not mutually exclusive and, if necessary, could be coupled into one bioreactor (Brindle and Stephenson, 1996). Once the research field has gained momentum, commercial interest might correspondingly follow.

4.16.3 History and Fundamentals of Biosolid Separation MBR 4.16.3.1 Historical Development Membranes have been finding wide application in water and wastewater treatment ever since the early 1960s when Loeb and Sourirajan invented an asymmetric cellulose acetate membrane for RO (Visvanathan et al., 2000). Many combinations of membrane solid/liquid separators in biological treatment processes have been studied since. The first descriptions of the MBR technology date from the late 1960s. The trends that led to the development of today’s MBR are depicted in Figure 5. When the need for wastewater reuse first arose, the conventional approach was to use advanced treatment processes. The progress of membrane manufacturing technology and its applications could lead to the eventual replacement of tertiary treatment steps by MF or UF (Figure 5(a)). Parallel to this development, MF or UF was used for solid/liquid separation in the biological treatment

process and thereby sedimentation step could be eliminated (Figure 5(b)). The original process was introduced by DorrOlivier Inc. and combined the use of an activated sludge bioreactor with a cross-flow membrane-filtration loop (Smith et al., 1969). By pumping the mixed liquor at a high pressure into the membrane unit, the permeate passes through the membrane and the concentrate is returned to the bioreactor (Hardt et al., 1970; Arika et al., 1966; Krauth and Staab, 1988; Muller et al., 1995). The flat-sheet membranes used in this process were polymeric and featured pore size ranging from 0.003 to 0.01 mm (Enegess et al., 2003). Although the idea of replacing the settling tank of the conventional activated sludge (CAS) process was attractive, it was difficult to justify the use of such a process because of the high cost of membranes, low economic value of the product (tertiary effluent), and the potential rapid loss of performance due to fouling. Due to the poor economics of the first-generation MBRs, apart from a few examples such as installations at the basement level of skyscrapers in Tokyo, Japan, for wastewater reuse in flushing toilets, they usually found applications only in niche areas with special needs such as isolated trailer parks or ski resorts. The breakthrough for the MBRs occurred in 1989, the process involved submerging the membranes in the reactor itself and withdrawing the treated water through the membranes (Yamamoto et al., 1989; Kayawake et al., 1991; Chiemchaisri et al., 1993; Visvanathan et al., 1997). In this development, membranes were suspended in the reactor above the air diffusers (Figure 5(c)). The diffusers provided the oxygen necessary for treatment to take place and scour the surface of the membrane to remove deposited solids. There have been other parallel attempts to save energy in membrane-coupled bioreactors. In this regard, the use of jet aeration in the bioreactor was investigated (Yamagiwa et al., 1991). The main feature of this process was that the membrane module was incorporated into the liquid recirculation line for the formation of the liquid jet such that aeration and filtration could be accomplished using only a single pump. Jet aeration works on the principle that a liquid jet, after passing through a gas layer, plunges into a liquid bath entraining a considerable amount of air. Using only one pump makes it mechanically simpler and therefore useful to small communities. The limited amount of oxygen transfer possible with this technique, however, restricts this process only to such small-scale applications. The invention of air-backwashing techniques for membrane declogging led to the development of using the membrane itself as both clarifier and air diffuser (Parameshwaran and Visvanathan, 1998). In this approach, two sets of membrane modules are submerged in the aeration tank. While the permeate was extracted through one of the sets, the other set was supplied with compressed air for backwashing. The cycle was repeated alternatively, and there was a continuous airflow into the aeration tank, which was sufficient to aerate the mixed liquor. Eventually, two broad trends have emerged in recent times, namely submerged MBRs and sidestream MBRs. Submerged technologies tend to be more cost effective for largerscale lower-strength applications, and sidestream technologies are favored for smaller-scale higher-strength applications. The sidestream MBR envelope has been extended in recent years by the development of the air-lift concept, which

Membrane Biological Reactors

577

Conventional activated sludge + MF(UF)

Prescreening Sludge withdrawal (a)

Biological tank

Settling tank

Sidestream MBR Effluent

Membrane

(b)

Submerged MBR (integrated)

(c)

Submerged MBR (separated)

(d)

Figure 5 Evolution of membrane use in conjunction with bioreactor.

bridges the gap between submerged and cross-flow sidestream MBR, and may have the potential to challenge submerged systems in larger-scale applications (Pearce, 2008b). The economic viability of the current generation of MBRs depends on the achievable permeate flux, mainly controlled by effective fouling control with modest energy input (typically r1 kW h1 m3 product). More efficient fouling-mitigation methods can be implemented only when the phenomena occurring at the membrane surface are fully understood. Detailed discussion on the technology bottlenecks and the design aspects are provided in Sections 4.16.4 and 4.16.5, respectively. It is worth noting that as the oxygen supply limits maximum mixed-liquor suspended solids (MLSSs) in aerobic MBR, anaerobic MBRs (AnMBRs) were also developed. The first test of the concept of using membrane filtration with anaerobic treatment of wastewater appears to have been reported by Grethlein (1978). The first commercially available AnMBR was developed by Dorr-Oliver in the early 1980s for high-strength whey-processing wastewater treatment. The process, however, was not applied at full scale, possibly due to high membrane costs (Sutton et al., 1983). The Ministry of International Trade and Industry (MITI), Japan, launched a 6-year research and development (R&D) project named Aqua-Renaissance ’90 in 1985 with the particular objective of developing energy-saving and smaller footprint water-

treatment processes utilizing sidestream AnMBR to produce reusable water from industrial wastewater and sewage. However, a high cross-flow velocity and frequent physicochemical cleaning was required to maintain the performance of such a high-rate MBR (Yamamoto, 2009). It was difficult to reduce the energy consumption significantly by adopting the sidestream operation using a big recirculation pump. On the other hand, commencing in 1987, a system known as anaerobic digestion ultrafiltration (ADUF) was developed in South Africa for industrial wastewater treatment (Ross et al., 1992). This process is currently in operation. Further details on AnMBRs can be derived from the comprehensive review by Liao et al. (2006). This chapter, however, focuses on aerobic MBRs.

4.16.3.2 Process Comparison with Conventional Activated Sludge Process Some important basic characteristics of CAS and MBR are compared in this section.

4.16.3.2.1 Treatment efficiency/removal capacity The MBR process involves a suspended growth-activated sludge system that utilizes microporous membranes for solid/ liquid separation in lieu of secondary clarifiers. The biological treatment in MBR is performed according to the principles

578

Membrane Biological Reactors

known from activated sludge treatment. However, higher suspended solids, biological oxygen demand (BOD), and chemical oxygen demand (COD) removals in MBR have been reported throughout the literature. With CAS, the colloidal fraction (that represents about 20% of the organic content of wastewater) has a residence time (hydraulic residence time (HRT)) in the range of few hours while with MBR, due to total SS retention, the residence time of this fraction (sludge retention time (SRT)) is in the range of several days. Thus, the biodegradation for this fraction is higher in MBR than in CAS. Some soluble compounds too, after being adsorbed on SS, can be retained in MBR and can be biodegraded to a better extent. Thus, some studies have ascribed the better removal of soluble COD in MBR to the fact that the effluent is particle free (Cote et al., 1997; Engelhardt et al., 1998; De Wilde et al., 2003). MBR produces quality effluent suitable for reuse applications or as a high-quality feedwater source for RO treatment. Indicative output quality includes suspended solids o1 mg l1, turbidity o0.2 nephelometric turbidity unit (NTU), and up to 4 log removal of virus (depending on the membrane nominal pore size). In addition, it provides a barrier to certain chlorine-resistant pathogens such as Cryptosporidium and Giardia. In comparison to the CAS process, which typically achieves 95%, COD removal can be increased to 96–99% in MBRs (Stephenson et al., 2000). Nutrient removal is one of the main concerns in modern wastewater treatment especially in areas that are sensitive to eutrophication. As in the CAS, currently, the most widely applied technology for N removal from municipal wastewater is nitrification combined with denitrification. Total nitrogen removal through the inclusion of an anoxic zone is possible in MBR systems. Besides phosphorus precipitation, enhanced biological phosphorus removal (EBPR) can be implemented, which requires an additional anaerobic process step. Some characteristics of MBR technology render EBPR in combination with post-denitrification as an attractive alternative that achieves very low nutrient effluent concentrations (Drews et al., 2005b).

4.16.3.2.2 Sludge properties and composition The presence of a membrane for sludge separation has many consequences. This influences the rheological properties and composition of the sludge. Defrance et al. (2000) observed in a sidestream MBR with high cross-flow velocity that MBR sludge was less viscous than conventional sludge. The same was observed by Rosenberger et al. (2002). Furthermore, with increasing shear rate, viscosity of the sludge decreases (Rosenberger et al., 2002), although in some cases, the activated sludge behaves as a Newtonian fluid (Xing et al., 2001). Defrance and Jaffrin (1999) found out that filtering-activated sludge from an MBR resulted in fouling that could be totally, physically removed, whereas filtration of CAS led to physically irremovable fouling. It is quite difficult to generalize information about sludge composition from different installations, since each installation promotes different types of activated sludge. This has its effect on the microbial community that can be found in an activated sludge system. Nevertheless, it is obvious that the presence of the membrane in an MBR system influences the biomass composition. Since no suspended solids are washed

out with the effluent, the only sink is surplus sludge. From a secondary clarifier, lighter species are washed out, whereas in an MBR they are retained in the system by the membrane. Furthermore, changes in SRT and higher MLSS concentrations might lead to changes in the microbial community. Microbialcommunity analyses have revealed significant differences between CAS system and an MBR and a higher fraction of bacteria was found in the nongrowing state in the MBR (Witzig et al., 2002; Wagner and Rosenwinkel, 2000).

4.16.3.2.3 Sludge production and treatment Small-scale laboratory studies revealed a great advantage of MBRs, that is, lower or even zero excess sludge production, caused by low loading rates and high SRTs (Benitez et al., 1995). When longer SRTs are applied, sludge production, of course, decreases in the MBR (Wagner and Rosenwinkel, 2000). However, the amount of excess secondary sludge produced in larger MBR installations operated under the practical range of SRTs is somewhat lower than or even equal to that in conventional systems (Gu¨nder and Krauth, 2000). Table 1 provides a general comparison of the sludge-production rates from different treatment processes. It should be noted that the primary sludge production in the case of the MBR is lower. The suited pretreatment for the MBR is grids and/or sieves, and in an average, screened water was observed to contain 30% more solids than settled water (Jimenez et al., 2010). MBR sludge treatment is almost the same compared to CAS systems. The dewaterability of waste-activated sludge from the MBR seems to pose no additional problem, compared to aerobic stabilized waste sludge from CAS systems (Kraume and Bracklow, 2003).

4.16.3.2.4 Space requirements One of the advantages of the MBR is its compactness, because large sedimentation tanks are not needed. An interesting parameter in this respect is the surface-overflow rates for the two systems. The overflow rate of a secondary clarifier is defined as the ratio of its flow and footprint, that is, the volume of water that can be treated per square meter of tank. In practice, values around 22 m d1 are used. For an MBR filtration tank, an overflow rate can also be estimated from the permeate flux and the membrane-packing density within the

Table 1

Sludge production in case of different treatment processes

Treatment process

Sludge production kg (kg BOD)1

Submerged MBR Structured media biological aerated filter Trickling filter Conventional activated sludge Granular media BAF

0.0–0.3 0.15–0.25 0.3–0.5 0.6 0.63–1.06

Data from Stephenson T, Judd S, Jeferson B, and Brindle K (2000) Membrane Bioreactors for Wastewater Treatment. London: IWA. Gander MA, Jefferson B, and Judd SJ (2000) Membrane bioreactors for use in small wastewater treatment plants: Membrane materials and effluent quality. Water Science and Technology 41: 205–211, and Metcalf and Eddy, Inc. (2003) Wastewater Engineering – Treatment and Reuse, 4th edn. New York: McGraw-Hill.

Membrane Biological Reactors

tank. Following this method, Evenblij et al. (2005a) showed that with an average permeate flux of 15 l m2 h1, the overflow rates of the membrane tanks are in the range 25–62 m d1 which is up to 3 times higher than the overflow rate of a conventional secondary clarifier. Compared to an average overflow rate of 22 m d1 with a secondary clarifier, the space consumption for sludge-water separation in an MBR is 10–60% lower when flux is 15 l m2 h1 and 50–80% lower when flux is 25 l m2 h1. A further reduction in footprint is caused by the higher MLSS concentration that can be applied in an MBR. This estimate however did not take into account backflushing or relaxation periods, which reduce the overflow rate. Nevertheless, full-scale MBR plants also manifest these space-saving characteristics. For instance, Brescia WWTP, in Italy, which is the world’s largest MBR retrofit of an existing conventional plant, gives a full-scale example of a ratio of 2 when comparing area needed by CAS and MBR (Brepols et al., 2008).

4.16.3.2.5 Wastewater treatment cost

Relative cost (1994 cost equals 1)

The high cost connected with MBR is often mentioned in discussions about applicability of MBR. However, it is not easy to make a general economical comparison between MBR and CAS systems. First of all, the reference system should not simply be an activated sludge system, but a system that produces an effluent of the same quality. Moreover, an MBR is a modular system, that is, easily expandable, which is often mentioned as an advantage of the system. However, this makes the system less competitive with conventional systems, since these become relatively less expensive per population equivalent (p.e.) at larger scale. It should be noted that although the equipment and energy costs of an MBR are higher than systems used in conventional treatment, total water costs can be competitive due to the lower footprint and installation costs (Pearce, 2008b; Lesjean et al., 2004; Cote et al., 2004; De Wilde et al., 2003). MBR costs have declined sharply since the early 1990s, falling typically by a factor of 10 in 15 years. As MBR technology has become accepted, and the scale of installations has increased, there has been a steady downward trend in membrane prices (Figure 6), which is still continuing. This is particularly notable with the acceptance of the MBRs in the municipal sector. The uptake of membrane technology for municipal applications has had the affect of

1.0 Membrane cost (per unit flow rate) 0.8

579

downward pressure on price. A detailed holistic cost comparison may reveal reasonably comparable results between the cost of the MBR option versus other advanced treatment options, especially if land value is considered. Studies show that depending on the design and site-specific factors the total water cost associated with MBR may be less or higher than the CAS-UF/MF option. For example, a cost comparison by the US consultant HDR in 2007 showed that MBR was 15% more expensive on a 15 million liters a day (MLD) case study, whereas a study by Zenon in 2003 gave MBR 5% lower costs (Pearce, 2008a). The differences were due to the design fluxes assumed and the capital charge rate for the project. Neither study allocated a cost advantage from the reduced footprint, which could typically translate to a treated water cost saving of up to 5%. It is interesting to evaluate the development in cost estimates over the past several years. Davies et al. (1998) made a cost comparison for two wastewater treatment plants (WWTPs), with capacities of 2350 and 37 500 p.e. With the assumptions they made (e.g., a membrane lifetime of 7 years) they conclude that depending on the design capacity (i.e., 2 times DWF to be treated) MBR is competitive with conventional treatment up to a treatment capacity of 12 000 m3 d1 (Table 2). Engelhardt et al. (1998) after carrying out pilot experiments also made a cost calculation for an MBR with a capacity of 3000 p.e., designed for nitrification/denitrification and treatment of 2*DWF. Investment costs were estimated at h3104 000 (including pretreatment) and operational cost at h194 000 yr1. Adham et al. (2001) made a cost comparison between MBR oxidation ditch followed by membrane filtration and CAS followed by membrane filtration. They concluded that MBR is competitive with the other treatment systems (Table 3). Chang et al. (2001) report experiments with low-cost membranes. The effect of membrane cost on the investment cost is considerable, but operational problems hinder further application of low-cost membranes. A drawback of the applied membranes is its limited disinfecting capacity. Van Der Roest et al. (2002a) described a cost comparison between an MBR installation and a CAS system with tertiary sand filtration. The calculations were carried out for two new WWTPs with the aim of producing effluent with low

Table 2 Capital and operating cost ratios of MBR and conventional activated sludge (CAS) process assuming a capacity of 2*(dry weather flow) Parameter

Cost ratio (MBR:ASP)

Capital cost 2350 p.e 37 500 p.e Operating costs per year

0.63 2.00

2350 p.e 37 500 p.e

1.34 2.27

0.6 0.4 0.2 0.0 (1994)

(1995)

(1997)

(1999)

(2000)

Figure 6 Sharp cost decline of membranes for MBR (cost of Zenon membranes as an example).

Data from Davies WJ, Le MS, and Heath CR (1998) Intensified activated sludge process with submerged membrane microfiltration. Water Science and Technology 38(5): 421–428.

580

Membrane Biological Reactors

concentrations of nitrogen and phosphorus. Almost the same investment costs and 10–20% higher operating costs, depending on the capacity of the plant, for MBR were estimated (Table 4). Cost differences between an MBR and a traditional WWTP concerning manpower, chemical consumption, and sludge treatment were noted to be minimal. WERF (2001) summarized operating and water-quality data obtained over 1 year from two MBR pilot plants located at the Aqua 2000 Research Center at the City of San Diego (California) North City Plant. Preliminary cost estimates of the MBR technology were also developed. MBRs demonstrated that their effluent was suitable to be fed directly into an RO process from a particulate standpoint with silt density index (SDI) values averaging well below 3. The MBR effluent water quality was superior to the quality of a full-scale tertiary conventional WWTP. The preliminary cost estimate in this report was performed for a 1 million gallons a day (mgd) scalping facility (WWTP drawing a designated amount of flow from the sewer system; excess sewage flow is treated at another plant located at the end of the sewer line). This facility produced an effluent suitable as feedwater for an RO process. Based upon this estimate, the present value was estimated as $0.81 m3, $0.96 m3, and $1.16 m3 for the MBR process, oxidation ditch with MF, and oxidation ditch with conventional tertiary lime pre-treatment, respectively. Therefore, the MBR process was reported as the most cost-effective alternative for water reclamation where demineralization or indirect drinking water-production (RO) is required. McInnis (2005) reported a detailed comparative cost analysis of two membrane-based tertiary treatment options: (1)

Table 3 Capital and total cost ratios of MBR and tertiary MF following alternative biological processes Alternatives

Oxidation ditch-MF CAS-MF

Cost ratio (MBR:alternative) Capital

Total per year

0.91 0.85

0.89 0.9

Data from Adham S, Mirlo R, and Gagliardo P (2000) Membrane bioreactors for water reclamation – phase II. Desalination Research and Development Program Report No. 60, Project No. 98-FC-81-0031. Denver, CO: US Department of the Interior, Bureau of Reclamation, Denver Office.

MBR and, (2) CAS process followed by MF (CAS/MF). According to that study, irrespective of design flow rate, the MBR entails slightly higher unit capital costs as compared to CAS/ MF process, while, depending on the design flow rate, the operation and maintenance costs (O&M) of the former are higher than or comparable to that of the latter. Comparative O&M cost breakdown revealed that MBR entails less labor cost, considerably higher power and chemical consumptions and slightly higher membrane cost, other costs remaining virtually the same. In the CAS/MF process, labor cost induces the highest cost, while in case of the MBR process, labor and electrical power-consumption costs are almost similar. Overall, the MBR imposes slightly higher capital and operating/ maintenance cost over that of CAS/MF. Cote et al. (2004) explored two membrane-based options available to treat sewage for water reuse, tertiary filtration (TF) of the effluent from a CAS process, and an integrated MBR. These options were compared from the point of view of technical performance and cost using ZeeWeed immersed membranes. The analysis showed that an integrated MBR is less expensive than the CAS-TF option. The total life cycle costs for the treatment of sewage to a quality suitable for irrigation reuse or for feeding RO decreased from 0.40$ m3 to 0.20$ m3 as plant size increased to 75 000 m3 d1. It was also shown that the incremental life-cycle cost to treat sewage to indirect potable water-reuse standards (i.e., by UF and RO) was only 39% of the cost of seawater desalination. A recent market research report (BCC Research, 2008) estimated the capital cost of a 50 000 gallons per day (gpd) (190 m3 d1) plant at US$350 000, a 100 000 gpd plant at US$500 000, and a 500 000 gpd plant at US$2 million. For systems of 1 mgd (million gallons per day) and larger, capital costs start at US$3.5 million (Table 5). The largest percentage of new system installations, 93%, continue to fall into the 5000–500 000 gpd range (most of those, about 57% of them, have capacities of less than 25 000 gpd), 2% of installations range from 0.5–1 mgd, and 5% of them are larger than 1 mgd. Tables 2–5 list cost values reported during the period 1998–2008. Obviously, the data from the initial stage of the MBR development holds little relevance today. However, these are listed here to provide a general trend of cost-data evolution.

4.16.3.2.6 Comparative energy usage Table 4 Capital and total cost ratios of MBR and tertiary sand filtration following CAS Parameter

Cost ratio

Capital cost 10 000 p.e 50 000 p.e Operating costs per year

0.92 1.01

10 000 p.e 50 000 p.e

1.09 1.21

Data from Van Der Roest HF, Lawrence DP, and Van Bentem AG (2002a) Membrane Bioreactors for Municipal Wastewater Treatment (Water and Wastewater Practitioner Series: Stowa Report). London: IWA.

MBR provides an equivalent treatment level to CAS-UF/MF, but at the expense of higher energy cost since the efficiency of air usage in MBR is relatively low. The MBR process uses more Table 5

Capital cost of MBR depending on plant sizea

Plant size, gpd  103

Capital cost, US$  103

50 100 500 1000

350 500 2000 3500

a

1 m3 d1 ¼ 264.17 gpd. Data from BCC Research (2008) Membrane bioreactors: Global markets. Report Code MST047B, Report Category – Membranes & Separation Technology.

Membrane Biological Reactors

air, and hence higher energy than conventional treatment. This is because aeration is required for both the biological process and the membrane cleaning, and the type, volume, and location of air required for the two processes are not matched. Biotreatment utilizes fine air bubbles, since oxygen needs to be absorbed for the biological reaction step. In contrast, fouling control is best achieved by larger bubbles, since the air is required to scour the membrane surface or shake the membrane to remove the foulant. Accordingly, although the concept of MBR was first developed to exploit the fact that the biological wastewater-treatment process and the process of membrane-fouling control can both use aeration (Pearce, 2008b), the potential for dual-purpose aeration is strictly limited. Based on a survey of conventional wastewater-treatment facilities in the US, Metcalfe and Eddy, Inc. (2003) reported that the energy usage range was 0.32–0.66 kW h1 m3. Energy usage in wastewater treatment is somewhat lower in Europe, partly due to a greater consciousness for energy efficiency, and partly due to the fact that average BOD loading/ capita in the US is 20–25% greater than that in Europe (due to the use of kitchen disposal units). Long-term monitoring of wastewater-treatment systems has shown usages as low as 0.15 kW h1 m3 for activated sludge, increasing to 0.25 kW h1 m3 if a biological aerated filter (BAF) stage is included (Pearce, 2008a). Membrane filtration after conventional treatment is estimated to add 0.1–0.2 kW h1 m3 to the energy, equivalent to a total energy use for CAS-UF/MF of 0.35–0.5 kW h1 m3 in a new facility (Lesjean et al., 2004). Experience in large-scale commercial MBRs shows an energy usage of around 1.0 kW h1 m3, although smaller-scale facilities typically operate at 1.2–1.5 kW h1 m3 or higher (Judd, 2006). However, in comparison to these values, energy consumption of around 1.9 kW h1 m3 was reported in 2003 (Zhang et al., 2003) and up to 2.5 kW h1 m3 in 1999 (Ueda and Hata, 1999). This proves that there is a gradual improvement in MBR design (Figure 7). Further improvements in air efficiency and membrane-packing density are expected

Energy consumption, kW hr−1 m−3

3.0

2.0

1.0

to improve the current values in the future. Even so, it seems likely that MBR energy costs will continue to exceed those of CAS-UF/MF by 0.4 kW h1 m3 or more (Pearce, 2008a). However, the fact that membrane filtration after conventional treatment is estimated to add only 0.1–0.2 kW h1 m3 to the energy points out that the higher energy consumption of MBR over CAS-UF/MF is due to the difference in consumption in the respective biological processes. MBRs are generally operated at quite low F/M ratios (less than 0.2), or high MLSS concentrations, and this is one of the reasons for the excellent biodegradation efficiency, and high aeration cost as well. CAS plants, on the other hand, are operated at higher F/M ratios, implying lower oxygen need for biodegradation. Table 6 lists typical energy-use rates of different biologicalbased treatment combinations. Section 4.16.5 provides further information on energy comparison of the MBR formats.

4.16.3.3 Relative Advantages of MBR There are several advantages associated with the MBR technology, which make it a valuable alternative over other treatment techniques. The combination of activated sludge with membrane separation in the MBR results in efficiencies of footprint, effluent quality, and residual production that cannot be attained when these same processes are operated in sequence. The MBR system is particularly attractive when applied in situations where long biological solid-retention times are necessary and physical retention and subsequent hydrolysis are critical to achieving biological degradation of pollutants (Chen et al., 2003). The prime advantages of MBR are the treated water quality, the small footprint of the plant, less sludge production, and flexibility of operation (Visvanathan et al., 2000). First of all, the retention of all suspended matter and most of the soluble compounds within the bioreactor leads to excellent effluent quality capable of meeting stringent discharge requirements and paving the way for direct water reuse. The possibility of retaining all bacteria and viruses results in a sterile effluent, eliminating extensive disinfection and the corresponding hazards related to disinfection by-products. As the entire process equipment can be made airtight, odor dispersion can be prevented quite successfully. Since suspended solids are not lost in the clarification step, total separation and control of the SRT and hydraulic retention time (HRT) are possible enabling optimum control of the microbial population and flexibility in operation. The absence of a clarifier, which also acts as a natural selector for settling organisms, enables sensitive, slow-growing

0.0 1999

2003 Year

2006

Figure 7 Gradual reduction in reported values of energy consumption by MBR. Data from Ueda T and Hata K (1999) Domestic wastewater treatment by a submerged membrane bioreactor with gravitational filtration. Water Research 33: 2888–2892; Zhang SY, Van Houten R, Eikelboom DH, et al. (2003) Sewage treatment by a low energy membrane bioreactor. Bioresource Technology 90: 185–192; and Judd S (ed.) (2006) The MBR Book: Principles & Applications of MBRs in Water & Wastewater Treatment. Oxford: Elsevier.

581

Table 6 Comparative typical energy consumption by different treatment options Treatment option

Energy use (kW h1 m3)

CAS CAS-BAF CAS-MF/UF MBR

0.15 0.25 0.35–0.5 0.75–1.5a

a

Power consumption range for large- to smaller-scale plants.

582

Membrane Biological Reactors

species (nitrifying bacteria, bacteria capable of degrading complex compounds) to develop and persist in the system (Cicek et al., 2001; Rosenberger et al., 2002). The membrane not only retains the entire biomass but also prevents the escape of exocellular enzymes and soluble oxidants creating a more active biological mixture capable of degrading a wider range of carbon sources (Cicek et al., 1999b). MBRs eliminate process difficulties and problems associated with settling, which is usually the most troublesome part of wastewater treatment. The potential for operating the MBR at very high SRTs without the obstacle of settling allows high biomass concentrations in the bioreactor. Consequently, higher-strength wastewater can be treated and lower biomass yields are realized (Muller et al., 1995). This also results in more compact systems than conventional processes, significantly reducing plant footprint and making it useful in waterrecycling applications (Konopka et al., 1996). The low sludge load in terms of BOD forces the bacteria to mineralize poorly degradable organic compounds. The higher biomass loading also increases shock tolerance, which is particularly important where feed is highly variable (Xing et al., 2000). The increased endogenous (autolytic) metabolism of the biomass (Liu and Tay, 2001) under long SRT allows development of predatory and grazing communities, with the accompanying trophiclevel energy losses (Ghyoot and Verstraete, 1999). These factors, in addition to resulting in lower overall sludge production, lead to higher mineralization efficiency than those of a CAS process. High molecular weight soluble compounds, which are not readily biodegradable in conventional systems, are retained in the MBR (Cicek et al., 2002). Thus, their residence time is prolonged and the possibility of oxidation is improved. The system is also able to handle fluctuations in nutrient concentrations due to extensive biological acclimation and retention of decaying biomass (Cicek et al., 1999a).

4.16.3.4 Factors Influencing Performance/Design Considerations This section sheds light on some important design considerations of MBR. More detailed information on some of these parameters is provided in Section 4.16.4.7, in relation to membrane fouling.

4.16.3.4.1 Pretreatment All MBRs require pretreatment, for example, screening and grit removal, to protect the membranes. Screening has historically been limited to 3 mm; however, hair and fiber can still pass through this size of the screen and become embedded or wrapped around the hollow fibers. The MBR providers have standardized their screen selections to a 2-mm traveling band, punched screen. Conversely, the flat-sheet membranes experience less problems with hair and fiber, and are standardized to a 3-mm screen. Further discussion regarding mechanical pretreatment is provided in Section 4.16.4.6.

4.16.3.4.2 Membrane selection and applied flux An MBR membrane needs to be mechanically robust, chemically resistant to high Cl2 concentrations used in cleaning, and nonbiodegradable (Pearce, 2008a). Clean-water permeability

is not as important in an MBR as in membrane-filtration applications, since the membrane transport properties are strongly influenced by the accumulation of foulant particles at the membrane surface. However, process flux in treating a wastewater feed is important since it directly affects capital cost, due to its effect on membrane area and footprint, and operating costs due to the effect of membrane area on chemical and air use. Most MBRs operate at an average flux rate between 12.5 and 25 l m2 h1, with Mitsubishi’s unit operating in the lower range. The key flux rates that determine the number of membranes required are associated with the peak flow rates. For plants with peaking factors of less than two, an MBR can handle the plant flow variation without having a significantly impact on the average design flux rate. Otherwise, equalization needs to be provided with either a separate tank at the head of the facility or within the aeration basin, allowing sidewater depth variations during peak flow.

4.16.3.4.3 Sludge retention time In the past, most MBR systems were designed with extremely long SRTs, of the order of 30–70 days, and very few were operated at less than about 20 days. Two reasons prompted such practice: (1) the drive to minimize sludge production or eliminate it all together and (2) the concern over the reduced flux resulting from short SRT operation, presumably due to the fouling effect of extracellular excretions from younger sludge. Currently, the selection of SRT is based more on the treatment requirements, and SRTs as low as 8–10 days can now be contemplated.

4.16.3.4.4 Mixed liquor suspended solids concentration From the point of view of bioreactor volume reduction and minimization of excess sludge, submerged MBR systems have been typically operated with MLSS concentrations of more than 12 000 mg l1, and often in the range of 20 000 mg l1. Hence, they offer greater flexibility in the selection of the design SRT. However, excessively high MLSS may render the aeration system ineffective and reduce membrane flux. A trade-off, therefore, comes into play. Current design practice is to assume the MLSS to be closer to 10 000 mg l1 to ensure adequate oxygen transfer and to allow for higher membrane flux. With larger systems, it is more cost effective to reduce the design MLSS because of the high relative cost of membranes when compared to the cost of additional tank volume.

4.16.3.4.5 Oxygen transfer At high MLSS concentrations, the demand for oxygen can be significant. In some cases, the demand can exceed the volumetric capacity of typical oxygenation systems. The oxygentransfer capacity of the aeration system must also be carefully analyzed. Submerged membranes are typically provided with shallow coarse bubble air to agitate the membranes as a means to control fouling. Such aeration provides some oxygenation, but at low efficiency. In compact systems, fine bubble aeration may be placed at greater depth below the membrane aeration; however, the combined efficiency and the bubblecoalescing effects require further consideration during design (Visvanathan et al., 2000).

Membrane Biological Reactors

The lower operating cost obtained with the submerged configuration along with the steady decrease in the membrane cost encouraged an exponential increase in MBR plant installations from the mid-1990s onward. Since then, further improvements in the MBR design and operation have been introduced and incorporated into larger plants. The key steps in the recent MBR development are summarized below:

• •

•

The acceptance of modest fluxes (25% or less of those in the first generation), and the idea of using two-phase bubbly flow to control fouling. While early MBRs were operated at SRTs as high as 100 days with MLSS up to 30 g l1, the recent trend is to apply a lower SRT (around 10–20 days), resulting in more manageable MLSS levels (10–15 g l1). Thanks to these new operating conditions, the fouling propensity in the MBR has tended to decrease and overall maintenance has been simplified, as less-frequent membrane cleaning is necessary.

Further discussion on these aspects is provided in the following sections.

4.16.4 Worldwide Research and Development Challenges 4.16.4.1 Importance of Water Reuse and the Role of MBR The need for pure water is a problem of global proportions. In the Earth’s hydrologic cycle, freshwater supplies are fixed and constant, while global water demand is growing (Howell, 2004; Bixio et al., 2006). With each passing year, the quality of the planet’s water measurably deteriorates, presenting challenges for the major users: the municipal, industrial, and environmental sectors. Increasing demand for water, and drought and water scarcity are now common issues facing many urban and rural communities around the world (Howell, 2004; Tadkaew et al., 2007; Jimenez and Asano, 2008). Water treatment has, therfore, become an area of global concern as individuals, communities, industries, countries, and their national institutions strive for ways to keep this essential resource available and suitable for use. Water recycling is a pragmatic and sustainable approach for many countries to mitigate or solve the problems of water supply. There is a growing interest in using nontraditional water resources by means of water reclamation and water recycling for long-term sustainability. It can be divided into two categories, internal domestic or industrial recycling and external recycling, where high-quality reclaimed water from a sewage treatment plant is used for aquifer recharge or irrigation. With the current focus on water-reuse projects and the role they play in the water cycle, the search for cost-competitive advanced wastewater-treatment technologies has never before been so important. Treatment technology for water recycling encompasses a vast number of options. A general paucity of legislative and socioeconomic information has led to the development of a diverse range of technical solutions (Jefferson et al., 2000). Membrane processes are regarded as key elements of advanced wastewater reclamation and reuse schemes and are included in a number of prominent schemes worldwide, for example, for artificial groundwater recharge, indirect

583

potable reuse, as well as for industrial-process water production (Melin et al., 2006; Bixio et al., 2008). Among the many treatment alternatives, MBRs, which combine membrane filtration and biological process for wastewater treatment, are seen to have an effective technology capable of transforming various types of wastewater into high-quality effluent exceeding most discharge requirements and suitable for a variety of nonpotable water-reuse applications such as flushing toilets and for irrigation (Tadkaew et al., 2007; Jimenez and Asano, 2008). In some cases, treated water can be applied to recharge groundwater to halt saltwater intrusion into coastal aquifers, abate subsidence in areas sinking due to overpumping groundwater, and support aquifer storage and recovery. Issues of water quality, water quantity, and aging/nonexistent infrastructure propel the market for MBRs. Escalating water costs due to dwindling supplies for communities and businesses also drive the growing acceptance of MBRs. Anticipated stricter environmental regulations are driving sales of MBRs to industry, municipalities, and are prompting maritime users to consider MBR technology (Jefferson et al., 2000; Jimenez and Asano, 2008). This is probably due to the effectively disinfected high-quality effluent and high performance in trace organic removal for safe and environmentally benign discharge that MBRs can offer. In practical terms, the process has many benefits, which make it suitable for the size of the systems applicable to recycling. The ability to run independently of load variation and produce no sludge are critical and highlight MBRs as possibly the most viable small-footprint, high-treatment option for water recycling (Jefferson et al., 2000; Melin et al., 2006; Tadkaew et al., 2007). Comparison with other technologies used for water recycling reveals that MBRs not only produce lower residual concentrations but do so more robustly than the alternatives (Jefferson et al., 2000; Melin et al., 2006). The favorable microbiological quality of the effluent of MBRs is a major factor in their frequent selection for water reuse, even if full disinfection cannot be expected, particularly considering the distribution and storage components of a full-scale system, which can be prone to regrowth of microorganism and contamination from various sources. However, the MBR effluent is adequate for many water-reuse applications with little residual chlorine disinfection for subsequent distribution. The MBR then does provide a dual layer of protection against pathogen breakthrough, greatly lowering the risk during operation. MBRS have the greatest efficacy toward water recycling, albeit contingent upon a loading rate constrained by the operable flux. Not only do they comply with all likely waterquality criteria for domestic recycling but they also produce a product that is visibly clear and pathogen free, both of which are likely to be key concerns in terms of public acceptability. There are some issues that still need to be addressed and these are highlighted throughout Sections 96.4.6 and 96.4.7 of this chapter.

4.16.4.2 Worldwide Research Trend Early development efforts in MBR technology were concentrated in UK, France, Japan, and South Korea, whereas extensive research in China and Germany began after 2000. Much

584

Membrane Biological Reactors

of the research in the newcomer countries is building on pioneering work from the UK, France, Japan, and South Korea. Three stages may be identified in the worldwide MBR research: 1. An entry-level stage spanning from 1966 to 1980, during which lab-scale research was mainly conducted. Membranes of that period had low flux and short life span due to undeveloped membrane-manufacturing technology. 2. A slow-to-moderate growth period from 1980 to 1995, when MBR technology was well investigated especially in Japan, Canada, and the USA. During this stage, new membrane-material development, MBR configuration design, and MBR operation were critically studied. The submerged MBR concept was put forward by Japanese researchers in 1989. 3. The rapid development stage started in 1995 and continues even now, when MBR technology underwent a rapid development prompted by deep understanding of the technology in research communities and by the installation of full-scale MBRs. Much of the published information on MBRs to date has mainly focused on bench or pilot-scale studies, performance results of treating a specific type of wastewater, and short-term operations. Regardless of the source of wastewater, whether it is municipal or industrial, very few publications involved fullscale studies for long-term operational periods. In a comprehensive review, Yang et al. (2006) grouped the available worldwide publications regarding MBR into six main research areas: (1) literature and critical reviews; (2) fundamental aspect; (3) municipal and domestic wastewater treatment; (4) industrial wastewater and landfill leachate treatment; (5) drinking-water treatment; and (6) others, which include gas removal, sludge treatment, hydrogen production, and gas diffusion. The fundamental research category was based on studies that exclusively looked at membrane fouling, operation and design parameters, sludge properties, microbiological characteristics, cost, and modeling. Studies, which focused on applied research and general reactor performance, were categorized by influent (feed) type (groups 3–6). Membrane fouling, which has been widely considered as one of the major limitations to faster commercialization of MBRs, has been investigated from various perspectives including the causes, characteristics, mechanisms of fouling, and methods to prevent or reduce membrane fouling. More than one-third of studies in the fundamental aspects group were found to deal with issues related to membrane fouling.

4.16.4.3 Modeling Studies on MBR Models that can accurately describe the MBR process are important for the design, prediction, and control of MBR systems. Due to the intrinsic complexity and uncertainty of MBR processes, basic models that can provide a holistic understanding of the technology at a fundamental level are of great necessity. Complex models that are also practical for real applications can greatly assist in capitalizing on the benefits of MBR technology. However, compared to experimental R&D, followed by commercialization of the technology, modeling studies for system-design analysis and performance prediction

are at a relatively preliminary stage. In an attempt to identify the required research initiatives in this regard, this section looks briefly into the state-of-the-art MBR modeling efforts. Effluent quality and the investment and operating costs are the primary concerns for any given wastewater treatment system. Therefore, model development should center on components for which water-quality standards have been set and parameters which are strongly correlated to cost. Ng and Kim (2007) put forward a few key model components and parameters for MBR modeling:

•

•

•

•

•

The ability to quantify individual resistance (i.e., resistance from cake formation, biofilm formation, and adsorptive fouling) as a function of the various influencing parameters is important in determining which parameters have the greatest influence on fouling and for designing and optimizing the system to achieve an economical balance between production and applied pressure. Determining the relationship between biomass concentration and other parameters can aid in identifying an optimal biomass concentration for operation, which can lead to significant economical savings. Aeration accounts for a significant portion of energy costs in the operation of MBR systems. The factors that influence oxygen requirement (wastewater and biomass concentration/growth rates) and the oxygen-transfer rate (MLSS concentration, MBR configuration, type of bubbles used, and specific airflow rate) should receive due consideration in the model to optimize aeration. Carbon and nutrient (nitrogen and phosphorous components) concentrations and their influencing factors (e.g., respective concentrations and growth rates of the various types of organisms and concentration of oxygen) should be incorporated into the models. Soluble microbial products (SMPs), which comprise a major portion of the organic matter in effluents from biological treatment processes and are potentially associated with issues such as disinfection by-product formation, biological growth in distribution systems, and membrane fouling, should be given proper consideration in models.

MBR models available in the literature can be broadly classified into three categories: biomass kinetic models, membranefouling models, and integrated models to describe the complete MBR process (Ng and Kim, 2007; Zarragoitia-Gonza´lez et al., 2008). Models describing biomass kinetics in an MBR include the activated sludge model (ASM) family (Henze et al., 2000), the SMP model (Furumai and Rittmann, 1992; Urbain et al., 1998; de Silva et al., 1998), and the ASM–SMP hybrid model (Lu et al., 2001; Jiang et al., 2008). The ASMs were developed to model the activated sludge process. The MBR process is the activated sludge process with the secondary clarification step replaced by membrane filtration; therefore, it is reasonable to use ASMs to characterize the biomass dynamics in an MBR system. However, their ability to describe the MBR process accurately has not been verified by in-depth experiments. Research suggests that SMPs are important components in describing biomass kinetics due to high SRTs in MBR systems. Accordingly, the SMP model demonstrated the capability of

Membrane Biological Reactors

characterizing the biomass with a reasonable-to-high degree of accuracy. Lu et al. proposed that the modified versions of ASM1 (Lu et al., 2001) and ASM3 (Lu et al., 2002), which incorporate SMPs, demonstrated fairly reasonable accuracy in quantifying COD and soluble nitrogen concentrations. Jiang et al. (2008) extended the existing ASM No. 2d (ASM2d) to ASM2dSMP with introduction of only four additional SMPrelated parameters. In addition to minimizing model complexity and parameter correlations, the model parameter estimation resulted in reasonable confidence intervals. Models describing membrane fouling include the empirical hydrodynamic model (Liu et al., 2003), fractal permeation model (Meng et al., 2005), sectional resistance model (Li and Wang, 2006), subcritical fouling behavior model (Saroj et al., 2008), and the resistance-in-series models that were presented as a part of the integrated models. Some of them are simply based on solid–liquid separation and simulate filtration processes (Chaize and Huyard, 1991; Gori et al., 2004). Other models consider specific physical approaches: cross-flow filtration (Cheryan, 1998; Hong et al., 2002; Beltfort et al., 1994) and mass-transport models (Beltfort et al., 1994; Bacchin et al., 2002). Nevertheless, membrane fouling is generally evaluated by employing the resistance-in-series model (Wintgens et al., 2003; Wisniewski and Grasmick, 1998) or, rarely, using empirical models (Benitez et al., 1995; De Wilde et al., 2003). The integrated models, basically, couple the kinetic models with the fouling ones (such as the resistance-in-series model) and they often consider the formation and degradation of SMPs (Ng and Kim, 2007). The models reported to date are valuable preliminary attempts, but require further improvements. For instance, the empirical hydrodynamic model is too simple to describe the membrane-fouling phenomenon, and the sectional resistance model lacks accuracy. Both the fractal permeation model and resistance-in-series model by Lee et al. (2002) provide good scientific insight, but specific experimental verification is necessary for general use of the models. The resistance-in-series model developed by Wintgens et al. (2003) shows the most promise, as it is fairly accurate, accounts for cleaning cycles, and can predict permeability changes over time. Further tests are needed to determine whether the model requires calibration or if the model parameters are applicable to other MBR systems. Recently, Zarragoitia-Gonza´lez et al. (2008) included the biological kinetics and the dynamic effect of the sludge attachment and detachment from the membrane, in relation to the filtration and a strong intermittent aeration in a hybrid model. The model was established considering SMP formation–degradation kinetic based on previous published models (Cho et al., 2003; Lu et al., 2001). A modification of Li and Wang’s model (Li and Wang, 2006) allows to calculate the increase of the transmembrane pressure (TMP), evaluating, at the same time, the influence of an intermittent aeration of bubbles synchronized with the filtration cycles on fouling control, and to analyze the effects of shear intensity on sludge cake removal. On the other hand, in order to describe the biological system behavior, a modified ASM1 model was used. The final hybrid model was developed to calculate the evolution of sludge properties, its relation to sludge cake growth, and the influence of sludge properties on membrane fouling.

585

A simple model for evaluating energy demand arising from aeration of an MBR was presented by Verrecht et al. (2008) based on a combination of empirical data for membrane aeration and biokinetic modeling for biological aeration. The model assumes that aeration of the membrane provides a portion of the dissolved oxygen needed for biotreatment. The model also assumes, based on literature information sources, a linear relationship between membrane permeability and membrane aeration up to a threshold value, beyond which permeability is unchanged with membrane aeration. An analysis reveals that significant reductions in energy demand are attained through operating at lower MLSS levels and membrane fluxes. The complete organic removal in MBR is due to all the inseries phenomena: biological degradation of biomass, biological filtration of cake layer, and final filtration of physical membrane. Di Bella et al. (2008) set up a mathematical model for the simulation of physical–biological wastewater organic removal for SMBR system. The model consists of two submodels: the first one for the simulation of the biological processes and a second one for the physical processes. In particular, regarding the biological aspects, it is based on the ASM concept. On the other hand, organic-matter removal due to filtration (the physical process) was described by simple models proposed in the literature (Kuberkar and Davis, 2000; Jang et al., 2006; Li and Wang, 2006). It is conceivable that several of the existing models, particularly the ASMs, require validation to determine their applicability for modeling the MBR process and to evaluate whether they can serve as a base for future MBR model development. Membrane fouling in MBRs is affected by the biotransformation processes in the system; therefore, a more effective integration of biomass kinetics and membrane fouling into the models is required. Moreover, examination of alternative empirical modeling approaches, such as the application of artificial neural networks, is worthwhile to establish a thorough link between inputs and outputs of MBR systems and to find phenomenological interrelationships among components and parameters (Ng and Kim, 2007).

4.16.4.4 Innovative Modifications to MBR Design Researchers have put forth different modifications to the conventional design of MBRs in order to enhance removal performance and/or mitigate membrane fouling. This section highlights some of such examples (Table 7). The commercialized MBR formats are discussed separately in Section 4.16.5.2.

4.16.4.4.1 Inclined plate MBR Theoretically, an infinite SRT provides a possibility of naturally achieving zero-excess sludge discharge from MBR under normal environment. It should, however, be noted that zeroexcess sludge production is just a theoretical concept which can only be obtained with a feed containing only solutes. In real life, sewage or industrial effluents contain nonbiodegradable suspended solids and colloids that accumulate in the reactor, continuously increasing the sludge concentration. Therefore, an immediate challenge encountered at infinite SRT is the extremely high sludge concentration

586

Membrane Biological Reactors

Table 7

Examples of innovative modifications to MBR design

Modified design

Main purpose

Selected reference

Inclined plate MBR

Omit excess sludge production and thereby realize long-term stable membrane filtration Derive the simultaneous advantages of efficient nutrient removal and mitigate membrane fouling (Chae et al., 2006 a,b;). Treatment of high-strength wastewater without encountering severe fouling Enhanced removal of recalcitrant compound and/or membrane fouling mitigation Obtain in one step indirect potable reuse standard effluent Indirect potable reuse along with energy demand reduction Obtain in one step indirect potable reuse standard effluent

Xing et al. (2006)

Integrated anoxic–aerobic MBR

Jet-loop-type MBR Biofilm MBR Nanofiltration MBR Forward osmosis MBR Membrane distillation bioreactor (MDBR)

produced in the bioreactor (Wen et al., 1999). Consequently, the method to achieve zero-excess sludge discharge translates into how to realize long-term stable membrane filtration of high-concentration sludge beyond the guideline value of 10– 20 g l1 recommended for submerged MBRs when applied to domestic wastewater treatment. In order to omit excess sludge production, Xing et al. (2006) proposed an innovative MBR design comprising an anoxic tank equipped with settlingenhancer inclined plates and a subsequent aerobic tank containing the membrane. The inclined plates together with intermittent air blowing (to blow off gaseous content generated by denitrification, etc.) proved to be quite effective in confining high MLSS sludge within the anoxic tank leading to an MLSS difference of 0.1– 13.1 g l1 between the aerobic and anoxic sludge. Consequently, the capability of MBRs in handling the extremely high MLSS challenge encountered especially at zero-excess sludge could be extended. Results indicated that at an HRT of 6 h, average removals of COD, ammonia nitrogen, and turbidity were 92.1, 93, and 99.9%, resulting in daily averages of 12.6 mg COD l1, 1.3 mg NH3–N l1, and 0.03 NTU, respectively.

4.16.4.4.2 Integrated anoxic–aerobic MBR In contrast to separate anoxic tanks for denitrification or creation of alternating anoxic/oxic conditions within the same tank by intermittent aeration, an integrated anoxic/oxic MBR, containing anoxic/oxic compartments in one reactor, was developed to derive simultaneous advantages of efficient nutrient removal (Chae et al., 2006a, 2006b) and mitigated membrane fouling (Chae et al., 2006a, 2006b; Hai, 2007; Hai et al., 2007; Hai et al., 2006b; Hai et al., 2008a). Under the optimal volume ratio of anoxic and oxic zones of 0.6 and the desirable internal recycle rate and HRT of 400% and 8 h, respectively, the average removal efficiencies of total nitrogen (T-N) and total phosphorus (T-P) were 75% and 71%, respectively (Chae et al., 2006b). Furthermore, comparison with sequential anoxic/oxic MBR under the same conditions revealed the membrane-fouling reduction potential of this specific design (Chae et al., 2006a).

Chae et al. (2006a,b), Hai et al. (2006b, 2008a)

Park et al. (2005), Yeon et al. (2005) Lee et al. (2006), Leiknes and Odegaard (2007), Ngo et al. (2008), Hai et al. (2008) Choi et al. (2002) Achilli et al. (2009), Cornelissen et al. (2008) Phattaranawik et al. (2008, 2009)

Working with a high-strength industrial wastewater, Hai et al. (2006a, 2006b, 2008a) demonstrated minimization of excess sludge growth and maintenance of less MLSS concentration in contact with the membrane at the aerobic zone by exploring a similar reactor design along with a strategy of splitting the feed through the two zones.

4.16.4.4.3 Jet-loop-type MBR The so-called high-performance compact reactor (HCR) which is a jet-loop-type reactor with a draft tube and a two-phase nozzle was coupled with a submerged membrane by Park et al. (2005). The HCR is able to deal with very high organic loading rates due to the high efficiency of oxygen transfer, mixing, and turbulence achieved. The significant amount of bubbles and turbulence present in the HCR can be beneficial in retarding fouling of the submerged membrane. The developed MBR showed much greater membrane permeability than the conventional MBR, promising very high potential for the treatment of high-strength wastewater without encountering severe fouling (Park et al., 2005; Yeon et al., 2005).

4.16.4.4.4 Biofilm MBR Membrane-coupled moving-bed biofilm reactor system, wherein the membrane is submerged within the same tank (Lee et al., 2006) or in an additional tank (Leiknes and Odegaard, 2007), has been extensively studied in association with different kinds of biocarriers. Powdered activated carbon (PAC) which also acts as an adsorbent is commonly added into the bioreactor as the biocarrier (Ng et al., 2006; Hai, 2007; Hai et al., 2008b). However, carriers made of inert materials, such as plastic (Leiknes and Odegaard, 2007) and sponge (Lee et al., 2006; Ngo et al., 2008), have also been used. Biomass granulation with shell-support media coupled with membrane separation is also worth mentioning in this context (Thanh et al., 2008). The mechanisms of enhanced removal and/or membranefouling mitigation depend on the specific design and the utilized biocarrier type. For example, in an integrated membrane-coupled moving-bed biofilm reactor using sponge as the biocarrier, frictional force exerted by the circulating

Membrane Biological Reactors

carrier on the submerged membrane reduced the formation of cake layer on the membrane surface and thus enhanced the membrane permeability (Lee et al., 2006). On the other hand, Leiknes and Odegaard (2007) demonstrated that operation under high volumetric-loading rates of 2–8 kg COD m3 d1and HRTs up to 4 h and maintenance of membrane fluxes around 50 l m2 h1 were possible by placing the moving-bed biofilm reactor prior to the submerged MBR. The specific purpose of the biofilm reactor in this case was to reduce the organic loading on MBR. Ng et al. (2006) contend that the improved membrane performance of the MBR with added PAC could be due to a number of factors including, PAC providing sink for some of the fouling components and the scouring action of PAC. Hai et al. (2008b) reported that simultaneous PAC adsorption within a fungiMBR treating dye wastewater resulted in multiple advantages including co-adsorption of dye and fungal enzyme onto activated carbon and subsequent enzymatic dye degradation.

4.16.4.4.5 Nanofiltration MBR The potential for using NF technology in wastewater treatment and water reuse is noteworthy. A new concept with the addition of RO membrane after conventional MBR has been recently developed to reclaim municipal wastewater. The new MBR-RO process demonstrated the capability of producing the same or more consistent product quality (in terms of total organic carbon (TOC), NH4, and NO3) and sustained higher flux compared to the CAS-MF-RO process in reclamation of domestic sewage (Qin et al., 2006). Choi et al. (2002, 2007), on the other hand, demonstrated the technical feasibility of a submerged NF-MBR. For the initial 130 days, the NF-MBR achieved high permeate quality (DOC concentration ¼ 0.5–2.0 mg l1) and maintained reasonable water productivity. With low electrolyte rejection, operation under a low suction pressure was possible, and electrolyte accumulation in the bioreactor, which may hinder biological activity, did not occur. The permeate quality, however, deteriorated to some extent (DOC concentration ¼ 3.0 mg l1) due to the deterioration of the cellulose membrane.

4.16.4.4.6 Forward osmosis MBR The forward osmosis (FO)–MBR is an innovative technique for the reclamation of wastewater, which combines activated sludge treatment and FO membrane separation with an RO posttreatment. FO membranes, either submerged or external, are driven by an osmotic pressure difference over the membrane. Through osmosis, water is transported from the mixed liquor across the semipermeable membrane into a draw solution (DS) with a higher osmotic pressure. To produce potable water, the diluted DS is then treated in an RO unit, and the concentrated DS is reused in the FO process. The FO-MBR is expected to have the same advantages as conventional MBRs; however, it has to deal with the most important drawback, that is, a high energy demand. In this system, FO membranes with structures comparable with NF or RO membranes are used instead of MF/UF membranes for the separation of suspended solids, multivalent ions, natural organic matter, and biodegradable materials. Since fluxes are generally lower and no internal fouling occurs, fouling of NF

587

or RO membranes, compared to that of the MF or UF membranes in conventional MBR, may be dealt with easily. The RO system after FO-MBR can be operated with higher fluxes because all the bivalent ions are removed in the FO-MBR. Recent studies have demonstrated high sustainable flux and relatively low reverse transport of solutes from the DS into the mixed liquor, along with very high removal performance (Achilli et al., 2009; Cornelissen et al., 2008).

4.16.4.4.7 Membrane distillation bioreactor A novel wastewater-treatment process known as the membrane distillation bioreactor (MDBR) incorporating membrane distillation in an SMBR operated at an elevated temperature was developed and experimentally demonstrated by Phattaranawik et al. (2008, 2009). The ability of membrane distillation (MD) to transfer only volatiles means that very high quality treated water is obtainable, with TOC levels below 1 ppm and negligible quantity of salts. A unique feature is that the MDBR allows for organic retention times to be much greater than the HRT. The TOC in the permeate was consistently lower than 0.7 mg l1 for all experiments. Stable fluxes in the range 2–5 l m2 h1 have been sustained over extended periods. The MDBR was described to have the potential to achieve in a single step, the reclamation obtained by the combined MBR þ RO process. It was also suggested that for viable operation, it would be necessary to use low-grade (waste) heat and water cooling. Several other emerging approaches are also noticeable in contemporary literature. These include hybrid MBR-CAS concept (De Wilde et al., 2009), anaerobic baffled reactor-MBR combination (Pillay et al., 2008), etc.

4.16.4.5 Technology Benefits: Operators’ Perspective The relative advantages of MBR over the CAS process were outlined in Section 4.16.3.3. This section highlights the technical benefits of MBRs cited by the operators: 1. high-quality effluent, ideal for post membrane treatments (e.g., NF and UF); 2. space savings, enabling upgrading of plants without land expansion; 3. shorter start-up time compared to conventional treatment systems; 4. low operating and maintenance manpower requirement (average of 1.7 working hours per MLD); and 5. (5) automated control.

4.16.4.6 Technology Bottlenecks MBR technology is facing some research and development challenges. The technology bottlenecks as reported in the literature include (Howell, 2002, 2004; Lesjean et al., 2004; Le-Clech et al., 2005a; Yang et al., 2006; Melin et al., 2006) 1. Membrane fouling. Further understanding the mechanisms of membrane fouling and developing more effective and easier methods to control and minimize membrane fouling. 2. Pretreatment. Effective methods to limiting membrane clogging and operational failures.

588

Membrane Biological Reactors

3. Membrane life span. Increasing membrane mechanical and chemical stability. 4. Cost. Further reduction of costs for maintenance and replacement of membranes, energy requirement, and labor requirements. 5. Plant capacity. Scaling up for large plants. 6. Exchangeability of modules. Module exchangeability between different brands (reduction of costs for replacement of membranes). Some other problems often encountered by the operators include (Leslie and Chapman, 2003; Adham et al., 2004; LeClech et al., 2005a; Yang et al., 2006)

• • • • • • • • • •

membrane fouling during permeate backpulsing, entrained air impacting suction-pump operation, bioreactor foaming, inefficient aeration due to partial clogging of aerator holes, no significant decrease of biosolid production, scale buildup on membrane and piping, corrosion of concrete, hand rails, and metallic components due to corrosive vapor produced during high temperature NaOCl cleaning, membrane delamination and breakage during cleanings, odor from screening, compaction, drying beds, and storage areas (although normally less than in CAS), and failure of control system.

Although the commercialization of MBRs has expanded substantially in the past 20 years, target markets have not been tapped to a large extent and new potential areas of applications are continually developing. The R&D challenges mentioned above, when tackled, will lead to a more competitive and mature market for MBR applications. Lesjean et al. (2004) contend that academic research is addressing only some of these issues. For instance, while many publications on fouling are being produced and some cost studies are conducted, no significant research efforts have addressed membrane life span, pretreatment, and scale-up issues. Academic researchers can expect interest from MBR companies and plant operators on these subjects, and should direct some of their research programs to address these needs. Among the challenges underscored by the experts, membrane fouling is one of the most serious problems that has retarded faster commercialization of MBR technology. The causes, characteristics, mechanisms of fouling, and methods to prevent or reduce membrane fouling are discussed elaborately in Section 4.16.4.7; Section 4.16.5.5 sheds light on the issue of exchangeability of modules. The remainder of the current section will be devoted to the issues closely related to membrane fouling and performance, that is, mechanical pretreatment and membrane integrity:

•

Pretreatment. Pretreatment is one of the most critical factors for ensuring a stable and continuous MBR operation. Due to membrane sensitivity to the presence of foreign bodies, fine prescreening of the feed (and sometimes of the mixed liquors) must occur. The type of sieve installed is very important with regard to the total screening of hair and fibers. Recent studies (Frechen et al., 2006; Schier et al., 2009) have shown sieves with smaller gap sizes and with

•

two-dimensional gap geometries to perform better. On the other hand, even intensive long-term pilot plant trials can fail to suggest the effective scale-up design of the sieve (Melin et al., 2006). If too many clogging problems occur, the original pre-screen systems are usually upgraded to finer screens. However, when both the influent and the mixed liquor are filtered with a fine prescreen, a large amount of trash is produced (up to 3.8 m3 per week for a 1.4 MLD plant) (Le-Clech et al., 2005a; Melin et al., 2006; Schier et al., 2009). It should be noted that the investment in pretreatment is of little use if the bioreactor is uncovered, in which case, different sorts of debris can easily enter the bioreactor. It is recommended to remove these items using a high-pressure water hose. However, many MBR users report that this type of manual cleaning causes membrane-fiber breakage. In order to keep the membrane effectively separated from the fibrous materials, Schier et al. (2009)proposed the following mechanical-treatment concept: conventional pretreatment including screen and grit chamber/grease trap to be placed before the biological tank, causing braid of hair and fibers formed therein to be removed by the sieve placed before the separate filtration chamber housing the membrane modules. Membrane integrity. A major problem facing MBR systems is the loss of membrane integrity, which leads to the permeate-quality deterioration and ineffective backwashing. When breakage occurs in a submerged hollow-fiber MBR system, continuous filtration may allow solids and particles to quickly clog the broken fiber. However, application of backwash would force the solids out of the fiber. Accordingly, once damaged, disinfection of the product water would be compromised and it would also cause the loss of the backwash efficiency; and the faulty membrane/module would need to be changed quickly.

Faulty installation is one obvious reason for membrane failure. Once under pressure, an incorrectly installed membrane module can be compressed. Other reasons associated with regular operation include frequent and/or extended contact between membrane and cleaning solution causing delamination of the membrane, scoring and cleaving of the membrane resulting from the presence of abrasive or sharpedged materials in the influent, and operating stress and strain occurring in the system due to fiber movement and membrane backwashing. A better understanding of the effect of membrane material, age, and fouling on membrane integrity may be gained from hollow-fiber-tensile test reported in the literature (Childress et al., 2005; Gijsbertsen-Abrahamse et al., 2006). Even flat-sheet membranes used in MBRs are not immune to occasional failure (Cornel and Krause, 2003). The construction of current flat-sheet MBR membrane panels is a labor-intensive, multistep operation. These are typically sandwich constructions with three separate layers. Two of them are pre-fabricated membrane layers, while the third one is a permeate drainage layer which is sandwiched between them. The three layers of the sandwich are held together by gluing or laminating techniques over their entire surface or just at their edges. Flat-sheet membranes have been found to be sensitive to breaking near the top

Membrane Biological Reactors

due to poor adhesion of the membrane to the support layer (Doyen et al., 2010).

4.16.4.7 Membrane Fouling – the Achilles’ Heel of MBR Technology Although MBR has become a reliable alternative to CAS processes and an option of choice for many domestic and industrial applications, membrane fouling and its consequences in terms of plant maintenance and operating costs limit the widespread application of MBRs (Le-Clech et al., 2006). Membrane fouling can be defined as the undesirable deposition and accumulation of microorganisms, colloids, solutes, and cell debris within pores or on membrane surface (Meng et al., 2009). It results from the interaction between the membrane material and the components of the activated sludge liquor, which include biological flocs formed by a large range of living microorganisms along with soluble and colloidal compounds. Thus, it is not surprising that the fouling behavior in MBRs is more complicated than that in most membrane applications. The suspended biomass has no fixed composition and varies with both feedwater composition and MBR operating conditions employed. Accordingly, although many investigations of membrane fouling have been published, the diverse range of operating conditions and feedwater matrices employed, and the limited information reported in most studies on the biomass composition in suspension or on the membrane, have made it difficult to establish any generic behavior pertaining to membrane fouling in MBRs. Three fouling phenomena need to be recognized and duly addressed:

• • •

Cake formation. This results from the balance of forces (shear stress at the membrane wall and filtration force) and is evidently linked to the biomass characteristics. Blockage of bundle of fibers. The bundle of fibers act as a deep bed filter (depending on biomass characteristics and structure of the bundle). Biofilm formation. This is not strictly dependent upon biomass characteristics as, very often, the microorganisms involved in the biofilm formation are not the dominant species in the biomass.

4.16.4.7.1 Fouling development Zhang et al. (2006a) proposed a three-stage history for membrane fouling in MBRs:

• • •

Stage 1. An initial short-term rise in TMP due to conditioning. Stage 2. Long-term rise in TMP, either linear or weakly exponential. Stage 3. A sudden rise in TMP, with a sharp increase in dTMP/dt, also known as the TMP jump.

When operating at fluxes well below the apparent critical flux of the MLSS, a slow steady rise in TMP (stage 2) is observed which eventually changes to a rapid rise in TMP (stage 3). For sustainable operation, the aim would be to limit the extent of stage 1, prolong stage 2, and avoid stage 3, since it could be difficult to restore.

589

4.16.4.7.2 Types of membrane fouling Definitions based on ease of removal and a variety of confusing terminologies have been proposed in the literature to describe fouling. For example, based on the ease of removal, some authors prefer to use the term ‘irreversible fouling’ to the fouling that can be removed by chemical cleaning but not by physical cleaning. Recently, Meng et al. (2009) proposed a somewhat changed definition and used the terms ‘removable’ and ‘irremovable’ for the fouling which is easily eliminated and which requires chemical cleaning, respectively. This chapter, however, uses the more direct terms – physically removable fouling and chemically removable fouling. The formation of a cake layer which can be described as a porous media with a complex system of interconnected interparticle voids has been reported as the major contributor to membrane fouling in MBRs (Jeison and van Lier, 2007; Ramesh et al., 2007). Such fouling is usually physically removable. Recently, a large number of scientific investigations have been performed in order to gain a better understanding of cake-layer formation and cake-layer morphology employing techniques such as confocal laser-scanning microscopy (CLSM), multiphoton microscopy, etc. (Yang et al., 2007; Hughes et al., 2006, 2007). During initial filtration, colloids, solutes, and microbial cells pass through and deposit inside the membrane pores. However, during the long-term operation of MBRs, the deposited cells multiply and yield extracellular polymeric substance (EPS), which clog the pores and form a strongly attached fouling layer. Chemical cleaning is usually required to remove such fouling. Evaluation of physically removable and chemically removable fouling propensity of MBR mixed liquor has been the focus of many studies to date (Field et al., 1995; Ognier et al., 2004; Pollice et al., 2005; Bacchin et al., 2006; Guglielmi et al., 2007; Lebegue et al., 2008; Wang et al., 2008b). Some of the definitions are based on the fouling components. The fouling in MBRs can be classified into three major categories: biofouling, organic fouling, and inorganic fouling, although, in general, all of them take place simultaneously during membrane filtration of activated sludge. Biofouling refers to the deposition, growth, and metabolism of bacteria cells or flocs on the membranes. Biofouling may start with the deposition of individual cell or cell cluster on the membrane surface, after which the cells multiply and form a biocake (Liao et al., 2004; Pang et al., 2005; Wang et al., 2005; Ramesh et al., 2007). Techniques such as scanning electron microscopy (SEM), CLSM, atomic force microscopy (AFM), and direct observation through the membrane (DOTM) have been extensively used to derive valuable information regarding floc/cell-deposition process and the microstructure or architecture of the cake layer. Certain studies have also analyzed the microbial community structures and microbial colonization on the membranes in MBRs (Chen et al., 2004; Jin et al., 2006; Jinhua et al., 2006; Zhang et al., 2006b; Miura et al., 2007; Lee et al., 2009) employing molecular techniques. Such studies reported that the microbial communities on membrane surfaces were quite different from those in the suspended biomass and initially a specific phylogenetic group of bacteria may play the key role in development of the mature biofilm. However, a temporal change

590

Membrane Biological Reactors

of microbial-community structure can take place due to the development of anoxic conditions in the cake layer. Organic fouling in MBRs refers to the deposition of biopolymers on the membranes (Meng et al., 2009). Due to the small size, the soluble biopolymers can be deposited onto the membranes more readily, but they have lower back-transport velocity in comparison to large particles (e.g., colloids and sludge flocs). Powerful analytical tools such as Fourier transform infrared (FTIR) spectroscopy, solid-state 13C-nuclear magnetic resonance (NMR) spectroscopy, and high-performance size-exclusion chromatography (HP-SEC) are usually utilized for identification of the deposited biopolymers (Kimura et al., 2005; Rosenberger et al., 2006; Zhou et al., 2007; Teychene et al., 2008) and studies have confirmed that SMP or EPS is the origin of organic fouling in MBR. Inorganic elements such as Mg, Al, Fe, Ca, Si, etc. and metals can enhance the formation of biofouling and organic fouling and can together form a recalcitrant cake layer (Lyko et al., 2007; Wang et al., 2008b). Inorganic fouling can form in two ways – due to concentration-polarization-led chemical precipitation and entrapment within biopolymer gel layer (Meng et al., 2009). Chemical cleaning agents such as ethylenediaminetetraacetic acid (EDTA) might efficiently remove inorganics on the membrane surface (Al-Amoudi and Lovitt, 2007); however, the fouling caused by inorganic scaling may not be easy to eliminate even by chemical cleaning (You et al., 2006).

Figure 8 lists the membrane-fouling parameters, while Figure 9 illustrates the interrelations and combined effect of those parameters. Some of the membrane characteristics and the parameters that influence the performance of the MBRs are discussed in the following: 1. Physical parameters.

•

Pore size and distribution. Studies revealed that the pore size alone could not predict hydraulic performances. The effects of pore size (and distribution of pore size) on membrane fouling are strongly related to the feedsolution characteristics and in particular the particlesize distribution. The complex and changing nature of

Membrane fouling

4.16.4.7.3 Parameters influencing MBR fouling All the parameters involved in the design and operation of MBR processes have an influence on membrane fouling (Le-Clech et al., 2006; Meng et al., 2009). While some of these parameters have a direct influence on MBR fouling, many others result in subsequent effects on phenomena exacerbating fouling propensity. However, three main categories of factors can be identified – membrane and module characteristics, feed and biomass parameters, and operating conditions.

Membrane characteristics • Physical parameters -Pore size and distribution -Porosity/roughness -Membrane configuration • Chemical parameters -Hydrophobicity -Materials

Mixed liquor characteristics

Feed

Biomass

Figure 9 Interrelations and combined effect of the membrane fouling parameters.

Feed–biomass characteristics

Operating conditions

• Nature of feed and concentration • Biomass fractionation • Biomass (bulk) parameters -MLSS concentration -Viscosity -Temperature -Dissolved oxygen (DO) • Floc characteristics -Floc size -Hydrophobicity/surface charge

• Aeration, cross-flow velocity • Sludge retention time (SRT) • Unsteady state operation

• Extracellular polymeric substance (EPS) • Soluble microbial products (SMP) Figure 8 Membrane fouling parameters at a glance.

Operating conditions

Membrane characteristics

Membrane Biological Reactors

•

•

the biological suspension present in MBR systems and the large pore-size distribution of the membrane generally used in MBR systems are the main reasons for the undefined general dependency of the flux propensity on pore size (Chang et al., 2002a; Le-Clech et al., 2003b). It is generally expected that smaller-pore membranes would reject a wider range of materials, and the resulting cake layer would feature a higher resistance compared to large-pore membranes. However, this type of fouling is easily removed during the maintenance cleaning than fouling due to internal pore clogging obtained in larger-pore membrane systems. The chemically removable fouling, due to the deposition of organic and inorganic materials onto and into the membrane pores, is the main cause of the poor longterm performances of larger pore-size membranes (Chang et al., 2001; He et al., 2005). However, the opposite trend is sometimes reported (Gander et al., 2000). The duration of the experiment and other operating parameters such as cross-flow velocity and constant pressure or constant flux operation have a direct influence on the determination of the optimization of the membrane pore size and are responsible for contradictory reports in the literature. Porosity/roughness. Membrane roughness and porosity along with membrane microstructure, material, and pore-size distribution were suggested as potential reasons for the different fouling behaviors observed (Kang et al., 2006; Ho and Zydney, 2006). For instance, a track-etched membrane, with its dense structure and small but uniform cylindrical pores, featured the lowest resistance due to pore fouling in contrast to the other membranes having interwoven sponge-like highly porous network (Fang and Shi, 2005). Other studies have pointed out the importance of pore-aspect ratio (mean major-axis length/mean minor-axis length) (Kim et al., 2004) or roughness (He et al., 2005) on fouling in an MBR. Membrane configuration. In submerged MBR processes, the membrane can be configured as vertical flat plates, vertical or horizontal hollow fine fibers (filtration from out to in) or, more rarely as tubes (filtration from in to out). Each of hollow-fiber and flat-sheet membrane types has specific footprint and air scouring and chemical cleaning requirement, which may favor one process over another for a given application (Judd, 2002; Hai et al., 2005). Nevertheless, hollow-fiber modules are generally more economical to manufacture, provide high specific membrane area, and can tolerate vigorous backwashing (Stephenson et al., 2000). For low-flux operation, hollow fibers are attractive due to their high packing density. A higher fiber-packing density would increase productivity; however, increasing the packing density may lead to severe interstitial blockage due to the impeded propagation of air bubbles toward the core, limiting their effect on fouling limitation (Kiat et al., 1992; Yeo and Fane, 2005; Sridang et al., 2005). However, Hai et al. (2008a) developed a spacer-filled module in order to utilize high packing density without encountering

591

severe fouling. Studies have also revealed the effects of other membrane characteristics including hollowfiber orientation, size, and flexibility ( Cui et al., 2003; Ognier et al., 2004; Chang and Fane, 2002; Lipnizki and Field, 2001; Zheng et al., 2003; Zhongwei et al., 2003). 2. Chemical parameters.

•

•

Hydrophobicity. The influence of the membrane hydrophobicity on the early stage of the fouling formation may be significant; however, this parameter is expected to play only a minor role during extended filtration periods in MBRs (Le-Clech et al., 2006). Once initially fouled, the membrane’s chemical characteristics would become secondary to those of the sludge materials covering the membrane surface. Nevertheless, because of the hydrophobic interactions occurring between solutes, microbial cells and membrane material, membrane fouling is expected to be more severe with hydrophobic rather than hydrophilic membranes (Madaeni et al., 1999; Chang et al., 1999; Yu et al., 2005a), although different results have also been reported (Fang and Shi, 2005). In many reported studies, change in membrane hydrophobicity often occurs with other membrane modifications such as pore size and morphology, which make the correlation between membrane hydrophobicity and fouling more difficult to assess. Materials. The large majority of the membranes used in MBRs are polymeric based. A direct comparison between polyethylene (PE) and polyvinylidene fluoride (PVDF) membranes clearly indicated that the latter leads to a better prevention of physically irremovable fouling and that PE membrane fouled more quickly (Yamato et al., 2006). Zhang et al. (2008b) studied the affinity between EPS and the three polymeric UF membranes, and observed that the affinity capability of the three membranes was of the order polyacrylonitrile (PAN)oPVDFopolyethersulfone (PES). Although featuring superior chemical, thermal, and hydraulic resistances, ceramic (Fan et al., 1996; Scott et al., 1998; Luonsi et al., 2002; Xu et al., 2003; Judd et al., 2004) and stainless steel (Zhang et al., 2005) membrane modules are not the preferred option for MBR applications due to their high cost (around an order of magnitude more expensive than the polymeric materials).

3. Feed–biomass characteristics.

•

Nature of feed and concentration. Fouling in the MBR is mostly affected by the interactions between the membrane and the biological suspension rather than wastewater itself (Choi et al., 2005). Nevertheless, the fouling propensity of the wastewater has to be indirectly taken into consideration during the characterization of the biomass, as the wastewater nature can significantly influence the physicochemical changes in the biological suspensions (Le-Clech, 2003b; Jefferson et al., 2004), which in turn may aggravate fouling.

592

•

Membrane Biological Reactors

Biomass fractionation. The many studies (Bae and Tak, 2005; Li et al., 2005a; Itonga et al., 2004; Lee et al., (2003); Lee et al., 2001a; Wisniewski and Grasmick, 1998; Bouhabila et al., 2001) that are available on the contribution of different fractions of the biomass to fouling usually report contradictory results. Although the relatively low fouling role played by the suspended solids (biofloc and the attached EPS) compared to those of the soluble and colloids (generally defined as soluble microbial products or SMP) is usually reported, the reported relative contribution of the SMP to overall membrane fouling ranges from 17% (Bae and Tak, 2005) to 81% (Itonga et al., 2004). These wide discrepancies may be explained by the different operating conditions and biological states of the suspension used in the reported studies (Figure 10). Although an interesting approach for studying MBR fouling, the fractionation experiments neglect any coupling or synergistic effects which may occur among the different components of the biomass.

•

•

4. Biomass (bulk) parameters.

•

MLSS concentration. Although the increase in MLSS concentration has often been reported to have a mostly negative impact on the MBR hydraulic performances (Cicek et al., 1999b; Chang and Kim, 2005), controversies exist (Defrance and Jaffrin, 1999; Hong et al., 2002; Le-Clech et al., 2003b; Lesjean et al., 2005; Brookes et al., 2006). The existence of threshold values above (Lubbecke et al., 1995) or below (Rosenberger et al., 2005) which the MLSS concentration has a negative influence was also reported. Figure 11 depicts the influence of shift in MLSS concentration on flux as reported in different studies. Nowadays, information on additional biomass characteristics (e.g., composition and concentration of EPS) is deemed necessary to furnish a comprehensive picture. On the other hand,

100

Variable: Membrane type

•

Hai et al. (2006a)showed that the extent of fouling was independent of MLSS concentration itself, and was rather more influenced by the efficiency of the foulingprevention strategies adopted. Viscosity. The importance of MLSS viscosity is that it modifies bubble size and can dampen the movement of hollow fibers in submerged bundles (Wicaksana et al., 2006). The net result of this phenomenon would be a greater rate of fouling. Increased viscosity also reduces the efficiency of mass transfer of oxygen and can therefore effect dissolved oxygen (DO) (Germain and Stephenson, 2005); fouling, as discussed later, tends to be worse at low DO. Critical MLSS concentrations have been reported in the literature (Itonga et al., 2004) above which, suspension viscosity tends to increase exponentially with the solid concentration. Temperature. Experiments conducted under moderate temperature usually report greater deposition of materials on the membrane surface at lower temperatures. Temperature may impact membrane filtration by increasing fluid viscosity, causing defloculation of biomass and higher EPS secretion, reducing biodegradation rate, etc. (Jiang et al., 2005; Rosenberger et al., 2006). Dissolved oxygen. The effects of DO on MBR fouling are multiple and may include changes in biofilm structure, SMP levels, and floc-size distribution (Lee et al., 2005). The average level of DO in the bioreactor is controlled by the aeration rate, which not only provides oxygen to the biomass but also tends to limit fouling formation on the membrane surface. Optimum aeration would result in lower specific cake resistance of the fouling layer featuring larger particle sizes and greater porosity (Kang et al., 2003; Kim et al., 2006). Therefore, in general, higher DO tends to lead to better filterability, and lower fouling rate.

Variable: Sludge type

Variable: SRT

80 60 40 20 0 (a)

(b)

Soluble

Colloids

Suspended solids

(c)

Colloid + soluble

Figure 10 Influence of different parameters (membrane type, sludge type, and SRT) on the relative contributions (in %) of the different biomass fractions to MBR fouling. Data from (a) Bae TH and Tak TM (2005) Interpretation of fouling characteristics of ultrafiltration membranes during the filtration of membrane bioreactor mixed liquor. Journal of Membrane Science 264: 151–160; (b) Meng F and Yang F (2007) Fouling mechanisms of deflocculated sludge, normal sludge, and bulking sludge in membrane bioreactor. Journal of Membrane Science 305: 48–56; and (c) Lee W, Kang S, and Shin H (2003) Sludge characteristics and their contribution to microfiltration in submerged membrane bioreactors. Journal of Membrane Science 216: 217–227.

Membrane Biological Reactors

Limiting or critical or stabilized flux, (l m−2 h−1)

120

irregular floc shape, and higher hydrophobicity (Meng et al., 2006).

(1)

100 80

(2)

(7)

60

(3) (4)

40 (5) 20

(6)

0 0

5

10

15

20

25

MLSS concentration, gl−1 Figure 11 Influence of shift in MLSS concentration on flux (fouling) as reported in different studies. Data from (1) Cicek N, Franco JP, Suidan MT, and Urbain V (1998) Using a membrane bioreactor to reclaim wastewater. Journal of American Water Works Association 90: 105–113; (2) Beaubien A, Baty M, Jeannot F, Francoeur E, and Manem J (1996) Design and operation of anaerobic membrane bioreactors: Development of a filtration testing strategy. Journal of Membrane Science 109: 173–184; (3) Madaeni SS, Fane AG, and Wiley D (1999) Factors influencing critical flux in membrane filtration of activated sludge. Journal of Chemical Technology and Biotechnology 74: 539–543; (4) Han SS, Bae TH, Jang GG, and Tak TM (2005) Influence of sludge retention time on membrane fouling and bioactivities in membrane bioreactor system. Process Biochemistry 40: 2393–2400; (5) Bouhabila EH, Ben Aim R, and Buisson H (1998) Microfiltration of activated sludge using submerged membrane with air bubbling (application to wastewater treatment). Desalination 118: 315–322; (6) Bin C, Xiaochang W, and Enrang W (2004) Effects of TMP, MLSS concentration and intermittent membrane permeation on a hybrid submerged MBR fouling. In: Proceedings of the IWA – Water Environment – Membrane Technology (WEMT) Conference. Seoul, Korea, 7–10 June; and (7) Defrance L and Jaffrin MY (1999) Reversibility of fouling formed in activated sludge filtration. Journal of Membrane Science 157: 73–84.

5. Floc characteristics.

•

•

593

Floc size. The floc-size distribution obtained with the MBR sludge is lower than the results generally obtained from CASP (Zhang et al., 1997; Wisniewski and Grasmick, 1998; Lee et al., 2003; Cabassud et al., 2004; Bae and Tak, 2005). Unlike in the CAS systems, the effective separation of suspended biomass from the treated water is not critically dependent on aggregation of the microorganisms, and the formation of large floc. However, independent of their size, biological floc play a major role in the secretion of EPS and formation of the fouling cake on the membrane surface. Hydrophobicity/surface charge. The direct effect of floc hydrophobicity on MBR fouling is difficult to assess. Conceptually, hydrophobic flocs would lead to high flocculation propensity, less secretion of EPS, and low interaction with the hydrophilic membrane (Jang et al., 2006). However, reports of highly hydrophobic flocs fouling MBR membranes can be found in the literature. For instance, the excess growth of filamentous bacteria, known to be responsible for severe MBR fouling, also resulted in higher EPS levels, lower zeta potential, more

6. Extracellular polymeric substances. The term EPS is used as a general and comprehensive concept for different classes of macromolecules such as polysaccharides, proteins, nucleic acids, (phosphor-)lipids, and other polymeric compounds which have been found at, or outside, the cell surface and in the intercellular space of microbial aggregates (Flemming and Wingender, 2001). EPS are the construction materials for microbial aggregates such as biofilms, flocs, and activated sludge liquors. The functions of EPS matrix are multiple and include aggregation of bacterial cells in flocs and biofilms, formation of a protective barrier around the bacteria, retention of water, and adhesion to surfaces (Laspidou and Rittmann, 2002). With its heterogeneous and changing nature, EPS can form a highly hydrated gel matrix in which microbial cells are embedded (Nielson and Jahn, 1999). Therefore, they can be responsible for the creation of a significant barrier to permeate flow in the membrane processes. Contemporary literature is replete with reports identifying EPS as a major fouling parameter (Chang and Lee, 1998; Cho and Fane, 2002; Nagaoka et al., 1996, 1998; Rosenberger and Kraume, 2002). On the other hand, since the EPS matrix plays a major role in the hydrophobic interactions among microbial cells and thus in the floc formation (Liu and Fang, 2003), it was proposed that a decrease in EPS levels may cause floc deterioration and may be detrimental for the MBR performances. This indicates the existence of an optimum EPS level for which floc structure is maintained without featuring high fouling propensity. Many parameters including gas sparging, substrate composition (Fawehinmi et al., 2004), and loading rate (Cha et al., 2004; Ng et al., 2005) affect EPS characteristics in the MBR, but SRT probably remains the most significant of them (Hernandez et al., 2005). A functional relationship between specific resistance, mixed liquor volatile suspended solids (MLVSS), TMP, and permeate viscosity, and EPS is believed to exist (Cho et al., 2005). 7. Soluble microbial products. SMPs are defined as soluble cellular components that are released during cell lysis, diffuse through the cell membrane, and are lost during synthesis or are excreted for some purpose (Laspidou and Rittmann, 2002; Li et al., 2005a). During filtration, SMPs adsorb on the membrane surface, block membrane pores, and/or form a gel structure on the membrane surface where they provide a possible nutrient source for biofilm formation and a hydraulic resistance to permeate flow (Rosenberger et al., 2005). Since direct relationships between the carbohydrate level in SMP (SMPc) solution with fouling rate (Lesjean et al., 2005), filtration index and capillary suction time (CST) (Greiler et al., 2005; Evenblij et al., 2005b; Tarnacki et al., 2005), critical flux tests (Le-Clech et al., 2005b), and specific flux (Rosenberger et al., 2005) have been clearly described, it reveals SMPc to be the major foulant indicator in MBR systems. However, controversy over the relative contribution of carbohydrate and protein portions of SMP to fouling exists (Evenblij and Van der Graaf, 2004; Drews et al., 2005a; Drews et al., 2006).

594

Membrane Biological Reactors

The operating conditions of MBrs are discussed as follows:

•

•

•

Aeration, cross-flow velocity. Since the origin of the SMBR, bubbling has been defined as the strategy of choice to induce flow circulation and shear stress on the membrane surface. Aeration used in MBR systems has three major roles: providing oxygen to the biomass, maintaining the activated sludge in suspension, and mitigating fouling by constant scouring of the membrane surface (Dufresne et al., 1997). However, an optimum aeration rate, beyond which a further increase has no significant effect on fouling suppression, has been observed on many occasions (Ueda et al., 1997; Le-Clech et al., 2003a, 2003b; Liu et al., 2003; Psoch and Schiewer, 2005b). It is also important to note that too intense an aeration rate may damage the floc structure reducing their size, and release EPS into the bioreactor (Park et al., 2005; Ji and Zhou, 2006), and thereby aggravate fouling. Solid retention time. SRT (and thereby the F/M ratio), which greatly controls biomass characteristics, is regarded as the most important operating parameter influencing fouling propensity in MBRs. Considering the advantages of this process over the conventional activated sludge process (CASP), the early MBRs were typically run at very long SRTs to minimize excess sludge (Liu et al., 2005; Gao et al., 2004; Nuengjamnong et al., 2005). But unlike in bench-scale studies employing simpler synthetic feed, the progressive accumulation of nonbiodegradable materials (such as hair and lint) in an MBR fed with real sewage definitely leads to clogging of the membrane module (Le-Clech et al., 2005b). Operating an MBR at higher SRT leads inevitably to increase of MLSS concentration (Zhang et al., 2006c). The increase in aeration intensity to retain high MLSS levels in suspension and maintain proper oxygenation may not be a sustainable option for the treatment process. In this scenario, the increased shear provided to control fouling could cause biofloc deterioration as well as cell lysis and enhanced EPS secretion, and lead to fatal fouling. On the other hand, at infinite SRT, most of the substrate is consumed to ensure the maintenance needs and the synthesis of storage products. The very low apparent net biomass generation observed can explain the low fouling propensity observed for high SRT operation in certain studies (Orantes et al., 2004). It is likely that there is an optimal SRT, between the high fouling tendency of very low SRT operation and the high viscosity suspension prevalent for very long SRT. Unsteady state operation. In practical applications, unsteady state conditions such as variations in operating conditions (flow input/HRT and organic load) and shifts in oxygen supply could occur regularly (Drews et al., 2005a). The start-up phase can also be considered as unsteady operation and data collected before biomass stabilization (including the period necessary to reach acclimatization) may become relevant in the design of MBRs (Cho et al., 2005). Such unsteady state conditions have also been defined as additional factors leading to changes in MBR fouling propensity. For instance, the addition of a spike of acetate in the feedwater significantly decreased the filterability of the biomass in an MBR due to the rise in SMP levels resulting from the feed spike (Evenblij et al., 2005a).

4.16.4.7.4 Fouling mitigation The complex interactions between the fouling parameters complicate the perception of MBR fouling and it is therefore crucial to have a complete understanding of the biological, chemical, and physical phenomena occurring in MBRs to assess fouling propensity and mechanisms and thereby formulate mitigation strategies. As membrane fouling increases with increasing flux in all membrane separation processes, the operating flux should be lower than the critical flux. When the operating flux is below the critical flux, particle accumulation in the region of membranes can be effectively prevented. However, due to physicochemical solute–membrane material interactions, the membrane permeability decreases over time, even when MBRs are operated in subcritical (below critical flux) conditions. Other preventative methods need to be considered to maintain stable operation of MBR systems (Figure 12). Fouling can be removed by various methods and they are as discussed herein: 1. Physical cleaning. The following methods are usually used in combination to remove membrane fouling:

•

Permeate backwashing. Membrane backwashing or backflushing refers to pumping permeate in the reverse direction through the membrane. Backwashing has been found to successfully remove most of the reversible fouling due to pore blocking, transport it back into the bioreactor, and partially dislodge loosely attached sludge cake from the membrane surface (Bouhabila et al., 2001; Psoch and Schiewer, 2005a; Psoch and Schiewer, 2006). Frequency, duration, the ratio between those two parameters, and its intensity are the key parameters in the design of backwashing and different combinations of these parameters have proved to be more efficient in different studies (Jiang et al., 2005; Schoeberl et al., 2005). Between 5% and 30% of the produced permeate is used for backwashing. This also

Removal of fouling

Limitation of fouling

• Physical cleaning --Backwashing --Air backwashing --Intermittent operation --Sonification and other energy-intensive processes

• Optimization of membrane characteristics

• Chemical cleaning --Maintenance cleaning --Intensive cleaning

• Optimization of operating conditions --Aeration --Other operating conditions --Membrane module design • Modification of biomass characteristics -Aerobic granular sludge -Coagulant/flocculent -Adsorbent/flux enhancers

Figure 12 Reported membrane fouling mitigation strategies at a glance.

Membrane Biological Reactors

•

•

•

affects operating costs as, obviously, energy is required to achieve a pressure suitable for flow reversion. Certain studies are, therefore, devoted to optimization of backwashing (Smith et al., 2005). Air backwashing. Air, instead of permeate, can also be used as the backflushing medium (Visvanathan et al., 1997; Sun et al., 2004). The invention of air backwashing techniques for membrane declogging led to the development of using the membrane itself as both clarifier and air diffuser. In this approach, two sets of membrane modules are submerged in the aeration tank. While the permeate is extracted through one of the sets, the other is supplied with compressed air for backwashing. The cycle is repeated alternatively, and there is a continuous airflow into the aeration tank, which is sufficient to aerate the mixed liquor. However, air backwashing may also present potential issues of membrane breakage and rewetting (Le-Clech et al., 2006). Intermittent operation. Intermittent operation or membrane relaxation can significantly improve membrane productivity (Yamamoto et al., 1989). During relaxation, back transport of foulants is naturally enhanced as loosely attached foulants can diffuse away from the membrane surface (Ng et al., 2005). Although some studies found it more important than backwashing for fouling removal (Schoeberl et al., 2005), recent studies tend to combine intermittent operation with frequent backwashing for optimum results (Zhang et al., 2005; Vallero et al., 2005). The economic feasibility of intermittent operation for large-scale MBRs has been the focus of certain studies (Hong et al., 2002); however, it seems rather an established operation mode nowadays. Sonification and other energy-intensive processes. Although sonification would be difficult to apply at a large scale due to the focused nature of the sonic energy, laboratory-scale studies have explored sonification for breaking down cake layers in MBRs, especially in case of ceramic membranes. Certain studies have confirmed the efficiency of application of sonification alone or in combination with backwashing for removing the cake layer (Lim and Bai, 2003; Fang and Shi, 2005). However, other studies report that fouling may even worsen due to pore blocking (Hai et al., 2006a). Attempts have also been made to control fouling or modify sludge by using ozone and electric field (Chen et al., 2007; Huang and Wu, 2008; Sui et al., 2008; Wen et al., 2008).

•

Maintenance cleaning with moderate chemical concentration (weekly) is applied to maintain design permeability and it helps to reduce the frequency of intense cleaning. This may be replaced by a more frequent

(e.g., on a daily basis) chemically enhanced backwash utilizing mild chemical concentration. Intensive (or recovery) chemical cleaning (once or twice a year) is generally carried out when further filtration is no longer sustainable because of an elevated TMP.

The MBR suppliers propose their own chemical cleaning recipes, which differ mainly in terms of concentration and methods, and often site-specific protocols are followed (Kox, 2004; Tao et al., 2005; Le-Clech et al., 2005b). Mainly, sodium hypochlorite (for organic foulants) and citric acid (for inorganics) are used as chemical agents. Some pitfalls of chemical cleaning are worth noting. The detrimental effect of cleaning chemicals on biological performance has been reported (Lim et al., 2005; Hai et al., 2007). It has also been mentioned that the level of pollutants (measured as TOC) in the permeate rises just after the chemical cleaning step (Tao et al., 2005). This raises concern especially in case of MBRs used in the reclamation process trains (i.e., e.g., upstream of RO) (Le-Clech et al., 2006). Chemical cleaning may also shorten the membrane lifetime and disposal of spent chemical agents causes environmental problems (Yamamura et al., 2007). The measures to limit fouling are discussed next. Recently, there have been a significant number of studies which focused on the ways to limit fouling. The proposed strategies include (1) improving the antifouling properties of the membrane, (2) operating the MBR under specific nonor-little-fouling conditions, and/or (3) pretreating the biomass suspension to limit its fouling propensity. They are discussed as follows: 1. Membrane modification.

•

2. Chemical cleaning. The effectiveness of physical cleaning tends to decrease with operation time as more recalcitrant fouling accumulates on the membrane surface. Therefore, in addition to physical cleaning, different types/intensities of chemical cleaning are applied in practice. A combination of the following types of cleaning is usually applied (Le-Clech et al., 2006):

•

595

•

Optimization of membrane characteristics. Many studies have shown that chemical modifications of the membrane surface can efficiently improve antifouling properties. Recent examples comprise (1) increasing membrane hydrophilicity by NH3 and CO2 plasma treatments (Yu et al., 2005a, 2005b) and ultraviolet (UV) irradiation (Yu et al., 2007), (2) TiO2 entrapped membrane (Bae and Tak, 2005), and (3) applying precoating of TiO2 (Bae et al., 2006), GAC (Hai, 2007), ferric hydroxide (Zhang et al., 2004), polyvinylidene fluoride-graft-polyoxyethylene methacrylated (PVDF-gPOEM) (Asatekin et al., 2006), polyvinyl alcohol (PVA) (Zhang et al., 2008a), etc. Improved performance in case of precoated membrane has been attributed to the adsorption of soluble organics on the precoat, limiting the direct contact between the organics and the membrane. Self-forming dynamic membrane-coupled bioreactors, utilizing coarse pore-sized substrates and allowing cake and gel layers to deposit on the surface, have been reported to obtain high flux and good removal in certain studies, although stable performance cannot be expected with such a filtration barrier (Wu et al., 2004). Membrane module design. The membrane module design by optimizing the packing density of hollow fibers or flat sheets, the location of aerators, the orientation of

596

Membrane Biological Reactors

fibers, and diameters of fibers (Chang and Fane, 2001; Chang et al., 2002b; Fane et al., 2002) remains another important parameter in the optimization of the MBR operation. In a specially designed module in which air bubbles were confined in close proximity to the hollow fiber (rather than diffusing in the reactor), higher permeability was obtained (Ghosh, 2006). Two major design approaches are adopted in case of the commercially available hollow-fiber bundles. One of these approaches relies on partitioning of bundles of fibers, which are fixed at both ends, to secure flow path of air bubbles introduced from the center of the bundle at the base, thereby leading sludge out of the module. In another approach, bundle of one-end free fibers are allowed to float freely under the scouring action of air bubbles introduced from the core of the bundle to avoid accumulation of sludge. In order to utilize high packing density without encountering severe fouling, a new approach to hollow-fiber module design was explored by Hai et al. (2008a). Spacer was introduced within usual hollow-fiber bundles with the aim of minimizing the intrusion of sludge into the module. The little amount of intruded sludge was then backwashed through the bottom end, while the sludge deposited on the surface was effectively cleaned by air scouring. In this way, efficient utilization of cleaning solution and air for backwashing and surface cleaning, respectively, were possible. Recent approaches such as novel fiber sheet (FiSh) membrane (Heijnen et al., 2009), multimodule flat-sheet concept (Kreckel et al., 2009), and vacuum rotation membrane (Alnaizy and Sarin, 2009; Komesli et al., 2007) are also noticeable.

•

3. Modification of biomass characteristics.

•

•

2. Optimization of operating conditions.

•

Aeration. As mentioned earlier, bubbling is an established strategy to induce flow circulation and shear stress on the membrane surface. The aeration intensity (air/permeate ratio, m3/m3) applied by MBR suppliers may vary between 24 and 50, depending on the membrane configuration (flat sheet vs. hollow fiber) and the MBR tank design (whether the membrane and aerobic zone combined into a single tank or not) (Tao et al., 2005; Le-Clech et al., 2006). However, recent large-scale studies revealed these original ratios to be quite conservative (Tao et al., 2005). The specific design of bubble size, airflow rate and patterns, and location of aerators have been defined as crucial parameters in fouling mitigation. As the energy involved in providing aeration to the membrane remains a significant cost factor in MBR design, efforts have been focused on optimization of aeration both from the points of view of fouling mitigation and reducing energy requirement. Recent developments in aeration design include cyclic aeration systems (Rabie et al., 2003), intermittent aeration (Yeom et al., 1999; Nagaoka and Nemoto, 2005), air pulsing (Judd et al., 2006), air sparging (Ghosh, 2006), improved aerator systems (Miyashita et al., 2000; Cote, 2002; Hai et al., 2008), etc.

Other operating conditions. The overall performance of the MBR is closely related to the choice of SRT value. Further optimizations of operating conditions through reactor design have been studied and include the addition of a spiral flocculator (Guo et al., 2004), vibrating membranes (Genkin et al., 2005), helical baffles (Ghaffour et al., 2004), suction mode (Kim et al., 2004) and high-performance compact reactor (Yeon et al., 2005), novel types of air lift (Chang and Judd, 2002), porous and flexible suspended membrane carriers (Yang et al., 2006), and the sequencing batch MBR (Zhang et al., 2006d). A reasonable flux rate without significant fouling is ideally expected. The concept of sustainable flux in MBRs was introduced from this point of view (Ng et al., 2005).

•

Aerobic granular sludge. In order to obtain higher biological aggregates in the bioreactor, aerobic granular sludge has also been used in MBR systems (Li et al., 2005b). With an average size around 1 mm, granular sludge increased the membrane permeability by 50%, but lower cleaning recoveries were observed (88% of those obtained with a conventional MBR). Such granular sludge may also not be stable under long-term operation (Hai, 2007). Coagulant/flocculant. Due to back transport and shearinduced fouling control mechanisms, large microbial flocs are expected to have a lower impact on membrane fouling. Based on this expectation, studies have explored addition of coagulants such as alum (Holbrook et al., 2004), ferric chloride, zeolite (Lee et al., 2001b), chitosan (Ji et al., 2008), etc. and have shown permeability enhancement. Pretreatment of the effluent is also possible and studies based on the pre-coagulation/ sedimentation of effluent before its introduction in the bioreactor revealed the fouling limitation offered by this technique (Itonga and Watanabe, 2004; Le-Clech et al., 2006). Adsorbent/flux enhancers. Lower fouling propensity is observed in MBR processes when biomass is mixed with adsorbents in that addition of adsorbents into biological treatment systems decreases the level of pollutants, and more particularly organic compounds (Kim and Lee, 2003; Lesage et al., 2005; Li et al., 2005c; Ng et al., 2006). In view of saturation of PAC during longterm studies, researchers have suggested periodic addition of PAC (Ng et al., 2005; Fang et al., 2006). Certain studies have proposed pre-flocculation and PAC addition (Guo et al., 2004; Cao et al., 2005).

A cationic polymer-based membrane performance enhancer (MPE 50) has been commercialized by Nalco recently. The interaction between the polymer and the soluble organics was reported as the main mechanism responsible for performance enhancement (Yoon et al., 2005). The potential impacts of coagulants or adsorbents on biomass community or biomass metabolism need to be taken into account (Iversen et al., 2009), and the discharge of some chemicals that are used as coagulants or adsorbents might be a potential environmental

Membrane Biological Reactors

risk. Such flux enhancers are probably best suited for solving occasional upsets rather than their continuous addition. Emerging fouling monitoring/control techniques such as interference of microbial intercellular communication by enzymatic degradation of signal molecules (Kjelleberg et al., 2008; Yeon et al., 2009), proteins and polysaccharides sensor for online fouling control (Mehrez et al., 2007), application of two-dimensional fluorescence for monitoring MBR performance (Galinha et al., 2009), etc., are worth noting.

4.16.5 Worldwide Commercial Application 4.16.5.1 Installations Worldwide The MBR process is an emerging advanced wastewater-treatment technology that has been successfully applied at an everincreasing number of locations around the world. MBRs were first developed 40 years ago and have been used commercially in Japan for almost 30 years. Since 1990, MBR technology has been adopted in North America and Europe, and it is now experiencing rapid growth in a wide variety of applications. In Asia, the drive in Japan was followed by an enthusiastic uptake in South Korea in the 1990s, and more recently by China. The highest growth rates are found in areas of greatest water stress for reuse applications, such as the southwestern US, China, Singapore, and Australia. The low footprint of the MBR is a significant driver for developed economies.

4.16.5.1.1 Location-specific drivers for MBR applications Howell (2004) stipulated the location-specific global drivers for MBR technology as follows: 1. Asia. MBR technology is being considered at many locations all over Asia, the main driver being water reclamation. Examples of settings vary from small-scale applications in Japan, where MBR product water is reused as toilet-flushing water in apartment blocks, medium-sized industrial applications in various countries, and large-scale municipal WWTPs in China. 2. Middle East. Clean-water shortages are the obvious driver for MBR applications in the Middle East, in treatment of both municipal as well as industrial (petrochemical) wastewater. 3. Europe. In Western Europe, water reclamation is not the main driver. In the UK, an important driver is compactness and strict discharge limits due to bathing wastewater requirements. In Germany and the Netherlands, important push factors are strict discharge requirements due to ecologically sensitive surface waters and the innovative character of the technological developments related to MBR. In Southern Europe, water reclamation can be considered as the main driver. 4. Northern America. In the US and Canada, MBR initiatives are predominantly driven by strict discharge requirements due to ecologically sensitive surface waters. At some locations, water reclamation is another important driver. In the US, where wastewater-treatment infrastructure lags behind population growth, MBRs are being increasingly implemented to make up the shortfall. Where there is

597

limited space to locate treatment plants, MBRs offer the potential to meet the needs of communities. 5. Australia. Stringent effluent-quality targets and water-reuse potential are obvious drivers for drought-stricken Australia.

4.16.5.1.2 Plant size Earlier MBR technology was favored in difficult applications or those applications where compactness was important and reuse was the target; and it usually involved smaller plants. As the demand for MBR technology grows globally, both the number of installations and the capacity of the installed plants are increasing dramatically. The most optimistic industry estimates suggest that up to 1000 new MBR plants will be built annually during the survey period. The size of the constructed plants has grown from facilities treating hundreds to thousands of gallons of wastewater per day to those treating tens of millions of gallons per day in just a few years. However, the most common capacity for current worldwide MBR installations ranges from the 50 000 gpd (200 m3 d1) to 500 000 gpd systems. The largest MBR plant in the world is set to be operational in 2010/11 in King County, Washington State. When completed, the facility will have an initial peak flow capacity of 495 000 m3 d1 (average 136 000 m3 d1), rising to a daily 645 000 m3 (average 205 000 m3) by 2040.

4.16.5.1.3 Development trend and the current status in different regions Figure 13 shows the regional share of total MBR plants as of 2003. Next, we discusss the trend of MBR growth in the three continents, Asia, Europe, and North America. 1. Asia. In the 1970s sidestream technology first entered the Japanese market. By 1993, 39 of such facilities had been reported for use in sanitary and industrial applications (Aya, 1994). The application of MBR in Japan concerned Europe 11%

N. America 16%

Asia 73% Figure 13 Regional share of total MBR plants (2003). Data from Pearce G (2008a) Introduction to membranes: An introduction to membrane.

598

Membrane Biological Reactors

small-scale installations for domestic wastewater treatment and reuse and some industrial applications, mainly in the food and beverage industries where highly concentrated flows are common. The domestic application often consists of so-called Johkaso or septic-tank treatment and inbuilding (office or domestic) wastewater-collection systems. In the early 1990s, the Japanese Government launched an ambitious 6-year research and development (R&D) project which led to a major technological and industrial breakthrough of the MBR process: the conception of submerged membrane modules, working with low negative pressure (out-to-in permeate suction), and membrane aeration to reduce fouling. This paved the way toward a significant reduction of capital and operation costs, due to the reduction and simplification of equipment and the abandonment of the energy-demanding sludge-recirculation loop. Since then, commercial MBRs proliferated in Japan, which had 66% of the world’s processes in 2000 (Stephenson et al., 2000). In Japan, although MBRs have long been used for industrial wastewater treatment or for reuse of wastewater in large buildings and so on, the introduction of municipal MBRs has lagged behind compared with other water-related fields. The first MBR for municipal wastewater treatment with an installed capacity of 2100 m3 d1 (total design capacity 12 500 m3 d1) in Japan started operation in March 2005, and this accelerated the introduction of MBRs in Japanese sewerage systems. Nine MBR plants, mostly small scale, for municipal wastewater treatment, are in operation at present (Table 8). In addition, there are several MBR plants currently in the design or planning stage. The number of MBRs for municipal wastewater is expected to increase in the near future and the technology will also play an important role in retrofitting and upgrading of existing treatment plants. The MBR technology saw an enthusiastic uptake in South Korea in the 1990s following its introduction in Japan. By 2005, the number of MBR plants rose up to more than 1300 (Namkung, 2008). The plants are mostly small, with more than 60% of the total plants having a capacity of less than 50 m3 d1. The plants were predominantly built on the submerged membrane technology (hollow fiber, 79%; flat sheet, 12%), while a meager 9% facilities utilized the tubular membranes in sidestream format. China has recently emerged as a strong MBR market. Hence, it would be interesting to cast light on the specific

Table 8

mode of development in that country. While the first paper on MBR was published in 1991, the emergence of a number of local and overseas companies developing MBR market in China accelerated with the funding of R&D projects by the Ministry of Science and Technology (MOST) in 1996 (Wang et al., 2008a). Since then, much progress has been achieved both in research and in practical applications of MBR in China. This is evident by the recent yearly publication rate of 35–40 English articles on MBR in China and the construction of a total of 254 plants for municipal (137) and industrial (117) wastewater treatment by 2008. The Chinese MBR market has the presence of a total of 33 companies or institutes, including famous overseas companies such as GE–Zenon Environmental Inc., Mitsubishi–Rayon (Japan), Toray (Japan), NOVO Environmental Technology (Singapore), and XFlow (Netherlands). Among these, only three companies provide flat-sheet MBR, and, interestingly, the worldwide renowned flat-sheet membrane provider, Kubota (Japan), was not found to be very active in the Chinese membrane market. Most of the plants in operation are medium scale or small scale in terms of treatment capacity, the number of plants with treatment capacity below 1000 m3 d1 totaling 225. The largest MBR plant with a capacity of 80 000 m3 d1 for municipal wastewater treatment and reuse is located in Beijing. Several other large MBR plants are also in the planning stage. Wang et al. (2008a) contend that the increasingly stringent discharge standards and the great need of water reclamation and reuse will further push forward the application of everlarger municipal MBR plants in China, especially in North China which has severe water shortage. 2. Europe. A market survey of the European MBR industry was performed by Lesjean and Huisjes (2008). They identified MBR plants constructed up to 2005, and about 300 references of industrial applications (420 m3 d1) and about 100 municipal WWTPs 4500 p.e. were listed. In Europe, the first full-scale MBR plant for treatment of municipal wastewater was constructed in Porlock (UK, commissioned in 1998, 3800 p.e.), soon followed by WWTPs in Bu¨chel and Ro¨dingen (Germany, 1999, 1000 and 3000 p.e., respectively), and in Perthes-en-Gaˆtinais (France, 1999, 4500 p.e.). In 2004, the largest MBR plant worldwide so far was commissioned to serve a population of 80 000 p.e. (in Kaarst, Germany). The installations thus grew from small WWTPs to very large WWTPs within a few

Municipal MBR plants in Japan

Name of plant

Total design capacity (m3d1)

Capacity at commissioning (m3d1)

Membrane format

Start of operation

Fukusaki Kobuhara Yusuhara Okutsu Daito Kaietsu Zyosai Heta Ooda

12 500 240 720 580 2000 230 1375 3200 8600

2100 240 360 580 1000 230 1375 2140 1075

Flat sheet Flat sheet Flat sheet Hollow fiber Flat sheet Hollow fiber – – –

2005 2005 2005 2006 2006 2007 2008 2008 2009

Membrane Biological Reactors

years. Nevertheless, the favored range for MBR systems still appears to be only 100–500 m3 d1 and 1000–20 000 p.e. for industrial and municipal wastewaters, respectively. The design capacity of the industrial units is more than an order of magnitude smaller than for the municipal WWTPs. Lesjean and Huisjes (2008) opined that, although the construction of very large MBR plants (4100 000 p.e.) were recently announced with much publicity, this will remain the exception in Europe, because of the lower lifecycle costs (Lesjean et al., 2004) of WWTP plants equipped with tertiary-membrane filtration (Figure 14). Although not representative of the market, the very large plants will attract much attention and thereby may contribute to the market expansion. The industrial market was the pioneer in the early 1990s, whereas the municipal market took off only in 1999. In 2002, 154 MBR units could be counted, among which 85% were for industrial applications. However, taking the installed membrane surface as an indicator of market share, for the period 2003–05, the municipal sector represented 75% of the market volume. Both municipal and industrial sectors saw a sharp increase in the following years, due to the commercial success and much lower capital and operating costs. By 2005, the market growth rate was linear with at least 50 industrial units and 20 municipal plants constructed per year. This progression rate is expected to sustain in the next years or may even further accelerate owing to the evolution and implementation of European and national regulations (Lesjean et al., 2006). The survey by Lesjean and Huisjes (2008) also demonstrated the predominance of the suppliers Kubota (Japan) and GE–Zenon. Their technologies based on submerged filtration modules have been outstandingly successful since 2002. In recent years, the European market can therefore be seen as a quasi-duopoly of two nonEuropean suppliers. In contrast, the most successful MBR technologies in the 1990s, based on sidestream configurations supplied by Wehrle, Norit X-Flow, Berghof, Rodia Orelis, etc., did not experience any significant market

599

growth over the last 3 years. This could explain the recent move of companies such as Wehrle and Norit to develop and commercialize novel low-energy airlift MBR systems. They argued that the industrial market has become mature: the MBR is considered as the best available technology by many industries. On the other hand, the municipal market is expected to witness further growth over the next decade under the combined effects of the acceleration of plant construction and the capacity increase. 3. North America. Full-scale commercial applications of MBR technology in North America for treatment of industrial wastewaters dated back to 1991 (Sutton, 2003). In the early 1990s, MBR installations were mostly constructed in external configuration. After the mid-1990s, with the development of SMBR system, MBR applications in municipal wastewater extended widely. In the past 15 years, MBR technology has been of increased interest both for municipal and industrial wastewater treatment in North America. The hesitancy on the part of North American municipalities to consider alternative treatment systems to the well-established conventional treatment options delayed the introduction of MBRs into the municipal arena. Industrial applications, particularly for high-strength, difficult-to-treat waste streams, on the other hand, allowed for the considerations of alternative technologies, such as MBRs (Yang et al., 2006). Nevertheless, currently, commercial application in treating industrial wastewaters does not constitute a high percentage of total full-scale MBR plants. Zenon occupies the majority of the MBR market in North America. In 2006, the North American installations constituted about 11% of worldwide installations. As in other places, in North America too, although plant capacities of MBR systems for municipal wastewater treatment are becoming larger, most of the plants in operation are medium scale or small scale in terms of capacity. The largest capacity MBR plant in operation is in Traverse City, MI at 26 900 m3 d1, and the two largest capacity plants under construction are in Johns Creek, GA at 60 000 m3 d1 and King County, Washington State at 136 000 m3 d1.

Capacity, p.e × 104

8

4.16.5.1.4 Decentralized MBR plants: Where and why? 6

4

2

0 1996

1998

2000

2002

2004

2006

2008

Year of commissioning Figure 14 Plot of capacity of randomly selected European MBR plants showing predominance of medium size plants (similar trend prevails worldwide). Data from Schier W, Frechen FB, and Fischer S (2009) Efficiency of mechanical pre-treatment on European MBR plants. Desalination 236: 85–93.

MBR technology can also provide decentralized small-scale wastewater treatment for remote or isolated communities, campsites, tourist hotels, or industries not connected to municipal treatment plants. In small communities, houses are spread out, the population density is low, and hence the use of an on-site system for an individual home or for a cluster of homes could be a cost-effective option. For emerging nations with vast unsewered areas, the population has practically no access to water sanitation, whereby wastewater is directly discharged into water bodies or reused for irrigation without treatment, thus spreading waterborne diseases and causing eutrophication and pollution of water resources. MBR technology could provide a decentralized, robust, and cost-effective treatment for achieving high-quality effluent in such instances. MBRs also offer excellent retrofit capability for expanding or upgrading existing conventional WWTPs.

600

Membrane Biological Reactors

Even when appropriate infrastructure for large-scale water recycling facility exists, the decentralized option may be preferable in some cases. This is because the cost of large-scale water-recycling applications remains high and often uneconomical due to the need to overhaul the existing waterdistribution systems. Large-scale water-recycling applications are, hence, currently somewhat restricted. Furthermore, there is a significant risk of cross-connection associated with the dual-reticulation network, which can seriously dampen public support. While the implementation of the large-scale water recycling is expected to take many years, decentralized water recycling can be applied much more readily. It is expected that MBRs can contribute to a significant increase in decentralized water reclamation and reuse activities. The discussion now centers on the limitations of traditional onsite treatment systems. A gradual but permanent reduction in per-capita water use through socially acceptable means is widely recognized by all stakeholders in the water industry as the strategic longterm sustainable solution to address the ongoing water shortage currently experienced by many countries (Tadkaew et al., 2007). Decentralized wastewater management is not a new concept. Tchobanoglous et al. (2003) defined it as the collection, treatment, and disposal/reuse of wastewater from individual dwellings, clusters of homes or isolated communities, industries, or institution facilities. Traditional decentralized treatment systems such as septic tanks were, in the past, widely used to treat small quantities of wastewater. Due to the likely toughening of environmental legislation in the near future, many of the currently operating wastewater treatments will no longer be acceptable and there will be a need to increase their efficiency significantly. Stricter regulations are found for especially sensitive areas, drinking-waterabstraction areas, and bathing waters. The problem of meeting existing and forecasted more-stringent new regulations affects especially small communities, hotels, and campsites in relatively remote areas without access to sophisticated WWTPs. A major obstacle of decentralized water recycling remains the lack of a suitable technology that can meet the strict and unique effluent criteria required for small-scale water treatment. Some essential requirements are high and reliable treated effluent quality, robustness, tolerance to variable contaminant loading, small footprint, and ease of operation and maintenance. We now discuss the advantages of MBRs in decentralized treatment. As discussed in Section 4.16.5.1.2, historically, the largest number of MBR applications was for a capacity of less than 100 m3 d1. This suggests that the application of MBRs for on-site decentralized system is possible and can offer the most advanced wastewater-treatment options in low-density areas at a cost lower than that of conventional large-scale pipeand-plant systems. Jefferson et al. (2000) argued that smallscale WWTPs constitute a potential growth market for the next millennium and urban sustainability through domestic water recycling is a major identified source for this development. Key advantages of MBRs for decentralized wastewater treatment and reuse are:

•

High and reliable treated effluent quality, small footprint, and high tolerance to variable contaminant loading.

•

•

Due to the robustness and modular nature of MBRs, smallscale MBRs can retain the superiority over conventional treatment methods such as septic tanks with regard to effluent quality, which has been very well documented in the literature (Fane and Fane, 2005). MBRs can be easily combined with other complementary treatment technologies such as UV disinfection and prescreening, which can further enhance the robustness of the treatment system and hence make it particularly suitable for water-recycling applications.

The MBRs for decentralized treatment are not without limitations. Besides the obstacles against widespread application of MBR, in general, the high capital cost can be seen as the key limitation of small-scale MBRs although currently there is very little information to substantiate this premise. Friedler and Hadari (2006) analyzed the economic feasibility of on-site graywater-reuse systems in buildings based on MBR systems. They found that on-site MBR systems became feasible when they were used for the treatment of wastewater incorporating several buildings together because cost was sensitive to building size. Therefore, the on-site MBR system for single building might be unfeasible. This could be a limitation of decentralized MBR systems. However, the true cost of water supply, which takes into account the externalities of resource depletion, was not used in their analysis. It is expected that as the demand for decentralized MBRs increases and membrane technology continues to develop, the use of on-site MBRs even for individual dwellings can be cost competitive in the near future. Some of the examples of worldwide decentralized MBRs are discussed next. The successful introduction of MBR systems into small-scale and decentralized applications has led to the development of packaged treatment solutions from most of the main technology suppliers. Sports stadia, shopping complexes, and office blocks are becoming typical end users, especially in areas of water stress (Stephenson et al., 2000; Melin et al., 2006; Tadkaew et al., 2007). The application of MBRs in Japan to date has predominantly concerned small-scale installations for domestic wastewater treatment. One of the earliest reported case studies is on graywater recycling facilities in the Mori building, Tokyo (Stephenson et al., 2000). The plant consists of a sidestream Pleiade MBR (Ubis) to treat the building flow of 500 m3 d1. The selection of an MBR over a traditional treatment process saved an area equivalent to 25 car-parking places. The treated graywater contained less than 5.5 mg l1 BOD and belowdetection level of suspended solids, colon bacilli, and n-hexane extract, enabling reuse of the graywater. Today, the main Japanese MBR providers such as Kubota or Mitsubishi Rayon offer commercial MBR package plants for on-site domestic water treatment. In Australia, small-scale MBR systems for graywater recycling at a single household level have been marketed by several companies such as AquaCell in New South Wales and BushWater in Queensland (Tadkaew et al., 2007). Commercially available systems in Europe include the package treatment plant Clereflo MBR (Conder Products, UK), designed to service populations up to 5000 and the ZeeMOD (Zenon Environmental Inc.) which is available for flow rates

Membrane Biological Reactors of up to 7500 m3 d1. Most of the manufacturers offer similar systems which means that effluent qualities of 5:5:5 (mg l1) (BOD: NH4-N:SS) are now routinely available to end users as standard treatment options (Melin et al., 2006). Households/ community units (4–50 p.e.) is a concept pioneered by Busse (Germany) in 2000 (Lesjean and Huisjes, 2008). This has become a very competitive market (at least eight products available in Germany). The units are mostly covered by maintenance contracts. The number of sales is expected to increase to address wastewater schemes of small and remote communities, although the revenue may remain marginal in the overall European MBR market. An example in USA is in eastern San Diego County, California, where expansion of an existing casino and development of a shopping mall required extension to the existing treatment facilities. The existing extended aeration system was converted to a ZeeWeed MBR allowing almost triple the capacity of the infrastructure (Melin et al., 2006). The scheme has been operational since July 2000 with the water quality meeting the California tertiary effluent standards for waterreclamation plants.

4.16.5.2 Commercialized MBR Formats As mentioned in Section 4.16.3.1, the first-generation MBRs in wastewater treatment used a sidestream format, in which feed was pumped from the bioreactor through an external membrane system. This approach was suitable for the early stage, small-scale applications for difficult-to-treat feeds. An alternative format was developed in the 1990s using modules submerged in the bioreactor tank, or in an adjoining compartment. This was much more cost effective for treating larger-scale flows with more easily treatable wastewater. The submerged format is available with modules either in a flat-sheet configuration or as hollow fibers or capillary membranes. Originally, the favored concept was to submerge the modules directly into the bioreactor for simplicity. However, in order to gain better control of the balance between the biological and filtration-treatment capacity, it is now more common to use the membrane in an external membrane tank (Brow, 2007). The external arrangement allows the size and design of the membrane tank to be optimized independently, with practical advantages for operation and maintenance. The sidestream approaches are also divided into two formats – the long-established traditional method of crossflow, now used only for the most difficult feeds, and the newer concept of airlift, which uses air to recirculate the feed and thereby significantly reduces energy demand. Both sidestream formats use tubular membranes.

4.16.5.3 Case-Specific Suitability of Different Formats The competing MBR formats based on submerged and sidestream configurations each have their own pros and cons for different application types and plant size. The energy cost for the aeration to control membrane fouling in the MBR is of an order similar to the microbiology aeration for an easy-to-treat feed, and increases by 2.5–3.0 times for the more difficult feed (Cornel and Krause, 2006).

601

Crossflow is more energy intensive – very high cross-flow velocities (up to 5–6 m3 h1) may be necessary to control the fouling; but for the more difficult feeds, it may be the only option that works reliably. Airlift is a more cost-effective way of improving mass transfer through the creation of slug-flow conditions in the lumen of the membrane tubes (Laborie et al., 1997), but there is a limit to how much air flow can be used while retaining slug-flow conditions. Airlift technology has a power cost similar to that of the submerged technology. In general, submerged MBR formats based on hollow fibers have been found to provide the most cost-effective solution for large-scale, easy-to-treat applications. Technology has been developed with optimized packing density and aeration bubble size to achieve stable performance at minimum energy use (Fane et al., 2005). However, this format can experience operational difficulties due to fibers becoming matted close to the potted ends, and therefore pretreatment and removal of hairs and fibers is essential. Hollow-fiber technology hence requires more instrumentation and control. The submerged MBR formats based on flat sheets have been found to be cost effective for similar types of wastewater, but due to higher air use and lower compactness, tend to be selected for small- to medium-scale duties. The flat-sheet format has operational advantages in terms of plugging and cleaning, and has been used in somewhat more difficult feeds. Flat-sheet systems have the advantage of relatively low manufacturing cost compared to hollow-fiber systems. However, packing density tends to be significantly lower than a hollow-fiber system (e.g., by a factor of 2.5–3 times). Therefore, flat-sheet systems tend to have a cost advantage for smallto medium-scale systems, whereas hollow fiber becomes more attractive at large scale due to the footprint advantage (Pearce, 2008b). The comparison is made more complicated, however, since aeration costs for hollow-fiber systems are often lower. This means that the most cost-effective solution for total treatment costs at medium scale is closely contested, and both approaches are found across the size range due to site-specific circumstances, which could favor either solution. Lesjean et al. (2004), taking into account the current knowledge, anticipated a future market share as follows: for municipal applications, it is expected that the hollow-fiber submerged configuration would be competitive for mediumto large-size plants. For small to medium sizes, flat-sheet technologies would have an advantage. However, in case of larger plants, or a plant refurbishment, the alternative membrane scheme (secondary/tertiary treatment followed by an MF/UF membrane filtration) is very likely to be cost competitive, unless high-cost land has to be purchased for the construction. This multi-barrier scheme will also be easier to control and to optimize because of the disconnection of the treatment steps. It will also be associated with the lowest risk in relation to the membrane operation, as the membranes will be operated under smooth hydrodynamic conditions in terms of particle matter, turbulence, and backwash re´gime. In a recent paper, Lesjean and Huisjes (2008) reiterated this expectation despite the present trend of large MBR plant construction. The airlift format has been developed as a low-energy alternative to the energy-intensive cross-flow sidestream format,

602

Membrane Biological Reactors

which has been used historically for the most difficult feeds. As mentioned earlier, the energy cost of crossflow prohibits it as a treatment option for any application other than small scale or where there is no other treatment option. However, the airlift has very low energy use, and may even undercut the energy requirements of the submerged options, due to the advantage of containment of the feed inside the tubular membrane (Van ‘T Oever, 2005; Futselaar et al., 2007). Since airlift eliminates operator contact and has good operational characteristics, it may as well make a major impact on the MBR market in the long run. Pearce (2008a, 2008b) argued that the airlift format may find applications throughout a broader range than the submerged formats. Figure 15 depicts the concept of airlift MBR.

4.16.5.4 MBR Providers 4.16.5.4.1 Market share of the providers The global market value of MBR is expected to rise up to US$500 million by 2013 from around US$300 million in 2008 (BCC Research, 2008). The MBR market is dominated by three companies, namely GE–Zenon, Kubota, and Mitsubishi Rayon Engineering (MRE). Only GE–Zenon and Kubota have a strong presence in Europe and North America, while MRE have until now mainly focused on sales in Asia. All these companies use submerged formats, with GE–Zenon and MRE Air release

Return to bioreactor

Permeate

Permeate backwash

Air injection

Airlift Feed supply Figure 15 The concept of airlift MBR.

using hollow-fiber membranes, and Kubota, flat-sheet membranes. Another three companies too have an international presence, namely Siemens–Memcor, Norit, and Koch-Puron, but the sales for these three companies makes up a small portion of the worldwide market. Among the latter three, Norit promotes the airlift format. Figure 16(a) shows the worldwide relative market share (in terms of installations numbers) for the three large players (Yang et al., 2006; Pearce, 2008b). The MBR market has several dozen regional or application specialists, quite a few of who use flat-sheet formats as adopted by Kubota: for example, Japan’s Toray and A3 from Germany. In addition to these international companies, there are a further 30 companies in the European Union (EU) market that have either significant regional presence, or an application focus, or a low-level international presence (Lesjean and Huisjes, 2008). Many of these companies are significant in the local markets, but individually, they have a small share of the international market. It is interesting to note that the MBR market has characteristics different from that of the UF/MF market. In UF/MF, there are 10–12 significant players with worldwide presence, with four market leaders, none of who dominate the market. Besides these companies, other regional players are relatively insignificant (Pearce, 2008a, 2008b). Zenon is long established in the market and has been one of the major companies promoting the MBR concept, and the use of PVDF membranes. The North American market is dominated by Zenon (Yang et al., 2006) as shown by the revenue share illustrated in Figure 16(b) and has many more opportunities in the municipal sector than in industry. Zenon leads the European market as well (Figure 16(c)). Kubota was one of the early pioneers of the MBR concept, encouraged by a Japanese Government initiative in the 1980s. They achieved a very large number of installations in small- to medium-scale systems, initially focusing on the residential/ commercial market in Japan and have approached export markets through exclusive partnerships. Kubota has a significantly greater number of plants than Zenon, with a slightly higher proportion of industrial plants. Many of Kubota’s installations in Japan and Korea are for small-scale municipal and domestic applications. Figure 17 shows the market characteristics of the two market leaders, Kubota and Zenon, illustrating the significantly different market strategies with regard to the size of plant targeted. Kubota is the strongest market player for industrial and small-scale municipal applications. MRE is a long-established supplier of MBR, with a very strong position in the relatively mature MBR market in Japan and Korea. There are a large number of references for this technology in Asia, but relatively few installations elsewhere. MRE also has a very large number of installations, with a higher proportion of industrial users, mostly with small flowrates. Koch Membrane Systems (KMS) is a long-established membrane manufacturer and membrane-systems company. In 2004, KMS acquired the MBR start-up company Puron, which had been founded in 2001. They introduced an approach to fiber potting different from that of the other hollow-fiber module providers.

Membrane Biological Reactors

603

15

17

68

(a) Worldwide (relative installation numbers % in 2006)

2 3

6

10

20

33

65

61

(b) North America (revenue % in 2003)

GE−Zenon

Kubota

(c) Europe (installed membrane surface % in 2005)

Mitsubishi−Rayon

Siemens−Memcor

Koch−Puron

Others (N. America: Mitsubishi, Norit; Europe: Norit, Wehrle and other EU and non-EU suppliers) Figure 16 Market share of the suppliers. Data from (a) Yang Q, Chen J, and Zhang F (2006) Membrane fouling control in a submerged membrane; (b) Pearce G (2008 b) Introduction to membranes – MBRs: Manufacturers’ comparison: Part 1. Filtration and Separation 45(3): 28–31; and (c) calculated from Lesjean B and Huisjes EH (2008) Survey of the European MBR market: Trends and perspectives. Desalination 231: 71–81.

Memjet product is characterized by high permeability and packing density, providing a competitive position for capital and operating costs. However, worldwide market share for MemJet MBR is not very significant, since the company tends to focus on selected regional markets (Yang et al., 2006; Pearce, 2008b).

100 Plant capacity

80

60

No. of plants

% 40

20

0 Kubota

GE−Zenon

Figure 17 Relative market share (number of plants and capacity) showing distinct market strategies of the two market leaders.

Memcor have extensive experience in the use of their products in wastewater polishing. Their very fine polypropylene (PP) fibers developed in the 1980s were inexpensive and flexible, but unfortunately had low chlorine tolerance (Judd et al., 2004). In the late 1990s, Memcor developed a PVDF fiber, and now use the PVDF fiber for their MBR product range. The

4.16.5.4.2 Design considerations The design of the reactor (including membrane, baffle, and aerator locations) and the mode of operation of the membrane are key parameters in the optimization of the system. The leading MBR providers propose several MBR designs. In each case, the process proposed is very specific. Not only are the membrane material and configuration used different, but the operating conditions, cleaning protocols, and reactor designs also change from one company to another. For example, the flat-sheet membrane provided by Kubota does not require backwash operation, while hollow-fiber membranes have been especially designed to hydraulically backwash the membrane on a given frequency. The MBR industry first developed in Japan with the use of chlorinated polyethylene (PE) flat-sheet membrane by Kubota, and PE fibers by MRE (Stephenson et al., 2000). The modified PE is characterized by reasonable strength, flexibility, wettability, and resistance to chlorine. Although PE is normally made as an MF membrane, it has relatively low permeability, so process fluxes of PE membranes tend to be at the

Membrane Biological Reactors

Table 9

air-usage efficiency. In addition, the companies using hollow fiber use intermittent aeration, for example, based on a timer in the case of Zenon, or in proportion to flow in the case of Koch–Puron. Memcor introduced a novel cleaning method by using a mixture of air and mixed liquor, instead of using only air bubbles, to scour the membranes. The air bubbles effectively scour the membranes and the semi-crossflow of mixed liquor along the membranes continuously delivers the refresh mixed liquor to the membrane surface, minimizing the solidconcentration polarization at the membrane surface and therefore reducing filtration resistance. These enhancements have significantly reduced air usage and therefore power cost.

4.16.5.4.3 Performance comparison of different providers Few large-scale studies based on comparison of the commercially available MBR systems have been conducted. The city of San Diego, California, and the research consultant, Montgomery Watson Harza, have been evaluating the MBR process through various projects since 1997, including feasibility of using MBRs to produce reclaimed water (Adham and Gagliardo, 1998, 2000), optimization of MBR operation, and parallel comparison and cost estimations of the four leading MBR suppliers (Adham et al., 2004). MBRs were evaluated for their ability to produce high-quality effluent and to operate with minimum fouling. In terms of hydraulic performances, it (8.5−12)

500

400 (17−24) (50−60)

(17−24)

0

Toray

(29)

Norit

Siemens−Memcor

Koch−Puron

100

(30−34)

GE−Zenon

(17−24)

(14−26)

Mitsu. (PVDF)

200

Mitsubishi (PE module)

300

Kubota

low end of the range. Consequently, PE membranes are very cost effective at small scale, but struggle to compete in largerscale systems. In the 1990s, PVDF became established in MBRs through the reinforced capillary fiber in Zenon’s ZW 500 module (Yamato et al., 2006). PVDF has impressive performance in terms of strength and flexibility, but is significantly more expensive as a polymer. Nevertheless, PVDF membranes can achieve substantially higher flux, thereby overcoming price disadvantage. Recently, MRE also developed a PVDF-based membrane system. This membrane, designated as SADF, promises to be very competitive in both capital and operating costs, and despite it having a lower packing density than the PE product, it operates at much higher flux. With several companies now offering PVDF products in both capillary and flat-sheet formats, this is the dominant membrane polymer in the MBR market (Pearce, 2008c, 2008d). The third significantly used membrane polymer in MBR is a reinforced PES, used by Koch–Puron. Although PES is an important polymer in water treatment, in wastewater applications, its lack of flexibility limits the possibility of using air scour. Reinforcing the capillary does allow air scour, but at the expense of permeability. The Puron product uses reinforced PES rather than the PVDF, favored by its rivals. However, its main distinguishing feature is that the membrane fibers are potted at only one end. This overcomes the problem of fouling below the potting interface by hairs and fibers, which is a problem for the other hollow-fiber technologies (Vilim et al., 2009). Norit is the one major MBR company that offers a system based on a sidestream format with tubular membranes rather than a submerged format. Crossflow is only used for smallscale applications, with feeds that are difficult to treat, whereas airlift is cost effective for larger-scale municipal applications (Futselaar et al., 2007). Table 9 summarizes the specifications of the membranes used by different suppliers and Figure 18 compares the packing density and applicable flux of the membranes. Each of the suppliers makes regular improvements in air usage, since this has an important impact on total water cost. For example, the flat-sheet suppliers now use 1.5-m panels, which reduce air flow by up to 30% compared to the original 1 m panel (Pearce, 2008c, 2008d). In addition, they also use double-deck stacks wherever possible, which further improves

Membrane packing density, m2 m−3

604

Figure 18 Packing density (bar chart, m2 m3) and flux (values within parentheses, l m2 h1) of membranes from different suppliers.

MBR supplier specificationsa

Company

Membrane material

Pore size, mm

Membrane format

Fiber/tube dia (id,od),mm

pH tolerance

Kubota Mitsubishi Mitsubishi GE–Zenon Koch–Puron Siemens–Memcor Noritb Toray

Cl2 PE PE PVDF PVDF PES PVDF PVDF PVDF

0.4 0.4 0.4 0.04 0.05 0.04 0.03 0.08

FS HF HF HF HF HF TUB FS

– 0.37, 0.54 11, 2.8 0.8, 1.9 –, 2.6 –, 1.3 –, 5.2 or 8.0 –

1–13 1–13 1–10 2–10.5 2–12 2–10.5 1–11 1–11

a

All the membranes have moderate hydrophilicity and high chlorine resistance. All the companies except Norit use submerged format; Norit supplies airlift sidestream MBRs. FS, flat sheet; HF, hollow fiber; TUB, tubular. b

Membrane Biological Reactors

was shown that all four processes were able to cope with flux rates exceeding 33 l m2 h1 and HRTs as low as 2 h. A 6-year development program has also been initiated for the introduction of MBR technology in the Netherlands market. Started in 2000, a comparative study of four 750 m3 d1 MBRs carried out by DHV water has been reported (van der Roest et al., 2002b). Three MBR plants, treating a design flow of 300 m3 d1 each, have been operated in parallel during 2003 and 2004 in Singapore (Le-Clech et al., 2006). A 4-year study, started in 2001, comparing the performance of Mitsubishi, Kubota, and Zenon MBR was conducted by the Swiss Federal Institute of Aquatic Science and Technology (EAWAG) (Judd, 2006). The Zenon MBR exhibited the most stable performance in the study. Although these studies have been conducted with the MBR systems running in parallel (with the same influent water), the MBR maximum flux, operating conditions and general design applied were those recommended by the suppliers, and therefore somewhat different for each system. This makes it difficult to make a fair comparison. Therefore, it is not possible to classify the MBRs as a function of their relative hydraulic performances, which need to be considered along with the cleaning protocols applied to each system. Mansell et al. (2004) performed measurements in which MS2 coliphage were seeded to the influent of a Kubota MBR (characteristic pore size 0.4 mm) and a Zenon MBR (characteristic pore size 0.04 mm). Permeate concentrations showed a log removal range of 3.2–7.4 for the Kubota installation and 5.32–7.5 for the Zenon installation. All of the heavy metals detected in the influent were removed to levels below detection limit, as well as the VOCs that were measured.

4.16.5.5 Standardization of Design and Performance-Evaluation Method The MBR market is very fragmented and exhibits many MBR filtration products with diverse geometries, module capacities, and operational modes (De Wilde et al., 2008; Lesjean and Huisjes, 2008). Although this situation promotes a competitive market, it is detrimental for the acceptance of the technology as a state-of-the-art process, and raises concern with potential clients or end users. From the point of view of the MBR operators, the possibility of interchanging filtration modules of different companies/suppliers would facilitate the replacement of the modules at the end of their life, and would reduce the risk of a supplier withdrawing from the market or releasing a new series of the product. In addition, the stakeholders in the industry employ various methods of membrane characterization and performance evaluation. This creates confusion among the users and prohibits fair comparison. Based on an extensive survey of the MBR industry, De Wilde et al. (2008) provided an overview of the market interests/expectations and technical potential of going through a standardization process of the SMBR technology in Europe. Due to the predominance of submerged filtration systems in municipal applications, the study focused only on this configuration. Two different aspects of standardization were considered:

•

standardization of MBR filtration modules toward interchangeable modules in MBRs and

•

605

standardization of MBR acceptance and monitoring test methods toward uniform quality-assessment methods of MBR filtration systems.

4.16.5.5.1 Standardization of MBR filtration systems In relation to the market expectations, about 20 potential technological, financial, economical, or environmental benefits/opportunities and drawbacks/threats of MBR module standardization for suppliers and operators were identified and mapped. It appeared that the number of advantages and disadvantages was quite balanced for both sides of the market, the main advantage perceived by the industry being that standardization should contribute to the growth of the MBR market. Other main advantages/opportunities are avoidance of vendor lock-in, price decrease, and increased trust and acceptance. Main disadvantages/threats for the end users are overdimensioning of civil constructions and supplementary works and costs to the peripherals during replacement. Main disadvantages for the module suppliers seem to be the higher competition, lower profit margins, and a limitation for innovative module producers to enter the market. From the technical point of view, the analysis showed that a standardization process common for both flat-sheet and hollow-fiber membranes/modules would not be realistic. In order to achieve interchangeability of filtration modules, not only should the prospect of pure dimensional standards for the module be considered, but also the design and mode of operation of the peripheral components, such as the filtration tank, pumps, blowers, and pretreatment should be borne in mind. More than 30 technical factors hampering or interfering with a standardization process were identified and quantified, and their relative potential for affecting the possible outcome was evaluated. For instance, four factors were grouped as the extremely high hindering factors: module dimensions, filtration tank dimensions, specific permeate production capacity, and specific coarse-bubble aeration demand. These factors are mainly the result of a completely different geometry and design of the filtration module and discussions for the standardization of MBR filtration systems should in essence focus on these factors. For each category, more or less the same number of obstacles lies ahead. Nevertheless, the nature of some of these obstacles or points of attention can be different. Some factors are specifically important for FS modules (e.g., flushing of air-supply pipes and design of a permeatecollection tank), and others for HF modules (e.g., type of prescreening, whether gravity filtration or any other type).

4.16.5.5.2 Standardization of MBR characterization methods The survey conducted by De Wilde et al. (2008) also revealed the respondents’ consensus in general on the positive impact of harmonization of membrane-acceptance tests at module delivery and monitoring methods on municipal MBR market growth. Some important parameters, for which a common definition and measurement protocol could be helpful, are mentioned below:

•

clearly defined and harmonized parameters to monitor membrane fouling, integrity, and aging;

606

• • • • • • • • •

Membrane Biological Reactors

a common definition of membrane lifetime for the guarantee clause; determination/definition of flux (operation and nominal design); common definition for sustainable peak hydraulic load; harmonized tests to check membrane performances over a defined period and under specific conditions; characterization method for membrane acceptance at module delivery; minimum requirements and technical methods to check membrane performance at plant commissioning; monitoring methods of normalized permeability in clear water, permeability in sludge, transmembrane pressure, and fouling rate; monitoring methods of sustainable flux and maximum flux; and operating conditions (biology and filtration systems) for warranty clauses.

It is interesting to note that, most of the newcomers in the market are developing their systems so that they can easily replace the products of the two main suppliers (Zenon–GE and Kubota). A standardization process driven by the end users could accelerate this evolution and contribute to the market development (Lesjean and Huisjes, 2008). Pearce (2008a, 2008b, 2008c, 2008d) also pointed out that, although the dimensions of the relatively newer Puron products are not identical to Zenon’s ZW 500d or MRE’s SADF, the elements are similar, and cassettes made from the elements could be used interchangeably. This begins to introduce retrofit possibilities into what hasuntil now been a fragmented market with no standardization.

4.16.6 Future Vision In addition to the alleviation of the technology bottlenecks illustrated in this chapter, a radical shift from the conventional concept of advanced wastewater treatment is deemed

Urine separation is also worthwhile to be considered

4.16.7 Conclusion MBR is a physicobiological hybrid process. The membrane provides a physical barrier for hygienically safe and clean water with the help of microbial–ecological treatment that can achieve good public acceptance. It is also well recognized by the experts that the clear membrane permeate makes post treatment easy; then, a variety of hybrid systems having the MBR as the core can be considered depending on the specific quality requirements of the reclaimed water . These advantages

A large amount of diluted organic wastewater (graywater)

To co-generation system A small amount of highstrength organic waste kitchen waste disposer-wastewater and toilet flushing)

imperative. In the context of sustainable water system, the advanced treatment must couple technologies to produce water of the required quality and realize material conversion from waste as well. The required quality does not always mean high quality. The quality comes from necessity. Membrane technology has the potential to be an on-demand quality provider just by separation. The conversion mainly comes from the biological reaction in the MBR. Three aspects of a sustainable society, namely, the low carbon society, sound material cycle society, and ecological society, are notable. From the point of view of sustainable water system, the advanced wastewater-treatment processes can be classified into the categories of energy saving (or productive), material productive, and ecologically oriented. The MBR technology might match more with the first two. However, present MBR technologies are still large energy consumers. Next-generation MBRs need to be developed to reduce the significant aeration requirement (by compact module design and sludge-concentration control techniques) and recover energy (e.g., by adding other organic wastes and combining anaerobic digestion for methane recovery). In line with the proposed definition of advanced treatment, the notion needs to be changed from organic wastewater treatment to water/biomass production by developing next-generation MBRs where the membrane acts as a separator of water and biomass and biomass is utilized for energy production. The concept is illustrated in Figures 19 and 20.

Anaerobic pretreatment Pretreatment Methane production

Biomass production from liquid organic waste (*)

Aerobic MBR

(*)

(A very small amount of residue) • Renewable energy utilization • IT-based maintenance service system • User participation in monitoring

(*) N,P recovery option

Figure 19 Next-generation MBR system: anaerobic combination for on-site small-scale advanced treatment.

Safe effluent

Membrane Biological Reactors

W.W.

Solid−liquid Solid-liquid separation

Solid concentration/ concentration/ Anoxic anoxic reaction reaction

Biosorption/ membrane separation/ aerobic reaction

607

Safe reclaimed water

Energy and/or material recovery process Other than biogas production, physicochemical treatments are also candidates for energy recovery, for example, supercritical water gasification of sludge−water mixture where the biomass sludge is utilized as energy source to produce hydrogen from water molecules (coupling clean energy production). Figure 20 Next-generation MBR system: renovation of existing wastewater-treatment plants.

make MBR a good device in water reclamation and/or advanced wastewater treatment. The continued push toward stricter discharge standards, increased requirement for water reuse, and greater than before urbanization and land limitations fuel the use of MBRs. However, there is room for improvement to utilize the potential of the MBR fully. The challenges will center on energy saving, ease of operation, simplified membrane cleaning and replacement strategies, and peak-flow management. The international adventure on R&D of MBR technologies continues.

References Achilli A, Cath TY, Marchand EA, and Childress AE (2009) The forward osmosis membrane bioreactor: A low fouling alternative to MBR processes. Desalination 239: 10--21. Adham S, DeCarolis J, and Pearce W (2004) Optimization of various MBR systems for water reclamation – phase III. Agreement No. 01-FC-81-0736, Program Final Report No. 103, April 2004. Denver, CO: US Department of the Interior, Bureau of Reclamation, Denver Office. Adham S and Gagliardo P (1998) Membrane bioreactors for water repurification – phase I. Desalination Research and Development Program Report No. 34, Project No. 1425-97-FC-81-3 0006. Denver, CO: US Department of the Interior, Bureau of Reclamation, Denver Office. Adham S, Gagliardo P, Boulos L, Oppenheimer J, and Trussel R (2001) Feasibility of the membrane bioreactor process for water reclamation. Water Science and Technology 43(10): 203--209. Adham S, Mirlo R, and Gagliardo P (2000) Membrane bioreactors for water reclamation – phase II. Desalination Research and Development Program Report No. 60, Project No. 98-FC-81-0031. Denver, CO: US Department of the Interior, Bureau of Reclamation, Denver Office. Al-Amoudi A and Lovitt RW (2007) Fouling strategies and the cleaning system of NF membranes and factors affecting cleaning efficiency. Journal of Membrane Science 303: 4--28. Alnaizy R and Sarin V (2009) Performance assessment of MBR applications for metropolitan wastewater treatment in the United Arab Emirates. In: Lesjean B and Leiknes T (eds.) Final MBR-Network Workshop, pp. 85–86. Berlin, Germany, 31 March–1 April. Berlin: MBR-Network Workshop. Arika M, Kobayashi H, and Hihara H (1966) Pilot plant test of an activated sludge ultrafiltration combined process for domestic wastewater reclamation. Desalination 48: 299--319. Asatekin A, Menniti S, Kang M, Elimelech M, Morgenroth E, and Mayes AM (2006) Antifouling nanofiltration membranes for membrane bioreactors from selfassembling graft copolymers. Journal of Membrane Science 285: 81--89. Aya H (1994) Modular membranes for self-contained reuse systems. Water Quality International 4: 21--22.

Bae TH, Kim IC, and Tak TM (2006) Preparation and characterization of foulingresistant TiO2 self-assembled nanocomposite membranes. Journal of Membrane Science 275: 1--5. Bae TH and Tak TM (2005) Interpretation of fouling characteristics of ultrafiltration membranes during the filtration of membrane bioreactor mixed liquor. Journal of Membrane Science 264: 151--160. Bacchin P, Aimar P, and Field RW (2006) Critical and sustainable fluxes: Theory, experiments and applications. Journal of Membrane Science 281: 42--69. Bacchin P, Si-Hassen D, Starov V, Clifton MJ, and Aimar P (2002) A unifying model for concentration polarization, gel layer formation and particle deposition in cross-flow membrane filtration of colloidal suspensions. Chemical Engineering Science 57: 77--91. BCC Research (2008) Membrane bioreactors: Global markets. Report Code MST047 B, Report Category – Membrane & Separation Technology. Beltfort G, Davis RH, and Zydney AL (1994) The behaviour of suspension and macromolecular solutions in crossflow microfiltration. Journal of Membrane Science 96: 1--58. Benefield LD and Randall CW (1980) Biological Process Design for Wastewater Treatment. Englewood Cliffs, NJ: Prentice Hall. Benitez J, Rodriguez A, and Malavar R (1995) Stabilization and dewatering of wastewater using hollow fiber membranes. Water Research 29: 2281--2286. Bixio D, Thoeye C, De Koning J, et al. (2006) Wastewater reuse in Europe. Desalination 187: 89--101. Bixio D, Thoeye C, Wintgens T, et al. (2008) Water reclamation and reuse: Implementation and management issues. Desalination 218: 13--23. Bouhabila EH, Ben Aim R, and Buisson H (2001) Fouling characterization in membrane bioreactors. Separation and Purification Technology 22: 123--132. Brepols C, Dorgeloh E, Frechen FB, et al. (2008) Upgrading and retrofitting of municipal wastewater treatment plants by means of membrane bioreactor (retrofit aoMBR) technology. Desalination 231: 20--26. Brindle K and Stephenson T (1996) The application of membrane biological reactors for the treatment of wastewaters. Biotechnology and Bioengineering 49: 601--610. Brookes A, Jefferson B, Guglielmi G, and Judd SJ (2006) Sustainable flux fouling in a membrane bioreactor: Impact of flux and MLSS. Separation Science and Technology 41: 1279--1291. Brookes PR and Livingston AG (1993) Point source detoxification of an industrially produced 3,4-dichloroaniline-manufacture wastewater using a membrane bioreactor. Applied Microbiology and Biotechnology 39: 764--771. Brookes PR and Livingston AG (1994) Biological detoxification of a 3chloronitrobenzene manufacture waste-water in an extractive membrane bioreactor. Water Research 28: 1347--1354. Brow PJ (2007) Ballyclare MBR WwTW: Cost & challenges of compliance with stringent standards. In: Proceedings of the 4th International Water Association Conference on Membranes for Water and Wastewater Treatment, CD ISBN: 978-186194-127-5. Harrogate, UK, 15–17 May. London: IWA Publishing. Cabassud C, Masse A, Espinosa-Bouchot M, and Sperandio M (2004) Submerged membrane bioreactors: Interactions between membrane filtration and biological activity. In: Proceedings of the Water Environment-Membrane Technology Conference. Seoul, Korea, 7–10 June. London: IWA Publishing. Cao JH, Zhu BK, Lu H, and Xu YY (2005) Study on polypropylene hollow fiber based recirculated membrane bioreactor for treatment of municipal wastewater. Desalination 183: 431--438.

608

Membrane Biological Reactors

Cha GC, Jeong TY, Yoo IK, and Kim DJ (2004) Kinetics characteristics of SMP and ECP in relation to loading rate in a MBR process. In: Proceedings of the Water Environment – Membrane Technology Conference. Seoul, Korea, 7–10 June. London: IWA Publishing. Chae SR, Ahn YT, Kang ST, and Shin HS (2006a) Mitigated membrane fouling in a vertical submerged membrane bioreactor (VSMBR). Journal of Membrane Science 280: 572--581. Chae SR, Kang ST, Watanabe Y, and Shin HS (2006b) Development of an innovative vertical submerged membrane bioreactor (VSMBR) for simultaneous removal of organic matter and nutrients. Water Research 40: 2161--2167. Chang IS and Lee CH (1998) Membrane filtration characteristics in membrane coupled activated sludge system – the effect of physiological states of activated sludge on membrane fouling. Desalination 120: 221--233. Chang IS, Lee CH, and Ahn KH (1999) Membrane filtration characteristics in membrane-coupled activated sludge system: The effect of floc structure on membrane fouling. Separation Science and Technology 34: 1743--1758. Chang IS, Gander M, Jefferson B, and Judd SJ (2001) Low-cost membranes for use in a submerged MBR. Transactions IChemE 79(B): 183--188. Chang IS and Judd SJ (2002) Air sparging of a submerged MBR for municipal wastewater treatment. Process Biochemistry 37: 915--920. Chang IS and Kim SN (2005) Wastewater treatment using membrane filtration – effect of biosolids concentration on cake resistance. Process Biochemistry 40: 1307--1314. Chang IS, Le-Clech P, Jefferson B, and Judd S (2002a) Membrane fouling in membrane bioreactors for wastewater treatment. Journal of Environmental Engineering ASCE 128: 1018--1029. Chang S and Fane AG (2001) The effect of fibre diameter on filtration and flux distribution-relevance to submerged hollow fibre modules. Journal of Membrane Science 184: 221--231. Chang S and Fane AG (2002) Filtration of biomass with laboratory-scale submerged hollow fibre modules – effect of operating conditions and module configuration. Journal of Chemical Technology and Biotechnology 77: 1030--1038. Chang S, Fane AG, and Vigneswaran S (2002b) Modeling and optimizing submerged hollow fiber membrane modules. AIChE Journal 48: 2203--2212. Chaize S and Huyard A (1991) Membrane bioreactors on domestic wastewater treatment sludge production and modelling approach. Water Science and Technology 23: 1591--1600. Charcosset C (2006) Membrane processes in biotechnology: An overview. Biotechnology Advances 24: 482--492. Chen CL, Liu WT, Chong ML, et al. (2004) Community structure of microbial biofilms associated with membrane-based water purification processes as revealed using a polyphasic approach. Applied Microbiology and Biotechnology 63: 466--473. Chen JP, Yang CZ, Zhou JH, and Wang XY (2007) Study of the influence of the electric field on membrane flux of a new type of membrane bioreactor. Chemical Engineering Journal 128: 177--180. Chen TK, Chen JN, Ni CH, Lin GT, and Chang CY (2003) Application of a membrane bioreactor system for opto-electronic industrial wastewater treatment – a pilot study. Water Science and Technology 48: 195--202. Cheryan M (1998) Ultrafiltration and Microfiltration Handbook. Lancaster: Technomic Publishing. Chiemchaisri C, Yamamoto K, and Vigneswaran S (1993) Household membrane bioreactor in domestic wastewater treatment. Water Science and Technology 27(1): 171--178. Childress AE, Daugherty J, Chen C, Le-Clech P, and Leslie GL (2005) Hollow fiber membrane integrity: Mechanics and modeling. Desalination 180: 5--14. Cho BD and Fane AG (2002) Fouling transients in nominally sub-critical flux operation of a membrane bioreactor. Journal of Membrane Science 209: 391--403. Cho J, Ahn KH, Seo Y, and Lee Y (2003) Modification of ASM No.1 for a submerged membrane bioreactor system: Including the effects of soluble microbial products on membrane fouling. Water Science and Technology 47(12): 177--181. Cho JW, Song KG, Lee SH, and Ahn KH (2005) Sequencing anoxic/anaerobic membrane bioreactor (SAM) pilot plant for advanced wastewater treatment. Desalination 178: 219--225. Choi H, Zhang K, Dionysiou DD, Oerther DB, and Sorial GA (2005) Effect of permeate flux and tangential flow on membrane fouling for wastewater treatment. Separation and Purification Technology 45: 68--78. Choi JH, Dockko S, Fukushi K, and Yamamoto K (2002) A novel application of a submerged nanofiltration membrane bioreactor (NF MBR) for wastewater treatment. Desalination 146: 413--420. Choi JH, Fukushi K, and Yamamoto K (2007) A submerged nanofiltration membrane bioreactor for domestic wastewater treatment: The performance of cellulose acetate nanofiltration membranes for long-term operation. Separation and Purification Technology 52: 470--477.

Cicek N, Franco JP, Suidan MT, and Urbain V (1999a) Effect of phosphorus on operation and characteristics of MBR. Journal of Environmental Engineering 125: 738--746. Cicek N, Franco JP, Suidan MT, Urbain V, and Manem J (1999b) Characterization and comparison of a membrane bioreactor and a conventional activated sludge system in the treatment of wastewater containing high-molecular weight compounds. Water Environmental Research 71: 64--70. Cicek N, Macomber J, Davel J, Suidan MT, Audic J, and Genestet P (2001) Effect of solids retention time on the performance and biological characteristics of a membrane bioreactor. Water Science and Technology 43: 43--50. Cicek N, Suidan MT, Ginestet P, and Audic JM (2002) Impact of soluble organic compounds on permeate flux in an aerobic membrane bioreactor. Environmental Technology 24: 249--256. Clapp LW, Regan JM, Ali F, Newman JD, Park JK, and Noguera DR (1999) Activity, structure, and stratification of membrane-attached methonotrophic biofilms cometabolically degrading trichloroethylene. Water Science and Technology 39(7): 153--161. Cornel P and Krause S (2003) State of the art of MBRs in Europe. In: Proceedings of the Conference on Application and Perspective of MBRs in Wastewater Treatment and Reuse. Cremona, Italy, 28–29 April. Cornel P and Krause S (2006) Membrane bioreactors in industrial wastewater treatment – European experience, examples and trends. Water Science and Technology 53(3): 37--44. Cornelissen ER, Harmsen D, de Korte KF, et al. (2008) Membrane fouling and process performance of forward osmosis membranes on activated sludge. Journal of Membrane Science 319: 158--168. Cote P (2002) Inverted Air Box Aerator and Aeration Method for Immersed Membrane. US Pat. 6,863,823. Cote P, Bersillon JL, and Faup G (1988) Bubble free aeration using membranes: Process analysis. Journal of Water Pollution Control Federation 60: 1986--1992. Cote P, Buisson H, Pound C, and Arakaki G (1997) Immersed membrane activated sludge for the reuse of municipal wastewater. Desalination 113: 189--196. Cote P, Masini M, and Mourato D (2004) Comparison of membrane options for water reuse and reclamation. Desalination 167: 1--11. Cui ZF, Chang S, and Fane AG (2003) The use of gas bubbling to enhance membrane processes. Journal of Membrane Science 221: 1--35. Daubert I, Mercier-Bonin M, Maranges C, Goma G, Fonade C, and Lafforgue C (2003) Why and how membrane bioreactors with unsteady filtration conditions can improve the efficiency of biological processes. Advances in Membrane Technology 984: 420--435. Davies WJ, Le MS, and Heath CR (1998) Intensified activated sludge process with submerged membrane microfiltration. Water Science and Technology 38(5): 421--428. Debus O and Wanner O (1992) Degradation of xylene by a biofilm growing on a gaspermeable membrane. Water Science and Technology 26(3–4): 607--617. Defrance L and Jaffrin MY (1999) Reversibility of fouling formed in activated sludge filtration. Journal of Membrane Science 157: 73--84. Defrance L, Jaffrin MY, Gupta B, Paullier P, and Geaugey V (2000) Contribution of various constituents of activated sludge to membrane bioreactor fouling. Bioresource Technology 73: 105--112. De Silva DGV, Urbain V, Abeysinghe DH, and Rittmann B (1998) Advanced analysis of membrane-bioreactor performance with aerobic–anoxic cycling. Water Science and Technology 38(4–5): 505--512. De Wilde W, Geenens D, and Thoeye C (2003) Do we really want to build MBRs for domestic wastewater treatment? In: Proceedings MBR 4. Cranfield, April 9. De Wilde W, Richard M, Lesjean B, and Tazi-Pain A (2008) Towards standardisation of the MBR technology? Desalination 231: 156--165. De Wilde W, Moons K, Bixio D, Thoeye C, and De Gueldre G (2009) Technical feasibility and optimal control strategy of dual (hybrid) MBR-CAS concepts for plant refurbishment. In: Lesjean B and Leiknes T (eds.) Final MBR-Network Workshop, pp. 41–42. Berlin, Germany, 31 March–1 April. Di Bella G, Mannina G, and Viviani G (2008) An integrated model for physical– biological wastewater organic removal in a submerged membrane bioreactor: Model development and parameter estimation. Journal of Membrane Science 322: 1--12. Dossantos LMF and Livingston AG (1995) Novel membrane bioreactor for detoxification of voc wastewaters – biodegradation of 1,2-dichloroethane. Water Research 29: 179--194. Doyen W, Mues W, Molenberghs B, and Cobben B (2010) Spacer fabric supported flatsheet membranes: A new era of flat-sheet membrane technology. Desalination 250: 1078--1082. Drews A, Vocks M, Iversen V, Lesjean B, and Kraume M (2005a) Influence of unsteady membrane bioreactor operation on EPS formation and filtration resistance. In:

Membrane Biological Reactors

Proceedings of the International Congress on Membranes and Membrane Processes (ICOM). Seoul, Korea, 21–26 August. Drews A, Evenblij H, and Rosenberger S (2005b) Potential and drawbacks of microbiology–membrane interaction in membrane bioreactors. Environmental Progress 24: 426--433. Drews A, Lee CH, and Kraume M (2006) Membrane fouling – a review on the role of EPS. Desalination 200: 186--188. Dufresne R, Lebrun RE, and Lavallee HC (1997) Comparative study on fluxes and performances during paper mill wastewater treatment with membrane bioreactor. Canadian Journal of Chemical Engineering 75: 95--103. Enegess D, Togna AP, and Sutton PM (2003) Membrane separation applications to biosystems for wastewater treatment. Filtration and Separation 40: 14--17. Engelhardt N, Firk W, and Warnken W (1998) Integration of membrane filtration into the activated sludge process in municipal wastewater treatment. Water Science and Technology 38(4): 429--436. Evenblij H and Van der Graaf J (2004) Occurrence of EPS in activated sludge from a membrane bioreactor treating municipal wastewater. Water Science and Technology 50: 293--300. Evenblij H, Geilvoet S, Van der Graaf J, and Van der Roest HF (2005b) Filtration characterisation for assessing MBR performance: Three cases compared. Desalination 178: 115--124. Evenblij H, Verrecht B, Van der Graaf JHJM, and Van der Bruggen B (2005a) Manipulating filterability of MBR activated sludge by pulsed substrate addition. Desalination 178: 193--201. Fan XJ, Urbain V, Qian Y, and Manem J (1996) Nitrification and mass balance with a membrane bioreactor for municipal wastewater treatment. Water Science and Technology 34: 129--136. Fane AG, Chang S, and Chardon E (2002) Submerged hollow fiber membrane moduledesign options and operational considerations. Desalination 146: 231--236. Fane AG and Fane SA (2005) The role of membrane technology in sustainable decentralized wastewater systems. Water Science and Technology 51: 317--325. Fane AG, Yeo A, Law A, Parameshwaran K, Wicaksana F, and Chen V (2005) Low pressure membrane processes – doing more with less energy. Desalination 185: 1585--1591. Fang HHP and Shi X (2005) Pore fouling of microfiltration membranes by activated sludge. Journal of Membrane Science 264: 161--166. Fang HHP, Shi X, and Zhang T (2006) Effect of activated carbon on fouling of activated sludge filtration. Desalination 189: 193--199. Fawehinmi F, Lens P, Stephenson T, Rogalla F, and Jefferson B (2004) The influence of operating conditions on EPS, SMP and bio-fouling in anaerobic MBR. In: Proceedings of the Water Environment-Membrane Technology Conference. Seoul, Korea, 7–10 June. London: IWA Publishing. Flemming HC and Wingender J (2001) Relevance of microbial extracellular polymeric substances (EPSs). Part I. Structural and ecological aspects. Water Science and Technology 43: 1--8. Frechen FB, Schier W, and Wett M (2006) Pre-treatment of municipal MBR applications in Germany – current status and treatment efficiency. Water Practice and Technology 1(3). doi:10.2166/WPT.2006057 Friedler E and Hadari M (2006) Economic feasibility of on-site greywater reuse in multi-storey buildings. Desalination 190: 221--234. Field RW, Wu D, and Howell JA (1995) Critical flux concept for microfiltration fouling. Journal of Membrane Science 100: 259--272. Fuchs W, Braun R, and Theiss M (2005) Influence of various wastewater parameters on the fouling capacity during membrane filtration. In: Proceedings of the International Congress on Membranes and Membrane Processes (ICOM). Seoul, Korea, 7–10 June. Furumai H and Rittmann BE (1992) Advanced modeling of mixed populations of heterotrophs and nitrifiers considering the formation and exchange of soluble microbial products. Water Science and Technology 26(3–4): 493--502. Futselaar H, Schonewille H, De Ventec D, and Broensa L (2007) NORIT AirLift MBR: Side-stream system for municipal waste water treatment. Desalination 204: 1--7. Galinha CF, Carvalho G, Silva AF, et al. (2009). New developments in membrane bioreactors characterization and monitoring. In: Lesjean B and Leiknes T (eds.) Final MBR-Network Workshop, pp. 17–18. Berlin, Germany, 31 March–1 April. Gander MA, Jefferson B, and Judd SJ (2000) Membrane bioreactors for use in small wastewater treatment plants: Membrane materials and effluent quality. Water Science and Technology 41: 205--211. Gao M, Yang M, Li H, Wang Y, and Pan F (2004) Nitrification and sludge characteristics in a submerged membrane bioreactor on synthetic inorganic wastewater. Desalination 170: 177--185. Genkin G, Waite TD, Fane TG, and Chang S (2005) The effct of axial vibrations on the filtration performance of submerged hollow fibre membranes. In: Proceedings of

609

the International Congress on Membranes and Membrane Processes (ICOM). Seoul, Korea, 2005, 21–26 August. Germain E and Stephenson T (2005) Biomass characteristics, aeration and oxygen transfer in membrane bioreactors: Their interrelations explained by a review of aerobic biological processes. Reviews in Environmental Science and Biotechnology 4: 223--233. Ghaffour N, Jassim R, and Khir T (2004) Flux enhancement by using helical baffles in ultrafiltration of suspended solids. Desalination 167: 201--207. Ghosh R (2006) Enhancement of membrane permeability by gas-sparging in submerged hollow fibre ultrafiltration of macromolecular solutions: Role of module design. Journal of Membrane Science 274: 73--82. Ghyoot W and Verstraete W (1999) Reduced sludge production in a two stage membrane-assisted bioreactor. Water Research 34: 205--215. Gijsbertsen-Abrahamse AJ, Cornelissen ER, and Hofman JAMH (2006) Fiber failure frequency and causes of hollow fiber integrity loss. Desalination 194: 251--258. Gori R, Lubello C, and Caffaz S (2004) Sviluppo di un modello matematico per la simulazione di un reattore MBR applicato al trattamento di acque reflue industriali. In: Proceedings of ANDIS. Taormina, Italy, June 2004. Grethlein HE (1978) Anaerobic digestion and membrane separation of domestic wastewater. Journal of Water Pollution Control Federation 50: 754--763. Guglielmi G, Saroj DP, Chiarani D, and Andreottola G (2007) Subcritical fouling in a membrane bioreactor for municipal wastewater treatment: Experimental investigation and mathematical modelling. Water Research 41: 3903--3914. Gu¨nder B and Krauth KH (2000) Excess sludge production and Oxygen transfer in MBR. In: Proceedings of ATSV Conference. Netherlands, 8–9 February. Guo WS, Vigneswaran S, and Ngo HH (2004) A rational approach in controlling membrane fouling problems: Pretreatments to a submerged hollow fiber membrane system. In: Proceedings of the Water Environment – Membrane Technology Conference. Seoul, Korea, 7–10 June. London: IWA Publishing. Hai FI (2007) Enhancement of Dye Degradation and Mitigation of Membrane Fouling in a Membrane-Coupled Fungi Reactor Treating Textile Wastewater. PhD Thesis, University of Tokyo. Hai FI, Yamamoto K, and Fukushi K (2005) Different fouling modes of submerged hollow-fiber and flat-sheet membranes induced by high strength wastewater with concurrent biofouling. Desalination 180: 89--97. Hai FI, Yamamoto K, and Fukushi K (2006a) Development of a submerged membrane fungi reactor for textile wastewater treatment. Desalination 192: 315--322. Hai FI, Yamamoto K, and Fukushi K (2006b) Performance of newly developed hollow fiber module with spacer in integrated anaerobic–aerobic fungi reactor treating textile wastewater. Desalination 199: 305--307. Hai FI, Yamamoto K, Fukushi K, and Nakajima F (2008a) Fouling resistant compact hollow-fiber module with spacer for submerged membrane bioreactor treating high strength industrial wastewater. Journal of Membrane Science 317: 34--42. Hai FI, Yamamoto K, Nakajima F, and Fukushi K (2007) Textile effluent treatment by fungi MBR with sludge bed /GAC adsorption: Long-term performance of compact hollow-fiber module and overall treatment. In: Proceedings of the 4th International Water Association Conference on Membranes for Water and Wastewater Treatment, CD ISBN: 978–1-86194-127–5. Harrogate, UK, 15–17 May. London: IWA Publishing. Hai FI, Yamamoto K, Nakajima F, and Fukushi K (2008b) Removal of structurally different dyes in submerged membrane fungi reactor – biosorption/PACadsorption, membrane retention and biodegradation. Journal of Membrane Science 325: 395--403. Hardt FW, Clesceri LS, Nemerow NL, and Washington DR (1970) Solid separation by ultrafiltration for concentrated activated sludge. Journal of Water Pollution Control Federation 42(12): 2135--2148. He Y, Xu P, Li C, and Zhang B (2005) High-concentration food wastewater treatment by an anaerobic membrane bioreactor. Water Research 39: 4110--4118. Heijnen M, Winkler R, Vogg M, Roeder G, and Berg P (2009) Development of a novel fiber sheet membrane for MBR: The FiSh. In: Lesjean B and Leiknes T (eds.) Final MBR-Network Workshop, pp. 53–54. Berlin, Germany, 31 March–1 April. Henze M, Gujer W, Mino T, and van Loosdrecht M (eds.) (2000) Scientific and Technical Report Series, Vol. 9: Activated Sludge Models: ASM1, ASM2, ASM2d, and ASM3 London: IWA. Hernandez Rojas ME, Van Kaam R, Schetrite S, and Albasi C (2005) Role and variations of supernatant compounds in submerged membrane bioreactor fouling. Desalination 179: 95--107. Ho CC and Zydney AL (2006) Overview of fouling phenomena and modeling approaches for membrane bioreactors. Separation Science and Technology 41: 1231--1251. Ho WSW and Sirkar KK (eds.) (1992) Membrane Handbook. New York: Kluwer.

610

Membrane Biological Reactors

Holbrook RD, Higgins MJ, Murthy SN, et al. (2004) Effect of alum addition on the performance of submerged membranes for wastewater treatment. Water Environmental Research 76: 2699--2702. Hong SP, Bae TH, Tak TM, Hong S, and Randall A (2002) Fouling control in activated sludge submerged hollow fiber membrane bioreactors. Desalination 143: 219--228. Howell JA (2002) Future research and developments in the membrane field. Desalination 144: 127--131. Howell JA (2004) Future of membranes and membrane reactors in green technologies and for water reuse. Desalination 162: 1--11. Huang X and Wu J (2008) Improvement of membrane filterability of the mixed liquor in a membrane bioreactor by ozonation. Journal of Membrane Science 318: 210--216. Hughes D, Tirlapur UK, Field R, and Cui ZF (2006) In situ 3D characterization of membrane fouling by yeast suspensions using two-photon femtosecond near infrared non-linear optical imaging. Journal of Membrane Science 280: 124--133. Hughes DJ, Cui ZF, Field RW, and Tirlapur UK (2007) Membrane fouling by cell– protein mixtures: In situ characterisation using multi-photon microscopy. Biotechnology and Bioengineering 96: 1083--1091. Itonaga T, Kimura K, and Watanabe Y (2004) Influence of suspension viscosity and colloidal particles on permeability of membrane used in membrane bioreactor (MBR). Water Science and Technology 50: 301--309. Itonga T and Watanabe Y (2004) Performance of membrane bioreactor combined with pre-coagulation/sedimentation. Water Science and Technology: Water Supply 4(1): 143--149. Iversen V, Koseoglu H, Yigit NO, et al. (2009) Impacts of membrane flux enhancers on activated sludge respiration and nutrient removal in MBRs. Water Research 43: 822--830. Jang N, Ren X, Choi K, and Kim IS (2006) Comparison of membrane biofouling in nitrification and denitrification for the membrane bio-reactor (MBR). Water Science and Technology 53(6): 43–39. Jang N, Ren X, Cho J, and Kim IS (2006) Steady-state modelling of biofuling potentials with respect to the biological kinetics in the submerged membrane bioreactor. Journal of Membrane Science 284: 352--360. Janssen A, Van Agtmaal J, Van Den Broek WBP, et al. (2008) Monitoring of SUR to control and enhance the performance of dead-end ultrafiltration installations treating wwtp effluent. Desalination 231: 99--107. Jefferson B, Brookes A, Le-Clech P, and Judd SJ (2004) Methods for understanding organic fouling in MBRs. Water Science and Technology 49: 237--244. Jefferson B, Laine AL, Judd SJ, and Stephenson T (2000) Membrane bioreactors and their role in wastewater reuse. Water Science and Technology 41(1): 197--204. Jeison D and van Lier JB (2007) Cake formation and consolidation: Main factors governing the applicable flux in anaerobic submerged membrane bioreactors (AnSMBR) treating acidified wastewaters. Separation and Purification Technology 56: 71--78. Ji J, Qiu J, Wong FS, and Li Y (2008) Enhancement of filterability in MBR achieved by improvement of supernatant and floc characteristics via filter aids addition. Water Research 42: 3611--3622. Ji L and Zhou J (2006) Influence of aeration on microbial polymers and membrane fouling in submerged membrane bioreactors. Journal of Membrane Science 276: 168--177. Jiang T, Kennedy MD, Guinzbourg BF, Vanrolleghem PA, and Schippers JC (2005) Optimising the operation of a MBR pilot plant by quantitative analysis of the membrane fouling mechanism. Water Science and Technology 51: 19--25. Jiang T, Myngheer S, De Pauw DJW, et al. (2008) Modelling the production and degradation of soluble microbial products (SMP) in membrane bioreactors (MBR). Water Research 42: 4955--4964. Jimenez B and Asano T (eds.) (2008) Water Reuse: An International Survey of Current Practice, Issues and Needs. London: IWA. Jimenez J, Grelier P, Meinhold J, and Tazi-Pain A (2010) Biological modelling of MBR and impact of primary sedimentation. Desalination 250: 562--567. Jin YL, Lee WN, Lee CH, Chang IS, Huang X, and Swaminathan T (2006) Effect of DO concentration on biofilm structure and membrane filterability in submerged membrane bioreactor. Water Research 40: 2829--2836. Jinhua P, Fukushi K, and Yamamoto K (2006) Bacterial community structure on membrane surface and characteristics of strains isolated from membrane surface in submerged membrane bioreactor. Separation Science and Technology 41: 1527--1549. Judd S (2002) Submerged membrane bioreactors: Flat plate or hollow fibre? Filtration and Separation 39: 30--31. Judd S (ed.) (2006) The MBR Book: Principles & Applications of MBRs in Water & Wastewater Treatment. Oxford: Elsevier.

Judd S, Alvarez-Vazquez H, and Jefferson B (2006) The impact of intermittent aeration on the operation of air-lift tubular membrane bioreactors under sub-critical conditions. Separation Science and Technology 41: 1293--1302. Judd S and Jefferson B (eds.) (2003) Membranes for Industrial Wastewater Recovery and Re-Use. Oxford: Elsevier. Judd SJ, Robinson T, Holdner J, Alvarez-Vazquez H, and Jefferson B (2004) Impact of membrane material on membrane bioreactor permeability. In: Proceedings of the IWA – Water Environment – Membrane Technology (WEMT) Conference. Seoul, Korea, 7–10 June. London: IWA Publishing. Kang IJ, Lee CH, and Kim KJ (2003) Characteristics of microfiltration membranes in a membrane coupled sequencing batch reactor system. Water Research 37: 1192--1197. Kang S, Hoek EMV, Choi H, and Shin H (2006) Effect of membrane surface properties during the fast evaluation of cell attachment. Separation Science and Technology 41: 1475--1487. Kayawake E, Narukami Y, and Yamagata M (1991) Anaerobic digestion by a ceramic membrane enclosed reactor. Journal of Fermentation and Bioengineering 71: 122--125. Kiat WY, Yamamoto K, and Ohgaki S (1992) Optimal fiber spacing in externally pressurized hollow fiber module for solid liquid separation. Water Science and Technology 26(26): 1245--1254. Kim HY, Yeon KM, Lee CH, Lee S, and Swaminathan T (2006) Biofilm structure and extracellular polymeric substances in low and high dissolved oxygen membrane bioreactors. Separation Science and Technology 41: 1213--1230. Kim J, Jang M, Chio H, and Kim S (2004) Characteristics of membrane and module affecting membrane fouling. In: Proceedings of the IWA – Water Environment – Membrane Technology (WEMT) Conference. Seoul, Korea, 7–10 June. London: IWA Publishing. Kim JS and Lee CH (2003) Effect of powdered activated carbon on the performance of an aerobic membrane bioreactor: Comparison between crossflow and submerged membrane systems. Water Environment Research 75: 300--307. Kimura K, Yamato N, Yamamura H, and Watanabe Y (2005) Membrane fouling in pilotscale membrane bioreactors (MBRs) treating municipal wastewater. Environmental Science and Technology 39: 6293--6299. Kjelleberg S, Mcdougald D, Rasmussen TB, and Givskov M (2008) Quorum sensing inhibition. In: Winans SC and Bassler BL (eds.) Chemical Communication among Bacteria, pp. 393--416. Washington, DC: ASM Press. Komesli OK, Teschner K, Hegemann W, and Gokcay CF (2007) Vacuum membrane applications in domestic wastewater reuse. Desalination 215: 22--28. Konopka A, Zakharova T, Oliver L, Camp D, and Turco RF (1996) Biodegradation of organic wastes containing surfactants in a biomass recycle reactor. Applied and Environmental Microbiology 62: 3292--3297. Kniebusch MM, Wilderer PA, and Behling RD (1990) Immobilisation of cells on gas permeable membranes Physiology of Immobilised Cells, pp. 149--160. Amsterdam: Elservier. Kox LSDM (2004) Membrane bioreactor in Varsseveld, the Dutch approach. In: Proceedings of the IWA – Water Environment – Membrane Technology (WEMT) Conference. Seoul, Korea, 7–10 June. London: IWA Publishing. Kraume M and Bracklow U (2003) MBR in municipal wastewater treatment – operational experiences and design guidelines in Germany (in German). In: Proceedings of the 5th Aachen Membrane Conference, pp. 1–20. Aachen, Germany. Krauth KH and Staab KF (1988) Pressurized biomembrane rector for wastewater treatment. Hydrotop 94: 555--562. Kreckel B, Bruess U, Grelot A, and Tazi-Pain A (2009) Innovative European made flatsheet module system for MBR application. In: Lesjean B and Leiknes T (eds.) Final MBR-Network Workshop, pp. 55–56. Berlin, Germany, 31 March–1 April. Kuberkar VT and Davis RH (2000) Modeling of fouling reduction by secondary membranes. Journal of Membrane Science 168: 243--257. Laborie S, Cabassud C, Durand L, and Laine JM (1997) Flux enhancement by a continuous tangential gas flow in UF hollow fibres for drinking water production: Effects of slug flow on cake structure. Filtration and Separation 34(8): 887--891. Laspidou CS and Rittmann BE (2002) A unified theory for extracellular polymeric substances, soluble microbial products, and active and inert biomass. Water Research 36: 2711--2720. Lebegue J, Heran M, and Grasmick A (2008) Membrane bioreactor: Distribution of critical flux throughout an immersed HF bundle. Desalination 231: 245--252. Le-Clech P, Chen V, and Fane TAG (2006) Fouling in membrane bioreactors used in wastewater treatment. Journal of Membrane Science 284: 17--53. Le-Clech P, Jefferson B, Chang IS, and Judd SJ (2003a) Critical flux determination by the flux-step method in a submerged membrane bioreactor. Journal of Membrane Science 227: 81--93.

Membrane Biological Reactors Le-Clech P, Jefferson B, and Judd SJ (2003b) Impact of aeration, solids concentration and membrane characteristics on the hydraulic performance of a membrane bioreactor. Journal of Membrane Science 218: 117--129. Le-Clech P, Fane A, Leslie G, and Childress A (2005a) MBR focus: The operators’ perspective. Filtration and Separation 42(5): 20--23. Le-Clech P, Jefferson B, and Judd SJ (2005b) A comparison of submerged and sidestream tubular membrane bioreactor configurations. Desalination 173: 113--122. Lee J, Ahn WY, and Lee CH (2001a) Comparison of the filtration characteristics between attached and suspended growth microorganisms in submerged membrane bioreactor. Water Research 35: 2435--2445. Lee JC, Kim JS, Kang IJ, Cho MH, Park PK, and Lee CH (2001b) Potential and limitations of alum or zeolite addition to improve the performance of a submerged membrane bioreactor. Water Science and Technology 43: 59--66. Lee W, Jeon JH, Cho Y, Chung KY, and Min BR (2005) Behavior of TMP according to membrane pore size. In: Proceedings of the International Congress on Membranes and Membrane Processes (ICOM). Seoul, Korea, 21–26 August. Lee W, Kang S, and Shin H (2003) Sludge characteristics and their contribution to microfiltration in submerged membrane bioreactors. Journal of Membrane Science 216: 217--227. Lee WN, Cheong WS, Yeon KM, Hwang BK, and Lee CH (2009) Correlation between local TMP distribution and bio-cake porosity on the membrane in a submerged MBR. Journal of Membrane Science 332: 50--55. Lee WN, Kang IJ, and Lee CH (2006) Factors affecting filtration characteristics in membrane-coupled moving bed biofilm reactor. Water Research 40: 1827--1835. Lee Y, Cho J, Seo Y, Lee JW, and Ahn KH (2002) Modeling of submerged membrane bioreactor process for wastewater treatment. Desalination 146: 451--457. Leiknes T and Odegaard H (2007) The development of a biofilm membrane bioreactor. Desalination 202: 135--143. Lesage N, Sperandio M, and Cabassud C (2005) Performances of a hybrid adsorption/ submerged membrane biological process for toxic waste removal. Water Science and Technology 51: 173--180. Lesjean B and Huisjes EH (2008) Survey of the European MBR market: Trends and perspectives. Desalination 231: 71--81. Lesjean B, Leiknes T, Hochstrat R, Schories G, and Gonzalez A (2006) MBR: Technology gets timely EU cash boost. Filtration and Separation 43(9): 20--23. Lesjean B, Rosenberger S, Laabs C, Jekel M, Gnirss R, and Amy G (2005) Correlation between membrane fouling and soluble/colloidal organic substances in membrane bioreactors for municipal wastewater treatment. Water Science and Technology 51: 1--8. Lesjean B, Rosenberger S, Schrotter JC, and Recherche A (2004) Membrane-aided biological wastewater treatment – an overview of applied systems. Membrane Technology August 2004: 5–10. Leslie G and Chapman S (2003) Membrane bioreactors for municipal wastewater treatment – an Australian perspective. In: Proceedings of Australian Water Association Annual Conference, Perth, Australia, 6–10 April. Li H, Yang M, Zhang Y, Liu X, Gao M, and Kamagata Y (2005a) Comparison of nitrification performance and microbial community between submerged membrane bioreactor and conventional activated sludge system. Water Science and Technology 51: 193--200. Li X, Gao F, Hua Z, Du G, and Chen J (2005b) Treatment of synthetic wastewater by a novel MBR with granular sludge developed for controlling membrane fouling. Separation and Purification Technology 46: 19--25. Li YZ, He YL, Liu YH, Yang SC, and Zhang GJ (2005c) Comparison of the filtration characteristics between biological powdered activated carbon sludge and activated sludge in submerged membrane bioreactors. Desalination 174: 305--314. Li XY and Wang XM (2006) Modelling of membrane fouling in a submerged membrane bioreactor. Journal of Membrane Science 278: 151--161. Liao BQ, Bagley DM, Kraemer HE, Leppard GG, and Liss SN (2004) A review of biofouling and its control in membrane separation bioreactors. Water Environment Research 76: 425--436. Liao BQ, Kraemer JT, and Bagley D (2006) Anaerobic membrane bioreactors: Applications and research directions. Critical Reviews in Environmental Science and Technology 36(6): 489--530. Lim AL and Bai R (2003) Membrane fouling and cleaning in microfiltration of activated sludge wastewater. Journal of Membrane Science 216: 279--290. Lim BR, Ahn KH, Song KG, and Jin-Woo C (2005) Microbial community in biofilm on membrane surface of submerged MBR: Effect of in-line cleaning chemical agent. Water Science and Technology 51(6–7): 201–207. Lipnizki F and Field RW (2001) Pervaporation-based hybrid processes in treating phenolic wastewater: Technical aspects and cost engineering. Separation Science and Technology 36: 3311--3335.

611

Lipski C and Cote P (1990) The use of pervaporation for the removal of organic contaminants from water. Environmental Progress 9: 254--261. Liu R, Huang X, Sun YF, and Qian Y (2003) Hydrodynamic effect on sludge accumulation over membrane surfaces in a submerged membrane bioreactor. Process Biochemistry 39: 157--163. Liu R, Huang X, Chen L, Wen X, and Qian Y (2005) Operational performance of a submerged membrane bioreactor for reclamation of bath wastewater. Process Biochemistry 40: 125--130. Liu Y and Fang HHP (2003) Influences of extracellular polymeric substances (EPS) on flocculation, settling, and dewatering of activated sludge. Critical Reviews in Environmental Science and Technology 33: 237--273. Liu Y and Tay JH (2001) Strategy for minimization of excess sludge production from the activated sludge process. Biotechnology Advances 19: 97--107. Livingston AG (1993) A novel membrane bioreactor for detoxifying industrial wastewaters: Biodegradation of 3-chloronitrobenzene in an industrially produced wastewater. Biotechnology and Bioengineering 41: 927--936. Livingston AG (1994) Extractive membrane bioreactors: A new process technology for detoxifying chemical industry wastewaters. Journal of Chemical Technology and Biotechnology 60: 117--124. Livingston AG, Arcangeli JP, Boam AT, Zhang SF, Marangon M, and dos Santos LMF (1998) Extractive membrane bioreactors for detoxification of chemical industry wastes: Process development. Journal of Membrane Science 151: 29--44. Lu SG, Imai T, Ukita M, Sekine M, Fukagawa M, and Nakanishi H (1999) Fermentation wastewater treatment in a membrane bioreactor. Environmental Technology 20: 431--436. Lu SG, Imai T, Ukita M, Sekine M, Higuchi T, and Fukagawa M (2001) A model for membrane bioreactor process based on the concept of formation and degradation of soluble microbial products. Water Research 35: 2038--2048. Lu SG, Imai T, Ukita M, Sekine M, and Higuchi T (2002) Modeling prediction of membrane bioreactor process with the concept of soluble microbial product. Water Science and Technology 46(11–12): 63--69. Lubbecke S, Vogelpohl A, and Dewjanin W (1995) Wastewater treatment in a biological high-performance system with high biomass concentration. Water Research 29: 793--802. Luonsi A, Laitinen AN, Beyer K, Levanen E, Poussade Y, and Nystrom M (2002) Separation of CTMP mill-activated sludge with ceramic membranes. Desalination 146: 399--404. Lyko S, Al-Halbouni D, Wintgens T, et al. (2007) Polymeric compounds in activated sludge supernatant – characterisation and retention mechanisms at a full-scale municipal membrane bioreactor. Water Research 41: 3894--3902. Madaeni SS, Fane AG, and Wiley D (1999) Factors influencing critical flux in membrane filtration of activated sludge. Journal of Chemical Technology and Biotechnology 74: 539--543. Mansell B, Kuo J, Tang CC, et al. (2004) Comparison of two membrane bioreactors and an activated sludge plant with dual-media filtration: Nutrient and priority pollutants removal. In: Proceedings WEFTEC 04. New Orleans, USA, 2–6 October. McInnis AB (2005) Cost comparison of MBR and CAS/MF facilities. In: Proceedings WEFTEC 05. Washington, DC, USA, 29 October–2 November. Mehrez R, Ernst M, and Jekel M (2007) Development of a continuous protein and polysaccharide measurement method by sequential injection analysis for the application in membrane bioreactor systems. Water Science and Technology 56: 163--171. Melin T, Jefferson B, Bixio D, et al. (2006) Membrane bioreactor technology for wastewater treatment and reuse. Desalination 187: 271--282. Meng F, Chae SR, Drews A, Kraume M, Shin HS, and Yang F (2009) Recent advances in membrane bioreactors (MBRs): Membrane fouling and membrane material. Water Research 43: 1489--1512. Meng F, Zhang H, Li Y, Zhang X, and Yang F (2005) Application of fractal permeation model to investigate membrane fouling in membrane bioreactor. Journal of Membrane Science 262: 107--116. Meng F, Zhang H, Yang F, Li Y, Xiao J, and Zhang X (2006) Effect of filamentous bacteria on membrane fouling in submerged membrane bioreactor. Journal of Membrane Science 272: 161--168. Metcalf and Eddy, Inc (2003) Wastewater Engineering – Treatment and Reuse, 4th edn. New York: McGraw-Hill. Miyashita S, Honjyo K, Kato O, et al. (2000) Gas Diffuser for Aeration Vessel of Membrane Assembly. US Pat. 6,328,886. Miura Y, Watanabe Y, and Okabe S (2007) Membrane biofouling in pilot-scale membrane bioreactors (MBRs) treating municipal wastewater: Impact of biofilm formation. Environmental Science and Technology 41: 632--638. Mo H, Oleszkiewicz JA, Cicek N, and Rezania B (2005) Incorporating membrane gas diffusion into a membrane bioreactor for hydrogenotrophic denitrification of groundwater. Water Science and Technology 51(6–7): 357–364.

612

Membrane Biological Reactors

Modin O, Fukushi K, Nakajima F, and Yamamoto K (2008) Performance of a membrane biofilm reactor for denitrification with methane. Bioresource Technology 99: 8054--8060. Muller EB, Stouthamer AB, Verseveld HW, and Eikelboom EH (1995) Aerobic domestic waste water treatment in a pilot plant with complete sludge retention by cross-flow filtration. Water Research 29: 1179--1189. Nagaoka H and Nemoto H (2005) Influence of extracellular polymeric substances on nitrogen removal in an intermittently-aerated membrane bioreactor. Water Science and Technology 51: 151--158. Nagaoka H, Ueda S, and Miya A (1996) Influence of bacterial extracellular polymers on the membrane separation activated sludge process. Water Science and Technology 34: 165--172. Nagaoka H, Yamanishi S, and Miya A (1998) Modeling of biofouling by extracellular polymers in a membrane separation activated sludge system. Water Science and Technology 38: 497--504. Namkung E (2008) Current status and future of membrane process in Korea. In: Proceedings 11th RECWET Symposium Entitled ‘‘Water Environment Control by Membrane Treatment Technology’’. Tokyo, Japan, 2 December. Ng CA, Sun D, and Fane AG (2006) Operation of membrane bioreactor with powdered activated carbon addition. Separation Science and Technology 41: 1447--1466. Ng CA, Sun D, Zhang J, et al. (2005) Strategies to improve the sustainable operation of membrane bioreactors. In: Proceedings of the International Desalination Association World Congress. Singapore, 11–16 September. Ng ANL and Kim AS (2007) A mini-review of modeling studies on membrane bioreactor (MBR) treatment for municipal wastewaters. Desalination 212: 261--281. Ngo HH, Guo W, and Xing W (2008) Evaluation of a novel sponge-submerged membrane bioreactor (SSMBR) for sustainable water reclamation. Bioresource Technology 99: 2429--2435. Nguyen TQ and Nobe K (1987) Extraction of organic contaminants in aqueous solution by pervaporation. Journal of Membrane Science 30: 11--22. Nielson PH and Jahn A (1999) Extraction of EPS. In: Wingender J, Neu TR, and Flemming HCE (eds.) Microbial Extracellular Polymeric Substances, pp. 49--72. Berlin: Springer. Nuengjamnong C, Kweon JH, Cho J, Polprasert C, and Ahn KH (2005) Membrane fouling caused by extracellular polymeric substances during microfiltration processes. Desalination 179: 117--124. Nuhoglu A, Pekdemir T, Yielder E, Keskinlerand B, and Akay G (2002) Drinking water denitrification by a membrane bio-reactor. Water Research 36: 1155--1166. Ognier S, Wisniewski C, and Grasmick A (2004) Membrane bioreactor fouling in subcritical filtration conditions: A local critical flux concept. Journal of Membrane Science 229: 171--177. Orantes JC, Wisniewski C, Heran M, and Grasmick A (2004) Influence of total sludge retension on the performance of a submerge membrane bioreactor. In: Proceedings of the IWA – Water Environment Membrane Technology (WEMT) Specialty Conference. Seoul, South Korea, 7–10 June. London: IWA Publishing. Pang CM, Hong P, Guo H, and Liu WT (2005) Biofilm formation characteristics of bacterial isolates retrieved from a reverse osmosis membrane. Environmental Science and Technology 39: 7541--7550. Pankania M, Stephenson M, and Semmens MJ (1994) Hollow fiber bioreactor for wastewater treatment using bubbleless membrane aeration. Water Research 10: 2233--2236. Parameshwaran K and Visvanathan C (1998) Recent developments in membrane technology for wastewater reuse. In: Proceedings of the 6th International Conference on Advanced Wastewater Treatment, Recycling and Reuse. Milan, Italy, 14–16 September. Park S, Yeon KM, and Lee CH (2005) Hydrodynamics and microbial physiology affecting performance of a new MBR, membrane-coupled high performance compact reactor. Desalination 172: 181--188. Phattaranawik J, Fane AG, Pasquier ACS, and Bing W (2008) A novel membrane bioreactor based on membrane distillation. Desalination 223: 386--395. Phattaranawik J, Fane AG, Pasquier ACS, Bing W, and Wong FS (2009) Experimental study and design of a submerged membrane distillation bioreactor. Chemical Engineering and Technology 32: 38--44. Pearce G (2007) Introduction to membranes: Filtration for water and wastewater treatment. Filtration and Separation 44(2): 24--27. Pearce G (2008a) Introduction to membranes: An introduction to membrane bioreactors. Filtration and Separation 45(1): 32--35. Pearce G (2008b) Introduction to membranes – MBRs: Manufacturers’ comparison: Part 1. Filtration and Separation 45(3): 28--31. Pearce G (2008c) Introduction to membranes – MBRs: Manufacturers’ comparison: Part 2. Filtration and Separation 45(3): 30--32. Pearce G (2008d) Introduction to membranes – MBRs: Manufacturers’ comparison: Part 3. Filtration and Separation 45(4): 23--25.

Pillay S, Foxon KM, and Buckley CA (2008) An anaerobic baffled reactor/membrane bioreactor (ABR/MBR) for on-site sanitation in low income areas. Desalination 231: 91--98. Pollice A, Brookes A, Jefferson B, and Judd S (2005) Sub-critical flux fouling in membrane bioreactors – a review of recent literature. Desalination 174: 221--230. Psoch C and Schiewer S (2005a) Critical flux aspect of air sparging and backflushing on membrane bioreactors. Desalination 175: 61--71. Psoch C and Schiewer S (2005b) Long-term study of an intermittent air sparged MBR for synthetic wastewater treatment. Journal of Membrane Science 260: 56--65. Psoch C and Schiewer S (2006) Resistance analysis for enhanced wastewater membrane filtration. Journal of Membrane Science 280: 284--297. Qin JJ, Kekre KA, Tao G, et al. (2006) New option of MBR-RO process for production of NEWater from domestic sewage. Journal of Membrane Science 272: 70--77. Rabie HR, Cote P, Singh M, and Janson A (2003) Cyclic Aeration System for Submerged Membrane Modules. US Pat. 684,406. Ramesh A, Lee DJ, and Lai JY (2007) Membrane biofouling by extracellular polymeric substances or soluble mcirobial products from membrane bioreactor sludge. Applied Microbiology and Biotechnology 74: 699--707. Reij MW, Keurentjes JTF, and Hartmans S (1998) Membrane bioreactors for waste gas treatment. Journal of Biotechnology 59: 155--167. Rezania B, Oleszkiewicz JA, Cicek N, and Mo H (2005) Hydrogen dependent denitrification in an alternating anoxic–aerobic SBR membrane bioreactor. Water Science and Technology 51(6–7): 403–409. Rittmann BE (1987) Aerobic biological treatment. Environmental Science and Technology 21: 128--136. Rosenberger S, Evenblij H, Te Poele S, Wintgens T, and Laabs C (2005) The importance of liquid phase analyses to understand fouling in membrane assisted activated sludge processes-six case studies of different European research groups. Journal of Membrane Science 263: 113--126. Rosenberger S and Kraume M (2002) Filterability of activated sludge in membrane bioreactors. Desalination 146: 373--379. Rosenberger S, Kruger K, Witzig R, Szewzyk U, and Kraume M (2002) Performance of a bioreactor with submerged membranes for aerobic treatment of municipal waste water. Water Research 36: 413--420. Rosenberger S, Laabs C, Lesjean B, et al. (2006) Impact of colloidal and soluble organic material on membrane performance in membrane bioreactors for municipal wastewater treatment. Water Research 40: 710--720. Ross WR, Barnard JP, Strohwald NK, Grobler CJ, and Sanetra J (1992) Practical application of the ADUF process to the full-scale treatment of maize-processing effluent. Water Science and Technology 25(10): 27--39. Rothemund C, Camper A, and Wilderer PA (1994) Biofilms growing on gas permeable membranes. Water Science and Technology 29(10–11): 447--454. Saroj DP, Guglielmi G, Chiarani D, and Andreottola G (2008) Subcritical fouling behaviour modelling of membrane bioreactors for municipal wastewater treatment: The prediction of the time to reach critical operating condition. Desalination 231: 175--181. Schier W, Frechen FB, and Fischer S (2009) Efficiency of mechanical pre-treatment on European MBR plants. Desalination 236: 85--93. Schoeberl P, Brik M, Bertoni M, Braun R, and Fuchs W (2005) Optimization of operational parameters for a submerged membrane bioreactor treating dye house wastewater. Separation and Purification Technology 44: 61--68. Schrader GA, Zwijnenburg A, and Wessling M (2005) The effect of WWTP effluent zeta-potential on direct nanofiltration performance. Journal of Membrane Science 266: 80--93. Scott JA, Neilson DJ, Liu W, and Boon PN (1998) A dual function membrane bioreactor system for enhanced aerobic remediation of high-strength industrial waste. Water Science and Technology 38: 413--420. Semmens MJ (1991) Bubbleless aeration. Water Engineering and Management 138(4): 18--19. Semmens MJ and Gantzer CJ (1993) Gas transfer using hollow fiber membranes. In: Proceedings of the 66th Annual Conference and Exposition of the Water Environment Federation, pp. 365–406. Anaheim, CA, USA, 3–7 October. Smith CV, DiGregorio D, and Talcott RM (1969) The use of ultrafiltration membranes for activated sludge separation. In: Proceedings of the 24th Annual Purdue Industrial Waste Conference. Lafayette, IN, USA, 6–8 May. Smith PJ, Vigneswaran S, Ngo HH, Ben-Aim R, and Nguyen H (2005) Design of a generic control system for optimising back flush durations in a submerged membrane hybrid reactor. Journal of Membrane Science 255: 99--106. Sridang PC, Heran M, and Grasmick A (2005) Influence of module configuration and hydrodynamics in water clarification by immersed membrane systems. Water Science and Technology 51(6–7): 135--142. Stephenson T, Judd S, Jeferson B, and Brindle K (2000) Membrane Bioreactors for Wastewater Treatment. London: IWA.

Membrane Biological Reactors Sui P, Wen X, and Huang X (2008) Feasibility of employing ultrasound for on-line membrane fouling control in an anaerobic membrane bioreactor. Desalination 219: 203--213. Sun Y, Huang X, Chen F, and Wen X (2004) A dual functional filtration/aeration membrane bioreactor for domestic wastewater treatment. In: Proceedings of the IWA – Water Environment – Membrane Technology (WEMT) Conference. Seoul, Korea, 7–10 June. London: IWA Publishing. Sutton PM (2003) Membrane bioreactors for industrial wastewater treatment: The state-of-the-art based on full scale commercial applications. In: Proceedings of WEF 76th Annual Conference and Exposition. Los Angeles, CA, USA, 12–15 October. Sutton PM, Li RR, and Korchin SR (1983) Dorr-oliver’s fixed film suspended growth anaerobic systems for industrial wastewater treatment and energy recovery. In: Proceedings of the 37th Annual Purdue Industrial Waste Conference, pp. 667–675. Lafayette, IN, USA, 11–13 May. Tao G, Kekre K, Wei Z, Lee TC, Viswanath B, and Seah H (2005) Membrane bioreactors for water reclamation. Water Science and Technology 51: 431--440. Tadkaew N, Sivakumar M, and Nghiem LD (2007) Membrane bioreactor technology for decentralised wastewater treatment and reuse. International Journal of Water 3(4): 368--380. Tarnacki K, Lyko S, Wintgens T, Melin T, and Natau F (2005) Impact of extracellular polymeric substances on the filterability of activated sludge in membrane bioreactors for landfill leachate treatment. Desalination 179: 181--190. Tchobanoglous G, Burton FL, and Stensel HD (eds.) (2003) Wastewater Engineering: Treatment and Reuse. New York: McGraw-Hill Higher Education. Teychene B, Guigui C, Cabassud C, and Amy G (2008) Toward a better identification of foulant species in MBR processes. Desalination 231: 27--34. Thanh BX, Visvanathan C, Sperandio M, and Ben Aim R (2008) Fouling characterization in aerobic granulation coupled baffled membrane separation unit. Journal of Membrane Science 318: 334--339. Ueda T and Hata K (1999) Domestic wastewater treatment by a submerged membrane bioreactor with gravitational filtration. Water Research 33: 2888--2892. Ueda T, Hata K, Kikuoka Y, and Seino O (1997) Effects of aeration on suction pressure in a submerged membrane bioreactor. Water Research 31: 489--494. Urbain V, Mobarry B, de Silva V, Stahl DA, Rittmann BE, and Manem J (1998) Integration of performance, molecular biology and modeling to describe the activated sludge process. Water Science and Technology 37(4–5): 223--229. Vallero MVG, Lettinga G, and Lens PNL (2005) High rate sulfate reduction in a submerged anaerobic membrane bioreactor (SAMBaR) at high salinity. Journal of Membrane Science 253: 217--232. Van Der Roest HF, Lawrence DP, and Van Bentem AG (2002a) Membrane Bioreactors for Municipal Wastewater Treatment (Water and Wastewater Practitioner Series: Stowa Report). London: IWA. Van Der Roest HF, Van Bentem AGN, and Lawrence DP (2002b) MBR-technology in municipal wastewater treatment: Challenging the traditional treatment technologies. Water Science and Technology 46: 273--280. Van’T Oever R (2005) MBR focus: Is submerged best? Filtration and Separation 42(5): 24--27. Verrecht B, Judd S, Guglielmi G, Brepols C, and Mulder JW (2008) An aeration energy model for an immersed membrane bioreactor. Water Research 42: 4761--4770. Vilim D, Hlavinek P, Kubik J, and Hlustik P (2009) Development of containerized turnkey MBR plants. In: Lesjean B and Leiknes T (eds.) Final MBR-Network Workshop, pp. 57–58. Berlin, Germany, 31 March–1 April. Visvanathan C, Byung-Soo Y, Muttamara S, and Maythanukhraw R (1997) Application of air backflushing technique in membrane bioreactor. Water Science and Technology 36(12): 259--266. Visvanathan C, Ben Aim R, and Parameshwaran K (2000) Membrane separation bioreactors for wastewater treatment. Critical Reviews in Environmental Science and Technology 30: 1--48. Wagner J and Rosenwinkel KH (2000) Sludge production in membrane bioreactors under different conditions. Water Science and Technology 41(10–11): 251--258. Wang S, Guillen G, and Hoek EMV (2005) Direct observation of microbial adhesion to membranes. Environmental Science and Technology 39(17): 6461--6469. Wang Z, Wu Z, Maib S, et al. (2008a) Research and applications of membrane bioreactors in China: Progress and prospect. Separation and Purification Technology 62: 249--263. Wang Z, Wu Z, Yin X, and Tian L (2008b) Membrane fouling in a submerged membrane bioreactor (MBR) under sub-critical flux operation: Membrane foulant and gel layer characterization. Journal of Membrane Science 325: 238--244. Wen X, Sui P, and Huang X (2008) Exerting ultrasound to control the membrane fouling in filtration of anaerobic activated sludge – mechanism and membrane damage. Water Science and Technology 57(5): 773--779.

613

Wen X-H, Xing C-H, and Qian Y (1999) A kinetic model for the prediction of sludge formation in a membrane bioreactor. Process Biochemistry 35: 249--254. WERF (2001) Membrane bioreactors: Feasibility and use in water reclamation. Report of the Water Environment Research Foundation. Alexandria, VA: WERF. Wicaksana F, Fane AG, and Chen V (2006) Fibre movement induced by bubbling using submerged hollow fibre membranes. Journal of Membrane Science 271: 186--195. Wilderer PA, Brautigam J, and Sekoulov I (1985) Application of gas permeable membranes for auxiliary oxygenation of sequencing batch reactors. Conservation and Recycling 8(1–2): 181--192. Wisniewski C and Grasmick A (1998) Floc size distribution in a membrane bioreactor and consequences for membrane fouling. Colloids and Surfaces A: Physicochemical and Engineering Aspects 138: 403--411. Wintgens T, Melin T, Scha¨fer A, et al. (2005) The role of membrane processes in municipal wastewater reclamation and reuse. Desalination 178: 1--11. Wintgens T, Rosen J, Melin T, Brepols C, Drensla K, and Engelhardt N (2003) Modelling of a membrane bioreactor system for municipal wastewater treatment. Journal of Membrane Science 216: 55--65. Witzig R, Manz W, Rosenberger S, Kru¨ger U, Kraume M, and Szewzyk U (2002) Microbiological aspects of a bioreactor with submerged membranes for aerobic treatment of municipal wastewater. Water Research 36: 394--402. Wu Y, Huang X, Wen X, and Chen F (2004) Function of dynamic membrane in selfforming dynamic membrane coupled bioreactor. In: Proceedings of the IWA – Water Environment – Membrane Technology (WEMT) Conference. Seoul, Korea, 7–10 June. Xing CH, Qian Y, Wen XH, Wu WZ, and Sun D (2001) Physical and biological characteristics of a tangential-flow MBR for municipal wastewater treatment. Journal of Membrane Science 191: 31--42. Xing CH, Tardieu E, Qian Y, and Wen XH (2000) Ultrafiltration membrane bioreactor for urban wastewater reclamation. Journal of Membrane Science 177: 73--82. Xing CH, Yamamoto K, and Fukushi K (2006) Performance of an inclined-plate membrane bioreactor at zero excess sludge discharge. Journal of Membrane Science 275: 175--186. Xu N, Xing WH, Xu NP, and Shi J (2003) Study on ceramic membrane bioreactor with turbulence promoter. Separation and Purification Technology 32: 403--410. Yamagiwa K, Ohmae Y, Dahlan MH, and Ohkawa A (1991) Activated sludge treatment of small-scale wastewater by a plunging liquid jet bioreactor with cross-flow filtration. Bioresource Technology 37: 215--222. Yamamoto K (2009) Submerged MBR technology: An unfinished international adventure of 20 years. In: Lesjean B and Leiknes T (eds.) Final MBR-Network Workshop, pp. 83–84. Berlin, Germany, 31 March–1 April. Yamamoto K, Hiasa H, Talat M, and Matsuo T (1989) Direct solid liquid separation using hollow fiber membranes in activated sludge aeration tank. Water Science and Technology 21: 43--54. Yamamura H, Kimura K, and Watanbe Y (2007) Mechanism involved in the evolution of physically irreversible fouling in microfiltration and ultrafiltration membranes used for drinking water treatment. Environmental Science and Technology 41(19): 6789--6794. Yamato N, Kimura K, Miyoshi T, and Watanabe Y (2006) Difference in membrane fouling in membrane bioreactors (MBRs) caused by membrane polymer materials. Journal of Membrane Science 280: 911--919. Yang Q, Chen J, and Zhang F (2006) Membrane fouling control in a submerged membrane bioreactor with porous, flexible suspended carriers. Desalination 189: 292--302. Yang Z, Peng XF, Chen MY, Lee DJ, and Lai JY (2007) Intralayer flow in fouling layer on membranes. Journal of Membrane Science 287: 280--286. Yeh SJ and Jenkins CR (1978) Pure oxygen fixed film reactor. Journal of Environmental Engineering (ASCE) 14: 611--623. Yeo A and Fane AG (2005) Performance of individual fibers in a submerged hollow fiber bundle. Water Science and Technology 51(6–7): 165--172. Yeom IT, Nah YM, and Ahn KH (1999) Treatment of household wastewater using an intermittently aerated membrane bioreactor. Desalination 124: 193--203. Yeon KM, Cheong WS, Oh SH, et al. (2009) Quorum sensing: A new biofouling control paradigm in membrane bioreactor for advanced wastewater treatment. Environmental Science and Technology 43: 380--385. Yeon KM, Park JS, Lee CH, and Kim SM (2005) Membrane coupled high performance compact reactor: A new MBR system for advanced wastewater treatment. Water Research 39: 1954--1961. Yoon SH, Collins JH, Musale D, et al. (2005) Effects of flux enhancing polymer on the characteristics of sludge in membrane bioreactor process. Water Science and Technology 51: 151--157. You HS, Huang CP, Pan JR, and Chang SC (2006) Behavior of membrane scaling during crossflow filtration in the anaerobic MBR system. Separation Science and Technology 41: 1265--1278.

4.17 Anaerobic Processes DJ Batstone and PD Jensen, The University of Queensland, Brisbane, QLD, Australia & 2011 Elsevier B.V. All rights reserved.

4.17.1 4.17.1.1 4.17.1.1.1 4.17.1.1.2 4.17.1.1.3 4.17.1.1.4 4.17.1.2 4.17.1.3 4.17.1.4 4.17.1.5 4.17.2 4.17.2.1 4.17.2.1.1 4.17.2.1.2 4.17.2.1.3 4.17.2.1.4 4.17.2.1.5 4.17.2.2 4.17.2.2.1 4.17.2.2.2 4.17.3 4.17.3.1 4.17.3.2 4.17.3.2.1 4.17.3.2.2 4.17.3.3 4.17.3.3.1 4.17.3.3.2 4.17.3.4 4.17.4 4.17.4.1 4.17.4.2 4.17.4.3 References

Anaerobic Process Fundamentals Anaerobic Conversion Processes Hydrolysis Fermentation/acidogenesis Acetogenesis and methanogenesis from hydrogen Aceticlastic methanogenesis Physicochemical Processes and pH Temperature Inhibition and Toxicity Rate-Limiting Steps Selection and Design of Anaerobic Technology Anaerobic Digester Technologies High-rate anaerobic digestion Anaerobic ponds Fully mixed liquid digester Plug-flow liquid digesters Solid phase (leach bed) Digester Selection and Design for Specific Applications Domestic and industrial wastewater Sewage solids and activated sludge biosolids Interpretation and Operation of Anaerobic Systems Evaluating and Determining Controlling Mechanisms Performance and Process Indicators High-rate anaerobic reactors Sludge digesters Evaluating Substrate and Microbial Properties Activity testing Biological methane potential testing Advanced Model-Based Analysis Future Applications of Anaerobic Digestion Sewage Treatment and Nutrient Removal Nutrient Recovery Future Applications in Energy Generation and Transport

4.17.1 Anaerobic Process Fundamentals Anaerobic digestion is the biological conversion by a complex microbial ecosystem of organic and occasionally inorganic substrates in the absence of an oxygen source. During the process, organic material is converted mainly to methane, carbon dioxide, and biomass. Nitrogen released from converted organics is in the form of ammonia. Anaerobic processes for wastewater treatment have advantages over aerobic treatment in that there are no power requirements for air supply, production of sludges requiring treatment and disposal is much lower, and the methane production can be used for energy production. Aerobic processes are catabolically more favorable, yielding approximately 10 times the energy, with a correspondingly higher microbial yield (Madigan et al., 2009). For this reason, yields used for mixed heterotrophic processes are of the order of

615 615 616 618 619 620 621 623 624 625 626 626 626 626 627 627 627 627 627 628 631 631 632 632 632 633 633 634 635 636 636 636 637 637

0.63 gCODX gCODS1 (Henze et al., 2000) as compared to 0.05–0.1 gCODX gCODS1 for anaerobic processes. COD is the chemical oxygen demand and is a measure of organics. In this case, gCODX represents the biomass generated (in grams COD), while gCODS represents the substrate consumed (Batstone et al., 2002). This lower microbial yield results in decreased operating costs. The lower yield generally implies that extended solid-retention times are required to avoid washout of active biomass. This can be done either in parallel with an increased liquid retention time, or by separation of liquid and solid-retention times. Operation, design, and interpretation of engineered anaerobic processes have greatly advanced over the last 20 years. This improvement is based on a very good understanding of underlying concepts, which has allowed implementation of technology such that it will stably and reliably operate without intervention. The process itself has (1) multiple microbial

615

616

Anaerobic Processes

steps, mediated by different organisms; (2) different steps that can be rate limiting under specific conditions; (3) interaction with the physicochemical system, particularly weak acid and base inhibition of microbial processes, and (4) highly nonlinear behavior, particularly with respect to pH regulation and inhibition. Therefore, application of anaerobic technology needs careful thought, especially to achieve an optimally engineered process for a specific application. Fortunately, understanding of the underlying microbial and chemical processes is very good, both in the scientific and in engineering sectors. Good understanding of fundamentals, as outlined in this section, has allowed the use of anaerobic technologies in a wide variety of applications, as outlined in Section 4.17.2.

The different microbial groups mediating each step have been well characterized, and are from phylogenetically defined regions. As examples, all methanogenic organisms discovered so far are archaea, while acidogens and acetogens are largely bacteria. Aceticlastic methanogens belong to one of the two specific genera: Methanosaeta or Methanosarcina. As shown in Figure 1, under different conditions, different steps can be rate limiting. Specifically, for particulate or slowly degradable materials, hydrolysis is rate limiting. Under conditions of stress, or where the primary substrate is rapidly degradable, aceticlastic methanogenesis is normally rate limiting. The first condition normally results in decreased performance as undegraded substrate is washed out, while the second condition results in elevated, effluent organic-acid concentrations.

4.17.1.1 Anaerobic Conversion Processes 4.17.1.1.1 Hydrolysis Anaerobic digestion proceeds through a series of parallel and sequential processes by a variety of consortia as represented in Figure 1 (Batstone et al., 2002; Pavlostathis and GiraldoGomez, 1991). In contrast to aerobic digestion, where oxygen is an external electron acceptor, gaseous and dissolved products (largely methane and carbon dioxide) have the same combined carbon-oxidation state as the primary substrates. Thus, anaerobic digestion is largely constrained by the need to find appropriate internal electron acceptors. When this is impossible, hydrogen ions or bicarbonate must be used as electron acceptors via anaerobic oxidation to produce hydrogen or formate. This introduces thermodynamic constraints that bring in obligate syntrophic relationships between the electron producer and the methanogenic electron consumer (Schink, 1997). It is conceptually correct and convenient to group complex organics into carbohydrates, proteins, and lipids, and their soluble analogs of sugars, amino acids, and long-chain fatty acids (LCFAs). Any mixed organic stream can be represented by these components, while preserving full information of mass, energy density (or COD), and nitrogen content (Nopens et al., 2009). Anaerobic digestion processes consist of four main steps:

• •

•

•

Hydrolysis is an enzyme-mediated extracellular step which solubilizes particulates and substrates that cannot be directly utilized by the anaerobic organisms. Acidogenesis or fermentation is the conversion of soluble substrates such as amino acids and sugars, which can be converted largely without an external electron acceptor. The products are largely organic acids and alcohols. Syntrophic acetogenesis is the degradation of fermentation products to acetate using hydrogen ions or bicarbonate as an external electron acceptor. This process is coupled with hydrogen or formate utilizing methanogenesis, which maintains a low hydrogen or formate concentration. Acetoclastic methanogenesis is the cleavage of acetate to methane and carbon dioxide.

Processes such as homoacetogenesis (conversion of hydrogen and carbon dioxide to acetate), and its reverse, acetate oxidation to hydrogen and carbon dioxide, have not been included in Figure 1, but can be important in specific circumstances as outlined further in this chapter.

While the formal definition of hydrolysis is much stricter, as a digestion component, hydrolysis is a term that is used to refer to solubilization of complex particulate materials. The material can be regarded either as a mixture of the basic components (carbohydrates, proteins, and fats), or as a composite compound (e.g., homogeneous material such as activated sludge and yeast). Separate classification and analysis of composite material as a separate input was proposed in the International Water Association (IWA) Anaerobic Digestion Model No. 1 (Batstone et al., 2002), but this was found to be cumbersome, especially when representing both composites and primary aggregates (e.g., waste-activated sludge (WAS) and primary sludges), and the current trend is to represent all feed materials as a combination of carbohydrates, proteins, and fats (Nopens et al., 2009). There are three main pathways for enzymatic hydrolysis. 1. The organisms excrete enzymes into the bulk liquid where it adsorbs onto a particle or reacts with a soluble substrate (Jain et al., 1992). 2. The organism attaches to the particle and secretes enzymes into the vicinity of the particle. The organism benefits from the soluble substrates being released (Vavilin et al., 1996). 3. The organism has an attached enzyme which may double up as a transport receptor to the interior of the cell (Tong and McCarty, 1991). This method requires the organism to adsorb onto the surface of the particle. The actual mechanism used depends heavily on the nature of the material, reactor hydraulics, and solid concentration, but forms 1 and 2 in the list are variations on the same mechanism, and are the principal forms considered here. Steps in extracellular enzymatic hydrolysis include (Figure 2): 1.

2.

4.

Production of enzyme – production rate can decrease when there is excessive soluble substrate available (Ramsay, 1997). Steps 2, 3, and 6 are transport processes, which can be limited due to large particles, or in solid-phase systems due to inadequate carrier liquid. Adsorption processes that are limited by surface area.

Anaerobic Processes 5. 7.

•

Reaction rates that are limited by surface area and enzyme concentrations. Deactivation can be excessive when away from optimal temperature and pH.

•

While there have been complex models that include all of these functions (e.g., Humphrey, 1979), in practice, it is very difficult to properly validate these models, and the most commonly used model is the first-order one. The use of first-order models has been justified as ‘‘an empirical expression that reflects the cumulative effect of all the microscopic processes occurringy’’ (Eastman and Ferguson, 1981). First order (or slightly more complex) has also been found to be just as effective as more complex models (Vavilin et al., 1996). Hydrolysis commonly becomes rate limiting when

•

In a continuous mixed digester, without retained solids, hydraulic-loading rate becomes too high (there is not enough time to hydrolyze the solids). Mass-loading rate is generally not an issue, and higher concentrations allow higher loading rates. Normally, a minimum of 9 days of hydraulic-retention time is required for any significant degradation (see Section 4.17.2.2.2). Mixed carbohydrate feeds are among the slowest to degrade.

In a batch system, there is insufficient batch time. Batch digesters have a higher volumetric efficiency, due to kinetic considerations. In a plug-flow system, there is insufficient reactor volume. Plug-flow digesters are highly efficient on a volumetric basis. Time of contact with the active biomass can also be an issue if the system is not effectively mixed at the inlet.

Particularly for mixed systems (the most common form of digester), where hydrolysis is rate limiting, the hydrolysis rate determines the size of the digester. We now discuss the hydrolysis of various feed materials: 1. Hydrolysis of WAS. There has been a large amount of work investigating the rate and extent of WAS digestion, but only limited analysis of the actual mechanisms of cell solubilization as specific to activated sludge. It is a complex process, involving lysis of the cell, and subsequent degradation of both soluble and particulate cellular components (Aquino et al., 2008; Madigan et al., 2009). This is further complicated by the issue that microbial cells are naturally resistant to cell lysis by other cells, and that the cells are in flocs, with varying sizes. Degradability and hydrolysis rate have been extensively analyzed. As mentioned earlier, activated-sludge hydrolysis

Particulate carbohydrates, proteins, and lipids

Acidogens produce enzymes

Hydrolysis

Sugars and amino acids Fermentation acidogenesis

CO2 Alcohols and Long-chain organic acids fatty acids

NH3

Acetogenesis CO2

CO2 Hydrogen

Acetic acid

Hydrogenotrophic methanogenesis

Aceticlastic methanogenesis

Methane Figure 1 Key steps in anaerobic digestion processes.

617

Methane

CO2

May be rate limiting

618

Anaerobic Processes

4. Adsorption of enzyme onto surface

6. Transport of product to bulk

5. Reaction

2. Transport to bulk or local environment 1. Production of enzyme

7. Deactivation of enzyme 3. Diffusion from bulk to particle

Figure 2 Steps in enzymatic hydrolysis.

is an extremely complicated physical and chemical process that is, of necessity, represented as a first-order process (Eastman and Ferguson, 1981). Practical batch testing indicates that this complex material is well represented by first-order kinetics (Dwyer et al., 2008), while primary sludge (for example) has a far more complex kinetic profile, due to the presence of multiple primary substrates (Yasui et al., 2008). Extensive analysis also indicates that for untreated activated sludge, hydrolysis rates are relatively constant at approximately 0.1 d1 (Batstone et al., 2002; Eastman and Ferguson, 1981; Ge et al., 2010; Pavlostathis and Giraldo-Gomez, 1991). The degradability of activated sludge can be entirely related back to upstream sludge age, and longer sludge-age material will be less degradable (i.e., have a higher inert fraction; Ekama et al., 2007; Gossett and Belser, 1982). It is now widely accepted that material that is undegradable under aerobic conditions, is also largely undegradable under anaerobic conditions (Park et al., 2006; Speece, 2008). Therefore, material that is degradable under anaerobic conditions can be numerically calculated from the degradable fraction of the active aerobic biomass in the WAS (Ekama et al., 2007; Nopens et al., 2009). There is a wide range of pretreatment methods to increase sludgedegradability extent and rate (Aquino et al., 2008), and these are discussed further in the Section 4.17.2 of this chapter. 2. Hydrolysis of carbohydrates. Carbohydrates mainly originate directly or indirectly from plants. Generally, plant material is a mixture of cellulose (25–60%), hemicellulose (15– 30%), and lignin (15–20%) (Tong and McCarty, 1991). Straw, a commonly used feed material, consists of 70% cellulose and hemicellulose, 8% lignins, 15% mineral solids, and 7% other organic compounds (Hashimoto, 1986). The remainder is tannins, soluble sugars, and ash. The first two components are very similar and are digested anaerobically via similar mechanisms. Tong and McCarty (1991) list typical chemical compositions of lignocellulosic materials. Cellulose is made up of linear chains of D-glucose units. Hemicellulose is a branched polymer comprising several natural minor sugars. Ease of degradation depends on the nature (crystalline or amorphous) and chain length. Hemicellulose is of a shorter length (200 units), while cellulose can have a chain length of up to 10 000 units. Lignin is a dense three-dimensional polymer of aromatic molecules. It is hydrophobic and is linked by carbon as well as ether bonds. Conversion of lignin by anaerobic bacteria is unknown,

H

R

O

H

R O

H R O

N

C

C

N

C

N

H

H

C

C

C

H

Figure 3 Protein chain with amino acids linked by amide groups.

and high lignin contents (together with the presence of crystalline cellulose) generally restrict or prevent hydrolysis of the underlying cellulosic material (Yang et al., 2009). 3. Hydrolysis of proteins. Proteins are natural polymers of different amino acids joined together by peptide (amide) bonds. The backbone of a protein is a repeating sequence of one nitrogen and two carbon atoms (Figure 3). There are 20 amino acids found in nature. These are differentiated by the R group, which defines the function of the amino acid. A protein has three structural components: • Amino-acid composition and sequence (primary structure). • The three-dimensional shape as set by bond angles and hydrogen bonds forms a helical shape in complex proteins. This is the secondary structure. • The tertiary structure defines the macromolecular shape as set by bonding between di-sulfide groups and to a lesser extent, other inter-R bonding. There are two major areas of importance for hydrolysis processes. Amino-acid composition (primary structure) affects the products. The tertiary structure defines the proteins as either fibrous or globular. Fibrous proteins are structural materials such as keratin, which is protective, and collagen, which is connective. Globular proteins are often chemically functional and act as enzymes, hormones, transport proteins, or storage proteins. Hydrolysis of proteins can be rate limiting in the overall process, depending on ease of structure degradation (Pavlostathis and Giraldo-Gomez, 1991). Protein structure is one of the main factors affecting the rate of hydrolysis. Globular proteins are rapidly hydrolyzable, while fibrous proteins are difficult to hydrolyze (McInerney, 1988). In general, all proteins apart from the most rigid type of keratin (such as the outer layer of hair and fingernails) are hydrolyzable (Figure 4). There are three main groups of proteases: serine, metallo, and acid proteases which have alkaline (8–11),

Anaerobic Processes

619

The 1,3-specific lipases can only act at the outside bonds of the triglycerides, yielding 1,2-diacylglycerols and 2-monoacylglycerols. These glyceride esters are unstable and undergo acyl migration to 1,3-diacylglycerol and 1-monoacylglycerol. Subsequently, these can be degraded further by the 1,3-specific lipase to glycerol and free fatty acids. Fatty-acid-specific lipases catalyze the removal of a specific fatty acid, preferentially removing cis-D9-monounsaturated fatty acids. Other fatty acids are degraded very slowly, especially those containing an additional double bond between D1 and D9. Figure 4 Cow hair from an anaerobic reactor showing intact keratin (A) compared with degradation of interior by anaerobic organisms (B). Photograph by Dr Damien Batstone.

CH2 OH CH

OH

CH2-O-fatty acid CH-O- fatty acid

CH2 OH

CH2-O-fatty acid

Glycerol

Triglyceride

Figure 5 Glycerol and triglycerides.

neutral (6–8), and acidic (4–6) pH optimums, respectively (Ramsay, 1997). Enzyme production may be suppressed when readably biodegradable substrates such as glucose or amino acids are supplied (Patterson-Curtis and Johnson, 1989; Ramsay, 1997). 4. Hydrolysis of lipids. Lipids are glycerol bonded to LCFAs, alcohols, and other groups by an ester or ether linkage (Madigan et al., 2009). Fats and oils have all the alcohol groups esterified with fatty acids as shown in Figure 5 and these form the bulk of glyceridic material in mixed oils and fat with other glyceridic compounds, usually a result of processing. Hydrolysis is catalyzed by LCFA ester hydrolases, called lipases. These act at the lipid–water interface in enzymatic hydrolysis to degrade the insoluble reactant to soluble products. There is little work on degradation of lipids in anaerobic environments when compared with that on carbohydrate and protein substrates. Most of this has been focused on the rumen, reviewed by (McInerney, 1988). One particular characteristic of lipases is increased activity with insoluble rather than soluble lipids (Martinelle and Hult, 1994), indicating that the activity of lipases increases greatly when the concentration of triglycerides reaches saturation and forms a second phase. The lipases are adsorbed at the interface. As there is an adsorption mechanism, combined reaction and adsorption rate may be dependent on the surface area of the insoluble triglycerides. Bacterial lipases can be divided into three main types: nonspecific lipases, 1,3-specific lipases, and fatty-acid-specific lipases (Finnerty, 1988). Nonspecific lipases can hydrolyze any fatty acid triglyceride regardless of structure, acting at any of the fatty acids. These can completely hydrolyze the ester bonds acting equally at all alkyl sites.

4.17.1.1.2 Fermentation/acidogenesis Fermentation and acidogenesis refer to the same process of conversion of sugars and amino acids to simpler compounds (mostly acids and alcohols). Fermentation is commonly applied in biotechnology processes where the focus is on the product. Acidogenesis is applied in wastewater processes. In our opinion, fermentation is a more precise and preferred term. Fermentation is defined as the conversion of organics without an obligate external electron acceptor to produce both reduced and oxidized products. The two major groups of compounds subject to fermentation under anaerobic conditions are sugars and amino acids, which are discussed next. Fermentation of sugars. Anaerobic fermentation from sugars is likely the most widely applied biotechnology process worldwide. It is used to produce food products, renewable fuels, pharmaceuticals, and industrial chemicals. It is currently in focus for production of biofuels (e.g., ethanol and butanol). Historically, fermentation has been carried out by pure or specialized microbial cultures, which are constrained to produce specific products from sugars, based on their physiology and genetic capabilities. In anaerobic digestion processes, fermentation is mediated by mixed culture, and a wide range of potential products can be formed. Sugars ferment via the Embden–Meyerhof–Parnas (EMP) pathway to pyruvate, and subsequently to C3 products (propionate or lactate), or C2–C6 products via acetyl-CoA (Madigan et al., 2009; Figure 6). The most common products are shown in Figure 6, as determined in practical mixedculture fermentation tests (Ren et al., 1997; Temudo et al., 2008). Smaller amounts of additional compounds, including metabolic intermediates, are also often detected. Actual product mixes are regulated by a number of environmental conditions, including pH, gas-phase hydrogen concentration, temperature, and biomass retention time. It is reasonable to assume that hydrogen-rich reactions (e.g., production of acetate) would be enhanced at low hydrogen concentrations and production of alcohols enhanced at low pH (Ren et al., 1997). Regulation of mixed-culture fermentation is exciting, as it offers the possibility of producing fuels and industrial chemicals directly from raw feedstocks such as crop residues and straw. While a number of models have been proposed (Costello et al., 1991; Mosey, 1983; Rodrı´guez et al., 2006), none of these can effectively describe the mixture of products under dynamic conditions. The most promising current approach evaluates the thermodynamic driving forces under varying conditions (Rodrı´guez et al., 2006).

620

Anaerobic Processes 1glucose

4e−

2pyruvate

4e−

4e−