Advances in Agronomy, Volume 38

ADVANCES IN AGRONOMY VOLUME 38 CONTRIBUTORS TO THIS VOLUME MARTINALEXANDER ANN P. HAMBLIN N. J. BARROW S. SANKARA...

Author: Nyle C. Brady

29 downloads 1037 Views 19MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

ADVANCES IN

AGRONOMY VOLUME 38

CONTRIBUTORS TO THIS VOLUME MARTINALEXANDER

ANN P. HAMBLIN

N. J. BARROW

S. SANKARAN

MARIONF. BAUMCARDNER

LEROYF. SILVA

LARRYL. BIEHL

DONALD L. SPARKS

S. K. DE DATTA

ERICR. STONER

A. S. EL-SEBAAY

B . B . TRANCMAR

FEI HUAILIN

G . UEHARA

Guo XIANYI

0. VAN CLEEMPUT

R.

s. YosT

ADVANCES IN

AGRONOMY Prepared in Cooperation with the AMERICAN SOCIETY OF AGRONOMY

VOLUME 38 Edited by N. C. BRADY Science and Technology Agency for International Development Department of State Washington, D. C .

ADVISORY BOARD

H. J. GORZ,CHAIRMAN E. J. KAMPRATHT. M. STARLING

J. B. POWELL J. W. BIGGAR

M. A. TABATABAI R . A. BRIGGS,Ex OFFICIO, ASA Headquarters

1985

ACADEMIC PRESS, INC. Harcourt Brace Jovanovich, Publishers

Orlando San Diego New York Austin London Montreal Sydney Tokyo Toronto

COPYRIGHT @ 1985 BY ACADEMIC PRESS. INC. ALL RIGHTS RESERVED. NO PART OF THIS PUBLICATION MAY BE REPRODUCED OR TRANSMI?TED IN ANY FORM OR BY ANY MEANS, ELECTRONIC OR MECHANICAL, INCLUDING PHOTOCOPY, RECORDING, OR ANY INFORMATION STORAGE AND RETRIEVAL SYSTEM, WITHOUT PERMISSION IN WRITING FROM THE PUBLISHER.

ACADEMIC PRESS, INC. Orlando, Florida 32887

United Kingdom Edition published by

ACADEMIC PRESS INC. (LONDON) LTD. 24-28 Oval Road, London NWI 7DX

LIBRARY OF CONGRESS CATALOG CARD NUMBER 50-5598 ISBN 0-12-000738-X PRINT20 IN THE UNITED STATFS OF AMERICA

85868788

9 8 7 6 5 4 3 2 1

CONTENTS CONTRIBUTORS ............................................ PREFACE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

ix xi

REFLECTANCE PROPERTIES OF SOILS

Marion F. Baumgardner, LeRoy F. Silva, Larry L. Biehl, and Eric R. Stoner I. Soil Color in Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2 5 13 28 33 39

.................. ................. IV. Reflectance Properties of Soils in Their Environment . . . . . . . . . . . . . V. Applications of Soil Reflectance Measurements . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. Instrumentation for Reflectance Measurements 111. Effects of Soil Constituents on Soil Reflectance

APPLICATION OF GEOSTATISTICS TO SPATIAL STUDIES OF SOIL PROPERTIES

B. B. Trangmar, R . S. Yost, and G. Uehara

...........

I. Introduction . . . . . . . . . II. Nature of Soil Variability . 111.

IV. V.

VI. VII.

45 47 49 53 56 70 89 91

......................... Traditional Methods of Describing Soil Variability . . . . . . . . . . . Regionalized Variable Theory and Geostatistics . . . . . . . . . . . . . . . . . . Analysis of Spatial Dependence . . . . . . . . . . . . . . . . . . . . . . . . Interpolation by Kriging . . . . . . . . ....................... Perspectives: Future Use of Geosta n Soil Research . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ........

THE INFLUENCE OF SOIL STRUCTURE ON WATER MOVEMENT, CROP ROOT GROWTH, AND WATER UPTAKE

Ann P. Hamblin I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. Soil Structure: Components of the Soil-Pore System . . . . . . . . . . . . . 111. Stability of the Pore System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV. Water Flow in Agricultural Soils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Patterns of Root Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI. Water Uptake by Roots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V

95 96 107 114 127 144

vi

CONTENTS

VII . Speculation: Are We Measuring and Averaging at Consistent Scales? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII . summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

149 151 152

GASEOUS HYDROCARBONS IN SOIL

0. Van Cleemput and A . S . El-Sebaay I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. Formation. Transformation. and Importance of Gaseous Hydrocarbons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111. Environmental Factors Affecting the Evolution of the Gaseous Hydrocarbons in Soil ....................................... IV . Sampling and Analysis of the Gaseous Hydrocarbons . . . . . . . . . . . . . V. Some Physical and Chemical Properties of the Gaseous Hydrocarbons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

159 160 167 176 178 178 178

REACTION OF ANIONS AND CATIONS WITH VARIABLE-CHARGE SOILS

N . J . Barrow I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. The Development of Charge on Variable-Charge Surfaces . . . . . . . . . Ill . Adsorption on Variable-Charge Surfaces ....................... 1v. Rates of Adsorption and Desorption ........................... V . Transferring the Variable-Charge Models to Soils . . . . . . . . . . . . . . . . VI . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

183 185 186 207 211 227 228

KINETICS OF IONIC REACTIONS IN CLAY MINERALS AND SOILS

Donald L . Sparks I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I1. Methodologies Used in Kinetic Studies ........................ nI . Application of Chemical Kinetics to Soil Solutions . . . . . . . . . . . . . . . IV . Rate-Determining Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V . Kinetics Models ........................................... Kinetics of Ionic Exchange VI ............... " in Clav . , Minerals ....................

231 233 238 251 256 258

CONTENTS

VII . Kinetics of Ionic Reactions in Heterogeneous Soil Systems . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vii 261 264

ENHANCING NITROGEN FIXATION BY USE OF PESTICIDES: A REVIEW

Martin Alexander I . Introduction

........................................... .... ............................... I n . Free-Living Heterotrophs . . . . . . . . . . . . ........ IV . Blue-Green Algae in Flooded Soils ........................... V . Resistant Isolates .......................................... VI . Choice of Pesticides and Inocula .............................. VII . Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII . Summary and Conclusions ............................. References ................................................ II. Rhizobium

267 269 273 274 276 277 278 280 281

WEEDS AND WEED MANAGEMENT IN UPLAND RICE

S . Sankaran and S . K . De Datta I. I1 . III . IV . V. VI .

Introduction ............................................... Weed Flora of Upland Rice . . ......................... Ecology of Upland Rice Weeds ......................... Weed Competition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Land Preparation and Crop Establishment Techniques . . . . . . . . . . . . Fertilizer Application and Weed Management . . . . . . . . . . . . . . . . . . . VII . Soil Moisture-Herbicide Relationships in Upland Rice . . . . . . . . . . . . VIII . Weed Control Methods in Upland Rice ........................ IX . Yield Response of Rice to Herbicides and Herbicide Combinations . . X . Phytotoxicity of XI . Economics of W XI1 . Critical Research Needs . Appendix: Common Names and Chemical Formulas of Herbicides . . References . . . . . .

284 285 294 295 303 306 310 313 323 323 327 328 330 330

RICE-BASED CROPPING SYSTEMS AND THEIR DEVELOPMENT IN CHINA

Guo Xian Yi and Fei Huai Lin I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

II. Environmental Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

340 340

...

Vlll

CONTENTS

111. Division of Rice Belts ...................................... IV. Reformation and Achievements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Problems in Multiple-Cropping Systems ........................ VI. Approaches to Solving the Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

INDEX. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

345 350 353 358 368 369

CONTRIBUTORS Numbers in parentheses indicate the pages on which the authors' contributions begin.

MARTIN ALEXANDER (267), Department of Agronomy, Cornell University, Ithaca, New York 14850 N. J. BARROW (183), CSIRO Division of Animal Production, Wembley, Western Australia 6014, Australia MARION F. BAUMGARDNER ( l ) , Purdue University, West Lafayette, Indiana 47907 LARRY L. BIEHL (l), Purdue University, West Lafayette, Indiana 47907 S. K. DE DATTA (283), Department of Agronomy, International Rice Research Institute, Manila, Philippines A. S. EL-SEBAAY (159), Faculty of Agriculture, University of Ghent, B-9000 Ghent, Belgium FEI HUAI LIN (339), China National Rice Research Institute, Hang Zhou, Zhe Jiang, People's Republic of China GUO XIAN YI (339), Crop Breeding and Cultivation Institute, Chinese Academy of Agricultural Sciences, Beijing, People's Republic of China ANN P. HAMBLIN* (959, Western Australian Department of Agriculture, South Perth, Western Australia 6151, Australia S. SANKARAN (283), Department of Agronomy, International Rice Research Institute, Manila, Philippines LEROY F. SILVA ( I ) , Purdue University, West Lafayette, Indiana 47907 DONALD L. SPARKS (231), Department of Plant Science, College of Agricultural Sciences, University of Delaware, Newark, Delaware 1971 7 ERIC R. STONER' ( I ) , Cornell University, Department of Agronomy, Ithaca, New York 14850 B . B . TRANGMAR (45), Soil Bureau, Department of Scientific and Industrial Research, Christchurch, New Zealand G. UEHARA ( 4 3 , Department of Agronomy and Soil Science, College of Tropical Agriculture and Human Resources, Universizy of Hawaii, Honolulu, Hawaii 96822 'Present address: CSIRO, Dryland Crops and Soils Research Program, Wembley P . O . ,Western Australia 6014. Australia. 'Present address: Cornell UniversityiTropSoils, EMBRAPAKPAC. Caixa Postal 70.0023, 73.300 Planaltina, D.F., Brazil.

ix

CONTRIBUTORS

X

0. VAN CLEEMPUT (159), Faculty of Agriculture, University of Ghent,

B-9000 Ghent, Belgium

R. S . YOST (43,Department of Agronomy and Soil Science, College of Tropical Agriculture and Human Resources, University of Hawaii, Honolulu, Hawaii 96822

PREFACE This volume continues the international focus of Advances in Agronomy. Scientists from six countries are among the authors of the nine papers included. They remind us of the universality of agronomic problems and of the attempts of scientists to solve them. Four of the contributions are concerned directly or indirectly with the physical properties of soils. One describes how the spectral properties of soils and their vegetation are being used in satellite imagery programs. A second summarizes our knowledge of spatial variability of soils-a topic of both practical and scientific significance. Recent investigations on the effects of soil structure on water movement and root growth are covered in a third. The fourth is concerned with the quantity and nature of hydrocarbons in the soil air, a subject of considerable importance, especially in soils that are not too well aerated. Chemical reactions in soils and clays are reviewed in two other chapters. One focuses on the reactions taking place on variably charged surfaces, a relatively understudied subject and one of special significance in some soils of the tropics. The application of chemical lunetics of soil systems is covered in a second, which reviews our understanding of the mechanisms of ionic reactions in soils and clays, and the rates at which they occur. The surprisingly positive role of certain pesticides on nitrogen fixation in soils is reviewed in another contribution. Research suggests that certain soil organisms that prey on or compete with the nitrogen-fixing bacteria are controlled by these pesticides, thereby freeing the bacteria to fix more nitrogen. Of the two chapters that specifically consider crop production, one focuses on weed control, namely research on several methods of weed control in dryland (upland) rice. This information will be especially helpful to scientists in developing countries where weeds severely restrict dryland rice production. The second crop production chapter provides a brief review of rice-based cropping systems in the People’s Republic of China. It gives a description of certain extremely intensive farming systems used in China, some of which have been modified in recent years in response to economic and agronomic factors. Our appreciation is expressed to these crop and soil scientists for providing their interesting reviews.

N. C . BRADY

xi

This Page Intentionally Left Blank

ADVANCES IN AGRONOMY. VOL 38

R EFLECTAN CE PROP ERTl ES OF SOILS Marion F. Baumgardner. LeRoy F. Silva. Larry L. Biehl. and Eric R . Stoner2y3 Purdue University. West Lafayette. Indiana 2Cornell University. Department of Agronomy. Ithaca. New York

I . Soil Color in Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A . Visible Reflectance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Nonvisible Reflectance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Need for Quantitative Reflectance Measurements . . . . . . . . . . . . . . . . . I1. Instrumentation for Reflectance Measurements . . . . . . . . . . . . . . . . . . . . . A . Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Geometrical Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I11. Effects of Soil Constituents on Soil Reflectance . . . . . . . . . . . . . . . . . . . . . A . Moisture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Organic Matter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C . Particle Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Ironoxides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E. Mineral Composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . F. Soluble Salts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . G. Parent Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . H . Other Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV. Reflectance Properties of Soils in Their Environment . . . . . . . . . . . . . . . . . A . Atmospheric Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Physical Surface Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Surfacecover . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D . Surface Expression of Subsoil Characteristics . . . . . . . . . . . . . . . . . . . E. Sensor Data Dimensionality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V . Applications of Soil Reflectance Measurements . . . . . . . . . . . . . . . . . . . . . A . SoilSurvey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Soil Degradation Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Soil Information Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2 2 4 4 5 5 8 11 13 14 20 21 23 25 26 27 27 28 28 29 30 31 32 33 33 38 38 39

Present address: Cornell UniversityflropSoils. EMBRAPA/CPAC. Caixa Postal 70.0023, 73.300 Planaltina. D.F., Brazil. 1

Copyright 0 1985 by Academic Press. Inc. All rights or reproduction in any form reserved.

2

MARION F. BAUMGARDNER ET AL.

I. SOIL COLOR IN PERSPECTIVE

A. VISIBLEREFLECTANCE In delineating differences between soils and in describing the characteristics of a soil profile, color is one of the most obvious and useful attributes for documenting these differences. Soil visible reflectance, or color, is a differentiating characteristic for many classes in all modern soil classification systems and is an essential part of the definitions for both surface and subsurface diagnostic horizons. Although most of the differentiating characteristics selected as diagnostic criteria in the soil classification process are verifiable by instrumental procedures, soil color is still most commonly determined by a human observer making a visual comparison between a given sample and the various color chips in an array of artificially produced colors, arranged according to hue, value, and chroma (Soil Survey Staff, 1975). Early in this century, Munsell, a Boston artist and art teacher, developed a color notation (Munsell, 1947). The “Munsell Book of Color” and Munsell standard color papers were made and published by the Munsell Color Company for many years (Nickerson, 1940).Munsell’s work was later to play an important role in the standardization of soil color designations. For more than 60 years soil scientists have been studying methods of improving the measurement of soil color. In 1920 the American Association of Soil Survey Workers (later the American Soil Survey Association, which merged with the Soil Section of the American Society of Agronomy to form the Soil Science Society of America) formed a committee to establish standards for soil colors (Rice et a/., 1941). During the 1920s and 1930s soil color research was conducted by scientists in the U.S. Department of Agriculture (USDA) and at several state experimental stations (Hutton, 1928, 1932; Kellogg, 1937; Shaw, 1937; O’Neal, 1923). In 1927 a laboratory devoted to the measurement of color and its application to the grading of agricultural products was developed in the U.S. Department of Agriculture in Washington, D. C. (Nickerson, 1946). T. D. Rice and other soil scientists used the laboratory facilities in the study of soil color problems in a search for better soil color measurements and descriptions. In this laboratory, under controlled lighting conditions, a set of 250 soil samples was carefully selected by Rice and his co-workers to cover the range of soil color. The 250 soil samples were measured in the color laboratory in terms of the Munsell color notations. In 1941 an important bulletin published by the USDA summarized the

REFLECTANCE PROPERTIES OF SOIL

3

results of soil color research of the preceding 20 years (Rice et al., 1941). During this period significant progress was made in two areas: (1) the selection of an array of soil samples which would cover the range of all possible soil colors and (2) development of color names or descriptions to describe the color of a soil. In an experiment in 1939 using the 250 soil samples from the USDA collection, more than 50 U.S. soil scientists and several from other countries, all with field experience, were asked to name and record the colors of these samples. Although the names used generally followed the nomenclature suggested in the Soil Survey Manual, there was little agreement on an exact color designation for each sample. In fact, the few observers who were requested to repeat the exercise were unable to duplicate completely their original color designations, though there were no great discrepancies. The next step was to find a system of color names that could be sufficiently standardized to be acceptable to color scientists, sufficiently useful to satisfy the needs of soil scientists, and sufficiently commonplace to be understood by the users of soil information. This search for standardized color names resulted in the tentative adoption of the names used in the Inter-Society Color Council/National Bureau of Standards (ISCC-NBS) method (Judd and Kelly, 1939). The assignment of an ISCC-NBS color designation to each of the samples in the USDA collection resulted in a total of only 56 color names. From this research 56 samples were prepared to represent the central color of each of the designations in the soil color range. Charts, carrying 56 different color chips, were published as the “Soil Color Name Charts,” which accompanied the “Preliminary Color Standards and Color Names for Soils” developed by Rice and his colleagues (Rice et al., 1941). With use of and experience with color chip matching, it naturally followed that refining and updating of the method would occur. During the 1940s a committee of the US. Soil Survey, chaired by E. H. Templin, replaced the earlier charts with a much wider selection of colors. The Templin committee used 202 regular Munsell standards instead of the 56 special colors in the early charts. They also made adjustments in the names (Pendleton and Nickerson, 1951). Today the standard “Munsell Soil Color Charts” are published on charts representing 7 different hues and containing 99 different standard color chips (Munsell Color, 1975). Great progress was made during the period from the 1920s to the 1950s in the standardization of methods of measuring and designating soil color. Today the Munsell soil color notations are widely used throughout the world. However, the fact remains that the designation of soil color as normally made in the field or laboratory is subjective and nonquantitative.

4


B. NONVISIBLE REFLECTANCE Remarkable improvements and changes have occurred over the past three decades in the development of laboratory and field instrumentation for observing and measuring physical and chemical phenomena. An area of development as it relates to soil color is an array of new instruments which scan a wide range of the electromagnetic spectrum and record quantitatively the intensity of energy radiating from a specific material or scene. In soil studies it is possible to measure soil reflectance in the laboratory or in situ and obtain spectral curves which plot intensity of reflectance in the ultraviolet, visible, and infrared portions of the spectrum. Although spectrometers have been used by analysts in the laboratory for many years, new designs of instruments have extended the use of spectroradiometers to many new applications. One of the driving forces of some of these applications has been the kinds of sensors which have been and are being designed for earth observation systems, primarily involving sensors on aerospace platforms. Since atmospheric attenuation severely limits the use of ultraviolet measurements from such platforms, this article will not discuss the application of ultraviolet radiation to the study of soils. On the other hand, the atmosphere is a relatively open window to the longer visible wavelengths and to infrared reflectance (Gates, 1962, 1963). For this reason special attention is given to visible and infrared (nonvisible) reflectance. With the increasing availability and continuing improvement of these spectroradiometers during recent years, there has been an expanding interest among soil scientists in developing techniques to obtain more precise quantitative reflectance (visible and infrared) measurements of soils (Baumgardner and Stoner, 1982; Cipra et al., 1971b; Condit, 1970, 1972; DaCosta, 1979; Gausman et al., 1977; Karmanov, 1970; Montgomery, 1976; Obukhov and Orlov, 1964; Stoner, 1979).

c. NEEDFOR QUANTITATIVE REFLECTANCEMEASUREMENTS Ever since soil science evolved into an important discipline for study and research, color has been one of the most useful soil variables in characterizing and describing a particular soil (Kohnke, 1968; Pendleton and Nickerson, 1951; Soil Survey Staff, 1975). The quantity and quality of soil components and the variable conditions under which soils are observed affect soil color. Our commonly used “measurement” of soil reflectance, usually confined to the visible, is at best semiquantitative. Both in the field and the laboratory the assignment of a specific soil to a specific Munsell notation or category is


5

subjective and is limited to the visible portion of the spectrum and by the number of Munsell color chips. Numerous studies in recent years have shown relatively high correlations between soil reflectance and certain other physical and chemical properties of soils (Baumgardner and Stoner, 1982; Da Costa, 1979; Montgomery, 1976; Pazar, 1983; Stoner, 1979; Stoner and Baumgardner, 1981). It has also been noted that the environmental conditions under which soils have been formed affect soil reflectance (Montgomery, 1976; Stoner, 1979). If these relationships among soil reflectance and chemical and physical properties can be established quantitatively and definitively for given environmental conditions, the capacity to extract useful soils information from sensor data obtained by current and future earth observation satellite systems will be greatly enhanced.

II. INSTRUMENTATION FOR REFLECTANCE MEASUREMENTS A. NOMENCLATURE Reflective optical radiation is defined as propagating electromagnetic energy with characteristic wavelengths between 0.4 and 3 pm. When +tical radiation interacts with a surface, a portion of that radiation is either absorbed in the material below the surface or is transmitted through the bulk of the material through another surface into another medium. The remainder of the radiation is said to be reflected from the surface. In general terms, the ratio of the reflected radiation to the total radiation falling upon the surface is defined as reflectance. This is contrasted to reflectivity, which is an intrinsic material property. Reflectance is the result of a measurement concerning the aforementioned ratio. In order to expedite the discussion of reflectance, it is convenient to introduce some radiometric terminology. Irradiance is the optical radiative power falling on a unit area of surface. It has the units of watts per square meter and is usually denoted by the symbol E . If the distribution of the power per unit area with respect to wavelength is being described, a related term called spectral irradiance is used. It has the units of watts/(m* - pm). The term most frequently used to describe reflected radiation is that of radiance, denoted by the symbol L. It has the units of watts/(m2 - sr), where sr is the abbreviation for the unit of solid angle, the steradian. The spectral quantity associated with radiance is called spectral radiance and carries the units of watts/(m2 - sr - pm).

6

MARION F. BAUMGARDNER ET AL. Normal to sample Incident flux

I

Viewed flux

FIG. 1. Geometric parameters describing reflection from a surface: 0, zenith angle; 4, azimuth angle; o,beam solid angle; a prime on a symbol refers to viewing (reflected)conditions.

Figure 1 illustrates the basic geometric relationships between incoming radiation and outgoing radiation using the previously described terminology. The reflecting properties of a surface are most precisely described using a parameter called the bidirectional rejectance distribution function (BRDF). The defining equation for the BRDF is

The angles are shown in Fig. 1. The BRDF is the ratio of a radiance to an irradiance; therefore, it has the units of sr-'. If the numerator and denominator of the expression are spectral quantities, then a spectral BRDF has been defined and is usually denoted by the symbol A. A careful examination of Fig. 1 reveals that the BRDF is the ratio of two differential solid angles. This is a mathematical abstraction that is closely realized by many physical situations in which the incident and reflective solid angles are small enough to approximate the differential case. The physically measured BRDF is therefore an average fr value over the parameter intervals. The incident and reflected solid angles, however, need to be small to obtain a good estimate of the true BRDF. The measurement of the BRDF is, however, a particularly difficult problem. It would be necessary to place a sensor at the surface to measure the incoming radiation and then take that sensor, or another sensor, and place it in the viewing position necessary to measure the reflected radiation. Although this represents a possible approach, an experimentally more convenient method uses a reflectance standard in the measurement procedure. The


Reference

7

Targef

FIG.2. Illustration of procedure for measuring the reflectance factor. The response of the sensor to a perfectly diffuse, ideal reference is recorded, and then the response of the sensor to a target of interest is recorded under the same illumination conditions.

geometry of this method is illustrated in Fig. 2. In this method a single sensor located in the viewing position is used to view the reflected radiation from a perfectly diffuse ideal reflector as well as that from the scene of interest. If the scene and the perfectly diffuse ideal reflecting surface are viewed and illuminated under identical conditions, the ratio of the two measurements is referred to as the reflectance factor of the scene (Nicodemus et al., 1977). If the geometry of the situation resembles that of Fig. 1, then the quantity we are measuring is referred to as the bidirectional reflectance factor (BRF). If differential solid angles are not assumed and real conical solid angles are the case, then the quantity being measured is called the biconical reflectance factor. If the conical solid angles for both incident and reflected radiation include all directions, i.e., the hemisphere, then the measurement is referred to as the bihemispherical rejectance factor. Also, the measurement may be referred to as the directional-hemispherical reflectance factor, hemisphericalconical reflectance factor, etc., depending on the instrumentation setup. In many real measurement situations the magnitudes of the solid angles involved are small enough so as to approximate the differential situation. For this reason, the results of the reflectance factor measurement are referred to as the bidirectional reflectance factor, whereas in reality they are actually the results of a biconical reflectance factor measurement. The principal advantage of the reflectance factor method of measurement is that the sensor can be kept in its viewing position and it is only necessary to

8


measure the radiation from the reflectance standard and that from the scene of interest within a short time period and then take the ratio to obtain the desired reflectance factor measurement. The method is amenable to both laboratory and field measurement situations. An important part of the measurement procedure just discussed is the reflectance standard. A perfectly diffuse reflecting surface is one that reflects equally in all directions. An ideal reflector is a surface in which all of the energy falling on the surface is reflected. Again, this is an abstraction, but physical surfaces have been prepared which approximate this ideal situation. Examples of some of the reflecting materials and coatings used for diffuse surfaces include magnesium oxide, barium sulfate (Grum and Luckey, 1968; Billmeyer et al., 1971; Shai and Schutt, 1971; Young et al., 1980), homopolymers and copolymers with fluorine substitution (Schutt et al., 1981), and canvas panels (Robinson and Biehl, 1979). Barium sulfate surfaces have received the widest acceptance. Specially prepared barium sulfate surfaces approximate a perfect diffuser at angles within 45" of a normal to the surface. The reflectance of properly prepared barium sulfate surfaces is over 0.9 for most wavelengths in the reflective spectrum. Departures from perfectly diffuse or ideally reflecting properties of the reflectance standard can be documented and accounted for during the analysis of the measurement data (Robinson and Biehl, 1979). Primary reflectance factor calibration measurements are usually made with tablets that are fabricated out of pressed barium sulfate powder. The results of these measurements are referred to standard data that have been made available by the manufacturers of this special barium sulfate reflectance powder. The resulting measurements are then used to calibrate painted barium sulfate reflectance standard panels. The pressed barium sulfate tablets can often be directly used in the laboratory, but for field situations 1.25 x 1.25 m painted panels are usually prepared. The panels are covered when not in use and are otherwise handled very carefully to avoid contamination from dust. B. GEOMETRICAL CONSIDERATIONS Figure 3 is a schematic diagram of the angles involved in a physically realizable measurement situation. In this case, the source and the sensor have real apertures which produce conical solid angles rather than differential angles as indicated in Fig. 1. In addition, the sensor has a field of view determined by its internal optics and indicated on the diagram by the angle p. The amount of reflected power gathered by the sensor is proportional to the square of the field of view, the sensor aperture area, the irradiance, the


9

n

source

'sensor field of view

'source

azimuth angle

FIG. 3. Schematic diagram of illumination and viewing angles involved in a physically realizable measurement situation.

irradiance angles, the sensor view angles, and, of course, the bidirectional reflectance distribution of the target. The angular relationships of the source, target, and sensor can significantly affect the measured reflectance of the target. For example, the reflectance measured by a sensor vertically viewing the soil in a plowed field with deep furrows will be higher when the solar azimuth angle is parallel with the furrow (no shadows) than when the solar azimuth is perpendicular to the furrows (shadows present). If the target is smooth, then the azimuthal orientation of the sensor (or source) is irrelevant. Usually laboratory samples of a soil are prepared in such a way as to be free of azimuthal variation. Most field observations are usually made at a sensor zenith angle of 0" and the data are taken at a variety of solar illumination angles. Laboratory observations are, again, usually made at a sensor zenith angle of 0" and an illumination zenith angle of 10 to 30". As previously stated, the signal power received by the sensor is proportional to the square of the field of view of the sensor. However, this is not the sole factor in the choice of the particular field of view that is going to be used in a measurement situation. As shown in Fig, 4, a narrow field of view may not properly integrate the geometric features of the scene into the signal received by the sensor. A "wide enough" field of view is necessary in order to represent properly the geometric features of a target to the sensor. Examples of such structure are the plant rows in a row crop and the roughness of a plowed field. However, if the field of view is "too wide," then it is difficult to characterize properly the sensor zenith angle. In field situations, an appropriate compromise is a field of view of approximately 15". In laboratory situations where the surface roughness of the target can be controlled, a narrow field of view such as 1" may be properly used.

10


FIG.4. Illustration of a field of view (FOV) which is large enough to represent properly the geometric features of the target and a FOV which is too small to integrate the geometricfeatures.

In some cases, the surface roughness of the target may be so extreme as to make it impossible to characterize properly the reflectance properties with one measurement. In this case, several observations of the target are made at a variety of illumination angles or viewing angles to characterize the target completely. In this way, shadowing effects which might otherwise obscure the reflectance properties of the target can be taken into consideration. This method is available only in a low-altitude field instrumentation situation. Extreme surface roughness may seriously complicate high-altitude observations from aircraft or satellites. Another factor that enters into the instrumentation geometric problem is that of the ratio of direct to diffuse illumination. On a very clear, lowhumidity day, most of the incoming illumination (in a field situation) is directly from the sun. On hazy days, a significant amount of incident illumination may be indirect due to atmospheric scattering. This is generally referred to as the diffuse component of the incident illumination. Moreover, the diffuse component is a function of the wavelength, being proportionally more at shorter wavelengths. If the diffuse component of incident illumination is a significant part of the total illumination, then it is difficult to characterize properly the zenith angle of the incident illumination. In a laboratory situation, the instrumentation system is arranged so as to negate the effects of diffuse component illumination. In a field situation, it is often possible to reduce diffuse effects by restricting data acquisition times to those in which atmospheric conditions do not give rise to extreme diffuse effects. If it is necessary to take data under hazy atmospheric conditions, then the diffuse component may be determined by shadowing the target (and standard) from direct illumination and making a reflectance factor measurement.


11

This diffuse component is then subtracted from the total illumination in order to obtain the reflectance factor associated with the direct illumination component.

C. INSTRUMENTATION Instruments that have been used to measure the reflectance of soil can generally be divided into two broad classes-spectroradiometers and multiband radiometers. The discussion that follows is a brief summary of the instrumentation. More detailed information can be found in Robinson and DeWitt (1983) and Zissis (1979). Multiband radiometers contain several optical filters to define the spectral bandpasses. These spectral bandpasses are selected to sample discrete portions of the optical spectrum, e.g., the Landsat multispectral scanner (MSS) or the thematic mapper (TM) bands. Multiband radiometers can be of two types-nonchopping (dc) or chopping (ac). The detectors in dc multiband radiometers are directly coupled to the output amplifiers. The most reliable dc multiband radiometers are limited to the spectral range of the silicon detector because detectors such as lead sulfide, which are sensitive beyond 1 pm, are not sufficiently stable. Many dc multiband radiometers being used to measure the reflectance of soils in situ contain the Landsat MSS spectral bands or the first four TM spectral bands. Some companies have built multiband radiometers that make it relatively easy for researchers to interchange different sets of optical filters. ac multiband radiometers contain a chopper in front of the detectors to allow the field of view of the detector to be alternately filled with the internal reference source and the target. The signal measuring the radiance from the target is now the ac signal. The dc signal, which is due to the instability of the detector, is removed. ac multiband radiometers generally use lead sulfide detectors to measure the radiant power in spectral bands from 1.0 to 3.0 pm, such as the last three reflective TM bands. Spectroradiometers.are distinguished from multiband radiometers because they measure flux in much narrower, continuous spectral bands. Spectroradiometers can also be chopping or nonchopping. Similar to multiband radiometers, the most reliable nonchopping spectroradiometers are limited to the silicon detector spectral range. In spectroradiometers, optical dispersion devices replace the simple optical filters. The dispersion device may be a segmented filter wheel, a circular variable filter (CVF), a prism, or grating. The differences in dispersion devices relate to wavelength resolution and spectral scan speed. Circular variable filter spectroradiometers that operate from 0.4to 2.4 pm have been widely

12


used for laboratory and in situ reflectance measurements. The wavelength resolution of these CVF instruments is generally 1-2 % of the wavelength. Spectroradiometers and multiband radiometers may also have an integrating sphere and an artificial light source attached to allow for the measurement of the directional-hemispherical reflectance factor. Many of the reflectance data cited in the literature are directional-hemispherical reflectance factor measurements. The systems described above have their own set of advantages and disadvantages. Multiband radiometers tend to be less costly to buy and operate. They are lighter and therefore easier to mount on simple pickup truck booms. Generally, one can obtain a set of multiband radiometer measurements much faster than a set of spectroradiometer measurements -fractions of a second compared to 10-120 sec. However, the spectral range and resolution of multiband radiometers are limited. For example, one can use reflectance measurements from a multiband radiometer with Landsat MSS bands to help interpret Landsat MSS data; however, the measurement could not be easily used to interpret TM data. Spectroradiometers, on the other hand, lend themselves well to obtaining sets of reflectance data which can be used to interpret many different sets of broadband satellite data, e.g., Landsat MSS and TM, the National Oceanic and Atmospheric Administration’s advanced very-high-resolution radiometer (NOAA AVHRR), and the Nimbus 7 coastal zone color scanner (CZCS). Spectroradiometers, however, tend to be costly to operate, require significant time to obtain a set of spectral measurements, and are cumbersome to move in the field. Recent advances in multilinear arrays, however, may negate some of these disadvantages in the near future. Instrumentation systems that have been used to measure the reflectance of soil can be divided into laboratory and in situ or field systems. The laboratory systems can be further divided into bidirectional reflectance factor systems and directional-hemispherical reflectance factor systems. Reflectance measurements were initially done in the laboratory using directional-hemispherical spectroradiometer systems. I n situ measurements were made of soil reflectance as field-worthy spectroradiometer systems were developed during the late 1960s and early 1970s. With the launch of the Landsat MSS scanners and the attention given to the Landsat MSS bands, multiband radiometers became a primary source of field measurements during the late 1970s and early 1980s. Laboratory measurements of reflectance also continued during this time. Instrument systems were developed to measure the bidirectional reflectance factor of soils in the laboratory (DeWitt and Robinson, 1976). Many instrumentation systems are now available to make reflectance measurements. However, to utilize and compare measurements from different systems, researchers need to follow well-defined calibration procedures and

13


describe adequately their measurements relative to illumination and viewing angle, sensor field of view, reference surface, illumination and viewing solid angles, and target surface preparation. Also, one needs to be cautious about comparing bidirectional measurements and directional-hemispherical measurements. Directional-hemispherical measurements are an integration of bidirectional measurements across all possible viewing angles.

Ill. EFFECTS OF SOIL CONSTITUENTS ON SOIL REFLECTANCE With the development of new laboratory and field spectroradiometers during the past two decades, it is now possible to measure quantitatively the effects of soil constituents on soil reflectance. One of the early studies to quantify soil reflectance and to define differences between soil reflectance spectra was conducted by Condit (1970, 1972). He made directional-hemispherical reflectance factor measurements over the range from 0.32 to 1.0 pm for 160 soil samples from 36 states in the United States. From these results he classified the spectral soil curves into three general types (Fig. 5). No attempt was made to relate quantitatively these spectral properties to other physical and chemical properties of the soils. By the mid-1970s laboratory and field instruments were available for obtaining bidirectional reflectance factor measurements over an extended range of the reflectance spectrum. Stoner and Baumgardner (1981) and Stoner et al. (1980a) reported the results of visible and infrared reflectance (0.52-2.38 pm) measurements for duplicate surface samples of more than 240 soil series obtained from 17 different temperature-moisture regimes in the 48 contiguous states in the United States. From these spectra and a limited 50 -

-__

.’

0 L

10 0

I

3

4

5

I

6 7 0 Wavelength (pm)

I

I

9

10

I

FIG.5. Soil spectral curve types defined by Condit (1970, 1972); reflectance is recorded as the directional-hemispherical reflectance factor.

14


1.2

1.6

2.0

2.4

Wavelength (pm)

FIG.6. Soil spectral curve forms defined by Stoner and Baumgardner (1981); reflectance is recorded as the bidirectional reflectance factor. curve a, Organic dominated; curve b, minimally altered; curve c, iron affected; curve d, organic affected; curve e, iron dominated.

number of reflectance spectra of soils from other parts of the world, they derived five soil spectral curve forms (Fig. 6 ) to which they established some general genetic, physical, and chemical relationships to each spectral curve form (Table I). This capability to obtain calibrated data across the visible and infrared reflectance spectrum provides an important new tool to soil scientists (Cipra et al., 1971a). The remainder of this section will present the results of observations made on those soil constituents which account for most of the variation in soil reflectance. A. MOISTURE It is a common observation that most soils appear darker when wet than when dry. This results from decreased reflectance of incident radiation in the visible region of the spectrum. Evans (1948), who presented reflectance curves for three soils in both the wet and dry state, found that the wet samples showed lower reflectance. Unfortunately, he provided no information about the soil series or moisture contents of the soils. Brooks (1952) reported 10% reflectivity of moist Yo10 fine sandy loam over the wavelength range of 0.4 to 2.5 pm but did not indicate the moisture content; for the dry condition, he reported a reflectivity of 30%. Kojima (1958b) studied the effect of moisture content on the color of 16 soils. His results, reported in Munsell color notation, also showed a decrease in reflectance with an increase in moisture. He made no reference to energy changes related to reflectance and moisture content. In “Soil Taxonomy,” the Soil Conservation Service standard for soil survey over much of the world, the range of change in soil color upon wetting is given as varying between 1/2 and 3 Munsell color steps (Soil Survey Staff, 1975). No formulas are proposed for predicting change in color between the

Table I Characteristics of Surface Samples of the Five Mineral Soils Represented by the Spectral Curve Forms in Fig. 6#

Reflectance curve form

Soil series Horizon sampled Soil subgroup Sample location Climatic zone Parent material Drainage class Textural class Moist soil color Munsell color Contents: Organic matter Iron oxide Moisture at 0.1 bar tension Mineralogy ~

a

~~

Iron

Organic affected (4

Iron

Organic dominated (a)

Minimally altered (b)

affected

Drummer AP Typic Haplaquoll Champaign Co., Ill., USA Humid mesic Loess over glacial drift Poorly drained Silty clay loam lOYR 211 Black

Jal All Typic Calciorthid Lea Co., N. Mex., USA Semiarid thermic Fine-textured alluvium or lacustrine Well drained Loamy fine sand lOYR 513 Brown

Talbott AP Typic Hapludalf Rutherford Co., Tenn., USA Humid thermic Clayey limestone residuum Well drained Silty clay loam 1.5YR 416 Strong brown

Onaway AP Alfic Haplorthod Delta Co., Mich., USA Humid frigid Glacial drift

(Not given) AP Typic Haplorthox Londrina, Parana, Brazil Humid hyperthermic Basalt

Well drained Fine sandy loam 7.5YR 312 Dark brown

Excessively drained Clay 2.5YR 316 Dark red

0.59% 0.03% 17.0%

1.84% 3.68% 28.2%

3.3% 0.8 1% 27.3%

2.28% 25.6% 33.1%

Mixed

Mixed

Mixed

Commonly kaolinitic

5.61 %

0.76% 41.1%

Commonly montmorillonitic ~

From Stoner and Baumgardner (1981)

(c)

dominated (e)

16


wet and the dry states. Soil surveyors correct for differences in moisture by comparing soil color with Munsell standards at the two soil moisture levels known as air dry andjeld capacity. The directions for a wet reading specify color at field capacity as the estimated color observed after moistening a sample and comparing the color with that of Munsell standards as soon as the visible moisture films have disappeared. Angstrom (1925) attributed this darkening effect of moisture in soils to internal total reflections within the thin water film covering soil particles. It was felt that a portion of the energy would not be reflected to space but would be re-reflected between the surface of the particle and the surface of the water film. Planet (1970) indicated that the reflectance difference of a soil between its dry and wet states could be determined if the following factors were taken into account: (1) variations in the index of refraction of the water due to dissolved soil constituents, reflectance being known to decrease with an increase in the index of refraction of the transmitting medium; (2) changes in the physical nature of soil particles by the presence of water; and (3) similarities in the indices of refraction of the soil and water leading to the Christiansen effect. Hoffer and Johannsen (1969) showed that moist soils had an overall lower reflectance than their dry counterpart in the 0.4-2.5-pm wavelength region. Bowers and Hanks (1965) noted a lowering in reflectance for Newtonia silt loam (Typic Paleudoll) at six increasing soil moisture contents over the wavelength range from 0.5 to 2.5 pm. Hunt and Salisbury (1971) found the spectral reflectance curve of montmorillonite to be similar to that of the Newtonia silt loam used in the Bowers and Hanks study. The spectral reflectance curve of montmorillonite was found to be dominated by very strong absorption bands at 1.4 and 1.9 pm, which were assumed to be caused by “bound” water typical of montmorillonite. Usually a weaker band was noted at 1.16 pm and was thought to be due to adsorbed water. The major spectral reflectance features of kaolinite were found to be very strong hydroxyl bands in the “near” infrared centered near 1.4 and 2.2pm (Hunt and Salisbury, 1971). As would be expected, the low amount of bound water present resulted in weakness of the band at 1.9 pm. Bowers and Smith (1972) showed that soil water content could be measured by transmitting a beam of 1.94-pm energy through a methanol-soil extract. Later Bowers et al. (1975) improved the design of a spectrophotometer to make such measurements. Obukhov and Orlov (1964) observed that the spectral curve does not change in appearance upon wetting of soil and that the ratio of the reflectance of moist soil to that of dry soil remained practically constant in the visible portion of the spectrum. It was also noted that the decrease in reflectance was greater upon wetting of forest soils containing little organic matter than upon


17

wetting of prairie soils high in organic matter. Condit (1970, 1972), in his study of 160 soils from 36 states in the United States, was able to identify three characteristic shapes of reflectance curves in the 0.32- 1.0 pm wavelength range. Although the curve shape was not noted to change between dry and wet soil reflectance readings, the soil moisture content was not reported for any of the soil samples. The need for carefully controlled moisture tension equilibria and soil moisture content determination in soil reflectance studies was emphasized by Beck et al. (1976). The shape of soil reflectance curves is affected by the presence of strong water absorption bands at 1.45 and 1.95 pm, and occasionally weaker water absorption bands at 0.97, 1.2, and 1.77pm. Specifically, these bands are overtones and combinations of the three fundamental vibrational frequencies of the water molecule which occur beyond 2.5 pm (Bowers and Smith, 1972). The band at 1.94pm, a combination of the v 2 and v 3 fundamental frequencies, is the most sensitive to water and has been found best for relating reflectance measurements to soil moisture content (Bowers and Hanks, 1965). An absorption band at 2.2pm was not identified in early studies but was later identified as a vibrational mode of the hydroxyl ion (Hunt and Salisbury, 1970). Absorption due to the hydroxyl ion also gives rise to a band at 1.45 pm, the same as that of liquid water. The appearance of the 1.45-pm band without the 1.95-pm band indicates that hydroxyl groups and not free water are present in the material. Sharp bands at 1.45 and 1.95-pm indicate that the water molecules are located in well-defined, ordered sites, while broad bands at these wavelengths indicate that they are relatively unordered, as is often the case in naturally occurring soils (Fig. 7). Weak absorption bands at 0.97, 1.2, and 1.77 pm correspond to the absorption bands observed in transmission spectra of water of a few millimeters in thickness (Lindberg and Snyder, 1972). These weak water absorption bands can be assigned in terms of water vibrations, with the additional combination of librations of the water molecule in the 1.77-pm band (Hunt et al., 1971a).

.8

1.2 1.6 2.0 Wavelength (pn)

2.4

FIG.7. Spectral curves from duplicate samples of a highly reflective soil annotated with prominent iron and water absorption bands; reflectance is recorded as BRF. -, sample 1; ---, sample 2.

18


Beck et al. (1976) found that of several factors studied, soil moisture had the greatest influence on soil reflectance at the one-third bar moisture tension level. Because of the impracticality of measuring soil reflectance in the field in the 1.95-pm water absorption band (also a region of strong atmospheric water absorption), reflectance in the 1.50-1.73-pm wavelength region was suggested as the best possibility for mapping water content in surface soils. Although the importance of soil moisture to reflectance was recognized by Montgomery and Baumgardner (1974) and Montgomery (1976), the contribution of this parameter to soil reflectance was not evaluated quantitatively because of the air dry state in which all of the soil samples were measured. Stoner (1979) found soil moisture content to be the most important variable for explaining reflectance differences in the 2.08-2.32-pm band, similar to one of the middle infrared bands of the thematic mapper scanner of Landsats 4 and 5. Bowers and Smith (1972) showed that a linear relation between absorbance and percentage soil water was adequate for moisture determination from air dry to the moisture equivalent. Peterson et al. (1979) assumed that the integrated effect of all factors contributing to soil reflectance would exhibit a constant shift in reflectance for a given soil observed under different moisture tensions. They compared the reflectance values at 0.71 pm for 15 different surface samples of Alfisols and Mollisols in Indiana. A plot of the bidirectional reflectance factor of soils equilibrated at 15 bar moisture tension, R,,, bar), versus the bidirectional reflectance factor of oven dry soils, Rs(ovendry), for these 15 soils gave an r2 value of 0.95. The equation for predicting RS(,bar) from Rs(ovendry) data at 0.71 pm is Rs(15 bar)

= 1.685 +

1.067Rs(oven

dry)

(2)

The equation for calculating the reflectance at 0.71 pm for soils at 0.3 bar moisture tension, R s ( 0 . 3 bar), from dry) data is & ( 0 . 3 bar)

= Oe709-k 0*487Rs(oven

dry)

(3)

This work was followed by a study of the relation of wetting to soil reflectance for 57 soils supplied by the Soil Conservation Service (SCS) from selected sites of soil series benchmark soils within the 48 contiguous states in the United States (Peterson, 1980). Samples were selected to give a wide range of soil reflectance values. Regression values of loss in reflectance upon wetting to 0.1 bar versus oven-dry reflectances varied with the wavelength band used. The r2 value for reflectance at 0.76 to 0.90 pm was 0.65, that for 0.45 to 0.52 pm was 0.79, that for 1.42 to 1.52 pm was 0.92, and that for 1.92 to 2.02 pm was 0.96 (Fig. 8). Thus, it was found that the effect of soil moisture on reflectance among soils with different chemical and physical properties could be generalized if


50 0

19

r

0 40 P

r 0

Y

v)

Y

30

n

2

20

9

s g 10 0;

I0

io

io

60

!50

$0

RS(oven dry) (%)

FIG. 8. Relationship between middle-infrared (1.92-2.02 pm) reflectance of oven-dry soil and middle-infrared reflectance of oven-dry soil minus middle-infrared reflectance of soil at 0.1 bar moisture tension. 2 = 1.92-2.02 pm; r2 = 0.96.

the soil moisture content were related to the energy with which it is held by the soil rather than expressed in percentage of the dry weight of the soil (Peterson et al., 1979). This approach has long been used by soil and plant scientists to relate in a universal way the relationships of soil moisture to plant growth and plant behavior. Soil scientists have capitalized on the fact that as soils become progressively drier below the field capacity level, the soil water is held predominantly in an ever thinner film on the particle surfaces and with an ever greater force. Awareness of this phenomenon has led soil scientists to describe soil water in terms of the work per unit mass of water that has to be done to change the soil water’s energy status to that of pure free water (Kohnke, 1968). The most common term used to express the free energy level of soil moisture is soil moisture tension, expressed in centimeters or bars, where 1 bar = 1 x lo6 dyn/cm2 or 0.98692 atm. Using such measurements the relation of soil moisture content to plant growth can be readily generalized. For example, all plants will vary closely to the soil water tension a t the “wilting point” or at approximately 15 bar, regardless of the kind of soil or the amount of water present by volume or weight. A basis for this assumption is the long-held view of soil scientists, deduced from indirect evidence, that in soils drier than field capacity the water films around the particles are of uniform thickness for all soils when they are at a uniform tension. Evidence of this was presented by Low (1980), who showed that at uniform spacing of clay particles in a water suspension the suction (tension) on the clay is the same for all clays. This has lately been verified by further evidence now being prepared for publication. This, together with the view that the thickness of

20


Ol.4

".8*

"

1.2 '

"

1.6

"

"

2.0 '

"

'

" 2.4

Wavelength (urn)

-,

FIG.9. Spectral (BRF) curves of a Typic Hapludalf soil at four different moisture tensions: oven dry; --, 15 bar; ---,0.3 bar; * - - , 0.1 bar.

the water on the surface of a particle influences reflectance (Rao and Ulaby, 1977; Strandberg, 1968), provides reason for expecting the influence of soil moisture on reflectance to be readily generalized on the basis of soil moisture tension (Fig. 9).

MATTER B. ORGANIC Soil organic matter content and thL composition of organic constituents are known to have a strong influence on soil reflectance. A general observation has been that as organic matter content increases, soil reflectance decreases throughout the 0.4-2.5-pm wavelength range (Hoffer and Johannsen, 1969). Baumgardner et al. (1970) found that organic matter content plays a dominant role in bestowing spectral properties to soils when the organic matter content exceeds 2.0%. As the organic matter content drops below 2.0%, it becomes less effective in masking the effects on reflectance of other soil constituents. Although it was not elaborated by Condit (1970, 1972), his Type 1 and Type 2 curves corresponded, respectively, to the reflectance curves of high surface organic content Mollisols and low surface organic content Alfisols (Cipra et al., 1971b). In a similar manner, curve forms described by Stoner and Baumgardner (1981) as organic dominated and organic affected owe their character to the elevated content (>2%) and decomposition state of organic matter (Fig. 6). Organic constituents, including humic and fulvic acid and nonspecific compounds including decomposing plant residues, are known to influence soil reflectance to differing degrees (Obukhov and Orlov, 1964; Vinogradov, 1981), although the contribution of each has been difficult to quantify. The decomposition state in organic soils has been observed to alter drastically nonvisible reflectance (Fig. $0).The high reflectance of fibric soil materials in the infrared region resembles the infrared reflectance of senesced leaves


21

Wavelength (pm)

FIG.10. Spectral (BRF) curves of three organic soils exhibiting significantly different levels of decomposition. Curve a, fibric; curve b, hemic; curve c, sapric.

(Gausman et al., 1975). This increased infrared reflectance has been attributed to tissue morphology in which an increased number of air voids provides more air-cell interfaces for enhanced reflection. Oxidation of organic matter in a soil sample with H z O z resulted in increased reflectance from 0.44 to 2.4 pm, although the difference in reflectance beyond 1.3 pm became very small (Bowers and Hanks, 1965). Mathews et al. (1973a) destroyed the organic matter in a 12.8% organic matter silty clay soil, with the resulting reflectance being increased greatly in the spectral region from 0.4 to 1.3 pm, while the reflectance actually decreased slightly in the region from 1.5 to 2.4 pm. Regression studies indicated that organic matter content could be related to soil reflectance by a curvilinear exponential function (Schreier, 1977). Mathews et al. (1973b) found that organic matter correlated most highly with reflectance in the 0.5-1.2-pm range, while Beck et al. (1976) suggested that the 0.90-1.22-pm range was best for mapping organic carbon in soils. Montgomery (1976) indicated that organic matter contents as high as 9.0 % did not appear to mask the contributions of other soil parameters to soil reflectance. Montgomery differed from Beck in recommending the visible wavelength region as the best for spectral measurement of organic matter content in soils. Stoner (1979) showed organic matter to be the single most important variable to explain reflectance differences in the spectral region 0.52-1.75 pm, while the strongest correlations occurred in the visible wavelengths. In laboratory measurements with a color-difference meter, Page (1974) found that nearly 80% of the total variation in organic matter content of 96 Atlantic Coastal Plain soils could be accounted for by reflectance. SIZE C. PARTICLE

In the literature one of the early reports of the effect of soil particle size on reflectance was made by Zwerman and Andrews (1940). Working with enameled surfaces, they stated that at a given wavelength a material of given

22


refractive index reflects light with an intensity that varies inversely as the particle diameter. Kojima (1958a), using a photocolorimeter, measured the change in soil color with change in particle size. He reported results in tristimulus coordinates and, in general, indicated an increase in the Y coordinate-the luminosity function-as particle size decreased. Aside from the reflectance differences which can be accounted for by differences in surface roughness and soil structure, soil particle size and shape, as well as the size and shape of soil aggregates resulting from mild crushing, appear to influence soil reflectance in varying manners. Rowers and Hanks (1965) measured the reflectance of pure kaolinite in size fractions from 0.022 to 2.68 mm diam. (coarse silt to very coarse sand particle size classes) and found a rapid exponential increase in reflectance at all wavelengths between 0.4 and 1.0 pm with decreasing particle size. The most notable increases in reflectance occurred at sizes less than 0.4 mm diam. (approximately medium sand particle size class and finer). It was felt that particles or aggregates larger than 2-3 mm diam. would have little influence on additional absorption of solar energy. Orlov (1966) found the reflectance of aggregates from 0.25 to 10 mm in diameter to vary little for Mollisol-type soils. However, for the fraction less than 0.25 mm diam. (fine sand particle size class and finer) reflectance increased, a fact that Orlov attributed to sharp changes in chemical composition of aggregates less than 0.25 mm diam. compared with coarser aggregates. Averaged reflectance spectra from a broad range of sandy soils exhibit the trend of increasing reflectance with decreasing particle size (Fig. 11). Surface roughness on a micro scale may be the determining factor in explaining changes in reflectance as a function of particle or aggregate diameter. Bowers and Hanks (1965) observed that as particle size decreased, the surface of kaolinite became smoother. Similarly, Orlov (1966) found that fine particles filled a volume more completely and gave a more even surface. Coarse aggregates, having an irregular shape, formed a complex surface with

O.tO'

.8 '

' ' 1.2

B

' 1.6 ' ' ' ' 2.0 '

I

Wavelength

*

' 214

(urn)

FIG.11. Spectral (BRF) curves of soils differentiated by predominant particle size. -, sands; - -, Fine loamy sands; ---, loamy sands; * .,loamy coarse sands.

-

Fine


23

a large number of interaggregate spaces. As light falls on large, irregularly shaped aggregates, most of the incident flux penetrates into light traps and is completely extinguished there. Hunt and Salisbury (1971, 1976a,b) and Hunt et al. (1971a,b, 1973a-c, 1974) measured the reflectance of a large number of minerals and rocks in four size fractions: 0-0.005,O-0.074,0.074-0.25, and 0.25- 1.2 mm. For silicate and carbonate minerals it was noted that the general effect of decreasing the particle size of the samples was to increase the reflectance at all wavelengths and to decrease the contrast of any well-resolved spectral features. Conversely to previously mentioned studies, in the case of oxides and sulphides the reflectance as a function of wavelength sometimes actually decreased with decreasing particle size. This phenomenon appeared to occur in materials of very low reflectance. As in other studies, however, it was found that only integral reflection varied with particle diameter, whereas the shape of the spectral curve remained the same. Regression studies by Montgomery (1976) and Montgomery and Baumgardner (1974) found silt content to be the single most significant parameter in explaining the spectral variations in soils. It was felt that the significance of silt content may have been attributable to the size of the silt particles relative to the reflective wavelengths. Beck et al. (1976), studying predominantly silty soils, concluded that the wavelength region from 1.50 to 1.73 pm was best for mapping clay content in surface soils. While fine silt content was found to contribute significantly to explaining reflectance differences in the 0.52-0.62pm band, contents of other particle sizes were less important in prediction equations for near- and middle-infrared wave bands (Stoner, 1979). Gerberman and Neher (1979) artificially added increasing increments of sand to a Harlingen clay soil (Entic Chromuslesterts). They measured the reflectance of a set of clay soil-sand mixtures consisting of samples containing from 0 to 100% sand. They found a linear relationship between the increasing sand content and increasing reflectance for wavelengths of 440, 540,640,720, and 860 nm.

D. IRON OXIDES The type and relative amotm of constituent iron oxides are known to influence the colors of red and yellow soils high in sesquioxide clays (Bigham et al., 1978). Predominantly yellow soils high in goethite were found to absorb more phosphate per unit weight than did otherwise similar red soils high in hematite. Soil spectral reflectances may be meaningful criteria for both taxonomic and management separations in highly weathered soils.


24

Obukhov and Orlov (1964) reported that soils with an elevated content of iron could be easily distinguished by the inflection characteristic for pure Fe,O,. They found the intensity of the reflection in the region from 0.5 to 0.64 pm to be inversely proportional to the iron content. Karmanov (1970) noted that the reflection intensity of iron hydroxides containing little water and having a dark brown-red color increased most strongly in the wave interval from 0.554 to 0.596 pm, while that of hydrous iron oxides increased most strongly in the spectral range from 0.50 to 0.54pm. Neither of these studies investigated iron oxide reflectance beyond the visible wavelengths. Most of the well-resolved electronic features of iron oxides in minerals and rocks can be attributed to transitions in the iron cations (Hunt et al., 1971a). Typically, the ferrous ion produces the band near 1.0pm due to the spinallowed transition between the E , and T2@quintet levels into which the D ground state splits in an octahedral crystal field. For the ferric ion, the major bands produced in the spectrum are a result of transition from the GA1, ground state to 4T1,at about 0.87 pm and to 4T2, at about 0.7 pm. Whereas only 1 % by weight of finely powdered hematite was found to alter a clayey, yellow Oxisol from lOYR to 5YR in color (Resende, 1976), as little as 0.0005 % of iron by weight was capable of producing a perceptible iron band at 0.87 pm in a highly transparent calcite mineral (Hunt and Salisbury, 1971). In addition to the ferrous iron band at 1.0 pm, another absorption band near 1.0 pm has been identified in a sample of gibbsite as a second overtone and combination of stretching modes of the hydroxyl radical (Hunt et al., 1971a). The iron absorption band at 0.87 pm is evident even in fine sandy soils with iron oxide coatings on sand grains (Fig. 12). Higher iron content soils reveal a broader absorption band at 0.87 pm contrasted with the narrow, yet distinct band in soils of lower iron content. Soils such as the Cecil series have been described as demonstrating an iron-affected curve form, typical of soils with 40 r

t

-

t

n .4

.8

I .6 Wavelength ( p m ) 1.2

2 .o

2.4

FIG.12. Spectral (BRF) curves of soils of different textures but exhibiting iron absorption bands. -.-, Fine sand, 0.20%Fe,O,; -..-, sandy loam, 0.64%Fe,O,; -, silty clay loam, 0.76%Fe,O,; --, clay, 25.6% Fe,O,.


25

1-4% Fe,O, (Stoner and Baumgardner, 1981). An even more distinctive curve form has been observed in soils with iron contents greater than 4%, typical of Oxisols (Fig. 6, curve e). The iron-dominated curve form resembles reflectance curves for magnetite (Hunt et al., 1971a) and is unusual in that relfectance actually decreases with increasing wavelength beyond the visible region. Montgomery (1976) found the free iron oxide content of soil to be significant in both the visible and infrared regions of the spectrum but observed that the significance of iron increased with increasing wavelength. The presence of organic matter did not diminish the contribution of iron to soil reflectance. Stoner (1979) also observed that higher correlations between reflectance and iron oxide contents occurred at the longer middle-infrared wavelength from 1.55 to 2.32 pm. Percentage iron in iron-organic complexes along with percentage carbon and exchangeable Mg and K were most significantly correlated with spectral measurements in a study by Schreier (1977). The narrowness of infrared iron absorption bands is incompatible with the broad infrared wavelength bands of the present Landsat satellites, and may render quantitative comparisons of reflectance with iron oxide levels in soils impractical. Crouse et al. (1983) examined the Landsat thematic mapper bands for their utility in using reflectance measurements to predict the presence of iron oxides. They found the relatively broad TM bands to be extremely useful in characterizing some properties of soils. However, they concluded that the TM bands are too broad to detect effectively the narrow iron absorption bands at 0.7, 0.9, and 1.0 pm.

E. MINERAL COMPOSITION Examination of the reflectance spectra of montmorillonite and kaolinite group clay minerals indicates that they owe the major features of their absorption spectra to the presence of water (Lindberg and Snyder, 1972). Very strong molecular water bands at 1.4 and 1.9 ,urn are due to the bound water typical of montmorillonites (Hunt and Salisbury, 1970). Strong hydroxyl bands centered at 1.4 and 2.2 pm are typical of kaolinite reflectance, with the lack of appreciable bound water resulting in only a weak band at 1.9 pm. The presence of a small amount of the ferrous ion results in a slight band near 1.0pm in kaolinite, while ferrous iron in a sixfold site in montmorillonite results in a strong band at 1.0pm. Mathews et al. (1973b) measured a consistently low reflectance response for wavelengths less than 1.7,um in a sample of illite and showed low absorption intensities for the water and hydroxyl bands when compared to kaolinite and nontronite

26


(montmorillonite group) samples. For most kaolinite and montmorillonite group minerals, a general decrease in reflectance occurs with decreasing wavelength beginning at about 0.7 pm (Lindberg and Snyder, 1972). Aside from the silicate clay minerals, sesquioxides typical of highly weathered soil regions present reflectance spectra dominated by ferric iron and occasionally by hydroxides. Samples of hematite, limonite, and goethite exhibit well-defined ferric iron absorption bands near 0.9 pm (Hunt et al., 1971a). Gibbsite, whose chemical formula is Al(OH),, shows a weak but relatively sharp band at 1.0pm, attributed to overtones of Al-O-H bending modes and combinations of stretching modes. Magnetite samples display overall low reflectance, changing very little throughout the visible and infrared. This opaque, spectrally featureless behavior is due to iron oxide and titanium conduction bands extending throughout the spectral range. Soil clays occur in intimate combination with other soil constituents. Mixed-clay mineralogies are more common than clay mineralogies predominated by single-clay types. Montgomery (1976) analyzed separately a group of soils with montmorillonitic mineralogy and noted little difference between statistical correlations of reflectance and soil properties for this group and for soils as a whole. DaCosta (1979) reviewed the possible contributions of sand and silt sizefraction mineralogy to reflectance of soils. Heavy mineral studies are basic to understanding stages in parent-material weathering sequences. Magnetic susceptibility of minerals varies greatly and plays an important part in classification of highly weathered tropical soils with elevated iron contents. F. SOLUBLE SALTS

Although there is a wealth of literature describing the physical and chemical characteristics of saline and alkaline soils, there are relatively few references to the effects of the quantity and quality of soluble salts on the reflectance properties of soils. Multiband photography has been used for both direct and indirect detection of soil salinity (Myers et a!., 1966; Meyer and Calpouzos, 1968; Carneggie et al., 1967; Driscoll, 1971; Everitt et al., 1977, 198 1). However, not until relatively recently have quantitative reflectance measurements been used to study soil salinity. When Stoner (1979) grouped and averaged 480 soil spectra by soil order, he found that of the 10 orders, Aridisols, generally high in soluble salts, had the highest average reflectance over the spectral range from 0.52 to 0.9 pm. Singh et al. (1977) reported successful delineation of salt-affected soils in the Ganges alluvial plain of North India by digital processing of Landsat multispectral data.


27

Al-Mahawili (1983) examined the reflectance characteristics of nonsaline, saline, and gypsiferous soils in the Mesopotamian Plain between the Tigris and Euphrates Rivers in Iraq. He found that saline soils (EC > 4 mmho/cm) gave lower reflectance values than did nonsaline, gypsiferous soils through the spectral range from 0.5 to 2.38 pm. Further, he found that leaching the saline soils, and thus significantly reducing the electrical conductivity, increased the reflectance.

G. PARENT MATERIAL Mathews et al. (1973a) demonstrated the inff uence of soil parent material on soil reflectance. Reflectance curves for soils developed from limestone, shale, and sandstone exhibited contrasting, characteristic shapes and were separable at all wavelengths. Schreier (1977) also noted that soil parent material seemed to affect the overall spectral reflectance of soils developed from deltaic, organic, marine alluvial, and outwash materials. Hunt et al. (1973b,c, 1974) observed that the overall reflectance intensities of igneous rocks decreased from acid to intermediate to basic, with ultrabasic forms always displaying a well-defined ferrous iron band near 1.0pm. These findings suggest that local geographic areas of similar parent material may best be studied separately when trying to relate soil reflectance to other soil parameters. Just such an approach was used successfully to separate drainage classes within parent material groups in Jasper County, Indiana (Weismiller et al., 1979). Prediction equations for reflective wave bands from the visible to the middle infrared frequently included parent material as a significant contributor for explaining reflectance differences (Stoner, 1979). In the 2.08-2.32-pm wave band this relationship was so strong that a parent-material variable was included in the stepwise regression procedure, second only to moisture content and before inclusion of organic matter.

H. OTHER CONSIDERATIONS Montgomery and Baumgardner (1974) reported that soil properties other than those soil constituents with obvious spectral character may be correlated with soil reflectance. This relationship is most pronounced for cation exchange capacity (CEC) of soils. Cation exchange capacity frequently reveals higher correlations with reflectance than do the contents of any of the particle size classes or even organic matter. For example, Stoner (1979) found that for 42 subhumid, frigid region soils, the CEC had a correlation of - 0.86

28


with the 1.55-1.75-pm band reflectance, while the highest correlation for organic matter content (transformed to natural log) was -0.77 in the 0.82-0.92-pm band and that for clay content was -0.67 in the 1.55-1.75-pm band. It is speculated that high correlations between reflectance and CEC are related closely to soil constituents that do exhibit spectral behavior, such as clay type and content and organic matter content. In these studies, CEC has usually been shown to be highly correlated with clay and organic matter contents.

IV. REFLECTANCE PROPERTIES OF SOILS IN THEIR ENVIRONMENT

A. ATMOSPHERIC EFFECTS The character of reflected light from a soil surface depends on the nature of the surface and the intensity and spectral quality of the radiation incident upon the surface. The soil surface may be irradiated by direct sunlight, scattered skylight, and thermal radiation from the surroundings. Scattered or diffuse radiation (skylight) may make up as much as 10-15 % of the sunlight on a clear day (Idso et al., 1966). Cloud light, or diffuse radiation which is the result of a complete overcast, selectively filters out longer-wavelength visible and infrared light. The distribution of solar energy at the earth's surface is also such that numerous absorption bands occur for 0,, O3 and H,O. Of special concern for remote sensing studies are the strong water absorption bands at 1.45 and 1.95 pm, which completely attenuate solar energy, rendering these wavelength regions useless for the study of soils in their environment (Fig. 7). Similarly, ultraviolet sensing of soils is made impossible by ozone absorption. Haze conditions, which do not produce a complete overcast, nonetheless scatter light by differing amounts in different wavelengths, tending to obscure visible wavelengths while allowing penetration of near-infrared wavelengths. Remote sensing of reflective soil properties is thus effectively limited to wavelengths between 0.4 to 2.4 pm under relatively haze-free conditions. Differences in laboratory instrument and airborne sensor configurations complicate comparisons of soil reflectance measured by these methods. Whereas most laboratory instruments utilize calibration standards viewed and irradiated identically to the target, airborne sensors must often rely on internal calibration sources. Solar radiation and atmospheric attenuation may vary from one airborne sensor flight to another.


29

Laboratory and field spectra of moist Alfisol and Mollisol surface soils measured with the same spectroradiometer and calibrated to a pressed barium sulphate standard exhibited characteristically shaped spectral curves for both soils (Stoner et al., 1980b). The spectral response for either soil measured in the field was about 1.5 times greater than the spectral response of laboratory-measured moist soils at any given wavelength from 0.52 to 1.75 pm. Lower moisture levels and the formation of a drier surface crust could easily account for the observed spectral differences, but importantly the ability to extend laboratory-measured soil spectra to field conditions was demonstrated. May and Petersen (1975) attempted to correct for solar radiation and atmospheric attenuation in comparing airborne MSS data to laboratory reflectance spectra of soil. Maps generated by supervised and unsupervised classification routines from laboratory and MSS-derived reflectances compared well with field survey maps. An agreement of 90% was obtained between the MSS-derived maps and the conventionally prepared soil maps. Cipra et al. (1980) found good agreement between Landsat spectral measurements from nonvegetated soil sites and laboratory spectroradiometric measurements of soil samples collected from these sites even without any correction of Landsat data for atmospheric effects.

B. PHYSICAL SURFACE CONDITIONS

Early remote sensing research in soils recognized the fact that soils often formed a surface crust that could make a soil appear dry when it was actually wet (Hoffer and Johannsen, 1969). Cipra et al. (1971b) found that crusted surfaces gave higher reflectance values in the 0.43-0.83-pm wavelength region than did soils with the crust broken. The lower reflectance of the disturbed soil was attributed to the rough surface, which presumably caused scattering of light as well as a shadowing effect. Surface roughness of a sandy Alfisol appeared to override the effects of moisture on reflectance (Johannsen, 1969). The sensor view angle of most reflectance-type measuring devices is normal to the surface being viewed, but important illumination angle effects commonly result from differences in solar elevation angle with the time of day and season of the year. Recently cultivated soils, aside from their generally higher surface moisture content in comparison with undisturbed soils, often exhibit a random geometry of reflecting surfaces whose overall reflectance may vary with illumination angle (Crown and Pawluk, 1974; Coulson and Reynolds, 1971). Tilled clay soil broken into aggregates several centimeters in size demonstrated marked differences in reflectance as a function of sun elevation (Coulson and Reynolds, 1971). A strong decrease in reflectance occurred with

30


increasing sun elevation, possibly caused by trapping of radiation among the coarse particles as the fraction of incident radiation entering the spaces increased with increasing sun elevation. Soils with well-defined structure in the plow layer were found to reflect 15-20% less light energy than structureless soils (Obukhov and Orlov, 1964). Difficulties in fully characterizing the moisture content and surface roughness of soils under various tillage treatments make this area one of the least understood areas of surface soil reflectance.

C.

SURFACE COVER

The spectral composition of the reflected radiation from soil is strikingly different from that reflected from plants (Gates, 1963, 1965). Single leaves exhibit absorption maxima in the blue and red regions at 0.47 and 0.68 pm, respectively, while the familiar green reflectance peak occurs at 0.55 pm. Total lack of pigment absorption and lack of appreciable absorption by liquid water results in strong near-infrared reflectance in healthy leaves from 0.7 to 1.3 pm. Major water absorption bands appear at 1.45 and 1.95 pm in leaves as they do in moist soils (Myers and Allen, 1968). Density, morphology, and condition of the geometrical arrangement of leaves in a plant canopy determine the extent to which green vegetative cover affects the reflectance from surface soils (Hoffer and Johannsen, 1969). Girard-Ganneau (1975) reported that up to a vegetative cover of 15 % the surface appears as soil, whereas in excess of 40 % cover, the spectral behavior resembles that of vegetation. Near-infrared-wavelength data from digitized photographs were used to estimate percentage ground cover in a maize canopy on both a Mollisol and an Alfisol (Stoner et al., 1976). Using aircraft MSS data, Kristof and Baumgardner (1975) found that the ratio between relative reflectance in the visible spectrum and the relative reflectance in the infrared spectrum could be used to characterize the seasonal variation which is intimately connected with changes in green vegetative cover. Soil patterns remained visible in spite of dense maize cover well into the growing season. Although dense vegetative canopies of crops or naturally occurring plant communities may mask the soils, it is important to realize that inherent fertility, drainage, and moisture-holding-capacity differences among soils tend to influence the vegetative growth on these soils. Thus, although the soil itself is eventually masked by plant canopies, the canopy varies in phenological and morphological characteristics with different soils (Westin and Lemme, 1978). In this way, green vegetative cover is especially important in


31

soil mapping of wild or uncultivated areas of native vegetation cover (Ranzani, 1969). Common seasonal components of remotely sensed ground scenes are nonsoil, nongreen vegetation residues of senesced vegetation or even snow and ice in temperate zones. Although topographic information may be obtained from snow-covered areas (Lewis et a/., 1975), generally the presence of snow cover only obscures the soil patterns of interest in soil mapping, and winter-collected data are usually avoided. However, it is not uncommon in cultivated regions to have a cover of crop residue on the soil surface at times of the year that would otherwise be ideal for obtaining remotely sensed data from soils (Stoner and Horvath, 1971). Senesced leaves behave differently in the near-infrared-wavelength region than do live, healthy leaves (Gausman et al., 1975). Whereas multiple leaf layers of healthy green leaves exhibit enhanced reflectance up to a stack of eight leaves, senesced leaves do not show increased infrared reflectance beyond two or three leaf layers. Aircraft and spacecraft reflectance measurements would not be expected to distinguish between different densities of senesced vegetation. Field spectroradiometric investigations showed that sugarcane crop residue littered on the soil surface had higher reflectance than bare soil, but that standing sugarcane crop residue had lower reflectance than bare soil (Gausman et al., 1975). Residue-covered soils for a variety of crops and grasses were best discriminated from bare soils with Landsat reflectance measurements from 0.5 to 0.6pm in the visible region of the spectrum. Further work by Gausman et nl. (1977) with wheat straw suggested that the near-infrared region from 0.75 to 1.3pm seemed better than the visible region or water absorption bands for distinguishing among reflectances of soil-tillage-straw treatments. Again, as in the case of green vegetative cover, indications are that the presence of nonsoil residue does not fully obscure detectable soil patterns when areas of similar residue cover are isolated and classified separately using airborne scanner data (Stoner and Horvath, 1971). Field spectroradiometric studies of maize residue cover on an Alfisol and a Mollisol provided evidence that the characteristic trends of spectral curves for these soils are not altered by residue cover or moisture differences (Stoner et al., 1980b).

D. SURFACE EXPRESSION OF SUBSOIL CHARACTERISTICS All soils have a specific internal drainage which is indicative of the local landscape position and broader climatic conditions under which they formed. Even for soils in which the marks of seasonal soil saturation may by

32


definition extend upward no higher than to horizons untouched by tillage equipment, the soil-forming processes involved exert their influence on the whole soil profile and often are evident in the surface soils. Soils grouped by internal drainage class display ever-decreasing soil reflectance with increasingly poorer drainage. Well-drained and moderately well-drained soils show very little difference in reflectance, as would be expected. Very poorly drained soils reflect considerably less than any of the other drainage classes at all wavelengths. Whereas the well-drained and moderately well-drained soils show evidence of ferric iron absorption at 0.9 pm, all three poorly drained soil classes lack the ferric iron absorption band. As a site characteristic integrating the effects of climate, local relief, and accumulated organic matter, soil drainage characteristics can be expected to be closely associated with reflectance properties of surface soils. Soil erosion monitoring on cultivated land is made possible because certain subsoil characteristics exert an influence on surface soil reflectance properties. Eroded land surfaces in a field are often obvious because of striking soil color changes. This usually results from removal of the original surface, exposing subsoil horizons, or more frequently incorporation of subsoil horizons into the plow layer. In soils with elevated iron contents in subsoil horizons, eroded soils in an erosion toposequence demonstrated a broad iron absorption band at 0.87 pm which was strong enough to influence reflectance magnitude in the Landsat MSS bands from 0.7 to 0.8 pm and 0.8 to 1.1 pm (Latz et al., 1984).

E. SENSOR DATADIMENSIONALITY Studies that utilize Landsat MSS data to estimate various agronomic parameters such as leaf area index and developmental stage of crops have attempted to account for the effect of soil background on the spectral signature obtained in the four MSS bands. Various transformations have been developed that are physically meaningful and that accomplish the desired effect of reducing the dimensionality of the highly correlated Landsat MSS bands. The so-called “tasseled cap” transformation of Kauth and Thomas (1976) succeeded in reducing the four-band Landsat MSS data to a two-dimensional data space in which bare soils were assumed to fall along a straight line parallel to an axis referred to as “brightness,” while vegetation would fall on a line perpendicular to the brightness line, referred to as “greenness.” Thompson et al. (1983), using the 481 soil spectra in the LARS data base, found that soils in greenness and brightness vector space are not parallel to brightness but have a slight slope that is significantly different from zero.


33

These results suggested that caution should be used in studies that develop relationships using a fixed soil background and attempt to extend the relationship to other regions and soils. Location of soil reflectance in greenness and brightness vector space was found to stratify soil organic matter contents into 0-2% and > 2 % groups with greater than 80% accuracy. Inclusion of two middle-infrared reflective wavelength bands to the complement of six reflective spectral bands on the thematic mapper sensor has added a distinct new dimension to the two-dimensional data space of Landsat MSS data. In what Crist (1983) refers to as the “thematic mapper tasseled cap space,” a distinct “plane of soils” can be seen, defined by a third feature which is a contrast between the sum of the two longer infrared thematic mapper bands (bands 5 and 7, 1.55-1.75 pm and 2.08-2.35 pm, respectively) and the sum of the visible and near-infrared bands. This feature responds to soil physical properties, particularly to soil moisture status, and has been tentatively termed “wetness.” This additional dimension affords the opportunity to extract more comprehensive soils-related information.

V. APPLICATIONS OF SOIL REFLECTANCE MEASUREMENTS One of the driving forces during recent years in the study of soil reflectance data has been the need to improve our capabilities to inventory and monitor soil resources. With the rapid emergence of aerospace sensors which can obtain soils reflectance data globally on a repetitive basis, the need for understanding the relationships between soil reflectance and other soil properties becomes more critical. This final section will briefly cite how resource scientists are using interpretations of soil reflectance data in three general areas: soil surveys, soil degradation assessment, and soil information systems.

A. SOILSURVEY

Soil reflectance in the form of black-and-white aerial photography was first used in 1929 to prepare base maps for a soil survey in Jennings County, Indiana (Bushnell, 1929). The results represented a significant improvement over the use of plane tables to draw base and soil maps. Since 1938 most soil surveys in the United States have been prepared with the use of black-andwhite aerial photographs as the base map (Soil Survey Staff, 1951).

Fig. 13. Spectral mapgenerated from digital analysis of Landsat MSS data for use in the soil survey of Jasper County, Indiana (Lacustrine area; Atlas sheet No. 68) (Weismiller et a!., 1979). The area covered in the figure is approximately 17.5 km. The original scale was 1: 15,840.


35

Fifty years after the first use of black-and-white aerial photographs for soil surveys, digital analysis of Landsat reflectance data was first used to prepare a spectral base map to assist in the detailed soil survey (1:15,840) of Jasper County, Indiana (Hinzel et al., 1980; Weismiller et al., 1979). Field surveyors found the spectral maps (Fig. 13) useful in delineating boundaries between soils and in assessing the homogeneity of soil map units. In the late 1960s and early 1970s, prior to the launch of Landsat 1, several soil scientists began to investigate the possibility of using MSS data as a tool for delineating differences in surface soils. Kristof (1971) and Kristof and Zachary (1971) reported that computer-implemented pattern recognition analysis of aircraft MSS data could be used to map some soil surface conditions over small areas with a reasonable degree of accuracy. Soon thereafter Kristof and Zachary (1974) reported partial success in delineating soil series in an Alfisol-Mollisol region by digital analysis of aerial MSS data. A new era in the availability of multispectral reflectance data for soil surveys began with the launch of Landsat 1 in July 1972. Since that data four more satellites in the Landsat series have been placed in sun-synchronous orbit for earth observation purposes (Table 11). One of the earlier reports on the use of Landsat data for soil survey was by Lewis et al. (1975). They Table I1 Specifications of the Multispectral Scanner and Thematic Mapper on Landsats 1-5

Landsat"

Sensor

1,2,3,4, 5

MSS

3 4, 5

TM

Spectral bands (pm)

Spatial resolution (m)

0.5-0.6 0.6-0.7 0.7-0.8 0.8- 1.1 10.4-12.6 0.45-0.52 0.52-0.60 0.63-0.69 0.76-0.90 1.55-1.75 2.08-2.35 10.40-12.50

80 80 80 80 240 30 30 30 30 30 30 120

'(launch date)/(termination of sensor operation): Landsat 1 , (07/72)/(01/78); Landsat 2, (01/75)/(01/80); Landsat 3, (03/78)/(09/83); Landsat 4, (07/82)/(MSS operational; TM 02/83); Landsat 5, (03/84)/(MSS and TM operational).

36


concluded that soil associations within the Sand Hills region of Nebraska can be interpreted on the basis of image patterns resulting from differences in vegetation and related drainage conditions and from topography enhanced by continuous snow cover and relatively low solar elevation angles. During the first 4 years of the Landsat program, the majority of the soil survey work using Landsat reflectance data was accomplished through visual interpretation of images to delineate soil boundaries (Frazee et al., 1974; Hilwig, 1976; Hilwig et al., 1974; Lewis et al., 1975; Parks and Bodenheimer, 1973; Seevers and Drew, 1973; Seevers et al., 1974; Steinhardt et a!., 1975; Westin, 1973, 1974; Westin and Myers, 1973). During this same period a limited number of papers reported the use of digital analysis of Landsat data for identifying and delineating soil differences (Baumgardner et al., 1973; Mathews et at., 1973a; May and Petersen, 1975). By 1980 the number of soil scientists having access to computer-implemented image processing capabilities had grown considerably. Two Landsat image mosaics (scale 1: 1,000,000) for the state of South Dakota were prepared by Westin and Frazee (1976) from 20 late spring 1973 Landsat scenes, each scene covering an area of 34,000 km2. They used one mosaic to prepare a land value map (soil associations keyed to land sale prices from 1967 to 1972); on the second they keyed soil associations to soil test results from the previous 25 years. Weismiller et al. (1977) used computer-implemented pattern recognition analysis of Landsat MSS data as an aid in the soil inventory of Chariton County, Missouri. They found that by combining digitized ancillary data (township, watershed, and physiographic boundaries) with Landsat MSS data, a more detailed delineation of soils could be obtained than with MSS data alone. Analyzing Landsat reflectance data obtained on 9 June 1973 over Clinton County, Indiana (Alfisol-Mollisol area), Kirschner et al. (1978) found a close correlation between spectral classes and soil drainage characteristics. They found the spectral map (scale 1:20,000) produced from digital analysis of Landsat reflectance data to be useful in assessing map unit composition, or homogeneity of the soils within a map unit. Several studies have reported positive results in the use of Landsat MSS data in delineating and mapping important soil differences in arid and semiarid regions (Horvath, 1981; Horvath et al., 1980, 1983, 1984; Kornblau, 1979; Kornblau and Cipra, 1983; Mimms, 1982). Horvath et al. (1984) examined the possibility of using Landsat reflectance data to determine important properties of Arizona rangelands. They found that satellite reflectance data added valuable information for the determination of soil map unit definition and distribution. They concluded that site characteristics (canopy cover, elevation) were more important in predicting


37

MSS reflectance of rangeland test sites than were the properties of soils, but vegetation and soil characteristics could be used in combination with reflectance data to delineate soil map units. Results have also been reported on the use of Landsat data as an aid to soil surveys in India (Hilwig, 1976; Hoore et al., 1982; Rao et al., 1982), Bolivia (Valenzuela, 1979), Spain (Hilwig et al., 1974), and the USSR (Andronikov and Liverovski, 1982). The launch of Landsat 4 in July 1982 made available for the first time Landsat TM reflectance data with 30-m spatial resolution. Since that launch several papers have reported the application of TM data to soil studies. Thompson et al. (1984) examined TM reflectance data acquired over Mississippi County, Arkansas, to determine the sensitivity of TM reflectance to soil properties under growing soybeans (Glycine rnax L.). They found that TM data provide information that is related to soil properties within a field. Thompson and Henderson (1984) examined reflectance data from an aircraft TM simulator and the Landsat 4 TM for their relative accuracy in separating soils from an intensely cultivated agricultural area. They used reflectance data from five different dates throughout the 1982 growing season obtained over Central Iowa. The results of their study indicated that the improved spectral and spatial resolution of TM (over MSS) data offers the potential to separate important soil properties even in regions with similar soils and under a dense corn (Zea mays L.) or soybean (Glycine max L.) canopy. A summary of the conclusions reported in the literature during the past decade suggest that the following aspects of and possibilities with satelliteacquired reflectance data provide a significant new tool to aid in soil surveys (Imhoff and Peterson, 1980; Kirschner et al., 1978; Longlois et a/., 1976; van Sleen, 1982; Weismiller et a]., 1977; Weismiller and Kaminsky, 1978; Westin and Frazee, 1976): 1. 2. 3. 4. 5. 6. 7. 8.

Synoptic view of survey area and surroundings Quantitative assessment of homogeneity of map unit Near orthographic quality of MSS and TM data Multispectral data set Repetitive coverage Digital format for tabular or image information Computer-implemented image processing possibilities Possibilities for registering, overlaying, and combining multiple data sets

The sources and uses of reflectance data from aerospace sensors have been summarized and reported by Baumgardner et al. (1983). Appropriate sources

38


I Survey type Suwey scale Size of mapping unit

-

Order of soil survey

5th order I 4thorder Reconnaissance1 :300,0001:125,0001:300,000 1:1,000,000 35-50 km2 500-500,000 ha

Kind of mapping unit Associations of phases Associations of of subgroupdgreat families of soil groups, suborders, series orders

Use in development planning

Common or potential remote sensing data sources

I

3rd order Semidetailed 1:32,0001:125,000 10-1 000 ha

I

1s t order Intensive 1: 10001: 12,000 0.5 ha or smaller Phases of soil series

-- -

-

Landsat

2nd order Detailed 1: 12,0001:32,000 1.O-1.6 ha

Associations of Consociations of phases of phases of soil soil series series

Resource inventory Proiect location

4

I

c

Feasibi!ity suweysManagement surveys

MSS and TM (images)

-

c

Landsat MSS t T M (digitall-Landsat

t -

NOAA 617 4

c

-

TM (digital)-

Aerial photography (high altitude) c Aerial photography (low altitude)

-

FIG.14. Use of soil reflectance data derived from aerospace sensors as an aid for preparing different orders of soil surveys. (Adapted from Baumgardner et nl., 1983.)

of reflectance data can be used in the preparation of different orders of soil survey (Fig. 14). B. SOILDEGRADATION ASSESSMENT

The global and repetitive coverage of polar-orbiting earth observation satellites offers the possibility of monitoring changes in earth surface phenomena. With the increasing human demands on the land resources of the earth, an appropriate application of repetitive Landsat reflectance data is to measure quantitatively the rates of soil degradation caused by wind erosion, water erosion, salinization, flooding, and other processes. Several limited studies have made use of Landsat data to inventory and monitor soil degradation (Bleeker, 1978a,b; Hellden and Stern, 1980; Latz et al., 1984; Mainguet et al., 1978; Mitchell and Ghorashian, 1978; Mitchell and Howard, 1978; Mitchell et al., 1978; Pacheco, 1978; Parada and Pinto, 1983). As high-quality, repetitive multispectral data from earth observation systems become more readily available, it should be possible to assess quantitatively changes in soil conditions caused by soil degradation. C. SOILINFORMATION SYSTEMS

The capability to acquire and store very rapidly large volumes of data about the earth’s surface has hastened the need for data management systems.

REFLECTANCE PROPERTIES O F SOIL

39

What is emerging for soils and other natural resources is an array of georeferenced information systems (GIs). The concept suggests that remotely sensed (e.g., reflectance, thermal, microwave), cartographic, climatic, chemical, physical, topographic, socioeconomic, and other data related to the earth’s surface can be digitized, registered, and overlaid such that all data are related to specific geographic coordinates on the earth’s surface. This data base can then be used for ready retrieval of information for resource decisionmaking and policy-making. Several papers have reported applications of the GIS concept to soil data management (Biehl et al., 1982; Dangermond, 1983; Hitchcock et al., 1975; Imhoff et al., 1982; Stoner et al., 1983). Soil scientists will be presented many new opportunities and challenges in the years ahead to utilize these new technologies to inventory and monitor our soil resources. REFERENCES Al-Mahawili, S. M. H. 1983. M.S. thesis, Purdue University, West Lafayette, Indiana. Andronikov, V. L., and Liverovski, Y. A. 1982. Ahstr. Trans. lnt. Congr. Soil Sci., J2th, New Delhi p. 132. Angstrom, D. 1925. Geograf. Ann. 7, 323. Baumgardner, M. F., and Stoner, E. R. 1982. Trans. lnt. Congr. Soil Sci., IZth, New Delhi 5, 419-441. Baumgardner, M. F., Kristof, S. J., Johannsen, C. J., and Zachary, A. L. 1970. Indiana Acad. Sci. Proc. 79,413-422. Baumgardner, M. F., Kristof, S. J., and Henderson, J. A,, Jr. 1973. Proc. Syrnp. Significant Results Obtained Earth Resources Techno]. Satellite-I, 1973 NASA SP-327, pp. 213-221. Baumgardner, M. F., Crosson, P. R., Dregne, H., Drosdoff, M., and Westin, F. C. 1983. In “Resource Inventory and Baseline Study Methods for Developing Countries” (F. Conant, P. Rogers, M. Baumgardner, C . McKell, R. Dasmann, and P. Reining, eds.), pp. 187-305. Amer. Assoc. Adv. Sci. 83-3. Washington, D.C. Beck, R. H., Robinson, B. F., McFee, W. H., and Peterson, J. B. 1976. Info. Note 081 176. Lab. Applic. Remote Sensing, Purdue Univ., West Lafayette, Indiana. Biehl, L. L., Bauer, M. E., Robinson, B. F., Daughtry, C . S. T., Pitts, D. E., and Silva, L. F. 1982. Proc. lnt. Symp. Machine Process. Remotely Sensed Data pp. 169-177. Bigham, J. M., Golden, D. C., Buol, S. W., Weed, S. B., and Bowen, L. H. 1978. Soil Sci. SOC.Am. J . 42, 825-830. Billmeyer, F. W., Jr., Lewis, D. L., and Davidson, J. G . 1971. Color Eng. May-June, 31-36. Bleeker, P. 1978a. The application of Landsat imagery to soil degradation mapping at 1:5,000,000 of Cambia, Guinea, Sierra Leone, and parts of Senegal, Liberia and Ivory Coast. Remote Sensing Unit, FAO, Rome. Bleeker, P. 1978b. The application of Landsat imagery to soil degradation mapping of Sierra Leone at 1:1,000,000. Remote Sensing Unit, FAO, Rome. Bowers, S. A,, and Hanks, R. 3. 1965. Soil Sci. 100, 130-138. Bowers, S. A., and Smith, S. J. 1972. Soil Sci. Sac. Am. Proc. 36,978-980. Bowers, S. A., Smith, S. J., Fisher, H. D., and Miller, G . E. 1975. Soil Sci. Sac. Am. Proc. 39, 391-393. Brooks, F. A. 1952. J . Meteorol. 9, 41-52.

40

MARION F. BAUMGARDNER E T A L .

Bushnell, T. M. 1929. Indiana Acad. Sci. Proc. 39,229-230. Carneggie, D. M., Poulton, C. E., and Roberts, E. H. 1967. The evaluation of rangeland resources by means of multispectral imagery. Annu. Prog. Rep., Earth Resources Survey Program, OSSAJNASA. Univ. Gal$, Berkeley. Cipra, J. E., Baumgardner, M. F., Stoner, E. R., and FJacDonald, R. B. 1971a. Soil Sci. SOC.Am. Proc. 35, 1014-1017. Cipra, J. E., Silva, L., and Hoffer, R. 1971b. Proc. Int. Symp. Remote Sensing Environ. 7th pp. 1509-1 518. Cipra, J. E., Franzmeier, D. P., Bauer, M. E., and Boyd, R. K. 1980. Soil Sci. SOC.Am. J . 44,80-84. Condit, H. R. 1970. Photogr. Eng. 36,955-966. Condit, H. R. 1972. Appl. Opt. 11, 74-86. Coulson, K. L., and Reynolds, D. W. 1971. J . Appl. Meteorol. 10, 285-1295. Crist, E. P. 1983. Proc. Int. Symp. Machine Process. Remotely Sensed Data pp. 357-363. Crouse, K. R., Henninger, D. L., and Thompson, D. R. 1983. Proc. Int. Geosci. Remote Sensing Symp. 1, 2.1-2.8. Crown, P. H., and Pawluk, S. 1974. Proc. Can. Symp. Remote Sensing, 2nd 1,450-462. DaCosta, L. M. 1979. Ph.D. dissertation, Univ. of Missouri, Columbia. Dangermond, J. 1983. Dig. Int. Symp. Geosci. Remote Sensing 1, 3.1-3.5. DeWitt, D. P., and Robinson, B. F. 1976. Tech. Rep. 091576. Lab. Applic. Remote Sensing, Purdue Univ., West Lafayette, Indiana. Driscoll, R. S. 1971. Forest Service Research Paper, RM-67. US. Dept. of Agriculture. Evans, R. M. 1948. “An Introduction to Color.” Wiley, New York. Everitt, J. H., Gerbermann, A. H., and Cuellar, J. A. 1977. Photogr. Eng. Remote Sensing 43, 1041 - 1047. Everitt, J. H., Gerbermann, A. H., and Alaniz, M. A. 1981. Photogr. Eng. 47, 1357-1362. Frazee, C. J., Rahn, P. H., Westin, F. C., and Myers, V. I. 1974. South Dakota Agric. Exp. Sta. Bull. 1276. Gates, D. M. 1962. “Energy Exchange in the Biosphere.” Harper, New York. Gates, D. M. 1963. Am. Sci. 51, 327-348. Gates, D. M. 1965. Proc. Symp. Remote Sensing Environ., 3rd, pp. 573-600. Gausman, H. W., Gerbermann, A. H., Wiegand, C. L., Learner, R. W., Rodriguez, R. R., and Noriega, J. R. 1975. Soil Sci. SOC.Am. Proc. 39, 752-755. Gausman, H. W., Learner, R. W., Noriega, J. R., Rodriguez, R. R., and Wiegand, C. L. 1977. Soil Sci. SOC.Am. Proc. 41, 793-796. Gerberman, A. H., and Neher, D. D. 1979. Photogr. Eng. Remote Sensing 45, 1145-1151. Girard-Ganneau, C. M. 1975. Docteur Ingenieur dissertation, University of Paris-Sud, Center &Orsay. Grum, F., and Luckey, G. W. 1968. Appl. Opt. 7, 2289-2294. Hellden, U., and Stern, M. 1980. Monitoring land degradation in Southern Tunisia: A test of Landsat imagery and digital data. Lab. Remote Sensing, Lund Univ., Sweden. Hilwig, F. W. 1976. 1 T C J . 1,26-42. Hilwig, F. W., Goosen, D., and Kateieris, D. 1974. ITC J . 3, 289-312. Hinzel, E. J., Weismiller, R. A., and Franzmeier, D. P. 1980. Tech. Rep. 080979. Lab. Applic. Remote Sensing, Purdue Univ., West Lafayette, Indiana. Hitchcock, H. C., Cox, T. L., Baxter, F. P., and Smart, C. W. 1975. Photogr. Eng. Remote Sensing 41, 1519-1524. Hoffer, R. M., and C. J. Johannsen. 1969. In “Remote Sensing in Ecology” (P.L. Johnson, ed.), pp. 1-29. Univ. Of Georgia Press, Athens. Hoore, J. L. D., H. S. Teotia, and R. Goombeer. 1982. Abstr. Trans. Int. Congr. Soil Sci., 12th New Delhi p. 134.


41

Horvath, E. H. 1981. Ph.D. dissertation, University of Arizona, Tucson. Horvath, E. H., Post, D. F., Lucas, W. M., and Weismiller, R. A. 1980. Proc. Znt. Symp. Machine Process. Remotely Sensed Data, 6th pp. 235-240. Horvath, E. H., Klingebiehl, A. A,, Moore, D. G., and Fosnight, E. A. 1983. USGS Open-File Rep. 83-880. Horvath, E. H., Post, D. F., and Kelsey, J. B. 1984. Soil Sci. SOC.Am. J . 48, 1331-1334. Hunt, G. R., and Salisbury, J. W. 1970. Mod. Geol. 1, 283-300. Hunt, G. R., and Salisbury, J. W. 1971. Mod. Geol. 2, 23-30. Hunt, G. R., and Salisbury, J. W. 1976a. Mod. Geol. 5, 211-218. Hunt, G. R., and Salisbury, J. W. 1976b. Mod. Geol. 5, 219-228. Hunt, G. R., Salisbury, J. W., and Lenhoff, C . J. 1971a. M o d . Geol. 2, 195-205. Hunt, G. R., Salisbury, J. W., and Lenhoff, C. J. 1971b. Mod. Geol. 3, 1-14. Hunt, G. R., Salisbury, J. W., and Lenhoff, C. J. 1973a. M o d . Geol. 4, 85-106. Hunt, G. R., Salisbury, J. W., and Lenhoff, C . J. 1973b. M o d . Geol. 4, 217-224. Hunt, G. R., Salisbury, J. W., and Lenhoff, C. J. 1973c. M o d . Geol. 4, 237-244. Hunt, G. R., Salisbury, J. W., and Lenhoff, C. J. 1974. M o d . Geol. 5, 15-22. Hutton, J. G. 1928. Proc. Int. Congr. Soil Sci., 1st 4, 164-172. Hutton, J. G. 1932. Report of the Color Standards Committee. Am. Soil Survey Assoc. Bull. 13, 179-180. Idso, S. B., Baker, D. G., and Gates, D. M. 1966.In“Advances in Agronomy 18” (A. G. Norman, ed.), pp. 171-218. Academic Press, New York. Imhoff, M. L., and Petersen, G. W. 1980. The role of Landsat products in soil surveys. Final Report for Contr. No. NAS5-25667. NASA, Goddard Space Flight Center. Imhoff, M. L., Petersen, G. W., Sykes, S. G., and Irons, J. R. 1982. Photogr. Eng. Remote Sensing 48, 133771342, Johannsen, C . J. 1969. Ph.D dissertation, Purdue Univ., West Lafayette, Indiana. Judd, D. B., and Kelly, K. L. 1939. U.S. NBS Jour. Res. ( R P 1239) 23, 355-385. Karmanov, I. I. 1970. Sou. Soil Sci. 4, 226-238. Kauth, R. J., and Thomas, G. S. 1976. Proc. Symp. Machine Process. Remotely Sensed Data pp. 41-51. Kellogg, C. E. 1937. Soil survey manual. Misc. Publ. No. 274, US.Dept of Agriculture. Kirschner, F. R., Kaminsky, S. A,, Weismiller, R. A,, Sinclair, H. R., and Hinzel, E. J. 1978. Soil Sci. SOC.Am. J . 42, 768-771. Kohnke, Helmut. 1968. In “Soil Physics,” pp, 39-54. McGraw-Hill, New York. Kojima, M. 1958a. Soil Plnnt Food 3, 204. Kojima, M. 1958b. Abstr. Soil Plant Food 3, 206. Kornblau, M. L. 1979. M. S. thesis, Colorado State University, Fort Collins. Kornblau, M. L., and Cipra, J. E. 1983. Remote Sensing Environ. 13, 103-1 12. Kristof, S. J. 1971. J . Soil Water Conserv. 26, 15-18. Kristof, S. J., and Baumgardner, M. F. 1975. Agron. J . 67, 317-321. Kristof, S. J., and Zachary, A. L. 1971. Proc. Symp. Remote Sensing Environ., 7th pp. 20952108. Kristof, S. J., and Zachary, A. L. 1974. Photogr. Eng. 40, 1427-1434. Latz, K., Weismiller, R. A,, Van Scoyoc, G. E., and Baumgardner, M. F. 1984. Soil Sci. SOC.Am. J . 48, 1130-1134. Lewis, D. T., Seever, P. M., and Drew, J. V. 1975. Soil Sci. SOC.Am. Proc. 39,330-335. Lindberg, J. D., and Snyder, D. G. 1972. Am. Mineral. 57,485-493. Longlois, K. H., Osterholz, L. C., and Kirschner, F. R. 1976. Indiana Acad. Sci. Proc. 85, 126. . Am. J. 44,667-676. LOW,P. F. 1980. Soil S C ~SOC.

42

MARION F. BAUMGARDNER ET A L .

Mainguet, M. M., Spiers, B., Canon, L. B., and Langeraar, W. D. 1978. Application of Landsat imagery to soil degradation assessment and mapping at 1:5,000,000: Niger, Upper Volta, Eastern Mali, Northern Nigeria, Chad, and Central African Empire. FAO, Rome. Mathews, H. L., Cunningham, R. L., and Petersen, G . W. 1973a. Soil Sci. SOC. Am. Proc. 37, 421 -424. Mathews, H. L., Cunningham, R. L., Cipra, J. E., and West, T. R. 1973b. Soil Sci. SOC.Am. Proc. 37, 88-93. May, G. A,, and Petersen, G. W. 1975. Remote Sensing Enuiron. 4, 21 1-220. Meyer, M. P., and Calpouzos. L. 1968. Photogr. Eng. Remote Sensing 36, 11 16-1125. Mimms, D. L. 1982. M.S. thesis, Pennsylvania State Univ., State College. Mitchell, C. W., and Ghorashian, K. 1978. The application of Landsat imagery to soil degradation mapping of Iran at 1:5,000,000. FAO, Rome. Mitchell, C. W., and Howard, J. A. (eds). 1978. The application of Landsat imagery to soil degradation mapping at 1: 1,000,000. FAO, Rome. Mitchell, C. W., Pacheco, R., and Howard, J. A. 1978. The application of Landsat imagery to the soil degradation mapping of Jordan, Syria, and Iraq at 1 :5,000,000. FAO, Rome. Montgomery, 0. L. 1976. Ph.D. dissertation, Purdue University, West Lafayette, Indiana. Montgomery, 0. L., and Baumgardner, M. F. 1974. Tech. Rep. 112674. Lab. Applic. Remote Sensing, Purdue Univ., West Lafayette, Indiana. Munsell, A. H. 1947. “A Color Notation,” 10th Ed. Munsell Color Co., Baltimore. Munsell Color. 1975. “Munsell Soil Color Charts.” MacBeth Division of Kollmorgen Corp. Baltimore. Myers, V. I., and Allen, W. A. 1968. Appl. Opt. 7, 1819-1838. Myers, V. I., Carter, D. L., and Rippert, W. J. 1966. J . Irrig. Drain. 92, 59-68. Nickerson, D. 1940. Opt. SOC.Am. J . 30, 375-386. Nickerson, D. 1946. Color measurement and its application to the grading of agricultural products. Misc. Publ. No. 580, U S . Dept. of Agriculture. Nicodemus, F. E., Richmond, J. C., Hsia, J. J., Ginsberg, 1. W., and Limperis, T. 1977. NBS Monograph 160. U S . Govt. Printing Office, Washington, D.C. Obukhov, A. I., and Orlov, D. S . 1964. Sou. Soil Sci. 2, 174-184. O’Neal, A. M. 1923. Soil Sci. 16, 275-279. Orlov, D. S . 1966. Sou. Soil Sci. 13, 1495-1498. Pacheco, R. 1978. The use of Landsat imagery for assessing soil degradation in Morocco. Proc. Int. SOC. Soil Sci. Rome. Page, N. R. 1974. Agron. J . 66, 652-653. Parada, N., and Pinto, S. 1983. Utilizacao de tecnicas de sensoriamento para a characterizacao de erosao do solo no sw do estado de Sao Paulo. Instituto de Pesquisas Espaciais, Sao Jose dos Campos, Brazil. Parks, W. L., and Bodenheimer, R. E. 1973. Proc. Symp. Signif: Results Obtained Earth Resources Technol. Satellite-1 N A S A SP-327 pp. 121-125. Pazar, S. E. 1983. M.S. thesis, Purdue University, West Lafayette, Indiana. Pendleton, R. L., and Nickerson, D. 1951. Soil Sci. 71, 35-43. Peterson, J. B. 1980. Use of spectral data to estimate the relationship between soil moisture tensions and their corresponding reflectances. Annu. Rep. O W R T Purdue Uniu. pp. 1-18. Peterson, J. B., Robinson, B. F., and Beck, R. H. 1979. Proc. Symp. Machine Process. Remotely Sensed Data pp. 264-273. Planet, W. G. 1970. Remote Sensing Enuiron. 1, 127-129. Ranzani, G. 1969. “Manual de Levantamento de Solos. Editora Blucher, Sao Paulo. Rao, R. G . S., and Ulaby, F. T. 1977. Remote Sensing Enuiron. 6, 289-301. Rao, K. V. S., Karale, R. L., and Singh, A. N. 1982. Abstr. Trans. Int. Congr. Soil Sci., 12th, New Delhi. p. 132.


43

Resende, M. 1976. Ph.D. dissertation, Purdue University, West Lafayette, Indiana. Rice, T. D., Nickerson, D., ONeal, A. M., and Thorp, J. 1941. Preliminary color standards and color names for soils. Misc. Publ. No. 425, U.S. Dept. of Agriculture. Robinson, B. F., and Biehl, L. L. 1979. Proc. Annu. Int. Tech. Symp., 23rd, SPIEE, Bellington, Washington. 196, 16-26. Robinson, B. F., and DeWitt, D. P. 1983. I n “Manual of Remote Sensing” (R. N. Colwell, ed.), pp. 293-333. Amer. SOC.of Photogr., Falls Church, Virginia. Schreier, H. 1977. Proc. Can. Symp. Remote Sensing, 4th I, 106-112. Schutt, J. B., Holben, B. N., Shai, C. M., and Henninger, J. H. 1981. Appl. Opt. 20, 2033-2035. Seevers, P. M., and Drew, J. V. 1973. Proc. Symp. Sign$ Results Obtained Earth Resources Technol. Satellite-I, NASA SP-327 pp. 87-89. Seevers, P. M., Lewis, D. T., and Drew, J. V. 1974. Proc. Earth Resources Technol. Satellitel, 3rd pp. 225-232. Shai, C. M., and Schutt, J. B. 1971. NASA X-762-71-266. Goddard Space Flight Center. Shaw, C. F. 1937. Soil Sci. SOC. Am. Proc. 2,431-436. Singh, A. N., Kristof, S. J., and Baumgardner, M. F. 1977. Lab. Applic. Remote Sensing, Purdue Univ. Tech. Rep. 1 1 1477. Soil Survey Staff, Soil Conservation Service, U.S. Dept. Agric. 1951. Soil survey manual. Agric. Handb. (18). Soil Survey Staff, Soil Conservation Service, U.S. Dept. Agric. 1975. Soil Taxonomy. Agric. Handbook 436. U.S. Govt. Print. Office, Washington, D.C. Steinhardt, G. D., Franzmeier, D. P., and Cipra, J. E. 1975. Indiana Acad. Sci. Proc. 84,463-468. Stoner, E. R. 1979. Ph.D. dissertation, Purdue Univ., West Lafayette, Indiana. Stoner, E. R., and Baumgardner, M. F. 1981. Soil Sci. SOC. Am. J . 45, 1161-1165. Stoner, E. R., and Horvath, E. H. 1971. Proc. Int. Symp. Remote Sensing Enuiron., 7th pp. 2109-21 13. Stoner, E. R., Baumgardner, M. F., and Swain, P. H. 1976. Agron. J . 68, 55-59. Stoner, E. R., Baumgardner, M. F., Biehl, L. L., and Robinson, B. F. 1980a. Purdue Uniu. Agric. Exp. Sta. Res. Bull. (962). Stoner, E. R., Baumgardner, M. F., Weismiller, R. A,, Biehl, L. L., and Robinson, B. F. 1980b. Soil Sci. SOC. Am. J . 44, 572-574. Stoner, E. R., Joyce, A. T., and Hogg, H. C. 1983. Dig. Int. Symp. Geosci. Remote Sensing 1, 8.1 -8.7. Strandberg, C. H. 1968. I n “Manual of Color Aerial Photography” (T. Smith and A. Abraham, eds.), pp. 3-11. Amer. SOC.of Photogr., Falls Church, Virginia. Thompson, D. R., and Henderson, K. E. 1984. Soil Sci. SOC.Am. J . 48, 1316-1319. Thompson, D. R., Pitts, D. E., and Henderson, K. E. 1983. Soil Sci. SOC.Am. J . 47, 542-546. Thompson, D. R., Henderson, K. E., Houston, A. G., and Pitts, D. E. 1984. Soil Sci. SOC.Am. J . 48, 137-142. Valenzuela, C . R. 1979. Estudio integrado de 10s recursos naturales del departamento de Oruro, La Paz, Bolivia, Programa ERTS/GEOBOL, pp. 240-384. van Sleen, L. A. 1982. Landsat data: Their use and accuracy for small scale soil surveys and their time and cost efficiency. Proc. Int. ConJ Remote Sensing Arid Semiarid Lands, Cairo. Vinogradov, B. V. 1981. Sou. Soil Sci. 11, 114-123. Weismiller, R. A,, and Kaminsky, S. A. 1978. J . Soil Water Conseru. 33, 287-289. Weismiller, R. A,, Kirschner, F. R., Kaminsky, S. A,, and Hinzel, E. J. 1979. Lab. Applic. Remote Sensing, Purdue Univ. Tech. Rep. 040179. Weismiller, R. A., Persinger, I. D., and Montgomery, 0. L. 1977. Soil Sci. SOC.Am. J . 41, 1 166- 1170. Westin, F. C. 1973. ERTS 1 imagery: A tool for identifying soil associations. Proc. Earth Survey Problems through Use Space Tech., Gen. Assembly, Comm. Space Res., Konstanz.

44


Westin, F. C. 1974. Proc. Earth Resources Technol. Satellite-1 Symp., 3rd pp. 183-204. Westin, F. C., and Frazee, C. J. 1976. Soil Sci. Soe. Am. J. 40, 81-89. Westin, F. C., and Lemme, G . D. 1978. Photogr. Eng. Remote Sensing 44, 315-325. Westin, F. C., and Myers, V. I. 1973. Proc. Symp. Sign$ Results Obtained Earth Resources Technol. Satellite-1, NASA SP-327 pp. 973-980. Young, E. R., Clark, K. C., Bennett, R. B., and Houk, T. L. 1980. Appl. Opt. 7, 3500-3505. Zissis, G . J. 1979. In “The Infrared Handbook” (W. L. Wolfe and G. J. Zissis, eds.). Env. Res. Inst. Michigan, Ann Arbor. Zwerman, C. H., and Andrews, A. I. 1940. J . Am. Ceram. SOC.23,93-102.

ADVANCES IN AGRONOMY. VOL. 38

APPLICATION OF GEOSTATISTICS TO SPATIAL STUDIES OF SOIL PROPERTIES B . B . Trangmar. R . S . Yost. and G . Uehara2

.

Soil Bureau Department of Scientific and Industrial Research. Christchurch. New Zealand 2Departrnent of Agronomy and Soil Science College of Tropical Agriculture and Human Resources University of Hawaii. Honolulu. Hawaii

I. Introduction

.........................................

I1. Nature of Soil Variability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

A . Systematic and Random Variation . . . . . . . . . . . . . . . . . . . . . . . . . B. Nested Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111. Traditional Methods of Describing Soil Variability . . . . . . . . . . . . . . . . . . A . Soil Classification and Soil Survey . . . . . . . . . . . . . . . . . . . . . . . . . B. Statistical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV. Regionalized Variable Theory and Geostatistics . . . . . . . . . . . . . . . . . . . . A . Development of Geostatistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Theory of Regionalized Variables . . . . . . . . . . . . . . . . . . . . . . . . . . V. Analysis of Spatial Dependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A . Autocorrelation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Semi-variograms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

VI . Interpolation by Kriging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Punctual Kriging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C . Block Kriging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D . Co-Kriging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E. Universal Kriging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . F. Kriging from Non-normally Distributed Data . . . . . . . . . . . . . . . . . . VII . Perspectives: Future Use of Geostatistics in Soil Research . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

45 47 47 47 49 49 51 53 53

54 56 56 57 70

70 71 75 80 85 88 89 91

I. INTRODUCTION The precision of statements that can be made about soil properties at any location depends largely on the amount of variation within the area sampled . As heterogeneity of soils increases. the precision of statements about their 45

Copyright 0 1965 by Academic Press. Inc. All rights of reproduction in any form reserved.

46

B. 9. TRANGMAR ET AL..

properties, behavior, and land use performance decreases. Soil scientists are restricted to limited observations of the earth‘s surface, necessitating the extrapolation of soil properties from locations where they are known to others where they are not known. The precision of such extrapolation is strongly influenced by the variability of soils both within sampling units and between locations. Similarly, the likelihood of successful transfer of land use experience from known to unknown sites is strongly influenced by the spatial and temporal variability of soil and other environmental parameters. Recognition of the importance of spatial variability on land use has led to study of soil heterogeneity, ranging from a global scale (Food and Agricultural Organization, 1974) to changes in structural and chemical composition of soil minerals on a microscale (Sawhney, 1977). The nature of soil variability is itself scale-dependent because the soil-forming factors and processes interact over many different spatial and temporal scales (Burrough, 1983a). As a result, the nature of variability identified by spatial studies of soil properties depends largely on the scale of observation, the properties in question, and the methodology used to conduct the investigation (Wilding and Drees, 1983). Soil classification and soil survey have traditionally been the most practical approaches to grouping similar and separating different soils on a regional scale. Variability of properties within soil mapping units and within smaller sampling units, such as fields, experimental plots, or pedons, is acknowledged and has often been described by classical statistical methods (Beckett and Webster, 1971; Wilding and Drees, 1978, 1983). Classical statistics assumes that the sampling unit mean is the expected value everywhere in the unit, with an estimation error expressed by the within-unit variance. This approach assumes that variability about the mean is random and contains no reference to the geographical distribution of differences within the sampling units. Several studies have shown that this “random” aspect of soil variability often contains a component that is spatially dependent (Campbell, 1978; Burgess and Webster, 1980a; Gajem et al., 1981; Vieira et al., 1981; Yost et al., 1982a; McBratney et al., 1982). This implies that within a given distance or range of spatial dependence differences in soil properties can be described as a function of their spatial separation. Traditional methods of soil classification and statistical analysis do not directly consider this aspect. Recent developments in statistical theory enable spatial relationships among sample values to be quantified and used for interpolation of values at unsampled locations. These developments are based on the theory of regionalized variables (Matheron, 1963, 1965). This theory (under the name of “geostatistics”) takes into account both the structured and random characteristics of spatially distributed variables to provide quantitative tools for their description and optimal, unbiased estimation. Geostatistical analysis

APPLICATION OF GEOSTATISTICS

47

of spatial variability has been extensively applied to estimation of ore reserves in the mining industry (Journel and Huijbregts, 1978; David, 1977; Clark, 19791, water resources research (Delfiner and Delhomme, 1973; Delhomme, 1978, 1979), soil science (Burgess and Webster, 1980a,b; Vieira et al., 1981; Yost et al., 1982a,b), and archaeology (Zubrow and Harbaugh, 1979). This article briefly reviews some of the traditional methods of describing soil variability, discusses geostatistical approaches to quantifying spatial dependence and their use for interpolation under different kinds of spatial variation, and, lastly, identifies some future applications of geostatistics to spatial studies in soil and agronomic research.

II. NATURE OF SOIL VARIABILITY A. SYSTEMATIC AND RANDOM VARIATION Spatial variation of soil properties contains systematic and random components. Wilding and Drees (1983) express systematic variability as gradual or distinct changes (or trends) in soil properties that can be understood in terms of soil-forming factors or processes at a given scale of observation. Sources of systematic variation may range from differences in topography, lithology, climate, biological activity, and age of soils in regional studies (Van Wambeke and Dudal, 1978) to differences in microfabric and physicochemical composition when soils are observed on a micro level (Blevins et al., 1970; Miller et al., 1971; Murphy and Banfield, 1978). Associated with systematic variation are differences in observed soil properties which cannot be related to a known cause. In addition, there are often spatial, temporal, and measurement sources of variation which cannot be discerned by the nature or scale of the investigation (Ball and Williams, 1968). This unexplained heterogeneity is termed “random” or “chance” variation by Wilding and Drees (1983) or “noise” by Webster and Cuanalo (1975) and Burrough (1983a). Clearly, as the soil body is studied in greater detail, heterogeneity that may have been previously defined as random variation may be found to contain a systematic component. B. NESTED EFFECTS

Soil variability is the product of soil-forming factors operating and interacting over a continuum of spatial and temporal scales. Processes which

48

B. B. TRANGMAR ET AL.

operate over large distances (e.g., climate) or long time periods (e.g., soil weathering) are modified by other processes which operate more locally (e.g., erosion and deposition of parent materials) or more frequently (e.g., weather). This nested nature of soil variation implies that the kind and causes of heterogeneity that are identified in variability studies depends largely on the scale or frequency of observation. Multistage sampling has often been necessary to effectively describe different orders of soil variation within field areas (Hammond et a!., 1958; Nortcliff, 1978; Burrough, 1983a). Similarly, many soil properties (e.g., temperature, moisture content) have been measured over many different time intervals to determine their temporal fluctuation. The change in spatial variability with increasing scale factor depends on the soil property in question and the soil factors determining spatial change (Wilding and Drees, 1983). Total variance will increase as sampling area increases (Beckett and Webster, 197l), but relative contributions of variance at different scales to the total variance follow no consistent pattern (Wilding and Drees, 1983). Much of the variability for some properties may occur over short distances within sampling units (McIntyre, 1967; Protz et al., 1968; Beckett and Webster, 1971), while for others a large proportion of the total variance is caused by long-range differences (Webster and Butler, 1976). The change in variability with increasing scale factor may be linear, curvilinear (upward or downward), or irregular, where different soil processes exert dominating effects over different spatial scales (Webster and Butler, 1976; Nortcliff, 1978; Burrough, 1983a). In their study of changes in soil properties over a range of sampling distances, Webster and Butler (1976) found that most of the within-field variance of phosphorus occurred within distances of 5 m; of bulk density and water content, over 18 m; of soluble potassium, over 56-180 m; of pH, over 56 and 180 m; and of morphological properties, over 5 and 180 m. Schafer (1979) found that in natural soils 50-75 % of the total variation in soil texture, color, root abundance, A horizon thickness, and depth to carbonates occurred at distances >500 m, while soils formed in recent mine spoils showed similar magnitudes of variation at 10 m. Differences in properties of natural soils were correlated to systematic variation of geomorphic factors, while soils from mine spoils were dependent on random variation in parent material composition as a function of mining and reclamation methods. Babaloa (1978) found that large variation of hydraulic conductivity in a 0.3-ha plot relative to a 92-ha field was caused by local changes in particle size distribution and bulk density. This short-range variability of soil hydraulic properties is considered particularly important by Wilding and Drees (1983) and Bouma (1983) because many soil processes are dependent on rates and directions of water movement and chemical transport. Beckett

-=


49

and Webster (1971) also acknowledged the importance of short-range variation by concluding that much of the variability within any soil sampling unit may be present within any square meter of it.

Ill. TRADITIONAL METHODS OF DESCRIBING SOIL VARIABILITY Soil classification, its field application through soil survey operations, and statistical analysis have been the most commonly used methods for describing soil variation. These methods have been comprehensively reviewed by Beckett and Webster (1971) and Wilding and Drees (1978, 1983), so only their main concepts are discussed here.

A. SOILCLASSIFICATION AND SOILSURVEY Soil classification and soil survey have been the most commonly used methods for partitioning field variation on a regional scale. By grouping soils that are similar and separating those that are different, this approach also forms the basis for establishing relationships between individual soils, predicting properties at unsampled locations, predicting soil behavior, and identifying potential uses (Buol et al., 1980). One of the assumptions made in soil classification systems, and in soil survey practice, is that soil differences can be adequately characterized by relatively few diagnostic properties. The diagnostic properties that are used to define categories and taxa of soil classification systems are ideally those which have greatest independence of variation from each other but which have high covariance with many other nondiagnostic properties. This results in the variance within taxonomic units measured over all properties being minimized with respect to their total variance (Norris, 1971). This condition applies to all levels of hierarchical classification systems. Successively lower levels of classification are designed to partition more of the total variation, thus creating more homogeneous classes. Several statistical studies have shown that the highest categories of soil classification systems tend to account for much of the variance in specific properties that is explained by classification and that lower categories only contribute small additional amounts to improving the homogeneity of taxa (Beckett and Webster, 1971; Yost and Fox, 1981; Trangmar, 1984). Small variance reductions at lower levels often result from correlation of diagnostic criteria at lower levels with

50


those of the more general, higher levels (Yost and Fox, 1983; Trangmar, 1984). Mapping the spatial distribution of soil taxonomic units involves systematically partitioning the landscape into soil mapping units that are reasonably homogeneous and that can be readily portrayed at the mapping scale used. Taxonomic impurity of soil mapping units is acknowledged, and the proportion of inclusions are commonly specified or estimated in soil survey reports. The proportion of inclusions allowed in simple soil mapping units ranges from 15% (Soil Survey Staff, 1951; Taylor and Pohlen, 1962) to 35% (Mapping Systems Working Group, 1981) before compound units are established. Constraints of time and sampling effort usually restrict the soil surveyor to only a few field observations per mapping unit, with the result that heterogeneity may often exceed the desired limits. In their review of this topic, Beckett and Webster (1971) concluded that simple mapping units might actually average only 50% purity. Burrough et al. (1971) found that mapping unit purity varied with map scale and observation density. Purity ranged from 45563% at a scale of 1:63,360 and 6 5 4 7 % at a scale of 1:25,000. This apparently high degree of taxonomic variability within mapping units is often diminished in importance because impurities often differ only in minor definitive features and do not require different management (Bascomb and Jarvis, 1976). If compound units are established, soil associations are used to portray groups of geographically associated soils (each of which is confined to a particular facet of the landscape) which occur in a predictable pattern (Dent and Young, 1981). Associations can be resolved into simple mapping units at a more detailed scale of investigation. If the soil pattern cannot be resolved because of its intricacy, it is mapped as a complex. Soil survey reports generally describe members of associations and complexes and indicate their relative proportions in the mapping unit. The areas of individual members within an association may be large enough to be managed separately from other members. Complexes generally have to be managed as complexes because of the small land areas covered by individual members (Cutler, 1977). The importance of variation in soil properties depends on the kind and intensity of land use of the area in question. Clearly, soil properties differ in their effects on different kinds of land use, and some specific chemical or physical properties may have more dominance than others. As a result, spatial variability of specific properties within mapping units is also of considerable interest to the map user. The variability of diagnostic properties of mutually exclusive taxonomic units is fixed and their distribution is truncated by limits between taxa. Not surprisingly, studies have shown that variability of such properties is smaller


51

within pedons than within series than within the corresponding mapping units (Beckett and Webster, 1971; Beckett and Burrough, 1971; Wilding and Drees, 1978, 1983). The variability of most properties is usually less within mapping units than between units (Wilding et al., 1965), although where variable levels of management have been applied, within-unit variation may exceed that between units (McCormack and Wilding, 1969; Beckett and Webster, 1971). The Benchmark Soils Project found that variability of soil properties within the same soil family of Soil Taxonomy was sufficiently low to support the hypothesis that soils of the same family have similar responses to similar management practices (Silva, 1984). Properties most affected by soil management (e.g., soluble phosphorus, exchangeable cations, sulphate-S, total sulfur) are commonly more variable than the morphological (e.g., color, A horizon thickness), physical (e.g., particle size, bulk density), and chemical (e.g., pH) properties used to define taxonomic units (Beckett and Webster, 1971; Adams and Wilde, 1976a,b; Wilding and Drees, 1978, 1983). In their summary paper, Wilding and Drees (1983) give mean coefficients of variation (CVs) for exchangeable calcium, magnesium, and potassium of 50-70% ranging up to 160%. They also concluded that the variability of physical properties such as Atterberg limits, particle size fractions, bulk density and water content (CVs of 10-53 %) is often much less than hydraulic conductivity (CVs of 50- 150 %) measured over the same area. As a result of such variation within sampling units, soil surveys cannot be expected to reliably predict variation of all properties, particularly those that are easily influenced by soil management.

B. STATISTICAL ANALYSIS

The application of statistics of soil variation has been summarized by Beckett and Webster (1971) and Wilding and Drees (1978, 1983). A more comprehensive treatment of the topic can be found in Webster (1977). Classical statistics assumes that the expected value of a soil property z at any location x within a sampling area is

+ E(X)

z(x) = /l

(1)

where p is the population mean or expected value of z and E ( X ) represents a random, spatially uncorrelated dispersion of values about the mean. Deviations from the population mean are assumed to be normally distributed with a mean of zero and a variance of (r2 (Sokal and Rohlf, 1969).

52


Many soil properties have skewed probability distributions and require transformation (e.g., natural log) to the normal distribution prior to statistical analysis (Cassel and Bauer, 1975; Wagenet and Jurinak, 1978). Other properties may be bimodally distributed (Smeck and Wilding, 1980),in which case each mode may be treated as a separate population for statistical analysis (Wilding and Drees, 1983). Because mean values are used for estimation of properties at unsampled locations within sampling units, statistics of dispersion (e.g., coefficients of variation, standard deviation, standard error, confidence limits) are used to indicate precision of the mean as an estimator. These statistics have been used extensively to document the variation of soil properties within sampling areas such as soil mapping units (Wilding et al., 1965; McCormack and Wilding, 1969; Adams and Wilde, 1976a,b),fields (Cassel and Bauer, 1975; Biggar and Nielsen, 1976), experimental plots (Jacob and Klute, 1956; Nielsen et al., 1973), and pedons (Smeck and Wilding, 1980). Analysis of variance and subsequent statistical testing has been a common method for comparing variation among sampling units (Jacob and Klute, 1956; Cassel and Bauer, 1975; McBratney et al., 1982). The influence of random sources on variance within sampling units has prompted much research into the sampling size required to estimate the sample mean at various levels of precision and confidence intervals (Ball and Williams, 1968; Beckett and Webster, 1971; Cassel and Bauer, 1975; Biggar and Nielsen, 1976; Adams and Wilde, 1976b). As within-unit variance increases, a proportionately larger number of samples is required to estimate the mean for a given level of confidence. Classical statistical procedures assume that variation is randomly distributed within sampling units. Actually, soil properties are continuous variables whose values at any location can be expected to vary according to direction and distance of separation from neighboring samples (Burgess and Webster, 1980a). By so varying, soil properties exhibit spatial dependence within some localized region. Estimation using the classical model cannot be improved on if the initial classification of a region into discrete sampling or mapping units accounts for all the spatially dependent variance (McBratney et al., 1982). However, spatial dependence of soil properties will usually occur in most sampling units. The classical model is inadequate for interpolation of spatially dependent variables, because it assumes random variation and takes no account of spatial correlation and relative location of samples. Several techniques which incorporate sample location to varying degrees have been used for interpolation of soil properties. These include proximal weighting (Van Kuilenburg et al., 1982), moving averages (Webster, 1978), weighted moving averages using inverse distance and inverse distance squared functions (Van Kuilenburg et al., 1982), trend surface analysis


53

(Watson, 1972; Whitten, 1975), and spline interpolation (Greville, 1969). These techniques are empirical, and although they may seem reasonable for many applications, they are theoretically unsatisfactory (Burgess and Webster, 1980a). Some provide good interpolation under optimal data configurations, but most give biased estimates that are not optimal; many do not provide estimates of the interpolation error and those that do, do not attempt to minimize that error (Burgess and Webster, 1980a).

IV. REGIONALIZED VARIABLE THEORY AND G EOSTATlSTlCS Recent developments in statistical theory enable spatial dependence of soil properties to be directly considered in interpolation. These developments are based on the theory of regionalized variables, which takes into account both the random and structured characteristics of spatially distributed variables to provide quantitative tools for their description and optimal, unbiased estimation. These tools can augment the more commonly used methods in analysis of soil variability.

A. DEVELOPMENT OF GEOSTATISTICS

Interpolation based on spatial dependence of samples was first used by D. G. Krige (1951, 1960) for estimation of the gold content of ore bodies in the mining industry of South Africa. Classical statistical interpolation procedures were considered inappropriate in the mining industry because they were biased and nonoptimal in that they did not take local spatial dependence into account during estimation. Interpolation procedures which considered local changes in ore content and grade were developed to obtain a method which would enable optimal sample placement to minimize the high cost of sampling mineral deposits. Krige’s practical methods were generalized and extended by Matheron (1963, 1965, 1969, 1970, 1971) into the theory of regionalized variables. This theory now forms the basis of procedures for analysis and estimation of spatially dependent variables. These procedures are known collectively as geostatistics. Blais and Carlier (1968) and Huijbregts and Matheron (1971) were among the first to apply kriging as an estimation procedure in mining engineering. Geostatistical theory continued to develop in the 1970s to include analysis of variables having lognormal (Rendu, 1979; Journel, 1980) or complex

54


(Matheron, 1976; Journel and Huijbregts, 1978; Jackson and Marechal, 1979) probability distributions and estimation in the presence of trends (Olea, 1974, 1975; Delfiner, 1976; Journel and Huijbregts, 1978). While the use of geostatistics has centered on the mining industry, it is now being used extensively in water resources research (Delfiner and Delhomme, 1973; Delhomme, 1978, 1979), soil science (Campbell, 1978; Burgess and Webster, 1980a,b; Vieira et al., 1981; Yost et al., 1982a,b), and archaeology (Zubrow and Harbaugh, 1979).

B. THEORY OF REGIONALIZED VARIABLES

Geostatistics are based on the concepts of regionalized variables, random functions, and stationarity. A brief theoretical discussion of these concepts is necessary to appreciate the practical application of geostatistics to the analysis of soil variation. Comprehensive coverage of regionalized variable theory and its geostatistical applications are given by David (1977), Journel and Huijbregts (1978), Clark (1979), and Royle (1980). 1. Regionalized Variables and Random Functions

A random variable is a measurement of individuals that is expected to vary according to some probability distribution law (Henley, 1981). The random variable is characterized by the parameters of the distribution, such as the mean and variance of the normal distribution. A regionalized variable z ( x ) is a random variable that takes different values z according to its location x within some region (Journel and Huijbregts, 1978). As such, a regionalized variable z(x) can be considered as a particular realization of a random variable 2 for a fixed location x within the region. If all values of z(x) are considered at all locations within the region, then the regionalized variable z(x) becomes a member of an infinite set of random variables Z(x) for all locations within the region. Such a set is called a random function because it associates a random variable 2 with any location x (Huijbregts, 1975). 2. Stationarity A random function Z(x) is said to befirst-order stationary if its expected value is the same at all locations throughout the study region, E[Z(x)] = m

(2)


55

where rn is the mean of classical statistics, and E[Z(x) - Z ( X + h)] = 0

(3)

where h is the vector of separation between sample locations. Second-order stationarity applies if the spatial covariance C(h) of each Z ( x ) and Z(x h ) pair is the same (independent of position) throughout the study region and depends on h:

+

C(h) = E[Z(x)

+

- ~ ] [ Z ( . X h) - m]

(4)

As h gets larger, C(h) decreases and the spatial covariance decays (Fig. 1). Stationarity of C(h)implies stationarity of the sample variance s2. The spatial covariance will approach the sample variance as the distance of separation tends to zero. Second-order stationarity does not apply if a finite variance and covariance cannot be defined, as in the case of trend phenomena (David, 1977), and a weaker form of stationarity called the intrinsic hypothesis must be assumed (Journel and Huijbregts, 1978). Second-order stationarity implies the intrinsic hypothesis, but not the converse. The intrinsic hypothesis requires that for all vectors of h, the variance of the increment Z ( x ) - Z(x + h) be finite and independent of position within the region, i.e., VAR[Z(x) - Z ( X

+ h)] = E [ Z ( x ) - Z(X + h)I2 = 2y(h)

(5)

Dividing by two yields the semi-variance statistic y(h). The semi-variance y depends on the vector of separation h. Ideally, y is zero at h = 0, but increases as h increases (Fig. 1).

FIG.1. Relationship between the spatial covariance C(h) and the semi-variogram statistic y(h). (From Journel and Huijbregts, 1978.)

56


V. ANALYSIS OF SPATIAL DEPENDENCE The concepts of regionalized variables and stationarity provide the theoretical basis for analysis of spatial dependence using autocorrelation or semivariograms. A. AUTOCORRELATION Autocorrelation functions express the linear correlation between a spatial series and the same series at a further distance interval (Vieira et al., 1981). Their definition assumes second-order stationarity, in which case the autocorrelation is expressed as

r(h) = C(h)/S2

(6)

where r(h) is the autocorrelation among samples at distance of separation, or lag, h. A plot of the autocorrelation values r(h) versus the lag is called the autocurrelogram. The maximum value of r(h) is 1 at zero distance ( h = 0), and values decrease with increasing h. The distance a at which r(h) no longer decreases defines the range over which samples of the variable are spatially dependent. Values of the autocorrelation function are normalized to the range from - 1 to 1 inclusive, making for easy interpretation of data values. The mean, variance, and autocorrelation function completely characterize the random function Z(x), where Z(x) is normally or lognormally distributed (Gajem et al., 1981). Autocorrelograms have been used to express spatial changes in fieldmeasured soil properties and the degree of dependency among neighboring observations (Webster, 1973, 1978; Webster and Cuanalo, 1975; Vieira et al., 1981 ; Sisson and Wierenga, 1981). Such information aids identification of the maximum sampling distance for which observations remain spatially correlated and can be used in designing soil sampling schemes (Vieira et al., 1981) or defining minimum cell size for interpolation by moving average techniques (Webster, 1978). Webster and Cuanalo (1975) used autocorrelation analysis of soil chemical properties sampled along transects to locate soil boundaries. Russo and Bresler (1981) found that ranges of spatial dependence for soil moisture characteristics decreased with depth, indicating greater continuity of these properties in topsoils than in subsoils. Spatial analysis of soil properties using autocorrelograms has been restricted to data sampled at regular spacings along transects (Webster and Cuanalo, 1975; Gajem et al., 1981) or grids (Vieira et al., 1981).


57

Soil properties which do not show second-order stationarity do not have finite variances over the distance between sample locations, making it impossible to define the autocorrelation function (David, 1977). This nonstationarity can be removed by detrending, but it is often more convenient to assume the intrinsic hypothesis and use semi-variograms for quantifying spatial dependence (Vieira et a/., 1981).

B. SEMI-VARIOGRAMS I . Assumptions and Dejinitions

Structural analysis of spatial dependence using semi-variograms can be made using weaker assumptions of stationarity than are necessary for autocorrelation. Semi-variogram analysis has the added advantage of defining parameters needed for local estimation by kriging (Section VI). The semi-variance statistic y(h) can be defined in terms of the variance s2 and spatial covariance C(h) of Z(x) if second-order stationarity applies, i.e., y(h) = s2 - C(h)

(7)

Alternatively, the weaker intrinsic hypothesis can be assumed (Section IV,B). The semi-variance y(h) describes the spatially dependent component of the random function Z. It is equal to half the expected squared distance between sample values separated by a given distance h, i.e., y(h) = E[Z(x) - Z(X

+ h)I2

(8)

Application of regionalized variable theory assumes that the semi-variance between any two locations in the study region depends only on the distance and direction of separation between the two locations and not on their geographic location. Based on this assumption, the average semi-variogram for each lag can be estimated for a given volume of three-dimensional space. The semi-variance at a given lag h is estimated as the average of the squared differences between all observations separated by the lag:

where there are N ( h ) pairs of observations. The semi-variogram for a given direction is usually displayed as a plot of semi-variance y(h) versus distance h (Fig. 2A).

58


A

a 1

ic

Sill

Maximum Variance in Data Set (k.)

0

/------

P

I

0

0)

0 c

Multiple, Equally SDoced Spaced Observations

.-L 0

.-I v)

._ E

-

Nugget Variance (C,)

v) I

I

I

Total Distance (h)

I

I

I

FIG. 2. (A) Idealized semi-variogram with zero nugget variance and (B) observed semivariograms for soil properties with nugget variance. (From Wilding and Drees, 1983.)

2. Parameters

The shape of the experimental semi-variogram may take many forms, depending on the data and sampling interval used. Ideally, the semi-variance increases with distance between sample locations, rising to a more or less constant value (the sill) at a given separation distance, called the range of spatial dependence, a (Fig. 2A). The sill approximates the sample variance s2 for stationary data. Samples separated by distances closer than the range are spatially related. Those separated by distances greater than the range are not spatially related because the semi-variance equals s2, implying random

59

APPLICATION OF GEOSTATISTICS Table I Parameter Values of Some Isotropic Semi-variograms for Soils and Related Data

Property PH

Exchangeable aluminum (medl00 g) Sodium content (meq/lO kg) Phosphorus sorbed (ppm) at 0.2 mg P/liter Sand (%) Bulk density (g/cc) Loam thickness (g/cc) Rice grain yield (g/mZ) Leaf phosphorus content (%) in sorghum

Sample spacing (m)

Range (m)

0.5 20

Nugget variance (% of sill)

Model"

4 320

4 23

S S

1,000 0.5 500

14,000 4 4,200

23 26 63

M S S

1.5

6

56

L

32,000

25

M

0.5 10

4 34

3 44

S S

0.2 20

6 100

30 24

L

0.5

18

55

S

Trangmar (1984) Vauclin et A/. (1983) Gajem et ~ l(1981) . Burgess and Webster (1980a) Trangmar (1984)

1.5

6

40

L

Trangmar (1982)

1,000

S

Reference Trangmar (1984) McBratney and Webster (1981b) Yost et al. (1982a) Trangmar (1984) Trangmar (1984) Burgess and Webster (1980a) Yost et a/. (1982a)

Semi-variogram model: S, spherical; M, Mitscherlich; L, Linear.

variation. The range also defines the maximum radius from which neighboring samples are drawn for interpolation by kriging (Section VI). Semi-variogram ranges depend on the scale of observation and the spatial interaction of soil processes affecting each property at the sampling scale used. Reported ranges of spatial dependence of soil properties vary from 0.6 m for 15-bar water sampled at 0.2-m intervals (Gajem et al., 1981) to 58 km for phosphorus sorbed at 0.2 mg P/liter sampled at 1-2-km intervals (Yost et a!., 1982a). Some ranges of semi-variograms for soil properties are given in Table I. An example of a well-structured semi-variogram is given in Fig. 3. Semi-variances may also increase continuously without showing a definite range and sill, thus preventing definition of a spatial variance, indicating the presence of trend effects and nonstationarity (Webster and Burgess, 1980; Gajem et al., 1981; Yost et al., 1982b). Other semi-variograms show a

60


,q

: :II! 0.

g .-

.

.

0. 12

7

.- 0.09

E

v)

I

I

I

,

,

0.00

0

3

6 9 Distance (km)

12

15

FIG.3. Example of a semi-variogram (for pH). (From Trangmar et al., 1984.)

complete absence of spatial structure, implying that there is no easily quantifiable spatial relationship between sample values at the sampling scale used. Ideally, the experimental semi-variogram should pass through the origin (Fig. 2A) when the distance of sample separation is zero. However, many soil properties have nonzero semi-variances as h tends to zero (Fig. 2B). This nonzero variance is called the “nugget variance” or “nugget effect” (Journel and Huijbregts, 1978). It represents unexplained or “random” variance, often caused by measurement error or microvariability of the property which cannot be detected at the scale of sampling. Some reported values of semivariogram nugget variances are given in Table I. The sum of the nugget variance C , and the spatial covariance C approximately equals the sill or sample variance sz for stationary data (Fig. 2B). The nugget variance can also be expressed as a percentage of the sill value (Table I) to enable comparison of the relative size of the nugget effect among properties (Yost et al., 1982a; Burrough, 1983a; Trangmar, 1984). Nugget variances of soil properties ranging from 0 (Vieira et al., 1981) up to 100% of the sill (Campbell, 1978; Luxmoore et al., 1981; Hajrasuliha et al., 1980) have been reported. A nugget variance of 0% of sill means that there is neither measurement error nor significant short-range variation present. The experimental semi-variogram exhibits pure nugget effect (100 % of sill) when y(h) equals the sill at all values of h. Pure nugget effect arises from very large point-to-point variation at short distances of separation and indicates a

61


total absence of spatial correlation at the sampling scale used. Increasing the detail of sampling will often reveal structure in the apparently random effects of the nugget and pure nugget variances (Burrough, 1983a). According to Journel and Huijbregts (1978), a pure nugget effect at all scales of sampling amounting to a single discontinuity at the origin is exceptional. If this occurs, it implies that the mean is the best estimator at every point in the study region. Part of the nugget variance may be caused by measurement and sampling error, so it also sets a lower limit to the precision of the sampling or measurement technique used (Burrough, 1983a).The size of the measurement error component is indicated if the nugget variance cannot be reduced by collecting additional samples at closer spacings. The true spatial component C of the sample variance is then also clearly defined (Fig. 2B). The magnitude of the nugget variance is important in kriging because it sets a lower limit to the size of the estimation variance and, therefore, to the precision of the interpolation. Figure 4 presents a set of idealized semi-variograms that commonly occur for soil properties. If short-range effects predominate, the semi-variogram has a large nugget variance (curve l), or if pure nugget effect occurs a straight line equal to the sill would be present. If a single, long-range process dominates, the semi-variogram is linear up to the sill, where it abruptly flattens out (curve 2). If several processes make important contributions to spatial dependence at different scales, the semi-variogram consists of several linear 50

i

40

20

10

I

0

1

2

4

I

I

I

6

8

10

1

12

I

14

L a g (h)

FIG. 4. Theoretical semi-variograms resulting from soil processes operating at different spatial scales.

62


portions, separated by marked slope changes at sampling intervals corresponding to the range of the soil process in question (curve 3). If several processes with similar contributions act over closely related scales, the resulting semi-variogram consists of a set of straight lines approximating a curve (curve 4). It is very difficult, if not impossible, to identify the relative contributions of each process for curves like type 4. 3. Estimation of Parameters

Parameters of experimental semi-variograms are commonly estimated using least squares regression, weighted for the number of sample pairs in each lag (Vieira et al., 1981; Yost et al., 1982a; Trangmar, 1984). This approach usually gives an adequate first approximation of semi-variogram model fitting against which the deviations of individual semi-variances from the overall structure can be assessed by critical review of the data. Minor errors in estimation of semi-variogram parameters make little difference to the reliability of interpolation because of the robustness of the kriging technique (David, 1977). The equations most commonly used to estimate parameters of isotropic or unidirectional semi-variograms are the linear equation (Burgess and Webster, 1980a; Hajrasuliha et al., 1980; Vauclin et al., 1983) as in Fig. 4, curves 1 and 2, and a segmented quadratic form known as the spherical model (Burgess and Webster, 1980a; Vieira et al., 1981;Van Kuilenburg et al., 1982; Vauclin et al., 1983; Trangmar, 1984) as in Fig. 4, curve 4. A Mitscherlich model was also used by Yost et al. (1982a) for estimating semi-variogram parameters. Segmented models such as the double spherical model of McBratney et al. (1982) have been used to estimate semi-variograms in which breaks in slope mark different ranges of spatial dependence associated with different soil processes (Fig. 4, curve 3). Other semi-variogram models that have been used in mining geostatistics (David, 1977), but which have not been used in soil science, include the De Wijsian (the linear model with the lag plotted on a log scale), the exponential (asymptotic convergence with the sill), and the “hole effect” model (for estimation of periodic semi-variances). The mathematical forms and detailed descriptions of the various models can be found in David (1977) and Journel and Huijbregts (1978). It is important to choose the appropriate model for estimating the semivariogram because each model yields quite different values for the nugget variance and range, both of which are critical parameters for kriging. The Mitscherlich and exponential forms have rarely been used because their infinite ranges imply very continuous processes (Journel and Huijbregts, 1978), which rarely occur in ore bodies or field soils. Yost et al. (1982a) found that an appropriate working range for the Mitscherlich form coincided with


63

the distance of separation at which the semi-variance equals 95 % of the sill. When fitted to the same experimental semi-variogram, the spherical model generally gives longer ranges and smaller nugget variances than the linear form but yields shorter ranges and larger nugget variances than the Mitscherlich form. Over intermediate lags there is little difference between the spherical or Mitscherlich model in estimating the semi-variance. 4. Sampling

Choice of configuration and minimum spacing of samples for semivariogram analysis has generally been based on the previous knowledge of variation within the study area, the objective of the study, and the costs of sampling and measurement. Sampling designs used for analysis of spatial dependence have included point samples collected along transects with regular (McBratney and Webster, 1981b; Gajem et al., 1981) or irregular spacings (Yost et al., 1982a), equilateral grids (Campbell, 1978; Burgess and Webster, 1980a; Hajrasuliha et al., 1980; Trangmar, 1982), equilateral grids with sampling at some shorter spacings in some “window areas” (Trangmar, 1984), and random sampling (Van Der Zaag et al., 1981; McBratney et al., 1982; Van Kuilenburg et al., 1982). Bulking of soil samples from within grid cells (Burgess and Webster, 1980a,b; McBratney and Webster, 1981a; Webster and Burgess, 1984) and areal measurements of crop parameters (Tabor et al., 1984; Trangmar, 1984) have also been made for semi-variogram analysis where spatial interpolation by block kriging is the study objective. McBratney and Webster (1983b) suggest that for soil mapping purposes transect sampling can be used to obtain a working semi-variogram to initially identify spatial dependence parameters. This could then be used to design an optimal sampling scheme for kriging (Section VI,C), if necessary, and would only require a fairly small proportion of the total sampling effort needed for kriging. They also suggest that if mean estimation variances or standard errors of within-sampling unit variation are required, then regular grid sampling may be the best strategy, with the interval determined by the number of observations that can be afforded. In our experience, it seems desirable to collect a number of samples at distances smaller than the smallest grid spacing to reliably estimate the semi-variogram at short lags and to reduce the size of the nugget variance (Trangmar, 1984).

5. Interpretation of Semi-variograms Analysis of spatial dependence using semi-variograms has contributed to our understanding of many aspects of soil variability, genesis, management, and interpretation. This section discusses some of these applications.

64


a. Isotropic and Anisotropic Variation. Soil properties are isotropic if they vary in a similar manner in all directions, in which case the semi-variogram depends only on the distance between samples, k. One semi-variogram applies to all parts of the study region and defines a circular range of spatial dependence about each sample location. Geometrical anisotropy occurs when variations for a given distance k in one direction are equivalent to variations for a distance kk in another direction. The anisotropy ratio k indicates the relative size of directional differences in variation. It characterizes an ellipsoidal zone of influence which is elongated in the direction of minimum variation. The direction of maximum variation is assumed to occur perpendicular to the direction of minimum variation (David, 1977). The anisotropy ratio would equal 1 and define a circular zone of influence if variation were the same in all directions, i.e., isotropic. Differences in slopes of individual semi-variograms computed in different directions reveal the presence or absence of anisotropic spatial dependence (Webster and Burgess, 1980; Burgess and Webster, 1980a; McBratney and Webster, 1981a, 1983a; Trangmar, 1984; Tabor et al., 1984). If anisotropy occurs, the semi-variogram computed in the direction of maximum variation will have the steepest slope, while that in the direction of minimum variation will have the lowest slope. Parameters of geometric anisotropic spatial dependence can be estimated by incorporating a directional component into the slope term of the semivariogram. This involves fitting a single equation which defines a continuous envelope of estimated semi-variograms for all directions between those of maximum and minimum variation. The anisotropic model used by Burgess and Webster (1980a), Webster and Burgess (1980), and Trangmar (1984) is

y(8, k) = C ,

+ [ A cos2(8 - I)) + B sin2(8 - +)]k

where y(0, h) is the semi-variance estimated in the direction 0 at distance of separation k, C, the nugget variance, I) the direction of maximum slope A (greatest variation), and B the slope of the semi-variogram at 90" to tj.The parameters A, B, and I)are generally estimated by least squares fitting of Eq. (10)to the pooled directional semi-variograms,with each semi-variance value being weighted by the number of pairs in each lag k. The anisotropy ratio is calculated as A/B. Slopes estimated by Eq. (10) from pooled directional semivariances compared closely with the slopes of the individual directional semivariograms for the data of Burgess and Webster (1980a) and Trangmar (1984). Figure 5 shows Eq. (10) fitted to semi-variances pooled from four directions. The direction of maximum variation is northeast to southwest and that of minimum variation is southeast to northwest.

65

APPLICATION OF GEOSTATISTICS 600-

DIRECTION + NE-SW 0 E-W 0 SE-NW

500-

A

S-N

4000 0

c 0

300-

.L

0

>

+

-1.200-

5

Ln

IOO-

0

3

9

6

Average

Distance

12

15

(km)

FIG.5. Geometric anisotropic model fitted to pooled directional semi-variances of sand content (%). (From Trangmar, 1984.)

An alternative linear anisotropic model which gives similar results is that of McBratney and Webster (1981a, 1983a), in which the square root of the slope term of Eq. (10) is used. Equations for applying the spherical model to anisotropic data are given in David (1977) but have yet to be applied to soil properties. The utility of anisotropic modeling lies in identification of changes in spatial dependence with direction which, in turn, reflect soil-forming processes. McBratney and Webster (1981a) found that the geometric anisotropy of peat thickness was related to the microtopography of the land surface prior to peat formation. The anisotropy was caused by directional differences in peat thickness across and up the slopes in the region. Trangmar (1984) found that the direction of maximum variation of particle size fractions occurred down the main axis of tuff fallout and deposition of alluvium. Anisotropy of pH and HC1-extractable phosphorus in the same area was caused by directional changes in the degree of soil weathering across geomorphic surfaces of different ages. Anisotropy ratios of up to 5.4 have been reported for soil properties, but directional differences of this magnitude are probably unusual for soils because most of the ratios are in the 1.3-4.0 range (Table 11). The relative degree of anisotropy between topsoils and subsoils in Table I1 does not show any clear pattern and probably depends on the particular properties and soil

66


Table 11 Anisotropy Ratios of Some Semi-variograms for Soils and Related Data Anisotropy ratio Property

Topsoil

Peat thickness (cm)

1.88

Stone content (%)

5.42

Sand

(%I

Silt (%)

PH HC1-extractable phosphorus (PPm) Electrical resistivity (am) Cotton petiole nitrate (ppm)

Subsoil

1.59

1.68

4.05 1.71

2.88 1.79

3.01 1.71

2.9 1 1.33

4.37 2.40 3.47

2.80 1.54 5.18

1.29 2.7

Reference McBratney and Webster (1981a) Burgess and Webster (1980a) McBratney and Webster (1983a) Trangmar (1984) McBratney and Webster (1983a) Trangmar (1984) McBratney and Webster (1983a) Trangmar (1984) Trangmar (1984) Trangmar (1984) Webster and Burgess (1980) Taboret al. (1984)

processes being studied. McBratney and Webster (1983a) were able to reliably fit one common anisotropic model to semi-variograms of sand and silt fractions in both topsoils and subsoils. All four semi-variograms had similar anisotropy ratios and directions of maximum variation, thus giving one simple linear model for four different variables. Soil properties which are highly correlated and whose auto-semi-variograms vary anisotropically often have anisotropic cross-semi-variograms (McBratney and Webster, 1983a). Similarly, properties whose auto-semivariances are isotropic tend to have isotropic cross-semi-variances (Vauclin et al., 1983). Zonal anisotropy is often expressed as nested semi-variogram structures in which the observed anisotropy cannot be reduced by a simple linear transformation of sample distance (Journel and Huijbregts, 1978). It may result in different sills or different forms of semi-variograms calculated for the same property in different directions (David, 1977). Zonal anisotropy is a common characteristic of properties showing geochemical or geophysical gradients caused by directional deposition of sediments or mineralization of ore bodies (Journel and Huijbregts, 1978). Zonal anisotropy can occur in


67

three dimensions and, although commonly observed in mineral deposits, it has not been described in the soils literature. The conceptual and mathematical models of zonal anisotropy are given in full by Journel and Huijbregts (1978). b. Trends. Many regionalized variables do not vary randomly but show local trends or components of broader regional trends. Quasi-stationarity (Journel and Huijbregts, 1978) can be safely assumed for interpolation purposes where there is a regional trend but local stationarity because the trend is more or less constant within the estimation neighborhood. Regional trends are indicated by semi-variances that increase with distance of sample separation and either do not approach a sill (Gajem et al., 1981) or have a sill which considerably exceeds the general variance s2 (Bresler et al., 1984). In this case, simple kriging is used locally and an appropriate radius for the kriging neighborhood is the distance at which the semi-variance intersects the general variance (David, 1977). In the case of overall stationarity but locally occurring trends, the stationarity assumptions of Section IV,B,2 break down and universal kriging must be used for local estimation. The stationarity assumptions are violated because the expected value of the random function 2 is not always constant within the neighborhood and is no longer equivalent to the mean, but to a general quantity of drift, m(x), which changes locally within the neighborhood. The significance of identifying locally changing drift lies in difficulties with kriging from nonstationary data (Section VI1,E). Local trends, or drift, are commonly identified by simply plotting values of the soil property as a function of distance or by examination of semivariograms (David, 1977; Webster and Burgess, 1980). Bresler et al. (1984) also analyzed residuals from the regression of soil property values on location to identify the presence of trends. Ideally, changing drift produces gently parabolic semi-variograms of the raw data which are concave upward near the origin (David, 1977). In practice, however, Webster and Burgess (1980) considered that the presence of short-range variation in most soils and noisy data over short lags generally makes local trend identification difficult. c. Periodic Phenomena. Periodicity of parent material deposition and repetition of land form sequences are often quoted sources of soil variation (Butler, 1959). Periodic variation is expressed in semi-variograms as a “hole effect” (Fig. 6), which is indicative of nonmonotonic growth of the semivariance with distance (Journel and Huijbregts, 1978). The hole effect can appear on models with or without sills. Periodic behavior in ore bodies is said to indicate a continuous process of mineralization and is often characteristic of a succession of rich and poor zones (David, 1977). The continuity of the process is indicated by the smooth shape of the hole-effect semi-variogram. The hole effect will usually be present only in certain directions because the

68


1.41.2-

--

1.0-

c ?.

.8-

.6.4.2 -

2

4

6

8

10 Lag

12

14

16

18

20

22

(h)

FIG.6. Typical shape of hole-effect semi-variogram indicating periodic soil variation.

periodicity of geologic and soil processes does not generally operate isotropically. Trangmar (1984) noted perodic behavior of a semi-variogram for sand content calculated perpendicular to the axes of two rivers which were 25 km apart. The periodicity was caused by repetition of deposition patterns away from each river. Barnes (198 1) obtained hole-effect electrical conductivity semi-variograms computed in directions perpendicular to irrigation lines. The electrical conductivity was low adjacent to the irrigation lines, where salts had been leached by the irrigation water, increased to a maximum equidistant between lines, where less water had been applied, and decreased again adjacent to the next line as water application and leaching increased. The natural variability of soil properties is generally nested (Burrough, 1983a) so that evidence of periodic semi-variogram structures at a given scale of observation may often be confounded with structures operating over different scales, the consequence of which is a dampening of the hole effect. d . Nested Variation. We have previously recognized that there may be many sources and scales of variability present in any spatial study of soil properties. In semi-variogram analysis, nested structures can be conveniently represented as the sum of a number of semi-variograms, each with its own range and nugget variance characterizing variability at a particular scale. The size of the nugget variance generally increases with sampling scale, largely due to the variance contributed by shorter-range processes. A single semi-variogram calculated over several discrete, independent, nested processes may consist of several near-linear portions which coincide with the range of the process in question (Fig. 4, curve 3). McBratney et al. (1982) obtained this type of semi-variogram for copper and cobalt in


69

Scotland and found that agricultural effects arising from field-to-field variation and farm-to-farm differences were responsible for short-range variation, while a geologic component caused a longer-range effect. A relatively new concept in dealing with nested variation of natural phenomena is that offractals (Mandelbrot, 1977), which provide a means for assessing the relative balance between processes operating over different spatial scales. Analysis of fractal behavior is similar to that of scaling (Warrick et al., 1977; Simmons et al., 1979), in which variations identified at one scale are statistically equivalent to those seen at other scales and are related simply by a scaling factor. An ideal fractal process operates singly, without interaction with other processes, over a discrete range of spatial scales. Burrough (1983a) applied fractal concepts to soils by using the slopes of the double log plot of semivariograms to obtain a measure of the relative balance between short- and long-range sources of soil variation. An ideal fractal process would have a double log semi-variogram with a slope of 1, but values for most soil properties exceed 1.S,indicating that soils are nonideal fractals (Burrough, 1983a). Structural analysis of semi-variograms using a nested, nonideal fractal approach (Burrough, 1983b) may provide useful information on the scales and relative importance of discrete sources of variation when such variation is caused by superimposed, independently acting soil processes. Such information can, in turn, be incorporated in the design of sampling schemes for properties with nested variation. e. Management Eflects and Soil Genesis. Management practices may considerably alter the inherent spatial structure of soil properties, and subsequently crop growth. Directional application of water, as in furrow irrigation, may impose considerable anisotropy of soil moisture content (Gajem et al., 1981) and subsequent uptake of nutrients such as nitrogen (Tabor et al., 1984). Similarly, nonuniform application of fertilizers may alter the spatial structure of nutrient uptake by crops (Trangmar, 1982). The spatial variability of soils may be altered when land use changes occur, such as conversion of forested land to arable farmland. In this regard, Trangmar (1984) found that topsoil removal by mechanical land clearance combined with subsequent burning of forest trash resulted in considerable short-range variation of topsoil acidity, exchange characteristics, and subsequent crop growth. Analysis of soil variability using semi-variograms has aided identification of soil mapping units (Campbell, 1978) and placement of mapping unit boundaries (McBratney and Webster, 1981b). Spatial dependence of soilforming factors such as rainfall (Yost et al., 1982a), parent material composition (McBratney et al., 1982) and deposition (Trangmar, 1984), and soil genetic processes, such as age and degree of weathering (Yost et al., 1982a; Trangmar, 1984), have also been quantified using semi-variogram analysis. Clearly, analysis of spatial dependence can aid identification of the

70


underlying spatial structure of many soil properties and contribute significantly to an understanding of the spatial effects of soil-forming factors and genetic processes.

VI. INTERPOLATION BY KRlGlNG A. GENERAL Kriging is a technique of making optimal, unbiased estimates of regionalized variables at unsampled locations using the structural properties of the semi-variogram and the initial set of data values. A useful feature of kriging is that an error term (estimation variance) is calculated for each estimated value, providing a measure of the reliability of the interpolation. The simplest forms of kriging involve estimation of point values (punctual kriging) or areas (block kriging) and assume that the sample data are normally distributed and stationary (Henley, 1981). Various other estimation procedures are available when sample data show departures from these assumptions (Fig. 7). Soil properties often exhibit lognormal or complex probability distributions, in which case lognormal or disjunctive kriging is more appropriate. Data showing weak forms of nonstationarity can be detrended or interpolated using universal kriging (Yost et al., 1982b; Hajrasuliha et al., 1980; Webster and Burgess, 1980). The concept of generalized covariances has been used in

Complex

Normal

Disjunctive kriging

Simple kriging

Stationary

7

?

Universal kriging

Generalized ? covariances

Simple

Local trends

drift

Severe anisotropy

Stationarity FIG.7. Kriging methods used under different conditions of stationarity and probability distribution of the data. (From Henley, 1981.)


71

the mining industry for interpolation in the presence of local trends (Matheron, 1973; Delfiner, 1976). Directional differences in variation can also be taken into account during interpolation by using the anisotropic semivariogram model to obtain the weights in the kriging system (Burgess and Webster, 1980a; Tabor et al., 1984). Detailed accounts of kriging theory and application are given in Journel and Huijbregts (1978) and David (1977). Less quantitative descriptions are provided by Clark (1979), Royle (1980), and Henley (1981). Only the main concepts of the various kriging procedures will be summarized here.

B. PUNCTUAL KRIGING Simple point estimation is probably the most common kriging procedure used in soil science to date (Burgess and Webster, 1980a; Vieira et a!., 1981; Van Der Zaag et al., 1981; Van Kuilenburg et al., 1982; Russo and Bresler, 1982; Trangmar, 1982; Trangmar et al., 1982). 1. Concepts of Punctual Kriging

Kriging is a means of local estimation in which each estimate is a weighted average of the observed values in its neighborhood. The interpolated value of regionalized variable z at location xo is n

qx,)

=

1 liZ(Xi) i=l

where n is the number of neighboring samples z(xi)and ;li are weights applied to each z(xi). The weights are chosen so that the estimate 2(x0) of the true value z(xo) is unbiased, i.e.,

E[~(x,)-z(x~)]= 0

(12)

and the estimation variance a: is minimized, i.e., r~," = VAR[S(x,)

-z(xo)]

= minimum

(13)

The weights placed on each neighboring sample sum to 1, and their unique combination for which 02 is minimized can be obtained when n

C

j= 1

Ajy(xi,xj) + p

= y(xi, xo)

for all i

(14)

The values y(xi,x j ) and ?(xi, xo) are the semi-variances, or preferably the covariances (second-order stationarity), between observed locations xi and xj

12


and between the observed location x i and the interpolated location x,, , respectively. These values are obtained from the semi-variogram of Z . The quantity p is the Lagrangian multiplier associated with the minimization of crz . Solution of the n + 1 equations of the kriging system [Eq. (12)] for each li and p enables the kriged estimate .2(xo)to be determined by Eq. (1 1) and the estimation variance to be determined by solving for n

c: =

C AiY(xi7 ~ 0 +) P

i= 1

(15)

The set of n + 1 simultaneous equations of the kriging system is most efficiently solved using matrix methods, as outlined by Burgess and Webster (1980a). The interpolated value at any unsampled location is the most precise possible from the available data and one that can be used with known confidence (McBratney et al., 1982). The estimation variance depends only on the semi-variogram and the configuration of the data locations in relation to the kriged points and not on the observed values themselves (Burgess and Webster, 1980a). Thus, it is a local error term giving more reliable estimation of interpolation precision at a given point than the global error terms of many other interpolation procedures (Giltrap, 1983a). Standard deviations or confidence limits can be calculated from the estimation variance, if necessary. The condition of unbiasedness ensures that kriging gives exact interpolation in that estimated values are identical to observed values when a kriged location coincides with an observed location. In such cases, the weights on neighboring samples are zero and the estimation variance will equal the nugget variance of the semi-variogram. This feature of kriging is particularly desirable when the nugget variance is small and observations have been made with negligible error. The appropriateness of the chosen semi-variogram model and the kriging assumptions of unbiasedness and minimum estimation variance can be tested by “jack-knifing” procedures (Efron and Gong, 1983). This involves deleting each sample in turn, then kriging it independently from all other points in the estimation neighborhood, and finally statistically testing the resulting mean prediction error for unbiasedness and the prediction mean square (variance of predicted minus observed values) for minimum estimation variance (Vieira et al., 1981; Vauclin et al., 1983; Tabor et al., 1984; Trangmar, 1984). Most other interpolation procedures do not have the optimality attributes of unbiasedness and minimum variance that kriging has for spatially dependent variables (Yost et al., 1982b). Only sample locations which are spatially related to the kriged location (i.e., within the range of spatial dependence) are used in kriging. The nearest


73

few samples are the most heavily weighted, with the result that the semivariogram needs to be accurate only over the first few lags. Little is gained by including distant points unless there are few samples close to the kriging location, which may be the case for randomly sampled data. If so, then estimation variances will be relatively large and additional sampling is probably necessary. Sample points occurring in clusters carry less weight than lone points, with the result that addition ofjust one additional sample in sparsely sampled regions can markedly reduce estimation variances in such regions. Sample locations lying between the kriged point and more distant samples screen the distant ones so that the latter have less weight than they otherwise would (Van Kuilenburg et al., 1982). Distances closer than the range may be used to define the maximum radius of the kriging neighborhood when the semi-variogram is well structured with long range and there are sufficient samples at short lags (Vieira et al., 1981). Reducing the kriging radius in such cases reduces computer costs and is an efficient practice if reliable interpolation is still obtained from fewer neighbors. The number of neighbor samples required for reliable estimation depends on the configuration of sampled to kriged locations and the degree of anisotropy and will vary among data sets. Reported numbers range from 7 (Vauclin et al., 1983) to 25 (Burgess and Webster, 1980a) for grid-sampled data and from 10 (Yost et al., 1982b) to 40 (Vieira et al., 1981) for irregularly spaced samples. A conservative estimate would seem to be about 20 to 25 neighbors to achieve reliable interpolation. Anisotropic spatial dependence can be readily taken into account in all forms of kriging simply by using the anisotropic semi-variogram model for estimation of the weights in Eq. (14) (Burgess and Webster, 1980a; Webster and Burgess, 1980; McBratney and Webster, 1983a; Tabor et al., 1984; Trangmar, 1984). Directional differences should be incorporated in the kriging of anisotropic data because it results in more precise interpolation than using the commonly assumed isotropic condition. 2. Application

Most of the early applications of kriging in soil science involved simple point estimation for isoproperty mapping (Burgess and Webster, 1980a; Van Der Zaag et al., 1981; Vieira et al., 1981) and use of estimation variances for designing sampling schemes for future kriging operations (Burgess et al., 1981; McBratney and Webster, 1981a; McBratney et al., 1981; Vieira et al., 1981; Trangmar et al., 1982). Point estimation has also been used in cokriging (McBratney and Webster, 1983a; Vauclin et al., 1983) and universal

74


kriging (Webster and Burgess, 1980; Hajrasuliha et al., 1981; Yost et al., 1982b). Punctual kriging for isoproperty mapping involves estimation of values for a fine grid of points through which isarithms (contours) or three-dimensional surfaces are drawn for spatial display as maps. Point interpolation has been carried out over many different spatial scales ranging from regional variation of soil chemical properties (Van Der Zaag et al., 1981; Yost et al., 1982b) to within-field variation (Burgess and Webster, 1980a) to microvariation of soil hydraulic properties (Vieira et al., 1981) and nutrient uptake by crops within experimental plots (Trangmar, 1982). Van Kuilenburg et al. (1982) compared the estimation precision for punctual kriging of soil moisture content with the variance of mean values for soil mapping units, and with proximal and weighted average interpolation. They concluded that kriging was the most precise estimator but that its efficiency should be balanced with the multipurpose utility of a soil map. This result emphasizes the role of geostatistics in augmenting conventional soil survey methods in describing the spatial variability of specific properties. Estimation variance maps of kriging from randomly sampled data have proved useful for identifying those areas where further sampling would provide the most additional information (Van Der Zaag et al., 1981). Trangmar (1984) was able to improve estimation precision by up to 40 % by collecting additional samples in areas of high estimation variance and then recomputing the semi-variogram and re-kriging the data. Despite the ease of computation and apparent utility of punctual kriging, it has some undesirable attributes. Punctual kriging is an exact interpolator and may produce local discontinuities where interpolated points coincide with sample locations. Isarithmic mapping often results in considerable local detail associated with these discontinuities, and broader regional soil patterns may be obscured (Burgess and Webster, 1980b; McBratney et a[., 1982). The position of this local detail is itself a sampling effect because it depends on the particular places at which samples were collected. Moving the origin or orientation of the sampling pattern could result in a different map of kriged values and estimation variances (Burgess and Webster, 1980b). So, results obtained by punctual kriging depend strongly on the sampling methodology used. The nugget or random variance is a component of the semi-variance at any lag and, although it does not influence the kriged value, it sets a minimum value to the estimation variance of each kriged location. As a result, punctual kriging may produce undesirably large estimation variances if the nugget variance is large. These shortcomings of punctual kriging can be avoided by interpolation over areas, using block kriging (Section VI,C), which results in smoother maps and smaller estimation variances.


75

c. BLOCKKRIGING Users of soils information are often interested in average estimates for discrete areas or blocks (e.g., management units) rather than point estimates obtained by punctual kriging. The “points” at which estimates are made by punctual kriging are actually volumes of soil of the same size and shape as the soil volumes (e.g., cores or pits) which were originally sampled. In many cases, one may wish to interpolate an average value for an area or block which is larger than the cross-sectional area of the soil volume actually sampled. Block kriging provides a method for achieving this and, at the same time, avoids some of the shortcomings of punctual kriging. I . Concepts

In block kriging, a value for an area or block with its centre at xo is estimated rather than values at points as in punctual kriging. The kriged value of property Z for any block V is a weighted average of the observed values xi in the neighborhood of the block, i.e., n

?(V) =

1 liZ(Xi)

i= 1

The only difference in Eq. (16) from the estimation equation (11) for punctual kriging is in the determination of the weighting coefficients. In the weighting procedure, the semi-variances between data points and the interpolated points of punctual kriging are replaced by the average semi-variances between the data points and all points in the block [?(xi, V ) ] . The optimum combination of the weights on sample locations is that for which the estimation variance is minimized. The minimum estimation variance for block V is

where ?(xi,V ) is the average semi-variance between the sample points x i in the neighborhood and those in the block V, y(K V ) is the average semivariance between all points within V (i.e., the within-block variance of classical statistics), and ,uV is the Lagrangian parameter associated with the minimization. The within-block variance includes the nugget variance component plus variance among any samples occurring within each block (Table 111). The estimation variance of block kriging is always less than that of punctual kriging because the within-block variance is removed from the error term [Eq. (17)]. Up to 20-fold improvements in average estimation precision have been achieved using block kriging compared to punctual kriging (Table

76

B. B. TRANGMAR ET AL. Table 111 Variances Associated with Punctual and Block Kriging of Some Soil Properties Property Stone content Statistic

Nugget variance Within-block variance Mean block kriging variance Mean punctual kriging variance Total sample variance sz Referenceb

(%S

Loam thickness (cm)'

Sodium content (medl0 kg)'

Total nitrogen*

10.0

187.0

8.7

0.07

3.2

15.1

369.2

10.5

0.09

3.9

0.6

32.5

0.8

0.04

0.9

12.8

320.0

10.7

0.10

4.0

79.3

786.7

15.3

0.14

4.3

I

1

1

2

2

(%)2

Lime requirement (ton/ha)2

'Variances determined on log transformed values.

* References: 1, Burgess and Webster (1980b); 2, Trangmar (1984). 111). Table I11 also demonstrates that variance for the sample mean obtained in the classical manner will always overestimate the interpolation error achieved by block or punctual kriging for spatially dependent properties.

2. Applications The most common use of block kriging has been for the production of isarithm maps of soil properties (Burgess and Webster, 1980b; McBratney et al., 1982; Trangmar, 1984). Experience indicates that block kriging produces smoother maps than punctual kriging by interpolating average values for blocks, with the effect of smoothing local discontinuities. Such smoothing effects can be seen by comparing the plots in Fig. 8. These effects are particularly desirable when regional patterns of variation are of more interest than local detail. Improvements in the estimation variance of block kriging over that of punctual kriging is most evident for properties with large nugget variances (Table 111). If the block size exceeds the minimum sample spacing, much of this short-range variation is included in the within-block variance term, resulting in small estimation variances.


77

FIG.8. Isarithm maps of cover loam thickness, Hole Farm, England, constructed from (A) punctual kriging and (B) block kriging of 400-m2cells. (From Burgess and Webster, 1980a,b.)

78


These desirable attributes of block kriging relative to punctual kriging make use of the former more prevalent in mining engineering (David, 1977), and this is likely to be the case in future spatial studies of soils. The relatively smooth isolines and maps produced by block kriging may appear similar to those obtained by simpler procedures, such as spline interpolation. However, the similarity of such maps is misleading because the purpose of block kriging is to provide accurate local estimates with known error for variables over discrete blocks of land of predetermined area. Such attributes are rarely embodied in other interpolation techniques (Burgess and Webster, 1980b). Block kriging has also been applied to interpolate spatial effects of crop response to variability imposed by soil management practices. Tabor et al. (1984) found that maps of block-kriged values for nitrate content of cotton petioles indicated a strong response to direction of planting rows and application of irrigation water. Trangmar (1984) used block kriging of growth and yield parameters of upland rice to demonstrate the inverse spatial relationship of soil aluminum saturation on crop growth within an experimental plot recently cleared of forest vegetation. These studies suggest that geostatistical analysis of environmental and crop parameters may be increasingly useful in evaluating the spatial effects of soil properties, pest and disease incidence, and other environmental effects on crop growth and in suggesting cost-effective strategies for amelioration of yield-reducing spatial effects. The adaptation of volume-variance relationships for estimation of ore recovery in mining (David, 1977; Clark, 1979) to the agronomic situation offers potential for interpretation of critical levels of soil nutrients for determination of fertilizer or amendment needs in the presence of spatial heterogeneity of nutrients. Block kriging of a very fine grid of cells forms the basis of interpolation procedures developed by Giltrap (1983a,b), which provide for prior stratification of the landscape into a number of land classes and can either restrict interpolation to cells within the same land class or allow interpolation across land classes using separately calculated autocorrelation functions. Using these procedures, Giltrap (1981) was able to rapidly and cheaply produce maps for many different soil properties at any scale smaller than that of the original interpolation grid. Such an approach has considerable potential to aid quantitative placement of mapping unit boundaries in soil surveys and areal estimation of inclusions within mapping units and to enable rapid generation of soil interpretive maps based on various combinations of fieldmeasured values of soil properties stored in a data base system. 3. Sampling for Punctual and Block Kriging

Much of the punctual and block kriging of soil properties has been done on data originally sampled for other purposes, often in a nongeometric design


79

(Yost et al.,’1982b; McBratney et al., 1982; Van Der Zaag et al., 1981; Van Kuilenburg et al., 1982). Kriging from randomly sampled data usually results in uneven estimation variance maps arising from the irregular sampling density. Overall estimation variance is not minimized and such designs are, therefore, nonoptimal from a kriging standpoint. The pooled value of estimation variances is minimized for any given sample if sampling is carried out on a grid basis, resulting in neighborhoods with the same number of samples. Estimation variances always increase along the margins of the study region, where the neighbors decrease in number irrespective of whether the data is grid or randomly sampled. Burgess et al. (1981) and McBratney and Webster (1983b) found that, given a previously determined semi-variogram, an equilateral triangular grid is the optimal sampling for kriging of isotropic data in terms of minimizing the interpolation error. They also concluded that square grids were almost as precise and were often more convenient. Rectangular grids with their longer intervals aligned in the direction of least variation are optimal where there is simple anisotropy. The anisotropy ratio defines the ratio of sampling intervals in the direction of maximum and minimum variation. Using these approaches, McBratney and Webster (1983b) obtained 3- to 9-fold gains in sampling efficiency over that estimated by classical statistical theory for simple random sampling. McBratney et al. (1981) and McBratney and Webster (198 la) provided sampling theory and a computer program for the design of optimal sampling schemes for kriging and mapping of soil properties given a previously determined semi-variogram and knowledge of any anisotropy . Given the semi-variogram and isotropic variation, Webster and Burgess (1984) showed that for block kriging, sampling of square grids with observations at the centers of a mosaic of cells within each block is the most efficient. They also concluded that bulking of samples from such a grid within blocks will always increase the estimation precision for additive properties compared to using individually measured samples at each cell center. Establishment of such sampling grids depends on knowing the semivariogram for the property of interest, which unfortunately is not generally the case until an initial sampling program has been carried out. McBratney and Webster (1983b) recommend that if the semi-variogram is not known and kriging over the whole region is the objective, then the best strategy is to sample on a regular grid, with the interval determined by the number of observations that can be afforded. One alternative approach is to sample along a series of transects to obtain the semi-variogram and then establish the sample pattern for kriging based on structural analysis of the semi-variogram. Another possible design in which the semi-variogram is unknown is that of Trangmar (1984), in which closely sampled transects were placed in four separate directions across a coarser-sampled square grid. The transects

80

B. B. TRANGMAR ET A L

enabled detection of anisotropy and establishment of the semi-variogram at short lags, while the coarser grid provided the basis for kriging interpolation.

D. CO-KRIGING The spatial distribution of any given property may often be closely related to that of other properties affected by the same regionalized phenomenon or spatial process. Such properties are said to be co-regionalized and are spatially dependent on one another. Co-kriging extends the principle of optimal estimation using regionalized variable theory from that of a single property to situations where there are two or more co-regionalized properties. Co-kriging is most efficiently used where one variable may not have been sampled sufficiently (due to experimental difficulties, high costs, etc.) to provide estimates of acceptable precision. Estimation precision can be improved by utilizing the spatial correlation between the undersampled (primary) variable and other, more frequently sampled covariables. 1. Concepts

The concepts of co-kriging discussed here assume only one covariable, but the equations are readily expanded to include additional covariables (Journel and Huijbregts, 1978). The co-regionalization cross-semi-variogram: i

n

where N ( h ) is the number of pairs of values separated by vector h (David, 1977). Variables z1 and z2 do not necessarily need to have the same number of samples, but the cross-semi-variogram is calculated using only the locations where both variables are measured. The spatial dependence of co-regionalized properties can also be determined using cross-correlograms (McBratney and Webster, 1983a). Unlike auto-semi-variances, cross-semi-variances can be negative if the relationship between z1 and z2 is negative. They may also be anisotropic, particularly if the auto-semi-variances are so (McBratney and Webster, 1983a). The co-kriged value of the undersampled, or primary, variable, 2,, is computed as a weighted average of the observed values of the covariable, z l , and z2 occurring in the estimation neighborhood of each kriged point. The


81

where Ali and AZjare the weights associated with z1 and z,, respectively, while n, and n2 are the number of neighbors of z 1 and z 2 involved in estimating i, at each location xo, respectively. The weights on observed values of z 1 and z , are chosen so that the estimate is unbiased with minimum variance, just as in auto-kriging. However, solution of the co-kriging system for the weights is obtained using the autosemi-variances and the cross-semi-variances of each z1 and z , with the kriged location xo. The resulting system of equations is more complex and more costly to compute than in simple auto-kriging. Solution of the co-kriging system also yields the co-kriging estimation variance for each interpolated location. The equations of the co-kriging system are presented in full by Journel and Huijbregts (1978), McBratney and Webster (1983a), and Vauclin et al. (1983). Co-kriging can be used for point estimation (punctual co-kriging) or block estimation (block co-kriging). In block co-kriging, average auto- and crosssemi-variances of samples within blocks are incorporated into the co-kriging system. The co-kriging system requires at least one sample point of both the primary variable and covariable properties within the estimation neighborhood. If the primary variable and covariable are present at all sampling sites in the neighborhood, then co-kriging yields the same estimate as auto-kriging of the primary variable alone. In such cases, co-kriging is unnecessary.

2. Application

Co-kriging has been applied only to point estimation of soil properties (McBratney and Webster, 1983a; Vauclin et al., 1983; Trangmar, 1984). McBratney and Webster (1983a) identified common anisotropic co-regionalization among particle size fractions and used it to co-krige topsoil silt content from more densely sampled subsoil silt and sand. Co-kriging using one and two covariables, respectively, reduced estimation variance in successively smaller increments relative to auto-kriging. Presumably, the increments in variance reduction gained by using more than one covariable must outweigh the increased complexity of the co-kriging system for practical uses in most situations. In other soils-related applications of co-kriging, Vauclin et al. (1983) interpolated available water and 1/3 bar water content using sand content as

82


the covariable, while Trangmar (1984) co-kriged available phosphorus using more densely sampled observations of total phosphorus. Maps of co-kriged values usually show the same broad pattern of variation as for auto-kriged values, but they tend to be more intricate because the covariables are sampled more densely. Co-kriging improves estimation precision only when there are few neighboring samples of the primary variable and the spatial correlation with the more frequently sampled covariable is taken into account. If several neighboring primary variables are present, they receive most of the weighting and the covariable merely adds another variance component to the estimation without resulting in much or any improvement to the overall estimation precision. The precision of cokriging decreases near the boundaries of the study region as samples of the primary variable and covariable decrease in number. Geometric configurations of the primary variable interspersed with a finer grid of covariable points provides the optimal sampling design for co-kriging (McBratney and Webster, 1983a; Vauclin et al., 1983). McBratney and Webster (1983a) demonstrate sampling configurations for different ratios of sampling intensity of the covariable relative to the primary variable and the subsequent spatial effects on estimation variance (Fig. 9). The position of maximum variance depends on the semi-variogram form and the strength of the cross-correlation, on the sampling interval, and on the sampling intensity ratio of covariable to primary variable. Once the semi-variogram is obtained, the maximum co-kriging variance corresponding to spacings of each variable can be determined by plotting co-kriging estimation variance values against sample spacings for a range of primary variable-to-covariable sampling ratios (Fig. 10). The optimal spacing and primary variable-to-covariable sampling ratio is that which acquires the desired precision at least cost. Anisotropy of the respective properties should be used to determine the relative dimensions of the sampling grids. Block co-kriging would yield smaller estimation variances and, therefore, narrower confidence limits than punctual co-kriging given the same sampling scheme (McBratney and Webster, 1983a). There are numerous situations in agronomic research where properties that are cheap or easy to measure are likely to be co-regionalized with others that are of importance but less easily determined. These might include interpolation of available water capacity from textural components (Vauclin et al., 1983), use of pH for interpolation of exchangeable cations and aluminum, and interpolation of textural components from field estimates and laboratory determinations (Hodgson et al., 1976). Co-kriging of soil moisture content from limited ground observations and the large amounts of data generated by remote sensing systems provides another possible use of coregionalized variables (Price, 1980). Similarly, co-regionalization of soil and

83

APPLICATION OF GEOSTATISTICS 0

0

I

I

.

Q\

0 /

\

/ \

\

I

\

\

.

\

\

/

I

\

\

/

.

/ /

I

0

SR=2

@

X

*

X

.

I

8’

e

\

.

X

i SR=8

x

0

,

’

. . . . * . . . , ./ . ‘. . . x

X

/

\

\

x

/

x

/

/

\

/

@

/

\

x

/

/

x

/

\

I

\

x

@ / .

SR=16

.

x

\

,

0

I

\

Q

.

x \

\

.

@/

x

\ \

‘6

SR=9

.

0 , . O \\ ’

‘

,’\

, x

\

SR=4

I’

Q

\

0

I

Y

.

/

/

/

0

/ /

/

0

.

,0

..... ...... . . . ., . . . .‘. . \ . . . . .‘a X

, /

e

\

\

/

\

/

/’\

x

X

\

/

/

/

0 ’ .

SR= 25

FIG.9. Locations of maximum co-kriging estimation variance for sampling schemes on square grids with different intensities for the covariable. Symbol 0 shows where both primary and covariables are observed; shows where only the covariable is observed. Maximum estimation variance can occur on the dashed lines and near the crosses. SR is the sampling ratio of covariable to primary variable. (From McBratney and Webster, 1983a.)

weather parameters with crop parameters could be used for spatial estimation of crop performance. The cost in time and money of obtaining a larger number of covariable measurements must be considerably less than for the variable of primary interest to make co-kriging viable in a planned experiment. McBratney and Webster (1983a) estimated that measuring a primary variable would need to

84

B. B. TRANCMAR ET AL.

0.10

/

/

/

/'

Sampling interval ( m )

FIG. 10. Graphs of punctual co-kriging variance against primary variable spacing for covariable-to-primary variable sampling ratios from 1 to 25. (From McBratney and Webster, 1983a.)

cost at least five times that of a covariable to make a design for co-kriging economically sound in their particular study. The success of co-kriging also relies on a well-structured cross-semi-variogram, which is not always present even if the auto-semi-variograms have strong structure. Despite these qualifying conditions, co-kriging has the potential to make considerable time and cost savings in mapping properties for which there are cheap, co-regionalized surrogates. Its potential should be further tested.


85

E. UNIVERSAL KRIGING Ordinary kriging is relatively well established, both theoretically and practically, and there are numerous applications in soil science and general agriculture. One of the troublesome assumptions required for ordinary kriging is that the data are stationary or specifically follow the “intrinsic hypothesis,” which requires that the expected value of the difference between any two samples depend on the distance between them but not on their location in the sampled region. Thus, the difference between any two samples the same distance apart should be generally similar throughout the region. This definition, however, does not indicate when a trend is sufficiently strong to require a method other than ordinary kriging. Universal kriging was developed to permit kriging in the presence of strong trends (Matheron, 1971; Dagbert and David, 1976; Delfiner, 1979). 1. Concepts

Universal kriging was designed to permit kriging in the presence of trends in the sample data, i.e.,

where E(Z(x)) is the expected value of the sample data, m(x) is the trend, J are the terms in the polynomial, and aiare the coefficients of the polynomial that describes the trend. The estimation of the trend has been approached several ways. In early work (Olea, 1974; David, 1977), geostatisticians fitted response surface equations to the entire region to remove the trends and then used the residuals to estimate the semi-variograms. This approach can give rise to unusual concave-shaped semi-variograms, which have smaller semivariances at the greater distances, rather than the usual case, where semivariances become larger and approach the sample variance. Estimates of the nugget variance and range from such semi-variograms are much less reliable so that the inference of soil property zones of influence and genetic origin, for example, becomes difficult, if not impossible. There is a problem, however, in determining the correct form of the trend. Because of the typically highly correlated errors, statistical tests of model adequacy are not reliable. There also is the question of whether all trends can be described by polynomial equations. A more recent approach to universal kriging has been to expand the matrices of the kriging equations and simultaneously estimate the weights

86


and remove the trend (be it linear, quadratic, cubic, etc.). With this approach the trend is removed from the neighborhood (the specific points used in the estimation) rather than from the entire region, as in the above case. This method also allows more flexibility because only the form of the trend is specified for each estimation neighborhood. This means that the order of the regression is determined, but the coefficients are independently determined for each neighborhood. Thus the estimation proceeds as a “moving neighborhood.” This approach is therefore much more flexible and should remove more variation than the single-equation approach of trend analysis. However, the questions remain, Was the trend correctly removed? Was it necessary to remove the trend in the first place? Further work has pointed out that trend removal is not always straightforward. Armstrong (1984) suggests that both of the above approaches suffer indeterminacy in both the drift and the semi-variogram, which precludes their accurate estimation unless additional constraints or data are determined. She suggests this indeterminacy is similar to that associated with the constant of integration. With both of the above approaches to universal kriging, one can increase the degree of the response surface equation from linear to quadratic and higher and it is not clear just where to stop. The problem is analogous to that of determining the appropriate degree for regression equations: how to determine when one is fitting the actual trend and when one is fitting local variation. 2. Application

It has been suggested that one can determine nonstationarity, and hence the need for universal kriging, from the appearance of the semi-variogram (Journel and Huijbregts, 1978; David, 1977). Specific evidence of nonstationarity is said to be displayed when the semi-variogram increases concave upward and does not level out to approach the population variance at large distance. The usual suggestion has been that if the semi-variance approaches but does not drastically exceed the overall variance at large distances, chances are that the data are stationary. Studies by Starks and Fang (1982), however, suggest that there may be more subtle distortion of the semivariogram caused by nonstationarity. They suggest that nonstationarity can result in bias which is not manifested solely by concave-upward-shaped semivariograms. In their simulation studies, both nugget variance and range were affected as well. They suggested the following to check for nonstationarity: 1. Visually inspect the data for large trends; 2. Compare kriged estimates with the actual data, i.e., “jack-knife’’ and compare the variance of actual kriged values with the estimation variances.


87

These authors point out that ordinary kriging is relatively robust to nonstationarity because (1) the ordinary kriged estimate is a good approximation for linear drift with small slope and (2) drift tends to increase the semi-variance, which causes a more conservative calculated estimation variance. Probably another factor is that nonstationarity seems to be expressed more at larger distances than at smaller ones. Ordinary kriging places particularly high emphasis on the closer points so that correct estimation of the semi-variance is most important at small distances, where nonstationarity probably is less frequent. The need to apply universal kriging or unbiased kriging of the kth order in agriculture has been difficult to determine, and more experience is needed. In one of the first applications of universal kriging to soil science, Webster and Burgess (1980) concluded that universal kriging appeared to be neither universally applicable nor of particular benefit. Some of our preliminary studies (Yost et al., 1982b) with large areas and clear trends in the data suggested that universal kriging resulted in very little improvement over ordinary kriging when comparing the observed and estimated points. Our results suggested that ordinary kriging is quite robust to the presence of even strong trends in the data. This comparison was obtained by “jack-knifing,” in which an observed data point was removed from consideration and then estimated from the surrounding data. The observed and predicted data were then compared. Such comparisons provide a good method of determining the accuracy and precision of alternative estimation procedures and are widely used in geostatistics. Faith and Sheshinski (1979) also attempted to determine the effect of ignoring trends in the data. Their results also indicate that ordinary kriging is apparently quite robust to the presence of drift. In both of the above cases the nonstationarity occurred at large distances, not at small distances. David (1977) pointed out that there can also be nonstationarity at small distances. Clearly there are many questions remaining about how and when to apply universal kriging to soils and soil phenomena. In general, however, universal kriging remains a methodology of geostatisticians, and its use requires expert assistance. Universal kriging is also known as unbiased kriging of the kth order (Journel and Huijbregts, 1978). This terminology derives from the concept of generalized covariances. The generalized covariances are presented in Matheron (1 973) and provide yet another approach to handling nonstationarity. The trends are removed from the data by successive differencing operations similar to the removal of nonstationarity in time series by successive differencing (Box and Jenkins, 1976). The differencing operation consists of subtracting Z(hi) from Z ( ~ Z , for + ~ all ) i. The procedure can be repeated as necessary to result in differences AZ which are stationary. The extent to

88


which the differencing also removes evidence of spatial dependence remains unknown. Matheron developed the procedure to ensure that the estimation variances are positive (Delfiner, 1976). Procedures and techniques have been described by Stark's et al. (1980) for use in coal mine contouring in Iowa. This method of handling trends in the data apparently serves well in mapping and contouring; however, the semivariogram is lost and with it the rich interpretations of the underlying soil-, geologic-, and soil genetic-related phenomena. Some of the earliest mapping and contouring procedures used by mining consultants (BLUEPACK and KRIGEPACK) used the generalized covariance approach. Evidence that the generalized covariance approach provides significantly more accurate contouring is scarce.

F. KRIGING FROM NON-NORMALLY DISTRIBUTED DATA Deviations from the assumptions of normality in regionalized variable theory has been a criticism of kriging (Henley, 1981). Transformation to normality prior to geostatistical analysis results in a nonlinear function of the original data, so that kriging estimates may not be made with minimum estimation variance and without bias. The serious of this violation increases as deviations from normality increase. Lognormal kriging of points or blocks has been developed for stationary, lognormally distributed data and disjunctive kriging for stationary variables of more complex probability distributions (Fig. 7). Kriging of nonstationary variables of complex distribution is still under theoretical development. I . Lognormal Kriging

Kriging of lognormally distributed data has been widely used in the mining industry (Rendu, 1979; Parker et al., 1979;Journel, 1980) and is beginning to receive more attention in soil science (Van Der Zaag et al., 1981; Yost et al., 1982a,b; Trangmar, 1984). It simply involves computation of semi-variograms and kriging on natural log-transformed values of the original data using the same procedures as for simple linear kriging. The log-kriged values and estimation variances can be re-expressed in terms of the original data using procedures outlined by Journel and Huijbregts (1978; Eqs. VIII-14, VIII-15). These authors concluded that lognormal kriging methods yield smaller estimation variances than simple linear kriging of the untransformed data if the original data is clearly lognormally distributed. This approach has been found to be a practical one (Rendu, 1979; Yost et al., 1982b), yielding reasonable estimates despite the minor violation of the kriging assumptions.


89

2.Disjunctive Kriging One of the innovative aspects of ordinary kriging was that it placed the data before the statistical model, therefore requiring a new model. Ordinary kriging also did not depend on the statistical distribution of the data. Disjunctive kriging is a specialized form of kriging which differs from ordinary kriging by requiring consideration of the statistical distribution of the data. If the data are normally distributed, however, one can obtain additional probabilistic information, i.e., one can determine proportions and state confidence limits about the prediction which are more difficult with non-normally distributed data. Transforming the data so it is normally distributed permits drawing on the large body of statistical inference. However, to obtain the normally distributed data required for disjunctive kriging one has to determine the necessary transformation to convert the data into bivariately normally distributed data. This process is achieved by application of Hermite polynomials to the original data. A serious disadvantage of disjunctive kriging is the transformation step in which the form of the transformation must be determined so that the resultant transformed data are normally distributed. Typically a 12th-order polynomial is used for the transformation. The polynomial must be estimated and then incorporated into the standard kriging equations. This results in considerably increased computation time. Similar to ordinary kriging, disjunctive kriging assumes stationarity. A recent study by Journel(l983) suggests another alternative to disjunctive kriging. He compared a multivariate Gaussian approach, which may be more straightforward to apply. As suggested above, by transforming the data into multivariate normal distribution, more powerful inferential statistics can be applied, leading to estimates of quantities, proportions, and grade-tonnage types of calculations. Currently these methods are too new and untested for general application to our needs. In our view, this area of research in geostatistical methods holds much promise for producing methods which will bring to fruition the application of geostatistics to agricultural problems.

VII. PERSPECTIVES: FUTURE USE OF GEOSTATISTICS IN SOIL RESEARCH The main contribution of geostatistics to soil research lies in structural analysis of soil variation and its use for local estimation. The semi-variogram provides a quantitative tool for relating the inherent structure of variation in specific properties to spatial effects of soil-forming factors and processes,

90


including management effects. The development of procedures for quantifying anisotropy, trends, and periodic phenomena gives geostatistics a versatility for dealing with natural phenomena that few other interpolation methods have. Procedures for quantifying nested variation (Burrough, 1983a,b) and computer production of maps at a range of spatial scales from a finely kriged grid (Giltrap, 1983a) represent new approaches to dealing with scale-related effects of soil processes and scientists’ perception of them. The initial emphasis in geostatistical interpolation has been on punctual kriging, due to its ease of computation. The relative benefits accruing from block kriging, such as smoother maps, smaller estimation variances, and easier use for design of sampling schemes, are now generally acknowledged and are likely to result in increased use of areal interpolation in the future. In designing spatial studies for estimation and mapping of properties for which there are cheap surrogates, the co-regionalization of properties and use of cokriging should be considered as a potential cost-saver in making field and laboratory measurements without loss in mapping precision. Geostatistics could be used in soil survey operations for structural analysis of soil variation to aid understanding of soil genesis and for analysis of reconnaissance data for defining future sampling populations and configurations both within and among different terrain units. The cost-effectiveness of geostatistics-based sampling schemes in practical soil survey operations needs to be field tested in different types of terrain for comparison with traditional sampling techniques. Kriging can augment general-purpose information contained in conventional soil maps by interpolation of interpretive data and specific measured or derived properties, which may vary independently of mapping unit boundaries. The ease of data manipulation, speed of computation, and precision of computer-generated maps based on kriging of soil properties make geostatistical techniques particularly desirable in the face of user demand for quick and reliable soil survey results (Giltrap, 1983a). The versatility and range of geostatistical software now available make spatial analysis of natural phenomena applicable to many areas of agronomic research. Block kriging appears to be particularly useful for estimating soil amendment requirements over areas the size of land management units. Adaptation of volume-variance relationships for estimation of ore recovery in mining (David, 1977; Clark, 1979) to the agronomic situation offers the potential for spatial interpretation of critical levels of soil constraints to crop production. Such an approach might be applied to using within-field variation of properties such as soil moisture content for improving the efficiency of irrigation water use, nutrient levels for fertilizer application, or soil chemical properties for amendment needs, such as liming. Analysis of the spatial response of crop growth to the variability of soil properties, such as nutrient uptake in response to variation of soil nutrient


91

parameters (Trangmar, 1982), available moisture (Tabor et al., 1984), or root penetration, may further contribute to the agronomists’ understanding of the role of spatial effects in soil-crop relations. Geostatistical analysis of the incidence of pest and disease attacks in crops might aid identification of spatial sources of such attacks. Identification of a spatially dependent component of “random” error may help further reduce the confounding effects of within-plot variability on treatment effects in agricultural experimentation. The use of spatial dependence in identifying optimal plot size and spacing of samples within plots has already been described by Vieira et al. (198 1). Geostatistical analysis of spatial variation in natural phenomena has a wide range of potential applications in soil and agronomic research. In applying geostatistics, it should be remembered that semi-variograms and kriging are tools constrained by their assumptions and, where these assumptions break down, other methods of spatial analysis may be more appropriate. REFERENCES Adams, J. A., and Wilde, R. H. 1976a. N . 2. J . Agric. Res. 19, 165-176. Adams, J. A., and Wilde, R. H. 1976b. N . Z . J . Agric. Res. 19, 435-442. Armstrong, M. 1984. J . Math. Geol. 16, 101-108. Babaloa, 0. 1978. Soil Sci. 126, 269-279. Ball, D. F., and Williams, W. M. 1968. J . Soil. Sci. 19, 435-442. Barnes, E. 1981. Ph.D. dissertation, Univ. of Hawaii, Honolulu. Bascomb, C. L., and Jarvis, M. G. 1976. J . Soil Sci. 27, 420-437. Beckett, P. H. T., and Burrough, P. A. 1971. Soil Sci. 22,466-489. Beckett, P. H. T., and Webster, R. 1971. Soils Fert. 34, 1-15. Biggar, J. W., and Nielsen, D. R. 1976. Water Resour. Rex 12, 78-84. Blais, R. A., and Carlier, P. A. 1968. Can. Inst. Min. Metall. 9 (Special Vol.). Blevins, R. L., Holowaychuk, N., and Wilding, L. P. 1970. Soil Sci. SOC.Am. Proc. 34, 315-331. Bouma, J . 1983. In “Pedogenesis and Soil Taxonomy. I. Concepts and Interactions” (L. P. Wilding, N. E. Smeck, and G. F. Hall, eds.), pp. 253-281. Elsevier, Amsterdam. Box, G. E. P., and Jenkins, G. M. 1976. “Time Series Analysis, Forecasting and Control.” Holden-Day, San Francisco. Bresler, E., Dagan, G., Wagenet, R. J., and Laufer, A. 1984. Soil Sci. SOC.Am. J . 48, 16-25. Buol, S. W., Hole, F. D., and McCracken, R. J. 1980. “Soil Genesis and Soil Classification,” 2nd Ed. Iowa State Univ. Press, Ames. Burgess, T. M., and Webster, R. 1980a. J . Soil Sci. 31, 315-331. Burgess, T. M., and Webster, R. 1980b. J . Soil Sci. 31, 333-341. Burgess, T. M.. Webster, R., and McBratney, A. 1981. J . Soil Sci. 32, 643-659. Burrough, P. A. 1983a. J . Soil Sci. 34, 577-597. Burrough, P. A. 1983b. J . Soil Sci. 34, 599-620. Burrough, P. A,, Beckett, P. H. T., and Jarvis, H. G. 1971. J . Soil Sci. 22,368-381. Butler, B. E. 1959. CSIRO Soil Publ. (14). Campbell, J. B. 1978. Soil Sci. Soc. Am. J . 42, 460-464.

92


Cassel, D. K., and Bauer, A. 1975. Soil Sci. Soc. Am. Proc. 39, 247-250. Clark, I. 1979. “Practical Geostatistics.” Applied Science Publ., London. Cutler, E. J. B. 1977.“Soil Resource Surveys, Interpretations and Applications.” Lincoln College Press, Christchurch, New Zealand. Dagbert, M., and David, M. 1976. Can. Inst. Mining Bull. Feb. David, M. 1977. “Geostatistical Ore Reserve Estimation.” Elsevier, Amsterdam. Delfiner, P. 1976. In “Advanced Geostatistics in the Mining Industry” (M. Guarascio, ed.), pp. 49-68. Reidel, Dordrecht. Delfiner, P. 1979. Bull. Centre Geostat. Morphol. Math. France (C-77). Delfiner, P., and Delhomme, J. P. 1973. I n “Display and Analysis of Spatial Data” (J. C. Davis and M. J. McCullough, eds.), pp. 96-114. Wiley, New York. Delhomme, J. P. 1978. Adu. Water Resour. 1, 251-266. Delhomme, J. P. 1979. Water Resour. Res. 15, 269-280. Dent, D., and Young, A. 1981. “Soil Survey and Land Evaluation.” Allen & Unwin, London. Efron, B., and Gong, G. 1983. Am. Stat. 37, 36-48. Food and Agricultural Organisation (FAO) 1974. “Soil Map of the World. Vol. I. Legend.” UNESCO, Paris. Faith, R., and Sheshinski, R. 1979. Dept. Stat., Stanford Uniu. Tech. Rep. (28). Gajem, Y. M., Warrick, A. W., and Myers, D. E. 1981. Soil Sci. SOC.Am. J . 45, 709-715. Giltrap, D. J. 1981. I n “Information Systems for Soil and Related Data” (A. W. Moore, B. G. Cook, and L. G. Lynch, eds.), pp. 75-82. PUDOC, Wageningen. Giltrap, D. J. 1983a. Geoderma 29, 295-311. Giltrap, D. J. 1983b. Geoderma 29, 313-325. Greville, T. N. E. 1969. “Theory and Applications of Spline Functions.” Academic Press, New York. Hajrasuliha, S. W., Baniabassi, N., Metthey, J., and Nielsen, D. R. 1980. Irrig. Sci. 1, 197-208. Hammond, L. C., Pritchett, W. L., and Chew, V. 1958. Soil Sci. Soc. Am. Proc. 22, 548-552. Henley, S. 1981. “Nonparametric Geostatistics.” Applied Science Publ., London. Hodgson, J. M., Hollis, J. M., Jones, R. A,, and Palmer, R. C. 1976. J . Soil Sci. 27, 411-419. Huijbregts, C. H. 1975. In “Display and Analysis of Spatial Data” (J. C. Davis and M. J. McCullough, eds.), pp. 38-53. Wiley, New York. Huijbregts, C. H., and Matheron, G. 1971. Can. Inst. Min. Metall. 12 (Special Vol.). Jackson, M., and Marechal, A. 1979. Proc. APCOM Symp., 16th. Jacob, W. C., and Klute, A. 1956. Soil Sci. SOC. Am. Proc. 20, 170-172. Journel, A. G . 1980. J . Math. Geol. 12,285-303. Journel, A. G. 1983. J . Math. Geol. 15,445-468. Journel, A. G., and Huijbregts, C. H. 1978. “Mining Geostatistics.” Academic Press, New York. Krige, D. G. 1951. J . Chem. Metall. Min. SOC.South Afr. 52, 119-139. Krige, D. G. 1960. J . South Afr. Inst. Min. Metall. 61, 231-233. Luxmoore, R. J., Spalding, B. P., and Munro, I. M. 1980. Soil Sci. SOC.Am. J . 45, 687-691. McBratney, A. B., and Webster, R. 1981a. Comput. Geosci. 7, 335-365. McBratney, A. B., and Webster, R. 1981b. Geoderma 26, 63-82. McBratney, A. B., and Webster, R. 1983a. J . Soil Sci. 34, 137-162. McBratney, A. B., and Webster, R. 1983b. Soil Sci. 135, 177-183. McBratney, A. B., Webster, R., and Burgess, T. M. 1981. Comput. Geosci. 7 , 331-334. McBratney, A. B., Webster, R., McLaren, R. G., and Spiers, R. B. 1982. Agronornie 2, 969-982. McCormack, D. E., and Wilding, L. P. 1969. Soil Sci. SOC. Am. Proc. 33, 587-593. McIntyre, G. A. 1967. J . Aust. Inst. Agric. Sci. 33, 308-320. Mandelbrot, B. B. 1977. “Fractals, Form, Chance and Dimension.” Freeman, London.

APPLICATION O F GEOSTATISTICS

93

Mapping Systems Working Group 1981. Land Resource Institute, Contribution No. 142. Agricultural Canada, Ottawa. Matheron, G. 1963. Econ. Geol. 51, 1246-1266. Matheron, G. 1965. “Les variables regionalisees et leur estimation. Une application de la theorie des functions aleatoires aux sciences de la nature.” Masson, Paris. Matheron, G. 1969. Cab. Cent. Morphol. Math. Fontainebleau 1. Matheron, G. 1970. I n “Geostatistics” (D. F. Merriam, ed.). Plenum, New York. Matheron, G . 1971. Cab. Cent. Morphol. Math. Fontainebleau 5. Matheron, G. 1973. Adv. Appl. Prob. 5,439-468. Matheron, G. 1976. I n “Advanced Geostatistics in the Mining Industry” (M. Guarascio, ed.), pp. 23 1-236. Reidel, Dordrecht. Miller, F. P., Holowaychuk, N., and Wilding, L. P. 1971. Soil Sci. Soc. Am. Proc. 35, 324-331. Murphy, C.P., and Banfield, C. F. 1978. J . Soil Sci. 29, 156-166. Nielsen, D. R., Biggar, J. W., and Erh, K.T. 1973. Hilgardia 42, 215-260. Norris, J. M. 1971. J . Soil Sci. 22, 69-80. Nortcliff, S. 1978. J . Soil Sci. 29, 403-417. Olea, R. A. 1974. J . Geophys. Rex 79,695-702, Olea, R. A. 1975. “Optimum Mapping Techniques Using Regionalized Variable Theory.” Kansas Geol. Survey, Lawrence, Ser. Spatial Anal. (3). Parker, H. M., Journel, A. G., and Dixon, W. C. 1979. Proc. APCOM Symp., 16th. Price, S . C. 1980. Water Resour. Res. 16, 787-795. Protz, R., Presant, E. W., and Arnold, R. W. 1968. Can. J . Soil Sci. 48, 7-19. Rendu, J. M. 1979. Proc. APCOM Symp., 16th. Royle, A. G. 1980. “Geostatistics.” McGraw-Hill, New York. Russo, D., and Bresler, E. 1981. Soil Sci. Soc. Am. J . 45, 682-687. Russo, D., and Bresler, E. 1982. Soil Sci. Soc. Am. J . 46, 20-26. Sawhney, B. L. 1977. I n “Minerals in Soil Environments” (J. B. Dixon and S. B. Weed, eds.), pp. 405-434. Soil Sci. SOC.Am., Madison, Wisconsin. Schafer, W. M. 1979. Soil Sci. Soc. Am. J . 43, 1207-1212. Silva, J. A. 1984. I n “A Multi-Disciplinary Approach to Agrotechnology Transfer” (G. Uehara, ed.), pp. 17-28. Univ. of Hawaii, Honolulu. Simmons, C. S., Nielsen, D. R., and Biggar, J. W. 1979. Hilgardia 47, 77-174. Sisson, J. B., and Wierenga, P. J. 1981. Soil Sci. Soc. Am. J . 45, 699-704. Smeck, N. E., and Wilding, L. P. 1980. Geoderma 24, 1-16. Soil Survey Staff, 1951. “Soil Survey Manual.” US. Dept. Agric. Handbook (18). Sokal, R. R., and Rohlf, F. J. 1969. “Biometry.” Freeman, San Francisco. Starks, T. H., and Fang, J. 1982. J . Math. Geol. 14, 309-319. Starks, T. H., Fang, J., and Chiu, C. 1980. Geostat. Report No. 1 . Coal Extraction and Utilization Research Center, Southern Illinois University, Carbondale. Tabor, J. A,, Warrick, A. W., Pennington, D. A,, and Myers, D. E. 1984. Soil Sci. Soc. Am. J . 48, 602- 607. Taylor, N. H., and Pohlen, I. J. 1962. Soil Survey Method. N. 2. Soil Bur. Bull. (25). Trangmar, B. B. 1982. Benchmark Soils News 6,4-5. Trangmar, B. B. 1984. Ph.D. dissertation, Univ. of Hawaii, Honolulu. Trangmar, B. B., Singh, U., Yost, R. S., and Uehara, G. 1982. Proc. Int. Soil Class. Workshop, 5th, Wad Medani, Sudan. Trangmar, B. B., Yost, R.S., Sudjadi, M., Soekardi, M., and Uehara, G. 1984. HITAHR Res. Ser. 26. Van Der Zaag, P., Fox, R. L., Yost, R. S., Trangmar, B. B., Hayashi, K., and Uehara, G. 1981. Proc. Int. Soil Class Workshop, 4th, Kigali, Rwanda.

94


Van Kuilenburg, J., De Gruijter, J. J., Marsman, B. A., and Bouma, J. 1982. Geoderma 27, 31 1-325. Van Wambeke, A,, and Dudal, R. 1978. I n “Diversity of Soils in the Tropics.” Am. SOC.Agron. Spec. Publ. 34, 13-28. Vauclin, M., Vieira, S. R., Vauchaud, G., and Neilsen, D. R. 1983. Soil Sci. SOC. Am. J . 47, 175-184. Vieira, S. R., Nielsen, D. R., and Biggar, J. W. 1981. Soil Sci. SOC.Am. J . 45, 1040-1048. Wagenet, R. J., and Jurinak, J. J. 1978. Soil Sci. 126, 342-349. Warrick, A. W., Mullen, G . J., and Nielsen, D. R. 1977. Soil Sci. Soc. Am. J . 41, 14-19. Watson, G. S. 1972. Geol. SOC.Am. Spec. Pap. 46, 39-46. Webster, R. 1973. Math. Geol. 5, 27-37. Webster, R. 1977. “Quantitative and Numerical Methods in Soil Classification and Soil Survey.” Clarendon, Oxford. Webster, R. 1978. J . Soil Sci. 29, 388-402. Webster, R., and Burgess, T. M. 1980. J . Soil Sci. 31, 505-524. Webster, R., and Burgess, T. M. 1984. J . Soil Sci. 35, 127-140. Webster, R., and Butler, E. 1976. Aust. J . Soil Res. 14, 1-24. Webster, R., and Cuanalo, H. E., de la C. 1975. J . Soil Sci. 26, 176-194. Whitten, E. H. T. 1975. I n “Display and Analysis of Spatial Data” (J. C. Davis and M. J. McCullough, eds.), pp. 282-297. Wiley, New York. Wilding, L. P., and Drees, L. R. 1978. I n “Diversity of Soils in the Tropics.” Am. Soc. Agron. Spec. Pub/. 34, 1-12. Wilding, L. P., and Drees, L. R. 1983. I n “Pedogenesis and Soil Taxonomy. I. Concepts and Interactions” (L. P. Wilding, N. E. Smeck, and G. F. Hall, eds.), pp. 83-116. Elsevier, Amsterdam. Wilding, L. P., Jones, R. B., and Schafer, G. M. 1965. Soil Sci. Soc. Am. Proc. 29, 711-717. Yost, R. S., and Fox, R. L. 1981. Soil Sci. Soc. Am. J . 45, 373-377. Yost, R. S., and Fox, R. L. 1983. Geoderma 29, 13-26. Yost, R. S., Fox, R. L., and Uehara, G . 1982a. Soil Sci. Soc. Am. J . 46, 1028-1032. Yost, R. S., Fox, R. L., and Uehara, G. 1982b. Soil Sci. Soc. Am. J . 46, 1033-1037. Zubrow, E. B. W., and Harbaugh, J. W. 1979. I n “Simulation Studies in Archaeology” (I. Hodder, ed.), pp. 109-122. Cambridge Univ. Press, London and New York.


THE INFLUENCE OF SOIL STRUCTURE O N WATER MOVEMENT, CROP ROOT GROWTH, A N D WATER UPTAKE Ann P. Hamblinl Western Australian Department of Agriculture South Perth. Western Australia, Australia

....... .......... .. .. A. Total Porosity; Capacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Pore-Size Distribution and the Moisture Characteristic . . . . . . . . . . . . C. Pore Continuity and Hydraulic Conductivity . . . , . . . . . . . . , . . . . . Stability of the Pore System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . :. A. The Concept of Stability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Clay and Aggregate Bonding in Agricultural Soils. . . . . . . . . . . . . . . . C. Organic Matter Bonding in Agricultural Topsoils. . . . . . . . . . . . . . . . Water Flow in Agricultural Soils . . , . . . . . . . . . . . . . . . . . . . . . . . . . A. Infiltration.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Redistribution within the Root Zone . . . . . . . . . . . . . . . . . . . . . . . Patterns of Root Growth. . . . , , . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Genotypic Variation . . . . , . . . . . . . . . . , . . . , . . . . . . . . . . . . . B. Environmental Influences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Water Upake by Roots. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Speculation: Are We Measuring and Averaging at Consistent Scales?. . . . . . .

I. Introduction . .

......

........

. .

.

. .

11. Soil Structure: Components of the Soil-Pore System. . . . . . . . . . . . . . .

111.

IV.

V.

VI. VII. VIII. Summary References

95 96 97 97 102 107 107 108 113 114 114 122 127 129 132 144 149 151 152

I . INTRODUCTION In recent years soil physics has been concerned with extending quantitative predictions of soil-water movement from defined, uniform conditions to the greater complexity of the “real world,” where heterogeneity of soil parameters occurs at many space and time scales. Concurrently, in plant physiology Present address: CSIRO Dryland Crops and Soils Research Program, Private Bag, Wembley P.O., Western Australia 6014, Australia. 95

Copyright CI 1985 by Academic Press, Inc. All rights of reproduction in any form reserved.

96

A N N P. HAMBLIN

efforts have increased to locate and quantify resistances to water flow in the soil-plant system, as has been reviewed by Taylor and Klepper (1978). The aim of much of this work has been to model water transport quantitatively to provide accurate solutions to water-use problems in agriculture and hydrology. The aim of this paper is to link these two topics by focusing on the role of the soil structure (the soil-pore system) through which both water and roots move. The principle reason for concentrating on soil structure is that, of the soil’s intrinsic physical properties, it is the one most easily, frequently, and widely altered, particularly by cultivation. Greater understanding of the role of soil structure, with both its inherent and induced variations, should improve our ability to manipulate deliberately the soil environment for more effective crop production and water management. Although the scope of this article is large, space considerations alone must make its treatment selective. The environments which have received most attention are temperate to subtropical, in the context of rain-fed arable agriculture.

II. SOIL STRUCTURE: COMPONENTS OF THE SOIL-PORE SYSTEM Almost any paper or book on soil structure written over the past 40 years commences with a reverential acknowledgment of the subject’s importance to soil physical conditions for crop growth. In the next breath, however, many of these works will confess the singularly intractable nature of the problem of characterizing those aspects of soil structure most relevant to plants. As in many branches of science, advances in understanding have frequently had to wait upon techniques for measurement and observation. In the case of soil structure, advances in colloid science and sedimentology led to more knowledge about the arrangement (and composition) of the solid soil particles at an earlier date than knowledge about the pores within and between them. Yet, as early as 1911, Green and Ampt, whose work on “the flow of air and water through soils” still provides the basis for many studies on water movement, commented that “the relations of the soil to the movements of air and water through it.. .are much less obscure if we direct our attention to the number and dimensions of the spaces between the particles rather than to the sizes of the particles themselves.” In recent years studies of soil structure have come around to their viewpoint and have concentrated on the soil-pore system. However, the true complexity of spatial variation and surface reactivity of soil structure is still seldom adequately quantified. We

THE INFLUENCE OF SOIL STRUCTURE

97

can understand this if we see how the pore system relates first to static, then dynamic, water behavior. A, TOTALPOROSITY; CAPACITY

Although it is most logical to describe the pore space as the void ratio e (volume of voids to volume of solids, V&), the porosity n (the volume of voids to bulk volume of soil, v,lV,)continues to be the most commonly used parameter in the agricultural literature. Since porosity is equal to 1.0 - pb/ pp(pb is the bulk density and pp the particle density) and since e = pp/pb - 1.0, the conversion can always be made by the reader. In the case of expanding clay soils, where pore volume and bulk volume change substantially with water content, this is advisable. Alternatively, for swelling soils, the soil weight, over a given depth and at the most swollen state, is used as the basis for comparative values in field situations. Reduction in e occurs with soil shrinkage as water is removed. Increased overburden pressure (loading such as occurs when the upper part of a dry profile re-wets) and compaction by traffic and cultivation implements are also responsible for reductions in pore volume. Increases in surface soil e values occur in freshly tilled seedbeds, during swelling of clay and peat soils (with vertical displacements of 0.1 to 0.2 m in the top meter), and through frost heave (with displacements of 0.1 to 0.5 m per meter of vertical movement). Freshly tilled topsoils may have e values of N 1.25, and soils high in organic matter with peat or litter layers may reach e values of 1.4. Subsoils generally have e values of 0.45 to 0.8, while cemented or indurated layers drop to e values of less than 0.25.

B. PORE-SIZE DISTRIBUTION AND THE MOISTURE CHARACTERISTIC Pores in soils range from 0.003-pm plate separations in clay particles to pipes, cracks, or channels tens of centimeters in diameter. Pore-size distribution is generally computed from the relationship between soil water content (0) and matric potential ($) by means of the equation: p g h = 2 y cos a/r

(1)

where p g h is the suction, equivalent to $ ( p is the density of water, g is gravity, and h is the hydrostatic suction); y and cos c1 are the surface tension and contact angle of water, respectively; and r is the equivalent cylindrical radius. Thus for large pores or pipes in which the fluid-air interface is not controlled by capillary (hydrophilic) forces, the moisture characteristic $(0) (Childs, 1940) cannot be used to infer pore size. The limiting diameter of a

98

ANN P. HAMBLIN

capillary pore in a normal temperature and pressure system which will support a meniscus is 3 x lo3 pm (3mm), equivalent to a potential of 0.1 kPa. Pore-size distributions determined from the draining function +(O) may contain several sources of error. In laboratory measurements sample size is often less than the minimum structural unit, particularly in the z dimension, where the depth is generally less than 0.03 m to reduce equilibration time. This biases the pore-size distribution (PSD) toward smaller pore classes. In fine-textured soils shrinkage of the sample upon desorption alters both the PSD and the pore shape, so that PSDs thus obtained are not representative of the proportion of the fine pores found in the fully hydrated field state. Lawrence et al. (1979) found a 10% collapse of “domain”-sized pores ( < 0.01 pm) (see Section III,B, 1) after water desorption, compared with values obtained by critical-point drying (Greene-Kelly, 1973) and mercury porosimetry in small clay aggregates. Although this alteration is significant in studies of flow in saturated subsoils, within the context of soil-water uptake by plant roots, we consider conventional water-desorption techniques to be more appropriate. Finally, it must be remembered that Eq. (1) assumes a regular, cylindrical pore shape, whereas the majority of soil pores have an Table I Description and Dimensions of Soil Pores

Reference Johnson et al. (1960)

Brewer (1964)

Greenland (1977)

ECD“ (w)

Capillary potential (kP4

< 75 75 to 1000 1000 to 2000 2000 to 5000 > 5000 -0.06 < -3000 -4 < - 6 x lo7 < - 6 x lo3 -6 x lo3 to -60 -60 to -0.6 > -0.6 < -30 -30 to -0.3

> -0.3 ~~

a

< -4

ECD, Equivalent cylindrical diameter.

Pore name Micropore Very fine Fine Medium Coarse Cryptovoid Ultramicrovoid Microvoid Mesovoid Macrovoid Bonding pore Residual pore Storage pore Transmission pore Fissure Pressure gradient pore Gravitational pore Channel-flow pore

99

THE INFLUENCE O F SOIL STRUCTURE Table I1 Pore Dimensions of Biological Origin or Significance Average pore diameter (pm) 2000-50,000 1500-8000 500-3500 2000-1 1,000 6000 300- 10,000 500- 10,000 100-1000 50- 100 20-50 5-10 1000 30 0.5-2 0.2-2 0.1

Biological significance Ant nests and channels Wormholes

Tap roots of dicotyledons Nodal roots of cereals Seminal roots of cereals Lateral roots of cereals 1st- and 2nd-order laterals Root hairs Root plus root hair cylinder in clover “Field capacity” (- 10 kPa) Fungal hyphae Bacteria Permanent wilting point ( - 1500 kPa)

Reference Green and Askew (1965) Barnes and Ellis (1979) Barley (1959) Ehlers (1975) Bouma et a/. (1982) Nye and Tinker (1977) and Russell (1977)

Caradus (1979) Griffin (1972)

irregular cross-sectional geometry (Brewer and Sleeman 1960) and fine clay pores are mainly slit shaped (Sills et al., 1973). There have been a large number of pore-size classifications over the past 25 years, some of which are given in Table I. The range in class limits used to describe macro, meso, and micro pores is an indication in itself of the arbitary nature of most classifications. In this review we use Greenland’s (1977) terminology because it is a physically based system with categories related to the dominant water process. While Greenland’s nomenclature is preferred, actual class limits will vary depending on mechanical analysis and composition. There are seldom unique, nonvarying functions which describe the material properties of each soil type, and it is probably fruitless to search for a single universal classification of pore sizes. The biological origin and significance of some of these pore sizes are indicated in Table 11. Since t+b(O), the moisture characteristic, and K(O), the hydraulic conductivity, are the two material properties which most completely describe the status of soil water we need to know how a change in soil structure manifests itself in the $(O) function. We would expect structures developed from primary soil particles to affect potentials greater than, say, - 50 to - 100 kPa (assuming a close packing density of spherical particles) in sands and greater than - 1.0 to - 1.5 MPa (-1 x lo3 to - 1.5 x lo3 kPa) in clays, where domain separations are 0.01 pm. Soil pores of > 50 pm ECD, which affect

-

100

ANN P. HAMBLIN

the higher potential range (equivalent to > - 6 kPa), are normally developed from interaggregate or interped pores. These pores are the ones most frequently altered by disturbance, including disturbance which may occur in sample collection and preparation. An early laboratory study of particular clarity which shows the change in $(O) with structural alteration was described by Elrick and Tanner (1955). 0.50 0.40

0.30 0.20 0.10

0.50 0.40 0.30

0.20 0.10 0

0.50

-

Miami Silt Loam Cu I t i v a t e d

u 0.001

0.01

0.1

1.0

MPa

FIG.1. 1/40)curves for undisturbed (m), sieved ( - 100 kPa range, corresponding to an alteration of the interped pore geometry, as we would anticipate. Although Elrick and Tanner do not themselves draw attention to the difference in $(O) between the virgin and cultivated sites of the same soil, that difference is large, with a 37 % difference in the water held between -0.01 kPa and - 50 kPa. Such a dramatic reduction in the transmission porosity frequently accompanies a change from native vegetation to agricultural land use. Unger (1975) demonstrated that the difference in water retention between undisturbed and sieved samples is equivalent to the effect of tillage. Although tillage recreates transmission pores, these are often transitory, and tillage-created cracks may collapse within the season as the result of raindrop impact, compression of soil by roots, and wetting-drying cycles (Hamblin and Tennant, 1981; Dexter, 1977). The effects of tillage on PSD and the $(O) function are not always consistent. Some studies comparing zero tillage with ploughing show a volume reduction of zero-tilled transmission pores within the topsoil (0-0.1 or 0-0.2 m, depending on cultivation depth) but equivalent volumes for both treatments in the subsoil (Douglas et al., 1980). Other studies show increased transmission porosity in surfaces of zero-tilled soils through the development of earthworm channels and other soil faunal pores (Ehlers, 1975; Lal, 1976; Barnes and Ellis, 1979). These conflicting views arise, I suggest, because we are not always comparing like with like. Soils which are measured in the first year or two of a changed management regime may not have stabilized, and measurements taken at different times in the year are therefore not comparable. Significant soil structural changes can occur within a few weeks in freshly tilled topsoils (Hamblin, 1982). Longer-term changes are reported after one or several years. An increased proportion of transmission pores in undisturbed soils depends on the number of biological active days in the year and the level of macrofaunal activity (Ehlers, 1973). Reduction in transmission porosity of ploughed soils depends on the frequency and intensity of tillage operations, the rainfall intensity, and the structural stability of the system (see Section 111). There are so many combinations of management and pore-geometry interaction that the literature in this area offers little consistent predictive information on specific tillage practices. Deep cultivation (subsoiling), for example, is designed to increase transmission porosity in massive or compacted subsoils, but the superimposition of different tillage systems on such deep-tilled soils may alter the results (Negi et al., 1981). Similarily, the crop species can interact with the soil to produce different pore sizes, depending on

102

ANN P. HAMBLIN

the geometry of the root system (tap roots may enlarge transmission pores greatly and yet compress the surrounding pore walls), the root density, and total depth. Fahad et al. (1982) studied the influence of crop species on soil structure in an experiment which compared continuous monoculture with various crops in rotations. There was a reduction in aggregate stability associated with the continuous monoculture which was correlated with reduced 8 at all values of II/, but particularly in the > -60-kPa range (transmission zone). Structural differences may have a significant influence when a range of soils within a textural group are compared. Conventionally, soil physicists and agronomists have tended to attribute most of the differences in the shape and position of the $(8) function to differences in clay and sand content. This view is oversimplified.A recent statistical analysis by Williams et al. (1983). found pedality (the presence or absence of visible ped structure) to be the most important distinguishing attribute within groups of similar moisture characteristics. While there were usually good positive correlations between the values of 8 at low potentials (such as -1500 kPa) and clay content, the correlations at - 10 to -0.1 kPa were generally much weaker. Greacen and Williams (1983) made an interesting study of 244 horizons of Australian soils to compare poorly and well-structured members of 14 textural groups. Between potentials of -10 and -1500 kPa [the available water capacity (AWC)], well-structured soils in the coarse- and medium-textured soils contained one-third to two times the water of comparable poorly structured soils. The greatest differences occurred in the fine sandy loam to loam classes. Similarly,studies on the effects of cultivation damage and subsoil compaction in silty soils (Hamblin and Davies, 1977; Ehlers, 1973) showed the special susceptibility of these soils to loss of transmission porosity.

CONDUCTIVITY c . PORE CONTINUITY AND HYDRAULIC Water flow in porous media depends upon a hydraulic gradient composed of the total difference in pressure potential between two points. The flux is dependent upon the product of this gradient and the water content, and the rate of flow is controlled by the pore geometry. The effect of pore geometry is described by the Darcy “constant” K , which, in experimental terms, is normally the unknown. Q (the flux) and d$/dz (hydraulic gradient) are the measured variables. In the one-dimensional, steady-state system, Q

=

- KdII//dz

where Q is the flux in cubic meters per second, K is the hydraulic conductivity in meters per second, and d$/dz is the hydraulic gradient over depth z.


103

Although an expression describing the pore structure thus sits centrally in the flow equation, this parameter is clearly not a constant. [A more realistic description might be obtained from modeling the PSD as bundles of different-sized capillary tubes and applying an appropriate model such as the Poiseuille-Hagen equation (see Section VI). Yet even this is obviously a simplification of the real complexity of the porosity of soils.] Variations in measured values of saturated conductivity ( K , ) at any one site may have coefficients of variation [CV = standard error divided by sample mean; as a percentage, i.e., (s/X)lOO] of 100-200 %, while CVs of 200-400 % are reported for K ( 0 ) (Warrick and Nielsen, 1980). In comparison, the CVs of static properties such as 6 and pb are generally less than 10 %. While the methods for measuring K , and K(6) are less precise than for other parameters, it is the inherent heterogeneity in the pore geometry both vertically in the profile and spatially in the landscape which accounts for most of this variation. The functional form of K ( 6 ) thus varies considerably depending on .the dimensions over which it is measured. For one-dimensional, steady-state flow, a logarithmic function based on geometric means has received a consensus (e.g., Bouwer, 1969; Nielsen et al., 1973). Laboratory determinations of K are generally made on undisturbed cores and values are obtained vertical to the soil surface. Simple field methods for K , and transient K ( 0 ) also measure one-dimensional K vertically. Hydrologic methods for saturated groundwater flows generally use pump methods, which measure horizontal flux. Such values of K cannot strictly be extrapolated to different spatial scales or vectors except where the material is isotropic, which is rare in agricultural soils. The theory, measurement, and implications of the anisotropy of pore structures on K have developed from many independent sources, but much of the more recent work rests on that of Childs and his colleagues (Childs, 1952; Childs et al., 1957). In a field study of anisotropy, Childs et al. (1957) measured vertical ( K , ) and horizontal (Kh) contributions to flow in East Anglian fluvioglacial materials. In these recent depositional sediments, anisotropy ( K JKv) varied from to lo", with about half of the 13 sites being isotropic. Soils of similar texture varied substantially both in K , values (e.g., three alluvial clays had values ranging from to 10-9m/s) and anisotropy; the same three soils had K d K , values of 0.03, 200-70,000, and 1.0. We should expect K JKv values of less than 1 in soils with strongly marked vertical cracking (Vertisols, Natric Ultisols) and K JKv values of greater than 1 in soils with laminar bedding planes and high proportions of platy particles such as silt. However, our assumptions on the preferred flow vector can be at fault. Bouma et al. (1981) give an interesting example of the hydraulic conductivity surrounding pipe drains in massive clay subsoils. Pipe installation was predicted to have resulted in the smearing of the layer beneath the

104

A N N P. HAMBLIN

pipes, and backfilling to have given a more open structure above the pipes, but measuring K above (a) and below (b) the pipes ( K , , Kb), they found the Kb values to be the same as the surrounding undisturbed soil, whereas the K , values were generally less than K,. While considerable progress has been made in the number of measurements and methods used to obtain K(B) over the past two decades, structural heterogeneity has hampered data collection and reduced the utility of many of the measurements made. Alternative indirect approaches for obtaining K(8) which attempt to reduce the measured variability will be discussed in Section IV. Pore continuity has particular significance in agricultural environments, where so many management practices (tillage, land clearance, root-crop harvesting, and landforming) tend to disrupt naturally continuous pores and produce one or more structurally discontinuous layers. I consider this has great significance in crop-water relations and is not, as yet, always appreciated. A typical example of the effect this has upon the hydraulic gradient is

0 -10 -20 -30 -40 0

-10 -20 -30 -40

Hydraulic Head (kPa)

Frc.2. Hydraulic gradients at 0.1 (0) and 0.01 ( 0 )pm/s flow rates for undisturbed, wheat-fallow, and wheat-pea rotations of a Walla-Walla silt loam. (After Allmaras et al., 1982.)

given by Allmaras et al. (1982), and is shown in Fig. 2. A reduction in the proportion of larger pores in the top 30 cm of an old (50 year) cultivated silt loam gave a 10-fold reduction in K(8) at potentials greater than - 10 kPa compared with undisturbed grassland. This reduction in K(B) was accompanied by longer periods of wet, anaerobic conditions in winter and spring, increased nitrogen loss, lower pH, and lower biological activity. Steady-state hydraulic gradients of the long-cultivated sites showed a double inflection, which is probably very typical of many agricultural soils.


105

In recent years much attention has been focused on the role of more or less vertical, continuous large pores in both saturated and unsaturated water flow. These have been generally termed “macropores”, as in a review by Beven and Germann (1982) which summarizes their hydrological significance. Earthworm channels tend to be randomly distributed, with spatial densities as high as 900m-2 (Bouma et al., 1982) in temperate Europe or 500 m - 2 in Mediterranean climates (Barley, 1959). They have been reported as continuous to 0.7 m by Ehlers (1975) and 1.6 m by Bouma et al. (1982), but Barnes and Ellis (1979) noted considerable annual variation in depth depending on the depth and duration of wetted soil profiles. Despite considerable swelling and shrinkage movement of the clay soil studied by Barnes and Ellis, the wormholes persisted for several years. Cultivation has been found to reduce the numbers of both holes and worms very drastically (Ehlers, 1975). One of the major factors in long-term inprovement of many zero-tilled soils has been the development of wormhole transmission pores. Earthworms are not found in alkaline soils, however, nor are they abundant in soils with low organic matter, especially where there is little surface plant litter. We might therefore anticipate more widespread effects of large biopores from root channels than from wormholes. Persistence of channels created by crop roots is a common observation, although it is seldom quantified. Many incidental references to new roots growing down old root channels occur (e.g., Ehlers et al., 1983). Where such channels do not become hydrophobic from detached, lignified, cortical cells, they could have a pronounced effect on infiltration and through-drainage. Some of the pioneering work on the effect of root growth on pore structure was carried out by Barley (1953, 1959). Barley considered the effect of earthworm channels less significant than that of root channels in a red brown earth, since the number of root channels exceeded wormholes of > 500 pm diameter by a factor of 10 in the topsoil. Later data (Barley, 1970) for cereal root numbers in the same soil suggest that at 0.5 m root channels would outnumber wormholes by lo3. Barley argued that the large diameter of wormholes reduced their role in water movement (though not their significance to gas exchange), as they would not fill until the soil (in this case a red brown earth or Haploxeralf) was near saturation. The functioning of these larger transmission pores in soils is not yet completely resolved. It is central, however, to the concept of preferred pathway flow, and we return to it in Section IV,B,l. In passing, it is interesting to note that when Sedgley and Barley (1958) measured the effects of grass roots on the hydraulic properties of a sandy loam, they found a reduction in K ( 0 ) at Ic/ = - 3.0 kPa when roots had been grown in the soil. They interpreted this as compression by the roots which had reduced pore space. Nevertheless, Sedgley and Barley’s work may be open to an alternative explanation, as put forward in a study by Gish and

106

A N N P. HAMBLIN

Jury (1983). These authors found the effect of root channels (with both living and decayed’roots in them) was to increase the proportion of immobile water, that is, water which has entered “dead-end” or stagnant pores, when solute flow was measured. The proportion of immobile water increased overall from < 10 % in precropped soil columns to 23 % where living roots were present to 39% in the case of decayed roots. Thomas and Phillips (1979) quoted examples where more than 50 % of new water additions into wet soil bypass the matrix water, rather than displacing it, because of preferred pathway transport. Although solute transport falls outside the scope of this review, it is worth noting that many predicted outflow curves of solute transport can only be made to “fit” observed values by designating some of the volume to stagnant pores. In field studies on solute transport, such as that of Omoti and Wild (1979), where fluorescent or colored dyes were used, preferential pathways are frequently a marked and characteristic feature ahead of the general wetting front. Intuitively we expect root channels to increase the hydraulic conductivity at $ > - 0.3 to - 6 kPa, equivalent to ECDs of 1000 to 50 pm, which would reduce the number of occasions when surface soils are waterlogged by transient perched water. One could argue, however, that because cultivation severs the continuity of root systems and probably blocks the upper openings of root channels, these effects will not be found in many cultivated soils (Greenland, 1977). Logically, therefore, it is in zero-tilled or pasture soils that continuous root channels should have their most pronounced effects on K(0). Yet many of the comparative studies in the literature between ploughed and undisturbed soils do not bear this out. This sometimes occurs because there are other interactions masking the effects of roots, such as the muchincreased role of earthworms in mulched, direct-drilled soils in West African (La1 et a/., 1980) and British (Barnes and Ellis, 1979) environments. In other cases zero-tilled crops have not grown so vigorously and have produced less root growth. Hamblin and Tennant (1981) found that a reduced proportion of large pores on a zero-tilled loamy sand in Western Australia (where there were no earthworms) was reflected in the slower movement of water through the topsoil of the undisturbed treatment and in reduced root growth associated with higher soil strength. This situation persisted unchanged over a 5-year continuous cropping period (Hamblin et a!., 1982) with no development of a significant number of continuous root channels or increased flow rates over time. Continuous vertical fissures created by shrinkage are at their most pronounced in soils of high smectite content. They are rare in soils dominated by “low-activity’’ clays (that is, those soils of low cation capacity and/or charge density) such as Oxisols and Ultisols. Soils with clay contents of greater than 30 % have a continuous clay matrix, and their physical proper-


107

ties are determined by clay type rather than clay content. In Vertisols these fissures may extend to over 1 m and do not always close completely, even when soils are fully hydrated (Warkentin, 1982). Minimum horizontal spacings of such cracks vary from 0.5 m (Chan, 1981), with larger spacings developing at depth. The surfaces are often “self-mulching” when subjected to wetting-drying cycles, producing large numbers of stable peds of 1000 to 5000 pm diameter. They thus possess a distinct trimodal pore system of large noncapillary cracks, small capillary cracks between micropeds, and storage pores of /(es (6) D(8) is expressed in terms of an exponential function of 8 (Brutsaert, 1979):

116

ANN P. HAMBLIN

where y(p) is a function of the slope of the wetting front for which Brutsaert suggested a value of N 8.0. However, if D(8) is not exponential with water content, this value is not a constant. In Eqs. (6) and (7) S is measured at initial water content Oi and after t(O,), while 0, and 8, are the saturated and residual water contents, respectively, and 8, is separately determined. A recent treatment of modern infiltration theory which describes these principles in more detail is given by White et al. (1982). They show that successful prediction of the movement of the wetting front, the time to incipient ponding, and the time-dependence of surface 8 can be made for the conditions of constant-rate rainfall, nonsealing surfaces, and uniform Bi. This is sufficient where sprinkler irrigation or rainfall simulators are csed to obtain infiltration characteristics ( I , i, etc.) of a stable, homogeneous soil. Variable-rate rainfall and sealing surfaces are more complex to treat, although by assuming the surface to be already sealed, as a layered soil, the latter problem has been tackled successfully (Section IV,B,l). 1. Field Infiltration Afected by Unstable Structures In a well-known paper written in 1940, Horton, an engineer and an experimentalist, placed considerable stress on the physicochemical processes which accompany rain falling on soil surfaces. He described three “exhaustion phenomena” which lead to a decline in the asymptotic values of i, including the decrease in effective pore size with soil crumb breakdown and the puddling of soil, which reduces porosity and “colloidal swelling,” which also reduces pore space. In addition, he considered the significance of entrapped air in immiscible displacement ahead of the wetting front and the

I ”O

F

A A A

0.6

J

F

M

A

M

J

J

A

A

S

O

N

D

Month

FIG.4. Seasonal variations in infiltration (m/s x drainage basin (Texas). (Adapted from Horton, 1940).

from the North Condho River


117

forces associated with rainstorms of different drop dimensions and velocities. Figure 4 shows his data for the seasonal variation in infiltration capacity for sandy soils in a Texas drainage basin. He ascribed these differences to the increased biological activity of the May to October period and, in particular, to the production of wormholes. Horton suggested that wormholes would operate before the whole soil was saturated (at suctions equivalent to the wormhole diameters) because of the development of surface detention in depressions, which pond before the main extent of the surface becomes ponded. Many of Horton’s hypotheses have been validated by more recent work, although current infiltration theory has been able to extend far beyond his notion of a single controlling factor (soil surface versus soil suction). His emphasis on time and space variables are, however, still as relevant today as they were 40 years ago. Prediction of cumulative infiltration and time to ponding in soils which are prone to structural deterioration and crusting remains a central problem of applied soil physics and hydrology. Considering infiltration into an existing crust, Hillel and Gardner (1970) developed a model in which the hydraulic resistance of the crust was related to the conductivity of the underlying transmission zone. They later utilized this theory as a method for obtaining field values of K ( 0 ) and D(0) by deliberately imposing a crust of kaolin, gypsum, or a porous ceramic plate on the soil and supplying water at a negligible head to it. At the scale of field or catchment hydrology infiltration has repeatedly been modeled using equations based on those of Green and Ampt. However, most assume a unique, single-value function of K and rl, (or rl, transformed in terms of 0). Wilson et al. (1982), for example, compared three models of increasing sophistication, which included increasing numbers of terms for air resistance and viscous drag. They found that the most complex model predicted infiltration reasonably well for dry soils, but all three models overpredicted time to ponding for wet soils. In field validation tests they found surface sealing occurred early during infiltration with wet soils, which was not adequately predicted by any of the models. In soils with roughened or unstable, bare surfaces, an “incipient ponding” condition often occurs early in a rainstorm in which free water lies in small depressions or furrows but there is no continuous water film. Surface ponding (rl, = 0) now occurs if rl, is averaged spatially over large enough unit areas, though a complete free-water film (overland flow) will not form if transmission pores, fissures, and holes exist which can drain the water at perhaps two to three orders of magnitude faster than K,. In dried, cracking-clay soils, two phases of infiltration and profile wetting thus occur (see Section IV,A,2). It is useful at this juncture to compare the rainfall rates and K values we may encounter in nature. Table I11 shows in schematic form, some equivalent

118

ANN P. HAMBLIN

Table I11 Comparison of Rainfall and Hydraulic Conductivity Rates Average rainfall rate (mmih)

Popular description

Equivalent rate at ground Ws)

0.1 1 .O 5.0 10.0 20.0 50.0 100.0

Mist Drizzle Light rain Steady rain Heavy rain Downpour Cloudburst

2.8 x 2.8 x 1.4 x 2.8 x 5.6 x 1.4 x 10-5 2.8 x

Approximate K, for various soils WS)

lo-’

Name Massive clays Cracking clays Sandy clays Loams Silt loams Loamy sands Fine to medium sands

rates for rainfall and K,, with the rainfall in units of millimeters per hour, units frequently used in infiltration studies although less widely used in soil physics or soil-plant water studies since rainfall events are not uniform for periods of as long as an hour. Surface sealing is a worldwide problem affecting freshly tilled soils which are exposed to the full force of growing-season rains. In drier regions the subsoil may contain little stored water for plant growth at cultivation time, and sealing can seriously reduce the proportion of total rain stored in the soil either by run-off (Epstein and Grant, 1967) or by increased soil evaporation from surface-ponded water (Hamblin, 1984). Infiltration rates may be reduced by several orders of magnitude in the crust. McIntyre (1958) showed that crusts characteristically may have a surface “skin” with a very low conductivity (e.g., K , = 5 x lo-’ m/s) overlying a crust ( K , = 5 x lo-* m/s) of disaggregated material, while the underlying soil may have a K , value of m/s. The susceptibility of the soil to surface sealing can therefore be determined by tests such as Emerson’s (1967), but how do such stability tests relate to rainfall parameters such as storm duration, intensity, and droplet energies? These forces may be very substantial: Ghadiri and Payne (1981) demonstrated that localized impact stresses from raindrops range from 2 to 6 MPa in the rebound corona for about 50 ms as a droplet hits a surface. Edwards (1982) developed empirical functions relating runoff probability to soil surface condition (crusted or not crusted at start of rain). The probability of crusting given the particular soil type and rainfall was assessed from the cumulative rainfall. Uncrusted soils were arbitarily defined as receiving < 13 mm of rain since the last tillage operation. Observed and predicted runoff values differed because of the creation, and then destruction, of a crust.


119

Unfortunately, this method seems limited in general application by the size of the data base required and the interaction with unpredictable cultural practices. Whisler et al. (1979) used a modeling technique in which the soil physical properties of the surface were time dependent, since the soil immediately starts draining when V, < K,. In their solution they used published values of K , surface flux, porosity, and Oi, 0,, and 8, for a crust and underlying soil. They then allowed a finite depth of the soil surface to adjust from bulk soil to crust values during infiltration, with parameters controlling the rate and amount by which K , decreased up to, and after, crust formation. Their model, however, still requires input parameters for the type of crust developing on each soil. There is a substantial gap in our knowledge of the field behavior of many soil types to varying rainfall conditions. The Food and Agriculture Organization (F. A. O., 1979) has used simple textural and organic matter formulas as crusting indices for global assessment purposes. The ratio of coarse silt to fine silt plus clay and organic matter was found to be in good agreement with susceptibility to crusting.

2. Infiltration with Surface-Connected Transmission Pores It can be argued that, provided Vo is less than K , and the surface of the soil does not slake and crust, transmission pores take no part in infiltration. This may apply to coarse-and medium-textured soils, and even in clay soils with cracks or cultivated shear planes, if V, is substantially less than K , (see Table 111). Clothier and White (1981) used the scale relationships of D and K converted to 8 and S, previously referred to in Section IV,A, to compare constant-flux infiltration at small negative pressures and by conventional ponded methods to identify the contribution of different-sized pores to the infiltration of sandy field soils. At a value of -0.4 kPa (ECD = 740 pm) they found 10-fold smaller values for S than those obtained by the ponding method. In a further study on the effect of pore geometry on the shape of D(0) curves, Clothier and White (1982) compared water content profiles from the same soil before and after disturbance at the same small suction by constantflux infiltration. While the disturbed samples yielded the expected exponential D(8) function, the undisturbed soil had a near linear function (constant with respect to 8). Similar low dependence of D on 8 has been reported by Hamblin (1982) for undisturbed sandy loamy and sandy clay loam topsoils and by Hamblin and Tennant (1981) for a loamy sand. Clothier and White found their disturbed soil contained far fewer pores greater than 200 pm in diameter than in the undisturbed soil. They called these “biopores” and considered them partly responsible for the lower diffusivities of the unm2/s, disturbed soil. [D* (linear diffusivity) of disturbed soil was 1.3 x

120

ANN P. HAMBLIN

whereas the undisturbed D* = 4.1 x m2/s]. These results support my earlier comment that structural alteration changes the material properties of soils, and hence all their derived functions. They also call into question the assumption that such functions have a single form [such as an exponential form for D(O)]. A theoretical treatment of infiltration in which noncapillary transmission pores are present and operate has been considered by Scotter (1978), Beven and Germann (1981), and Germann and Beven (1981). Transmission pores which are open to the soil surface but closed at the base are more common in nature than was traditionally expected or allowed for in classical soil physics. This has led to reports in the literature of unexpectedly short response times to water content changes deep in soil profiles after rain (Nulsen, 1980), even in circumstances where ponding did not apparently occur first. However, it is likely that ponding develops transiently and then ceases relatively soon, as surface flow is channeled into the large (noncapillary) pores. Beven and Germann (1981) and Bouma et al. (1982) described such situations theoretically and experimentally. They suggested that K(O) can increase by as much as four orders of magnitude over a pressure drop of only 10 kPa. Beven and Germann (1982) suggested that, from examples in the literature, rainfall intensities of 1- 10 mm/h would be sufficient to generate macropore flow in previously dry soils, but they stressed that antecedent rainfall, initial soil water content (OJ, and actual storm intensities would give rise to great variation in response. Germann and Beven (1981) suggested a statistical approach with their theory. This would provide field data on the proportion of the saturated volume flux density contributed by their “macropores” (Qma), where

cA xn where c = pg/8p ( p is the density, g is gravity, and p is the viscosity), a combination of terms for the flow of water in tubes from the Poiseuille equation [see Eq. (13)], A is the total cross-sectional area containing n vertical large pores, in which the porosity emais a proportionality constant: Q ma

= - 2

n ema= - Vp Ah

(9)

where Vp is the volume of each tube and h is its height. The notion of a proportionality “constant” still oversimplifies the real situation, and the authors noted that their exponent (of 2) was only a first approximation and did not fit published data well. The volume of “drainage” pores has frequently been measured in the past as the total porosity minus the water-filled pore space at field capacity.

121


However, the term “field capacity,” meaning the water content at which measurable drainage ceases, is woefully inexact and justifiably open to criticism (Ritchie, 1981). Laboratory estimates of the drained upper limit taken from both disturbed and undisturbed samples may be seriously in error because the representative elementary volume (REV) contains transmission pores which are not included in the sample. Moreover, as we have seen, the rate at which the soil wets also affects the way it drains, if transmission pores are a significant feature of the soil. Germann and Beven’s analysis suggests that a lognormal relationship exists between the proportion of large pores (ems) and the large-pore volume flux density (Q,,). The proportion of transmission pores need not necessarily be high to have a marked effect. Pores greater than 3 mm in diameter in the cases examined by these authors and by Bouma and Dekker (1978) represented no more than 2 or 3 % of the total porosity. Studies which have compared near-saturation hydraulic conductivities of soils with and without the transmission-pore contribution have shown a velocity ratio for difference in K ( 8 ) of four times (Ritchie et al., 1972) to several hundred times (Nulsen 1980), depending on Oi, V,, and the degree of swelling and closure of cracks during the period of measurement. The work of White et al. (1983) has yielded some interesting considerations on transmission-pore infiltration. From rapid in situ measurements of S, Bi, 8,, and K O using a rainfall simulator (sprinkling infiltrometer), they found a marked inflection in the surface-monitored $ / S curve at approximately -0.35 kPa (equivalent to a mean exclusion value for noncapillary pores). 0.1

0.2

-E

0.4

1

0.6 0

0.8 1

I Microdepression Chromic Vertisol (Houston Black Clay with Gilgai)

.o

u 0.2 m

Typic F luvaquent (Fine Clavev. Mixed, Mesic)

FIG.5. Schematic representationsof a cultivated chromatic Vertisol (Houston Black Clay) [redrawn from Ritchie et al. (1972)l and a Fluvaquent (alluvial clay) [redrawn from Bouma and Dekker (1976)l demonstrating bi- and triscalar pedality.

122

ANN P. HAMBLIN

There was also a high correlation with the reduction in coefficients of variation in values of S at = -0.35 kPa, again suggesting that transmission-pore hydraulic behavior is highly variable, being the major component in the very large CVs for K , and K(8). Hoogmoed and Bouma (1980) laid stress on a “short-circuiting” process (Beven and Germann’s “channeling”) during infiltration into cracking clays, where water moves within an interconnected transmission-pore structure which is not closed at the base, bypassing the matrix peds, particularly during high-intensity rainfall events (Bouma et al., 1977). Horizontal infiltration into peds was predicted to be larger per time increment than vertical movement m/s compared with 11 x lop6 m/s). under lower rainfall rates (3 x Such polymodal soil structures are characteristic of Vertisols and calcareous clay soils. Typical patterns are shown in Fig. 5 for a Vertisol from Ritchie et al. (1972) and for a Fluvaquent from Bouma and Dekker (1978). They give rise to some interesting considerations for water flow for plant roots (Section V1,A). Note that both examples show a marked discontinuity of the pore structure in the plough layer, which had a pronounced effect on the hydraulic conductivity in each case.

+

B. REDISTRIBUTION WITHIN THE ROOT ZONE

Infiltration, particularly into dry soil, is dominated by suction gradients (capillary potential gradients), but redistribution within the whole volume of the soil is a consequence of both capillary and gravity forces. Flow equations thus include both D(8) and K(8). The development of general and specific forms of the combined Darcy flow and continuity equations and their translation into diffusion terms is well covered in all recent works on soil physics (Marshall and Holmes, 1979; Hillel, 1980). In the case of a fallow soil with no alternative sink by plant roots, the balance between wetting and draining may be taken as an imaginary plane through the profile. In soils of higher K(8) values (mainly sands), this plane remains recognizable but moves down the profile in a wave-like progression. In slower-draining soils the movement proceeds more uniformly through the wetted region, resulting in a continually decreasing flux at the transition plane. The effect of gravity can only be ignored at small length scales (close to alternative sinks such as the upward flux of water vapor to the soil surface and to large cracks and the flux of the liquid water to plant roots) or in the case of small infiltration quantities into some dry soils. Hillel (1977) has published a number of computer simulations for water redistribution, with simultaneous evaporation and intermittent rainfall, for typical clay, loam, and sand soils. The clay soil, as we would expect, redistributes water over the


123

smallest depth but retains the highest water content in the upper parts of the profile per unit time. Structural influences on unsaturated redistribution are most noticeable where there is vertical or areal heterogeneity. I . Vertical Heterogeneity

The effect of vertical discontinuity on water flow has been studied most in relation to textural contrasts. For example, Hillel and Taplaz (1977) predicted water movement for sand over clay systems, while Clothier et al. (1978) discussed the significance of variations in water flow through fine- over coarse-textured soils. Structural variations which have developed from tillage operations are not always so easy to distinguish in the field, although penetrometer readings may identify compacted traffic pans and visual scoring of changes in structure near the surface can give a surprisingly sensitive indication of hydraulic behavior. Impedance to vertical water flow results in perched water tables, abrupt inflections in potential gradients, and anaerobic zones distinguished by mottling (see Section V,B,2). Such layering of hydraulic properties will nearly always result in some check or stress on crop growth, either through gross differences in water availability to different parts of the profile or through differential structures and inhibition of root growth into discretely segregated parts of the whole soil volume. This increases the risk of the crop’s inability to adapt to other sudden environmental stresses. There is conflicting opinion, and a shortage of unambiguous evidence, on the role of “short-circuiting” or preferred pathway flow in unsaturated soils in which the pores do not reach the surface. In principle, all pores greater than 30 pm in diameter should drain above $ = - 10 kPa and, as we recall from Table 11, pores less than 30pm in diameter are smaller than most longitudinally continuous holes created by roots, worms, or other biological agents, Yet, there are occasional examples of very rapid water redistribution at water contents substantially less than the drained upper limit, which suggest that quite small vertically continuous pores are filled very rapidly, perhaps as the result of local ponding of water with pockets of soil. This could occur at the base of the plough layer, or in sand lenses, or in soils which are initially moist but with negligible flux gradients. Quisenberry and Phillips (1976) attempted to identify the contribution of transmission-pore flow to solute transport and unsaturated water redistribution in a silt loam with an organic Ap horizon. Figure 6a shows a characteristic redistribution profile for tracer C1- (with chloride recalculated as concentration in the soil water rather than total soil content, as shown in the original profile) and Oi,which was less than the upper limit to available water (ULAW). The equivalent displacement or piston-flow profile (x,xi),which would have occurred after one hour’s redistribution if there had been no preferred pathway flow, is

124

ANN P. HAMBLIN CI‘ Concentration in Soil Water

Water Content (v/v) 0.26 0.34 0.42 0.48

Coefficient of Variation

ppm

0

1000

2000

0

40

80

120

160%

0.2

-E -

0.4

5 Q

; 0.6 0.8 1.o

b

a

F1c.6. (a) Water contents and tracer chloride concentration in the applied water for a cultivated Huntingdon silt loam, at 1 hr (V) and 30 hr after application of 178-mm water. Percentages refer to percentage water and chloride recovered. (Redrawn from Quisenberry and Phillips, 1976.) (b) Coefficients of variation for Chloride concentration down the profiles on undisturbed (m) and cultivated (0)Huntingdon silt loam (Quisenberry and Phillips, 1976).

(v)

indicated. The applied water and C1- had moved 0.7 m in the first hour, and most of the applied water had moved through the profile in 36 hr. The effect of tillage was noticeable as a bulge in water and Cl- profiles at 0.15 to 0.2 m, quantified by plotting the CVs of Cl- concentration with depth for the tilled and nontilled soils (Fig. 6b). The tilled surface soil (0.15 m) showed a gradual increase in CV to the depth of the cultivation but then a sudden jump in CV (from 60-120%) below that, whereas the untilled soil showed a smaller increase to 0.1 m and had CVs of only 40% in the subsoil. The authors take this inflection in CV distribution with depth to indicate changes in pore continuity, with breaks at 0.15m at the base of the ploughed soil. This analysis, utilizing the change of CV with depth to identify structural heterogeneity, is similar to that of White et al. (1983). It offers scope in identifying the contribution of preferred pathway flow at different values of 8. It should be more widely used, especially in attempting to distinguish flows resulting from textural versus structural differences. Soils having natural differences of either texture or structure with depth are particularly difficult to manage in relation to their water status for crop growth. Many agricultural soils have finer-textured subsoils than topsoils, such as the “textural B horizons” of red brown earths (Haploxeralfs and Rhodoxeralfs), Pullman clay loams in Texas (Torrenic paleustolls), clay with flint soils in Europe, and tropical Alfisols (Tropuldalfs). Their subsoil hydraulic conductivities are frequently too slow for good drainage in wet environments or for rapid enough redistribution of water in seasonally dry

THE INFLUENCE O F SOIL STRUCTURE

125

climates to allow deep root penetration. Roots then become confined to surface layers, with a host of subsequent problems which give rise to yield depression or even total crop loss. Land management techniques for reducing vertical heterogeneity is a practical remedy in these cases. In drought-stressed environments the relationship between crop growth and water use may be known well enough for the production value of each millimeter of soil water to be calculated. Deliberate manipulation of soil structure by landforming, mulching, subsoiling, or chemical stabilization can improve water-use efficiency so greatly that it pays for itself. While the use of gypsum for unstable soils and subsoiling (deep ripping, chiseling) are both widely used as individual amelioration treatments, a notable example from the Riverina irrigation district in South Eastern Australia has made simultaneous use of four methods to improve the physical properties of a duplex red brown earth (Rhodoxeralf) for stone fruit production. These soils are prone to crusting and have low subsoil conductivities, because of both their high clay content and low Ca: Mg ratios. Cockcroft and Tisdall(l978) reported a system of soil improvement in which gypsum was injected into the subsoil during deep ripping, and the top 0.6 m was then cultivated with incorporated straw and green manures. The surface between trees was covered with straw mulch to help suppress weeds, reduce crusting and temperature fluctuations, and increase earthworm numbers. Tisdall (1978) found the number of earthworms increased on improved plots from 150 to 2000 per m2, with an associated four-fold increase in pores emptying at - 4 kPa. Infiltration of 50 mm of sprinkler irrigation water was accomplished in 1 min compared with 83, and yields of stone fruit increased from 18 to 75 ton/ha. The example is also interesting in relation to the role of earthworm channels in an unsaturated water environment. Clearly, the channels functioned despite the “nonponded” surface. However, the soil was never allowed to dry beyond - 3.0 kPa, and therefore many of the narrower earthworm holes could have been filled with water much of the time. 2. Areal Heterogeneity Areal heterogeneity at scales up to tens of meters occurs in many tropical and subtropical soils which have developed over prolonged weathering and erosional sequences on old landforms. They have high CVs (35-100%) for particle size distribution and depth of horizon (Wilding and Drees, 1978) as a consequence. Many Alfisols and Ultisols are mosaics of different mechanical composition (with associated changes in base status, pH, and organic carbon content), whereas some Vertisols and other Alfisols develop “gilgai” phenomena with regular repeat sequences in relative heights of 0.1 to 0.5 m in a hummock and depression sequence related to differential swelling and

126

ANN P. HAMBLIN

shrinkage of surface and subsoil clays (Hallsworth et al., 1955; Knight, 1980). Normally, variations in microrelief can be leveled by grading and cultivation, but they tend to reappear over periods of 2-10 years. TaIsma and van der Lelij (1976) measured field infiltration and water movement in such a soil which had already been leveled. Large-diameter ring infiltrometers gave a bimodal distribution of infiltration rate, associated with different sizes of transmission pores. The author’s attempts to assess mean hydraulic conductivities from flux gradients and infiltration rates were hampered by the large variation in field-measured properties on plots which were only 30m2 in area. The very large CVs of soil hydraulic properties, and the preoccupation of soil physicists today with problems posed by heterogeneity, may be brought into focus when stated in the terms of the previous example. How may the soil physicist predict water availability for plant growth in a soil of known spatial and temporal heterogeneity? Measurements in the field must be made at point sources. Should they then be spatially averaged to give a single value for the field, since the crop yield will eventually be averaged, and if so, how? Spatial and temporal soil variability are now frequently analyzed by procedures originally developed in geostatistics. Geostatistical techniques are described in detail in a number of papers by Webster [e.g., Webster (1975); Burgess and Webster (1980)l and are reviewed by Warrick and Nielsen (1980). A paper by Sisson and Wirenga (1981) comparing autocorrelation functions with trend-series analysis (spectral analysis) has brought to light some interesting points about soil structure, water movement, and scale. The authors measured steady-state infiltration on a 40-m2 field plot with three sizes of ring infiltrometer (0.05-,0.25-, and 1.25-m diameter rings). The spatial distribution of the data was most erratic, with the greatest variation in the smallest ring size, as we might expect. Log transformation of the data did not completely stabilize the variances. The 0.05-m-ring data were not well described by the log transformation, being the most skewed, though the 1.27m-ring data were adequately described when log transformed. The most variance among observations occurred at larger distances between samples. Interestingly, the soil, a typic Torrifluvent (a fine, silty clay loam 0.7 m over fine sand) was not considered markedly heterogeneous. A complementary approach, which derives from applied physics, attempts a simplification of the soil system by “reducing” it to the smallest number of variables which are able to be “scaled,” so that specific physical properties become scaled models of a general “reduced” system. Scaling of “similar media” proceeds from describing a microscopic characteristic length of the matrix and a macroscopic characteristic length of the whole profile depth, and then multiplying measured properties by the appropriate scaling factor, K(e), or D(e). An example of the practical to produce scaled values of +(@),


127

use of scaled K , values was given in Section IV, for field measurement of nonponded infiltration. It is particularly valuable where properties of high intrinsic variability, such as K , can be reduced to equivalent properties of low variability, such as 8. Thus scaling allows a reduction in the number of parameters measured and in particular those with the largest CVs. The theory, illustrated with a range of examples, has been most clearly expressed by Miller (1980).

V. PATTERNS OF ROOT GROWTH Root morphology is as much a product of genotypic and environmental interaction as canopy morphology, but the sheer labor involved in obtaining satisfactory root measurements has hampered progress of below-ground features such that we still have a somewhat hazy notion of what may be the best root configuration for different environments and species. Some root descriptions we find in the literature leave us with the false impression of a static, fully formed root system. The best of the descriptive examples are equal to the painstaking and detailed root distribution maps of Weaver (1926) of a range of species recorded in the Great Plains of the United States. In the life of an annual crop, however, the root system of each plant expands and interleaves with its neighbors, extending into deeper horizons and subdividing more and more finely into smaller soil volumes. The primary root system ages, and some parts may die while newer laterals become the active uptake zones for water and nutrients. Water tends to be withdrawn to lower potentials from the surface downward, but intermittent wetting of the soil blurs this sequence, and flux densities and directions may change repeatedly. Root extension and proliferation increase rapidly in zones of moist (- - 5 to -50 kPa) soil, but when some parts of the soil dry to potentials below -2 MPa or are flooded for more than a few days, the roots within them die unless connected to better watered or aerated zones. In this case they may be quiescent until further wetting or draining initiates new lateral branching. Downward growth rates of annual crops range from less than 2 mm/day (Greacen and Hignett 1976), where soil temperatures are low and soil strengths are high, to greater than 80 mm/day in warm, moist conditions (Kaspar et al., 1978). However, in a growing season of, say, 180 days, with the normal sigmoid growth function and little new growth in the maturation phase, we can see that root systems of annual crops will seldom extend more than 3 m in depth, and in well-watered, temperate climates the maximum

-

128

ANN P. HAMBLIN

depth is often only 1.5 m. Nevertheless, the proliferation of roots in nonrestricted environments, such as well-fertilized topsoils, can be prodigious. Graminaceous species may commonly have total root lengths of 2-6 x lo4 m perm2 of ground area (Nye and Tinker, 1977), more than half of which occurs in the surface soil. Are such profuse root concentrations advantageous to the crop’s water and nutrient uptake, or do they represent a wasteful squandering of assimilates similar to the production of nonfertile tillers in cereals? The answer may well depend on the uniformity of water supply to the crop during its life cycle. In some drought-affected environments, where crops grow on stored water, it has been demonstrated that restriction of either the roots or the shoots can give a more equitable water balance between vegetative and reproductive growth in wheat (Passioura, 1972; Islam and Sedgley, 1981). However, the root-shoot ratios of many droughted crops are larger than those of well-watered crops (Schultz, 1974); and ~

Root Length LVrnrW3x lo4

0

1

2

3

4

5

6

7

0.4

Q

1.2

:::r 1.6 1.2

Spring Wheat (Australia)

Oats (Germany)

Frc. 7. Root-length densities (I+, m/m’) at anthesis for winter wheat grown in a sandy loam over gravely clay [redrawn from Gregory et al. (1978a)], spring wheat in a loamy sand [redrawn from Hamblin et al. (1982)], and oats in a loess [redrawn from Ehlers et al. (1981a)l.


129

drought tolerance in some species has been correlated with those cultivars which have longer, fine-branched root systems (OToole and Soemartono, 1981 ; Fischer and Turner, 1978). Nevertheless, the possibility of genetic selection of genotypes with root characteristics which will make more efficient use of water in the soil seems a daunting task in view of the enormous plasticity displayed by many root systems over a range of environments and the similarities in overall root morphologies between species, let alone cultivars. Figure 7 shows the distribution of root-length density (Lv;m/m3) for winter wheat grown in a sandy loam over gravely clay in southern England, a spring wheat grown in a loamy sand in Western Australia, and oats grown in a podzolic loess in Germany. Both the spring wheat and the oat crop were grown on sites which contained a traffic pan and were part of experiments comparing tillage systems. Roots from the spring and winter wheats were extracted by coring and washing techniques and grid counted after removal of dead material. The oat roots were counted directly from cut faces of the soil profile and a conversion factor was applied to transform root numbers to lengths. Yet, for all of this, there are probably more similarities than differences betpeen distribution, and in the case of the two wheats many of the parameters are very similar. This may be consoling to theoretical modelers of water extraction by roots, but how much scope does this allow for detection of specific or varietal difference in root parameters designed to improve crop water use?

A. GENOTYPIC VARIATION Criteria for optimal root configuration have been more thoroughly identified for nutrient than for water uptake. Admirable treatments of the nutrient aspect have been given by Barley (1970) and Nye and Tinker (1977). This may be due to the difficulty of identifying which parts of the root system are taking up water. Some authors have computed that less than a tenth of the total root length need be involved; Passioura (1980) calculated active root length was 30 % in young wheat plants and Weatherley (1975) found less than 1.O % were necessary to satisfy the uptake equations he used. Moreover, most measurements of field-grown roots do not consider the fraction of the root system which is suberized and relatively impervious to water. Histochemical staining techniques are available which may help to discriminate these portions [e.g., Ward et al. (1978)], but more development is needed for unambiguous detection of viable tissue from field environments. From field evidence of root growth and water uptake, which will be discussed in detail in Se=tion VI, there is some evidence to suggest that water is most rapidly taken

130

A N N P. HAMBLIN

up from younger roots in the 10-50-cm zone behind the root tips. This could often represent less than 50% of the root length. However, Taylor and Klepper (1975) could find no difference in root water uptake rates in cotton when different parts of the soil-plant system were compared at equivalent potentials-a point not always considered in other studies. Identification and selection of specific root morphological or anatomical characters for improving water-use efficency are thus hampered by lack of knowledge on the functioning of the root system in realistic field situations. Differences in root characters of cultivars have been demonstrated, but many parameters measured on whole-root systems [as, for example, the work of OBrien (1979) on spring wheats] are correlated with shoot development and are profoundly influenced by the test environment. There is good evidence for cereals (Gregory et al., 1978a; Klepper et al., 1984) to show that tiller development may be synchronized with root axis production, and the number of roots on a tiller may be predicted from the number of leaves on the same tiller. Selection of specific features with a realized heritability, such as root-hair length in clovers (Caradus, 1979) or xylem diameter in wheat (Richards and Passioura, 198l), for well-defined environmental stress situations is clearly a more realistic goal. Root characters which are adaptations to environments having substantially too little or too much water have been increasingly studied in recent years. When water deficiency is a regular part of the crop’s environment, either during vegetative growth or as an inevitable, increasing, drought stress during maturation, “aridopassive” species maintain their transpiration rate as near to the potential maximum for as long as possible, usually until less than one-third of the available soil water is left (Fischer and Turner, 1978). These authors suggest that adjustments to achieve this may include rapid root growth into wetter soil zones (Ritchie, 1973) or increased root density ( L J . An ability for rapid lateral root initiation and for deep root elongation would thus be advantageous. The depth-of-rooting argument is most convincingly demonstrated in soils of low water storage capacity [high K ( 0 ) values], in which water redistribution may well occur faster than root extension, while rainfall continues. Species which continue to grow new subsoil roots after flowering, or have deeper root systems anyway, may have up to an additional 0.5-1.5 m of soil water storage to tap, which may be equivalent to 100-200 mm of additional water during high evaporation periods. Hamblin and Hamblin (1985) compared the maximum rooting depths of forage and grain legumes in a Mediterranean climate on three sandy soils in the same latitude which differed markedly in growing-season rainfall. They found no significant differences between maximum root depth within species between sites, but highly significant differences between species


131

at each site. Lupin roots grew to 2 m, whereas pea and clover roots only grew to an average depth of 0.7 m. Differences in rooting depth have also been found between varieties. Kaspar et al. (1978) concluded from a review of soybean data that new root extension continued to occur in soybeans as late as the seed-fill stage, and that this frequently benefited seed yield in drier than average seasons, despite the additional root sink for assimilates. They found differences of 0.3 to 0.4 m in maximum root depth between seven commercial cultivars over the flowering to pod-fill stages. With the advent of newer high-yielding wheat varieties, which incorporate dwarfing genes, concern has been expressed that root systems of these varieties might also be shortened, with undesirable consequences when they are grown in water- or nutrient-stressed environments. An early comparison of European semidwarf and taller-stemmed winter wheat varieties was carried out by Lupton et al. (1974). They concluded that there was no significant difference in root density ( L J ,total root length per unit area (La), or dry root weight at any stage between the semidwarf and tall varieties. Cholick et al. (1977) made a similar comparison with varieties selected for American rain-fed conditions and drew similar conclusions for a wet year. However, in a drier-than-average year there was a significant difference in soil-water depletion curves in favor of a semidwarf variety, though in this instance all the varieties grew roots to depths of 3.0 m. Unfortunately, detailed leaf- or root-water data were not given to allow a more complete interpretation of reasons for this difference. Recent studies in upland rice have highlighted some of the problems in the correct identification and selection of “deep-rooting’’ characteristics. Jones et al. (1979) reported that semidwarf upland rice varieties appeared less drought resistant when compared with older, taller upland varieties when grown in acid, coarse-textured Brazilian Latosols. The lowland rice parents from which the semidwarf varieties had been selected had significantly lower L, values, especially in the subsoils, when compared with the upland varieties. However, the Brazilian subsoils were all very much more acid than the Asian soils from which the lowland strains originated, and sensitivity to aluminium toxicity was a confounding feature. Subsequent work by Mambani and La1 (1983a,b) has confirmed the relationship of drought resistance in upland rice to subsoil root density, using less acid environments. Even then, the question remains as to why some cultivars should exhibit more pronounced downward root elongation than others when their heights do not differ significantly. Downward extension is markedly affected by soil mechanical impedance, the soil water status, and the turgor pressure in the plant. Genotypic differences in the rate of root extension in compacted soils may provide the answer to this question.

132

ANN P. HAMBLIN

B. ENVIRONMENTAL INFLUENCES 1. Mechanical Impedance

Mechanical impedance is probably the most ubiquitous of physical constraints to the unhindered growth of crop roots through soil. Soil compaction restricts root growth, particularly in weakly pedal, silty, and sandy soils, as a result of tillage and traffic (Barnes et al., 1971). Additionally, many soil types contain indurated horizons such as fragipans, which are naturally resistant to root penetration. The pressure applied by growing root tips to the soil is approximately radial and was shown by Greacen et al. (1968) to be most conveniently measured by a cone-tipped cylindrical rod or probe. The root elongation rate has an exponential function with penetrometer resistance in a given soil. As bulk density decreases in any one soil, a suite of functional curves develop further and further from the intercept (Eavis, 1972). The rate of root elongation also decreases with increasing penetrometer resistance as soil temperature and aeration decrease (Greacen et al., 1968). Soil mechanical strength is dependent upon water content (or water potential). The sheer strength of soil (z) with normal loading (a,) depends upon the cohesive (coulombic) forces (C) and the angle of internal friction

($1: T =

C

+ 6, tan

(10)

Coulombic forces are dependent on absorbed water layers and surface area; therefore C increases with clay content and clay surface area and decreases at high values of 8. The angle of internal friction depends on the angularity of particles and their packing arrangement, as well as on 8. At high values of 8, primary particles are lubricated and slide past each other. Some soils develop secondary cementation as they dry, through the evaporation of soluble salts and the dehydration of hydrous metal oxides, which form films around primary and aggregated particles. Plinthite and fragipan layers, which are soft when wet and very hard when dry, are examples of this. Roots cannot normally grow into rigid pores narrower than their own diameters (Wiersum, 1958). When such pores are encountered roots may be able to exert sufficient pressure to expand the pores; otherwise they will be deflected. The root tip is then buckled, often with expansion of the vertical cross-sectional diameter (Camp and Lund, 1964) and a proliferation of root hairs behind the tip. Considerable vertical pressure can be exerted by root tips. Pfeffer’s (1893) experimental values of 10 MPa have been verified by many more recent workers. Radial pressures are normally only half or less the axial pressure, but they act over a much larger surface area of soil. Root-


133

created pores, which are conspicuous in many otherwise massive clay subsoil peds, have been formed by such pressures. Nevertheless, Russell and Goss (1974) and Abdalla et al. (1969) measured reductions of 20% in the root elongation rate at applied pressures as low as 5 kPa, and 80 % reductions at 50 kPa, the response curve being exponential. These are much greater reductions than would be obtained with equivalent osmotic pressures in the soil solution. However, their experimental system, of a flexible-membrane pressure cell filled with ballatoni beads, may have given rise to larger pressures at the root tip (through arching of the bead bed) than measured at the external wall of the membrane. The problem of accurate measurement of the forces encountered by roots and the relative strength measured by metal probes (penetrometers) has been extensively treated by Barley and Graecen (1967) and Graecen et al. (1968). Critical values at which root growth is inhibited range from 1 to > 4 MPa penetrometer resistance, depending on the soil's mechanical composition, the pore water pressure (total potential), and the plant species. Ehlers et al. (1983) also pointed out that soil structural differences down the profile can lead to differences in critical penetrometer values. They found that a value of 3.6 MPa inhibited oat root growth in the tilled Ap horizon, but that the critical value rose to 5 MPa in the untilled subsoil because a continuous pore system had developed in these horizons, from old root channels and worm holes, which roots could penetrate but which could not be sensed by the penetrometer. The effects of mechanical impedance on root and shoot growth differ under field conditions from those observed in controlled environments. Most laboratory experiments must perforce be constrained by scale considerations to observations on the seedling or the young vegetative stages of plants, yet in the case of crop-water relations the influence of the size of the root system is most apparent when leaf areas are greater than 1 m2/m2 and the soil water uptake demand exceeds 2 or 3 mm/day. In addition, controlled environments cannot normally mimic thermal or energy gradients in the soil, plant canopy, and air-canopy interface. Thus one or a few plants growing in a container may have a very different distribution of pressure potentials, a different water flux regime, and soil temperatures with larger diurnal fluctuations than in the field. This may produce quite abberant growth rates and very different root-shoot ratios from those of field-grown crops. Finally, the root tip and meristematic region are the site of hormone production and maximum nutrient uptake; complex feedback interactions occur with the shoots when the environment of the root apices is varied, the outcome of which is not yet physiologically predictable. Greacen (1977) emphasized the distinction between roots growing through sands and fine-structured soils, where pores are of smaller dimensions than those of the roots, and root growth through coarse-structured, fine-textured

134

ANN P. HAMBLIN 5.0 4.5 16

4.0

14

3.5

-

-

12

3.0

-

5

10

2.5

E

a,

.ul

5m C

2 ;j

m + m m

8

2.0

g

6

1.5

4

1.o

2

0.5

7

n 5

em

0

I

1

1

I

I

I

0.1

0.2

0.3

0.4

0.5

0.6

2

0.7

Fractional Soil Volume Intercepted by Root Cylinder

FIG.8. The interaction of aggregate size (circles) and aggregate strength (diamonds) on the o), fractional soil volume intercepted by the cylinder (root plus root hair cylinder) for wheat (0 and peas (a+).(Constructed from data in Dexter, 1978.)

soils, where roots tend to be restricted to regions of lowest resistance (major crack planes), despite the reduction which this imposes on lateral branching. Dexter (1978) developed a model to predict the probability of a root either entering or being deflected by an aggregate, depending on its strength and size. The disposition of aggregates was defined in terms of the scales encountered in a tilled topsoil. Although Dexter tested his model for nutrient rather than water uptake, the results are relevant in the present context. With the soil water potential held constant, the degree of root branching was determined by aggregate size. The optimum ped structure depended on the strength of aggregates and the plant species. Figure 8 is constructed from Dexter’s data and shows the effects of aggregate size and strength on the computed fractional soil volume which is intercepted by the root-plus-roothair cylinders for some crop species. Because increasing soil strength removed the effect of aggregate size, Dexter hypothesized that greater root branching occurred since the beds of aggregates were composed of progressively smaller size. His model is stochastic and is based upon the statistical probability of each root having either a penetrating or deflecting angle of incidence to each aggregate as it approaches it, in a statistically generated arrangement of solids and voids.


135

The effect of. ped orientation on root penetration was considered by Gardner and Danielson (1964). They found fewer root apices penetrated a wax slab set in the soil the further the slab was tilted from the horizontal. Whiteley and Dexter (1983) have extended this work to assess the influence of ped surface incidence-angle and critical ped strength on the rate of root elongation in several crop species growing in cracking clays. They found that where roots were forced to grow in cracks at very different angles to the preferred geotropic direction, extension rates were reduced to a half or a third of the unrestricted rate, even when cracks were wide, soil water potentials high, and soil strength less than 2 MPa. However, when cracks were oriented at 45" to 90" from the horizontal, elongation rates were consistently higher, irrespective of crack width, than for roots growing through soil peds. Even at the high water potentials used (- 1 to 20 kPa) ped strength was a significant barrier to penetration. This approach breaks new ground in quantifying the interactions between soil structure and root growth. However, it still relies heavily on relationships established in controlled environments with seedlings. Many field studies have been made on the effect of traffic pans on rootgrowth parameters and the effect of subsoiling on crop yields. Loosening of compacted layers often increases some aspect of root and (or) shoot growth significantly, but the effect on yield is more variable. Russell (1956), for example, found 53 % of subsoiled clay sites in England gave yield increases, 6 % gave negative yield responses, and 3 1 % showed no difference. On sandy soils the figures were 32,26, and 42 %, respectively. In the drier environment of Southwest Western Australia, on the other hand, Jarvis (1983) found positive responses to ripping of traffic pans on sandy soils in 88 % of cases, but in only 10% of clay-loam sites. Such simple yield data do not provide sufficient clues as to why the soil loosening was or was not effective. Subsoils may have very different nutrient or pH reactions from topsoils, and subsoiling may invert soil layers, diluting nutrient accumulation at the surface. Waterlogging is a problem frequently related to mechanical impedance in clay soils, and drainage may be improved by subsoiling, whereas substantial loosening of sandy soil may cause seeding machinery to sink into soft surfaces and sow crops too deeply. There are enough pointers in the literature, however, to show that yield reduction in the presence of compacted soil layers is associated with either or both water and aeration stress (Unger, 1979: M.A.F.F., 1975). Fine sands and loamy sands, particularly those composed of bimodally distributed sand sizes (Bodmin and Constantin, 1965; Coughlan et al., 1978) are both readily compacted and also have low water and nutrient capacities. For example, in the Mississippi River basin extensive coarsetextured alluvial soils occur, which develop severely restrictive traffic pans with bulk densities of 1.75-2.0 ton/m3. Restriction to penetration frequently

136

A N N P. HAMBLIN

inhibits cotton and soybean roots to a total soil depth of only 0.25 m of soil in which there may be, at best, only 30 mm available water-barely enough to supply a leafy crop for a week (Camel, 1980). In other cases impedance is not so severe, but still reduces growth significantly. Rowse and Stone (1981) found deep loosening of a sandy clay loam increased the rate of root elongation and gave significantly greater root lengths in the layers beneath the loosened top one-half meter of the profile compared with the same depths in shallow-ploughed treatments. More water was extracted from the deeper parts of the profile and less from the top 0.3 m in the loosened treatments. The authors argued that the pattern of root distribution in the loosened soil led to lower total plant resistance to water flow and hence higher uptake and transpiration rates, but the data were circumstantial. In the experiments previously referred to in Fig. 7, where root densities were compared, the amounts of water taken up from the traffic pan layers of the loessial German soil by oats and the Australian loamy sand by wheat were smaller than in the adjacent horizons. Yet, in the case of the oats, the root density in the 0.2-0.3-m zone was not significantly different between two tillage treatments, despite significant differences in bulk density, unsaturated hydraulic conductivity, and penetrometer resistance. Ehlers et al. (1981b) noted, however, that roots from this layer were much distorted and thickened and suggested they might exert a far higher radial resistance to water uptake. Other physiological alterations would have accompanied such anatomical differences; however, they were not considered in their discussion. 2. Interactions between Mechanical impedance, Water Status, and Aeration The interaction of soil strength, bulk density, and water content and their relations to root impedance have been treated in many texts, but nowhere better than in the review by Barley and Greacen (1967). The relationship between penetrometer resistance, water potential, and soil pore space is not so well defined in many field soils as in the idealized description in Section V,B,l. Soils with high stone or gravel contents give spuriously high values for penetrometer resistance, while vertically or areally heterogeneous soils have high CVs, which may simply reflect differences in water content between adjacent layers or peds. Normal averaging procedures of penetrometer values obtained from such soils mask small-scale variations in soil strength and water status which markedly influence root growth distribution. McIntyre and Tanner (1959) and Hewitt and Dexter (1983) have reported positively skewed penetrometer resistance distributions. Figure 9 shows variations which can occur in the depth of penetration into a columnar-structured xeric Alfisol (red-brown earth) taken at 0.1 m increments in a rectangular grid after 50mm of rain had fallen on a previously dry soil. There is a difference of


137

crn

1 7 < 6 c m 18-24cm

m 6 - 1 2 c m 24-30cm

u 1 2 - 1 8 c m -30-36crn

FIG.9. Variations in depth of penetration into a newly wetted, previously dry red-brown earth (Rhodoxeralf) across a 1.8 x 0.6 rn grid. (From Hamblin, 1984.)

0.3 m in the depth to wetting-front, at which point the penetrometer’s resistance exceeded 4.5 MPa (Hamblin, 1984). When roots grow against materials which offer mechanical resistance, the area of contact between roots and solid-plus-liquid phases increases. The proportion of root surface exposed to gaseous oxygen is thus reduced. At the same time, there is an increase (albeit small in proportion to the total respiration rate) in the oxygen demand (Greenwood, 1968). In situations where soils are already very wet, this reduction in air-filled pore space may be critical to the oxygen status of the root, particularly if the root compresses the soil as it passes through. I suggest that this may occur frequently in wet clay soils, which have small median pore sizes. Because the diffusivity of 0, in air is some lo4 times greater than in water, the existence of a few continuous gas-filled pores to the soil surface may suffice to provide sufficient aeration when root and soil respiration rates are low (as in winter conditions in Europe, when temperatures are low and plants are relatively small). However, the critical distance for O2 diffusion is seldom the length of that continuous pathway to the soil surface; rather, it is the distance through a water-filled soil volume from a root surface to the closest aerobic pore cylinder. Greenwood (1968) has discussed the influence of pore size on aeration, an&pointed out that for equivalent volumes of air-filled pore space, oxygen diffusion into anaerobic peds occurs more rapidly the higher the proportion of small pores, since the average diffusion path length across the gas-liquid interface is less. He computed that a soil system composed of m would require gas channels where the radius of the channels rc = 5 x a gas space ratio (x) of 0.2 to maintain an oxygen uptake rate of 1.3 x lo-* s-l, whereas a soil of rc = 1 x m would only need an x ratio of 0.05 to maintain the same rate. In partially saturated soils (say, I,$, = - 1 kPa) the actual supply of oxygen to the roots is not well described by the bulk air-filled pore space (x). Nor is

138

ANN P. HAMBLIN

C rn

Critical Value for Root Growth - 1- -1 -1 -1 -1 -1 -1 -1 -1 I 1 1 1 1 1

W

>

6

0

4

8

12

16

20

24

28

% Pores > 30 urn Diameter

FIG.10. Oxygen diffusion rate for tropical Alfisols (squares) and Ultisols (circles) at 10 kPa (0.) and 50 kPa (00). (Constructed from data presented by Pla, 1978.)

the average oxygen concentration of the liquid phase very much more helpful. Primary roots may be occupying larger gas-filled pores, while laterals with smaller diameters are growing in pores which remain filled with water and drain orders of magnituide more slowly. Hence the much quoted value of x = 10 % necessary for unimpeded root growth is an oversimplification and probably derives from the coincidence of the zero value for the diffusion ratio DID, (where D is the self-diffusion rate of 0, in the porous medium and Do the unimpeded rate in air) with 10% air porosity in many soils tested in the laboratory (Currie, 1962). Data from Pla (1978) [quoted by La1 (1980)l shows the relationship between macroporosity, bulk density, and 0, diffusion rates in compacted tropical Alfisols and Ultisols (Fig. 10). Some soils were compacted to greater than 1.75 ton/m3 and yet still apparently retained adequate 0, diffusion rates for root growth (see Fig. lo), since oxygen diffusion rates of less than 3 x pg/m2 s are taken as the limiting value for restricting root growth. However, Armstrong (1980) has demonstrated that 0, flux is a more accurate measure of root 0, demand than is concentration. Currie (1984) attempted to relate the D/D,ratio to x in a silty clay loam, wetted and compacted to different porosities, but found no general relationship even for a single soil. It is probable that no such relationship can be found given the variations in pore shape, tortuosity, degree of swelling, and hysteresis which may occur as soils wet and drain. Nevertheless, it is instructive to observe the general shape of the curves which enabled Currie to identify the transition between inter- and intraaggregate gas diffusion (Fig. 11). As samples wetted and became swollen, the relationship between x and D/Do followed an exponential form (DID,= axb), but as the soil drained and the inter-aggregate pores became gas filled, the relationship was fitted by a fourth-order polynomial of x. Thus in a sequence

139


0.1

0.2

0.3

0.4

0.5

0.6

0.7

Fractional Air Content E (m3 m-3 1

FIG.11. The change in oxygen diffusion coefficient DID, (the self-diffusion of 0, in the porous medium over the self-diffusion in air) which occurs when intercrumb pores wet (curves 0 to 5) and drain (curves 6 to 11) at decreasing bulk densities. (From Currie 1984.)

of compactive states (where bulk density varied from 0.86 to 1.29 ton/m3), the transition between inter- and intra-aggregate 0, diffusion could be identified at the intersection of the two curves. This inflection was less abrupt in samples which had been compacted to lower total porosities. In nonswelling soils, - 5 kPa soil water potential adequately described the transition point, but swelling soils still had inter-aggregate pores which were imperfectly drained at -5 kPa. Spatial variation of aerobic-anaerobic zones in soils has not received the same attention from soil physicists as have other soil properties, but mottling zones in soils of fluctuating water tables provide us with some indication of their scale dimensions. An instructive study by Veneman et al. (1976) related the occurrence of different scales of mottling features to the duration of total waterlogging and periods of high water potential in a Wisconsin toposequence. They were able to distinguish three distinct mottling patterns which corresponded to varying periods of saturation. Spatial variation in anaerobiosis, with its associated differences in duration of waterlogging, would thus lead us to expect nonuniform root distributions in many clay soils, in compacted layers of arable soils, and in soils at the base of slopes or low-lying parts of the landscape. Veneman et d ’ s use of detailed morphological criteria provide a more rapid method of assessment than is possible using direct gas or water measurements, and, I believe, they should be more widely adopted. Selection of plants adapted for waterlogged conditions and the study of plant response to oxygen deficiency caused by waterlogging has been reviewed by Krizek (1982). Roots of nonhydrophylic species growing in anaerobic environments may have adaptive mechanisms such as adventitious

140

ANN P. HAMBLIN

roots at the base of the stem and aerenchyma (air spaces) in the root cortex (Drew and Lynch, 1980); they may also have fewer laterals and more secondary root per unit of root length. A close correletion between tolerance to waterlogging and aerenchyma development has been implied for many crops. However, the root morphology of many genotypes only adapts to waterlogging if it develops slowly over many days, whereas waterlogging damage often occurs during quite transient events. Tolerance also depends upon the developmental stage [as, for example, emergence and flowering, which are both particularly sensitive (Cannell et al., 1979)], the postwaterlogging stress environment, and the specific oxygen demands of the crop at the time. Of special significance is the increased respiration requirement of both roots and soil microflora during periods of rising or high temperatures. As a result, the selection environment and imposed waterlogging stress substantially influence the relative response of different cultivars, and few crop plants have been successfully selected which show greater tolerance to a range of waterlogging events, despite good correlation between both adventitious root production, aerenchyma, and tolerance. 3. Predictions of Root Distribution in Space and Time

Many models for water uptake and evapotranspiration are now available at several orders of complexity. All require terms to describe the root sink function, and in the great majority L,, the density of roots per unit volume of soil, is the sink term. As increasingly large time steps are considered in a model, the soil volume is often partitioned one-dimensionally into n layers and La (the total root length per unit of ground area) is distributed proportionally through the layers in a time sequence which follows an exponential or power function. All such models assume that roots are uniformly distributed in each soil volume, which is, by definition, homogeneous. I think this is a questionable assumption, when so many field data show us the predominance of markedly nonuniform root patterns. It may well be that the spatial heterogeneity of roots is of less consequence to plant growth than the spatial heterogeneity of hydraulic parameters is to water cycling, but the question should certainly be addressed. Patterns of root distribution plotted two-dimensionally by pit-wall techniques (Bohm, 1977) illustrate the effect of plant density and row spacing on L, in relation to depth. Wide-spaced plants have root distributions which most closely approximate to theoretical models such as those proposed by Lungley (1973) and Hackett and Rose (1972). These models predict branching rates, lengths of axes and laterals, and spatial distributions in relation to a regular chronosequence. Higher-density crops which compete for water early in their life cycles may “sample” different scales of minimum REV. Taylor (1980) and Mason et al. (1982) compared L,s for soybean grown on the same

141

THE INFLUENCE OF SOIL STRUCTURE Table IV

Two-Dimensional Distributions of Root Lengths in Soybean Isolines Grown on a Sharpbury Silty Clay Loam” Isoline Nonpubescent

Depth (m)

Ratio of Ratio of Half-way “in” to Half-way “in” to In row between row “between” row In row between row “between” row

At 47 days 0-0.15 0.15-0.30 0.30-0.60 0.60-0.90

1.57 0.68 0.82 0.20

0.59 0.56 0.12 0

La

7.4

2.1

0.85 0.71 1.51 2.01 0.76

0.78 0.76 1.13 1.11 0.15

(ZO)

Pubescent

2.7 1.2 6.8 0

1.25 0.39 0.68 0.1

0.50 0.20 0.19 0

4.8

1.6

1.82 1.27 2.85 3.33 1.90 34.5

1.12 1.82 3.12 4.33 2.01 36.2

2.5 1.9 3.6 0

At 78 days

0-0.30 0.30-0.60 0.60-0.90 0.90- 1.20 1.20-1.50 La

(ESO)

17.5

13.6

1.1 0.9 1.3 1.8 5.1

1.6 0.7 0.9 0.8 0.9

a Root lengths measured in lo3 m/m3. Species: Glycine max (L.) Merr. La measured in m/m2 x 10’. Adapted from Garay and Wilhelm (1983).

loess soil used by Bohm, at 0.25 and 1.0 m spacings, while maintaining the same plant densities. They found narrow row spacing increased L, in most soil layers but z,,, (the maximum depth of roots) was 12.5 % greater in the wide-spaced crop. Their data, like that of Garay and Wilhelm (1983) presented in Table IV, showed a number of points which deviated from idealized root-density distributions. The data in Table IV compare root distributions for two soybean isolines grown at 0.75-m row spacing at 47 days (when the root systems were still rapidly expanding) and at 78 days (when seed set had just begun). The soil, a silty clay loam (Typic Arqiudoll), had a well-developed ped structure and showed considerable swelling. Deviations from a monotonically decreasing distribution included L, increasing with depth in the more mature plants, higher L, values between than within rows in the pubescent line at maturity, and a decrease in L, in the top soil of the nonpubescent line at maturity. The two isolines showed considerable difference in root extension rate and La at maturity, despite little difference in their top growth. Figure 12 shows the distribution of irrigated soybean roots in three replicates of a Waukegan loam (Typic Hapludoll) at

142

ANN P. HAMBLIN

a

Distance t o Nearest Plant Row (m)

n

1-2

a

2

> 2.5

- 2.5

Root Density (rn3rW3x l o 4 )

b

Distance t o Nearest Plant Row

0

0.2

0.4

0.2

0

FIG.12. (a) Three replicates of soybean root-length density (L,) at 91 days, grown in a typic Hapludoll (Waukegan loam) with irrigation. (Redrawn from Arya et al. 1975.) (b) Actual and interpolated (---) flow patterns of water in the root zone of 5- and 10-week-old soybean plants, each at 4 days after an irrigation cycle. (Redrawn from Arya et al., 1975.)

maturity (Arya et al., 1975). Root densities ranged from 5.24 to 0.7 x lo4 m - 2 even at the same depth, and patterns of distribution show marked clumping. Nevertheless, the authors used an arithmetic averaging procedure to obtain values of L, for root-water uptake modeling. Curiously, they concluded that root density per se was not determining the water extraction rate despite marked observed deviations from the assumed equidistant flow patterns used to compute the fluxes, which are given in Fig. 12b. For far too long values of soil physical properties were assumed normally distributed, but, as we have seen, a substantial body of evidence now exists to demonstrate that lognormal, or even more skewed distributions, are common not only for hydraulic properties but also for soil strength parameters (Section V). It is very probable that root density, responding to spatial variation in these soil properties, is itself nonnormally distributed. Tests for


143

nonnormality. allow the data to be transformed for more meaningful estimates of means and variances. In the case of spatial distribution, these can be based on nearest-neighbor techniques and provide a basis for indices of dispersion. The theory is discussed in texts on quantitative ecology [e.g., Greig-Smith (1964)l. Barley (1970) and Baldwin et al. (1971, 1972) have used this approach to examine the influence of spatial distributions of roots on nutrient uptake. Barley’s treatment utilized the areal variance of continuous polygons centered about each root, and his conclusions were that a difference in uptake rate for nutrient ions would not be greater than 5 % between regular and the mosty irregular distributions. Baldwin et al., however, found a very large difference ( > 75 %) in uptake rates over similar time intervals between regular and clumped distributions. Their treatment was rather simpler, but is of particular interest to field scientists using the spatial mapping techniques of root measurement. They proposed three categories of distribution defined by the mean, variance, and mean nearest-neighbor distance of the root density. So far I have ignored the question of which plane of orientation should be taken for root sampling. Because of the marked geotropism exhibited by most root systems, it seems logical to intersect the vertical plane at right angles, plotting distributions longitudinally on a vertical axis. However, we note that marked variation also occurs in distribution density ‘within and between rows (Table IV). This requires sampling in the vertical plane at right angles to rows. Representative sampling should combine both planes, for example, vertical cores taken along appropriately spaced transects across rows. When root numbers are counted, they may be converted to root length by formulas such as that of Melhuish and Lang (1968) for random root distributions or that of Lang and Melhuish (1970) for strongly anisotropic distributions. Passioura (1985) has considered the problem of marked anisotropy of root distribution for water uptake from roots confined to cracks in polygonally cracking soils and roots confined to wormholes in fine-textured soils without major crack patterns. Where roots were confined to cracks in parallel slabs, hexagonal prisms or “spheres” of radii >0.15 m the time constant (the time required to withdraw water to a lower exhaustion value) increased over 20 times the value calculated for uniformily distributed roots, irrespective of L,, except where L, was very small. Such calculations, like those of Baldwin et al., serve to remind us that field soils may impose the rate-limiting step on water uptake at an earlier stage, or at higher plant water potentials, than we anticipate from experiments conducted in controlled environments and homogenized soils. It would also seem that there is a gross oversimplification in current assumptions that root distributions may be arithmetically averaged over

144

ANN P. HAMBLIN

equidistant horizontal layers of the soil depth for the purposes of water uptake studies. Much of the confusion and conflicting results on the location and magnitude of resistance to flow may, I think, stem from such relatively simple considerations involving inconsistencies in the spatial scales and averaging procedures used to measure and separate root, soil, and water properties.

VI. WATER UPTAKE BY ROOTS Many uptake models have developed from the single-root approach (e.g., Gardner, 1964). This is most simply states as

Q = ($s- $p)/(Rs + RP)

(11)

where Q, the flux in cubic meters per second, is equal to rl/, and $p, the soil and plant potentials (in meters), divided by R , and R , (in seconds per square meter), their equivalent resistances to hydraulic conductivity. The approach treats water uptake as radial into a centrally located root with a pressure gradient that is positive upward. Later developments, as, for example, Cowan (1965), solved Eq. (1 1) quasi-analytically for equidistant arrays of root cylinders, each bounded by an equivalent cylinder of soil of radius r : Q = dO/dt

1 r

= - d/dr(r DO

dO/dr)

DO is the desorption water diffusivity, so in this instance it must be obtained by pressure-outflow methods. The assumptions in the equation of uniform root distribution and uniform uptake along the length of each cylinder are not always valid, as we have seen, in the case of field crops. However, direct measurement of root potentials or resistance deep in the soil profile are not normally possible, and the allocation of the resistances in the flow pathway is inevitably a matter of approximation and guesswork. I shall only briefly summarize current thinking on resistances to water flow through roots, as several excellent reviews have dealt with its various aspects in recent years (Tinker, 1976; Landsberg and Fowkes, 1978; Taylor and Klepper, 1978; Molz, 1981). For convenience, the total resistance to water flow is often divided up into a number of steps: from the bulk soil to the soil near the root surface (pararhizal zone); across the soil-root interface (perirhizal zone); across the cortex to the xylem (radial); up the root xylem to the stem (axial); and up the shoot xylem to the leaf surface. Such convenient but arbitrary distinctions may provide mental straightjackets to the complexity of real-world processes but are a necessary first step. Taylor and Klepper (1978) have assessed much of the evidence for the relative magnitudes of axial


145

and radial resistances for intact and field-grown roots. Axial resistance R, is the fractional resistance to flow in the xylem. It is often calculated from the Poiseuille-Hagen equation: Q

= rtr4/8q AL = Al///R,

(13)

where Q is the flux (in cubic meters per second), r is the radius of the xylem in meters, q is the viscosity of xylem sap (in meter. seconds), AL is the length of the root (in meters), and A+ is the decrease in hydraulic potential in the direction of flow. Axial resistance is therefore likely to be greatest in narrow xylem vessels, such as those of graminaceous roots, and least in widediameter xylem vessels, such as those of tap-rooted species, where secondary xylem tissue greatly increases total xylem volume. Axial resistance can also increase with the age of the root and the depth of the roots in the soil (Ponsana, 1975). The increase in secondary tissue xylem of tap-rooted (dicotyledoneous) species may partially offset this, however. Taylor and Klepper (1975) found cotton had negligible xylem resistance, whereas Willatt and Taylor (1978) found substantial axial resistance in soybeans as they matured. Passioura (1983) has pointed out that either adverse or beneficial effects may accrue where large axial resistance occurs in the seminal roots of wheat, depending on the water regime in the soil and the climatic environment. Cornish (1981) has demonstrated that deep sowing of cereals can produce a similar increase in axial resistance as occurs with the seminal roots through lengthening of the coleoptile internode. Table V summarizes some published data on the range and maximum rates of specific root water uptake U , for different species (in cubic meters per meter per day). The data are either taken from field experiments or from rhizotron environments where the areal portions of the plant were exposed to the ambient atmosphere. Maximum rates of uptake of 10-l’ m’/s are common for cereals, but grain legumes seem to have higher uptake rates when grown in comparable or the same environments. When we recall that soil-water diffusivities are seldom less than m2/s even when the soil is at low matric potential, a perirhizal resistance will occur only if a continuous water film does not exist at the root-soil interface or if only a fraction of the root is taking up water in that soil volume. The increase in total resistance commonly reported as a plant grows may reflect, in part, the increasing proportion of the root system which is suberized and nonabsorbing (despite the fact that La also increases with plant age). Such compensatory (or sometimes synergistic) effects of wholeplant growth may explain some of the conflicting evidence in the literature on soil-plant resistances. Similarly, plant resistances which are calculated from averaged or random measurements of leaf water potential taken during the mid-day period do not represent the full non-steady-state conditions of the

146

ANN P. HAMBLIN

Table V Specific Water Uptake Rates for Various Species in Temperate and Subtropical Environments’ Species

Reference

Temperate grasses Oats Winter wheat Spring wheat

Welbank et a/. (1974) Ehlers eta/. (1981a) Gregory et al. (1978a) Greacen and Hignett (1976) H a m b h and Perry (unpubl.) Allmaras et al. (1975) Taylor and Klepper (1973) Willatt and Taylor (1978) Willatt and Olsson (1982) Allmaras et a/. (1975) Hamblin and Perry (unpubl.) Hamblin and Perry (unpubl.)

Maize Soybean

Field pea Lupin Cotton

BarYosefandLambert(1981) Taylor and Klepper (1975)

Water uptake rates

(I, are

Water uptake rate t x 10-l’ (average)

8 x 10-12-1x lo-” 2 x 10-13-2x 10-12 1 x 10-”-2 x lo-’’ 4 x 10-l3-lx lo-’’ 3 x 10-12-2.4 x lo-’’ 1 x 10-”-2.4 x lo-” 2.8 x 10-”-2.2 x lo-’’ 2 x 10-”-9 x lo-’’ 2 x 10-”-3 x lo-’’ 1 x 10-”-2 x 10-l’ 3 x 10-“-2 x 5 x 10-12-5x lo-”

5

200

I

P e

w

9 .g C

100

a

2

4

6

8

10

12

14

PH FIG. 2. Adsorption of a range of anions on goethite (redrawn from Hingston et al. 1972). Two samples of goethite were used and the level of addition of adsorbing ion differed between the different ions.

values of pH. This interpolation procedure has frequently been used in place of the technically more difficult method of keeping the pH constant. It has been widely used in studies on anion adsorption, perhaps because analogous adsorption-concentration curves are common in studies on soil. When a hydrolyzable metal reacts with an oxide, there is a decrease in pH. This is usually recorded as a release of protons. The effect could equally be produced by coadsorption of O H - ions, but the conventional terminology will be used here. Some representative values are tabulated by Schindler (1981). The observed values usually lie between 1 and 2 mole of protons per mole of metal adsorbed. These measurements are usually done in a nitrate or perchlorate medium in order to avoid complexities due to ion pairs with chloride, and hence the average charge on the ions in solution is close to 2. The difference between the protons released and the average charge on the ions in solution is the charge conveyed to the surface. Thus, we may describe the same phenomenon by either specifying the protons released or the charge conveyed to the surface. For hydrolyzable metals it is more common to specify protons released; for anions it is more common to specify the charge conveyed to the surface than to specify hydroxyls released. Whichever method of specification is used it is common to specify one number as if it were a characteristic of the adsorbate-adsorbent system. This is not so. For example, Schindler (1981) noted that the value for protons released tends to be higher at low pH; Hingston et al. (1972) showed that the charge conveyed by selenite to goethite depends on the ionic strength of the medium; Bowden

190

N. J. BARROW

et al. (1980b) showed that the charge conveyed by phosphate to goethite is high at low pH, passes through a minimum, and is high again at high pH; and Bolan and Barrow (1984) showed that the charge conveyed during adsorption varies in a complex way with the amount of adsorption. From the observations that reaction changes the charge, it also follows that the point of zero charge of the oxides will be changed by reaction either with hydrolyzable metals or with anions (for references, see Pyman et al., 1979). The reaction between adsorbate and adsorbent may also be followed in a third way. The electrophoretic mobility may be measured and, from this, the electrostatic potential in the slip plane or the zeta potential calculated. However the calculations of potentials from the observed mobility is difficult, especially when the particles are not spherical and, indeed, van Olphen (1977) recommended that results of such measurements be reported as mobilities rather than be converted to potentials.

B. MODELSUSED TO DESCRIBE ADSORPTION ON SURFACE

A VARIABLE-CHARGE

In this section the three models that have been used to describe adsorption will be compared. The models are similar in that, in all three, the complex distribution of ions near the surface is simplified and particular ions are assigned to mean planes of adsorption. The models differ in the number of No. of layers Layer

s

1

4

3

S

s

a

B

d

Change in potential with distance

Ions adsorbed

etc.

FIG. 3. Diagram representing the interface between the solution and the surface on which metal cations, or anions, may be adsorbed. In each case the bulk of the adsorbing material is to the left and the solution to the right. The diagram indicates the mean planes to which individual classes of ions are allocated and shows the change in the electrostatic potential )I with distance.

REACTION OF ANIONS AND CATIONS

191

planes that are specified and in the allocation of ions (Fig. 3). They also differ in the way the surface reactions are specified. 1 . Single-Layer Model

In the simplest model, protons, hydroxide ions, and specifically adsorbed ions such as phosphate are all assigned to one mean plane of adsorption. No explicit account is taken of the background electrolyte and, in a given electrolyte, a single value for the capacitance is used to characterize the surface. This capacitance gives the proportionality between surface potential and surface charge. It is apparently because the value of the capacitance is independent of charge (Sigg and Stumm, 1980) that this model has been called the constant capacitance model. However, this independence is also incorporated in most other models. This terminology is also confusing on another ground. If the identity or the concentration of the electrolyte is changed, a different value of the capacitance is used (Westall and Hohl, 1980). Hence, in this sense it is not constant. In this article, this model will be called the single-layer model. This model was introduced by Schindler and Stumm (for references, see Westall and Hohl, 1980). It has been used to describe the adsorption of anions on goethite by Sigg and Stumm (1980) and the adsorption of phosphate on oxides by Goldberg and Sposito (1984). It is a characteristic of this model that a surface site is treated like a chemical which can dissociate to give the three species in eqn. (l), that is, SOH:, SOH', and SO-. Sites can also react with species in solution to give products such as SOH + H,PO,

< SHPO,I + H20

+ Hf

(4)

This is one of the three equations used by Goldberg and Sposito (1984) in their model of phosphate adsorption. The various surface products are considered to be mutually exclusive and hence, given the concentration of the reactants and appropriate reaction constants, the concentration of the various products can be calculated. A complication, however, is that the reaction takes place at a charged surface and is affected by the electrostatic potential of the surface. Hence, in the case of Eq. (4), the intrinsic equilibrium constant K is related to the reactants by

K exp(F$/RT)

=

[SHPO,][H+] CSOHICH ,PO41

(5)

where $ is the surface electrostatic potential, F is the Faraday constant, R the gas constant, and T the temperature. Goldberg and Sposito (1984) regard the exponential term as a solid-phase activity coefficient that corrects for the

192

N. J. BARROW

charge of a surface species. The value of the electrostatic potential cannot be measured. However, the potential can be treated as a component in computer programs designed to calculate the species present as a result of chemical equilibria and the system of equations can be solved. The single-layer model was used to describe adsorption of fluoride, sulfate, silicate, and phosphate on goethite by Sigg and Stumm (1980). For fluoride the model described both adsorption and the charge on the surface. It is perhaps not surprising that the small fluoride ion could be considered as occupying the same adsorption plane as hydroxide ion. For sulfate, however, the fit of the model was poor. This could be because, for sulfate, the assumption of a common plane was inappropriate. For both silicate and phosphate it is difficult to judge the model, because many of the results seemed to involve removal of most of the ions from solution, that is, adsorption was mostly determined by the initial concentration of adsorbate. Although only a limited amount of data was presented by Goldberg and Sposito (1984), the single-plane model was able to describe the effects of pH and of phosphate concentration on phosphate adsorption fairly well. Four criteria may be suggested by which this, or any other model, may be judged. These are simplicity, comprehensiveness, precision, and correspondence with reality. The single-layer model is simple in concept. Provided it is restricted to one concentration of a given electrolyte, few parameters are needed to describe the charge components of the model. However, several parameters are needed to model adsorption of ions. Sigg and Stumm (1980) used five separate equations to describe reactions of phosphate with the surface, and hence five parameters were needed to represent the reaction constants of these equations. Goldberg and Sposito (1984) used three equations and hence three parameters. Therefore, in this respect the model is not especially simple. Further, the simplicity of the concepts may well be at the cost of comprehensiveness. Is the model able to provide a consistent description of the adsorption of a wide range of anions and cations and, at the same time, describe the change in charge produced by that adsorption? Current evidence suggests that it is not. In order to judge precision, the model needs to be compared with a wide range of data to test how precisely it can reproduce them. This has not been done. Correspondence with reality can never be complete, because all models must simplify the real situation. However, some aspects of the model can be compared with reality. For phosphate adsorption on iron oxides, two of the hydroxyl groups of the phosphate form links to the surface (Parfitt et al., 1976). The third hydroxyl group would not normally dissociate until high pH, but for adsorbed phosphate it appears to be half dissociated at about pH 6.7 (Nanzyo and Watanabe, 1982). Thus there are two surface species, both with two links to the surface. This does not correspond with the model of Goldberg and


193

Sposito (1984), which required three surface species, each with a single link to the surface.

2. Three-Layer Model This model differsfrom the single-layer model in that ions near the surface are allocated to three distinct planes. It was introduced by Yates et al. (1974), Davis and Leckie (1978), and Davis et al. (1978). There is an innermost layer containing adsorbed H + and OH- ions. These are responsible for the charge gSand experience the potential $ s . This layer is separated from a second layer that contains electrolyte ions (for example, N a + , Cl-) plus adsorbed ions (for example, Cu2', H2P0,). These ions are responsible for the charge ap and experience the potential $ p . A third layer contains the ions of the diffuse layer. The ions in the second layer are bound to surface groups, and equations for the reactions that are thought to be involved are then written. For example, for the electrolyte ions one might write SO-

+ Na'

SOH;

+ C1-

SO-Na'

(6)

S0H;Cl-

(7)

and +_

When the ions of adsorbates such as copper and phosphate are considered, more than one species of ion may be present in solution, for example, Cu2+ and CuOH'; H 2 P 0 , and HPOi-. They may also react with two sites to form binuclear bridging commands. Several equations might therefore be written for each adsorbed. However, the equations used have been analogous to the following: SOH + Cu2+

SO-Cu2+ + H +

(8)

and

+ CU" + H,O - 1

+ 2H+

(9) An intrinsic equilibrium constant is then defined for each reaction, and by rearranging the equation the concentration of product may be obtained. For example, from Eq. (8), we have SOH

[SOCU'] = K

SO-CUOH'

[SOH] [Cu2 '1 exp( - 2F$,/R T ) CH '1 exp( - F$s/RT)

where K is the intrinsic equilibrium constant. The exponential terms in Eq. (10) were introduced by assuming that the concentration of the reacting

194

N. J. BARROW

ions would be different at the surface from that in the bulk solution and that this difference would be described by the Bolzmann distribution. This approach has been criticized by Sposito (1983), who argued that it was both unnecessary and undesirable to involve the Boltzmann distribution. Rather, the exponential terms came from the effect of the electrical field on activity coefficients. As for the single-layer model, occupation of a site by one product is considered to exclude others. Hence, given the concentrations of the reactants and the reaction constants, the equilibrium concentrations can be calculated. Again this involves adaptation of computer programs designed for calculating speciation in solution to permit estimation of the terms for electrostatic potential. It is instructive to consider how this model attempts to match the charge conveyed to the surface and how it describes the effects of pH on adsorption. Reactions (8) and (9) convey, respectively, a charge of + 1 and a charge of zero to the surface and consequently release, respectively, one and two protons. Matching the observed release of protons therefore involves getting the right mixture of reactions (8) and (9) by choosing the appropriate values for the intrinsic equilibrium constants. In the three-layer model (and also in the single-layer model), the effect of pH can be thought of as operating through two mechanisms. One is via the mass-action effects of H + in Eqs. (8) and (9) [and in Eq. (4)]. This affects the total amount of adsorption-clearly the higher the pH, the more the reactions will go to the right and the higher the adsorption will be. The mass-action effects also affect the balance between Eqs. (8) and (9)-the lower the pH, the less Eq. (9) will be favored relative to Eq. (8). [We may also view Eqn. (9) as a hydrolysis of CuZf to CuOH' followed by adsorption of CuOH +. This hydrolysis reaction is favored by high pH.1 The other mechanism by which pH operates is through the exponential terms in Eq. (10) [and Eq. ( 5 ) ] . For example, in Eq. (10) a negative value for the potential t,bp would give rise to a large value for the exponential term and favor adsorption of the positively charged copper ions, as would be expected. However, the potential t,bs would also be negative, and so the effect would be partly offset by the effect of electrostatic potential on the Ht ions. The three-layer model was shown by Davis and Leckie (1978) to describe the effects of pH on adsorption of the following cations: lead on aluminium oxide, cadmium on titanium dioxide, and copper and silver on iron oxyhydroxide. The effect of concentration of adsorbate was not tested, nor was the change in charge. Davis and Leckie (1980) used the model to describe adsorption on iron hydroxide of one level of sulfate and four levels of selenate and the effects of ionic strength on chromate adsorption. The match to the chromate data was not good. James et al. (1980) used the model to describe

195


the effect of pH on adsorption on titanium oxide of one level of cadmium, the charge and zeta potentials in the absence of cadmium and the surface charge in the presence of one level of cadmium. The three-layer model is less simple and more comprehensive than the single-layer model. It requires at least two parameters for any given adsorbate in order to specify equations that are analogous to Eqs. (8) and (9). In addition, Davis and Leckie (1980) introduced a further parameter to permit an adsorbed anion to cover more than one surface site. Its precision has not, as yet, been adequately tested in a wide range of data. And, as the adsorption equations are specified in a similar way to those of the single-layer model, it is likely that the surface products do not correspond to those shown by direct examination of the surface. 3. Four-Layer Model a. Description of the Model. This model was developed by Bowden et al. (1980b) from a simpler model described by Bowden et al. (1977) and in earlier papers. It originates from the same “school” as earlier simple models proposed by Hingston. These earlier models are still widely quoted (Mott, 1981; Schindler, 1981). However, the four-layer model is now regarded (Hingston, 1981) as a more useful mechanistic explanation. It differs in two main ways from the three-layer model. First, and obviously, an extra layer is introduced and ions such as phosphate and copper are allocated to this layer. Although we speak of a four-layer model, the position of the extra layer is not fixed. It is envisaged that some adsorbed ions, for example, fluoride, would reside, on the average, closer to the surface than, say, sulfate ions. Ions are therefore postulated to differ not only in their affinity for the surface but in their mean position when adsorbed. Hence, if widely differing ions were present, we would have to specify further layers, for example, separate layers for fluoride, phosphate, and sulfate. The second difference from the other models is the way in which a surface site is envisaged and, as a consequence, the way the adsorption equations are derived. In this model a neutral site is considered to be OH

/ \

Me

OH2

Hence the hydrolysis reaction is written as /OH: Me \

OH 2

OH , I Me’ \

OH,

+ H+

OH-

’

Me ‘OH

+2H+

(11)

196

N. J. BARROW

This difference has important consequences when considering the reaction with ions. Suppose a neutral site were to react with a monovalent anion (A-), displacing an hydroxyl group and thus causing no change in charge: OH

A

+ A-

Me’

\

I Me’ + OH‘OH,

OH2

(12)

One of the remaining protons could then dissociate to give a negative site. Thus the charge conveyed to the surface could be somewhere between zero and one, depending on the proportion of the reacting sites from which a proton was lost. Furthermore, occupation of the sites is not mutually exclusive. The fact that a site has reacted with the anion does not prevent it from varying its charge by gaining or losing a proton. This is illustrated for phosphate in the equations given by White (1980). A different approach is therefore needed to specify the amount of adsorption produced. Consequently Bowden et al. (1977) did not write a set of reaction equations. Rather, they argued that the electrochemical potential of an ion could be defined by an expression that includes a term for the electrostatic potential it experiences plus a term for its chemical activity. This equation can then be arranged to relate activity to electrostatic potential. They then assumed that the surface activity is equal to the ratio of the occupied sites to the vacant sites. This gives rise to an equation that can be written as A, =

’

N , ki ai exp( -zt+baFIR T ) 1 kiaiexp(-z$,F/RT)

+

where A , is the adsorption of ion i, N , the maximum adsorption (in the same units), ki a binding constant, and ai the activity in solution. Equation (13) can be understood if it is compared to the familiar Langmuir equation which, using similar terms, would be written as A

=N,

kc/( 1 + kc)

(14)

Two major differences are obvious. First, the Langmuir equation, as commonly used in soil science, does not specify individual ions. Thus for phosphate at, say, pH 7, it implies that reaction is with all the phosphate in solution, whereas reaction must be with phosphate ions. Second, Eq. (13) contains the exponential terms. These may be thought of as acting as a multiplier for the binding constant ki. Thus, for an anion, z is negative, and hence the value of the exponential term will be small if $a is negative. This has the same effect as decreasing the value of the binding constant. As a result, for negative surfaces adsorption of anions is decreased, but not prevented. Similarly, for a cation, z is positive and a negative value of $a will then


197

increase the value of the product with the binding constant and so increase adsorption. Thus, for this model, the electrostatic properties of the surface are very important in determining the extent of the reaction. They are therefore considered further. b. Effects of Electrostatic Potential in the Four-Plane Model. In general, high pH will produce a low value for the electrostatic potential t+ha and will favor adsorption of cations and decrease adsorption of anions. This effect will be influenced by the zero point of charge of the oxides-the lower the zero point, the more likely that the oxide is negatively charged and, other things being equal, the less likely it is to favor adsorption of anions and the more likely to favour adsorption of cations. This effect was noted earlier (Section II1,A) in that the manganese oxide (birnesite) with a point of zero charge at about pH 2 was a more effective adsorber of cations than the iron oxides. In this model there is a further way in which the electrostatic potential can be altered-the position of the adsorbed layer relative to the s layer may be changed. If the adsorbed layer is placed close to the s layer, the changes in potential with pH in that layer are exaggerated. At low pH and therefore b

0

i

1001

I 4

P

6

I 8

1

I 10

FIG.4. Adsorption of hypothetical anions on goethite. The goethite was assumed to have the same properties as that used for adsorption of phosphate by Bowden et al. (1980b). It is assumed that the ions differ in the position of the mean plane of adsorption and hence in the capacitance between the mean planes of adsorption. The five different positions, relative to the planes s and /I, are shown diagrammatically. (a) monovalent anion with pK, equal to 6; (b) divalent anion with pK, well below 4 and pK, equal to 6. Values for the binding constant and for the concentrations were chosen to give near-maximum adsorption for line 1. It was assumed that the divalent ion would occupy twice the area of the monovalent and so have half the maximum adsorption.

198

N. J. BARROW

positive potential, the potential is greater and so adsorption of anions is favored (Fig. 4); at high pH and therefore negative potential, the value is lower and so adsorption of anions is decreased. The net effect is to twist the adsorption envelope so that the closer the adsorption plane is to the s plane, the more steeply adsorption of anions decreases with increasing pH (Fig. 4).An analogous effect occurs with cations but the direction of the effectis opposite-the closer the adsorption plane is to the s plane, the more steeply adsorption increases as pH increases (Barrow et al., 1981a). The location of the adsorption plane also affects the way the charge on the adsorbed ions is balanced. If, for example, the adsorbed ions were located in a plane close to the s plane, then charge would be largely balanced by appropriate changes in the s plane. If an anion were absorbed, the charge would be balanced by displacement of an OH- ion [as in Eq. (12)]. In soil science surface charge is, in effect, defined as the charge balanced by electrolyte ions. This would not have changed if the charge balance had been entirely in the s plane and the change in charge would be recorded as zero. If, on the other hand, the adsorbed ions were located in a plane further out from the s plane, and thus close to the electrolyte ions, the charge on the adsorbed ions would be more likely to be balanced by changes in the retained electrolyte ions and the proton in Eq. (12) would be more likely to dissociate. This would therefore be recorded as a large net change in surface charge. The effect of the location of the adsorption plane on the surface charge is illustrated for a divalent anion in Fig. 5. Thus, in this model the location of the adsorption plane determines the steepness of the effect of pH and also the magnitude of the charge conveyed to the surface. This is analogous to choosing the balance between the various reaction equations for the three-plane model. This analysis also shows the importance of having a range of observations in testing such models. One needs not only the effects of pH on adsorption but also the effects of adsorption on charge if one is to test a model rigorously. Nevertheless, we can illustrate how the assumed position of the plane of adsorption can qualitatively explain some of the observations. Thus Hingston (1970) found that the charge conveyed to the surface by sulfate adsorption was greater than that for phosphate adsorption. This could be explained by assuming that the mean plane for sulfate adsorption was further from the s plane than that for phosphate adsorption. Further, the charge conveyed to the surface by fluoride adsorption was found to be small (Hingston, 1970), and this would be explained by assuming that the small fluoride ion resided close to the s plane. Because adsorption changes the net charge, it also changes the electrostatic potential $a, that is, there is a feedback mechanism. Each increment of adsorbate is adsorbed onto a surface with a different charge and a different potential from that of the previous increment. This negates a basic principle

REACTION O F ANIONS AND CATIONS

199

I on adsorbed (pmol m.’)

FIG.5. Effect of the location of the plane of adsorption on the charge conveyed to the surface by a hypothetical divalent anion. Conditions as for Fig. 4b at pH 6. For curve a, the adsorption plane is assumed to be close to the s plane (position 1 of Fig. 4); for curve b it is assumed to be far from the s plane (position 5 of Fig. 4).

of the Langmuir equation and explains why this equation should not be used to describe adsorption of ions by soil. This feedback mechanism is also important in determining the effect of level of addition of adsorbate on the charge conveyed to the surface. Because each increment of adsorbate is adsorbed onto a different surface from the previous increment, the way the charge on the adsorbate is balanced varies with level of addition. This is illustrated in Fig. 5. If the charge conveyed to the surface were independent of level of addition, the plot of net charge against adsorption would be a straight line. Figure 5 shows that the slope of this line varies and passes through a minimum when the net charge is zero. Hence the increment of charge per increment of adsorption has a minimum when the net charge is zero (lower half of Fig. 5). It was shown by Bolan and Barrow (1984) that this was consistent with published observations on adsorption of phosphate on iron oxides. The qualitative explanation is that when the net charge is positive, a larger proportion of the charge on the adsorbate ions is balanced by displacement of electrolyte anions. The change in charge per anion adsorbed is therefore large. When the net charge approaches zero, there are few electrolyte ions to displace and the charge is increasingly balanced by displacement of OH- ions from the s plane. The change in charge per anion adsorbed is therefore small. When the net charge is negative, the charge is balanced by increases in the electrolyte cations retained and so the change in charge is large again.

200

N. J. BARROW

c. Effects of Activity and Valency of the Absorbing Ion. Two other characteristics affect the amount of adsorption predicted from Eq. (13). One is the activity of the adsorbing ion and the other is its valency. For many adsorbates it has been found that one of the ion species present in solution appears to be adsorbed so much more strongly than the others that all the observations can be described by considering only that ion. This is not, however, a necessary property of the model and, for some metals adsorbing onto goethite from chloride solutions, it was found necessary to assume that MeC1' ions were adsorbed in addition to MeOH' ions (Barrow et al., 1981a). Nevertheless, it is an important property of the model and effectively explains the effects of pH on adsorption of anions. This is illustrated for a hypothetical set of monovalent anions in Fig. 6a. In each case, an adsorption maximum is observed just below the p K of the conjugate acid. The explanation is that, for anions, two of the terms of Eq. (13) have opposing effects. Because the electrostatic potential decreases as pH increases, the value of the exponential term of Eq. (13) decreases. This effect tends to decrease adsorption. It is opposed by increased dissociation of the acid HA. If we assume that the dissociated species A- is the adsorbing species, then at pH values below the pK, the concentration of this ion, though low, increases 10fold for each unit increase in pH. For monovalent anions, this increase suffices to more than counterbalance the decrease in potential, and so adsorption increases. Once the pK is reached, however, there is little further scope for increases in the concentration of A-, so the effect of the electrostatic potential predominates and adsorption decreases. The effect is slightly different when a divalent species adsorbs because the z term in Eq. (1 3) then 80-

Curve pK,

1 2

4 6

4

10

3

8

-

Curve pK,

1

7

$$AT

1

E

3 Lc

c

5 2

20-

n.

24

I

O4

I

34

I

I


20 1

has a value of -2, and hence the exponential term has a greater effect. In other words, the effect of changing the electrostatic potential is greater. The result is that adsorption decreases with increasing pH, with a slight increase in slope near the pK, of the acid (Fig. 4b). These differences between monovalent and divalent anions correspond to those observed by Hingston et al. (1972) (Fig. 2). In contrast, for hydrolyzable cations the two terms of Eq. (13) have additive effects. As pH is increased, the electrostatic potential decreases, and this favors adsorption of positively charged ions. If, in addition, we assume that MeOH’ ions are adsorbed, the increasing concentration of these ions, together with the increasingly favorable potential, combine to give a sudden increase in adsorption as the pH is increased, that is, an “adsorption edge” (Fig. 6b). In other words, the product kiai exp( -z$,F/RT) of Eq. (13) rapidly increases with increasing pH. This can occur well below the pK, for hydrolysis, and thus at a pH at which the proportion of MeOH’ ions is small. However, it is the product that must be large, not just the concentration; provided the ki term and the exponential term are large enough, this will be so. Differences in the relative position of the adsorption edge for different adsorbents (Fig. 1) can be ascribed to differing values of the binding constant k, for particular combinations of adsorbate and adsorbent. The amount by which the adsorption edge precedes the pK depends partly on the value of ki. It is worth reiterating that it also depends on the experimental conditions and the method of plotting. The adsorption edge can be varied over about two pH units by changing the proportions of adsorbate and adsorbent and by plotting adsorption as a proportion of the amount of adsorbate added rather than as a proportion of the available surface occupied (Benjamin and Leckie, 1981). The idea that it is the MeOH+ ion that is adsorbed is an effective explanation of the correlation between the adsorption edge of different metals and the pK, for hydrolysis. It would also be expected that it would be easier to remove such ions from their sheath of water molecules than to remove the divalent Me2+ ions (James and Healy, 1972). In the course of this description of the four-plane model, it has been shown that it qualitatively describes many of the observations on adsorption of anions and cations. However, rigorous testing of a mathematical model such as this requires that it also explain the observations quantitatively. This is discussed in the next section.

,

4. Methods of Evaluating the Four-Layer Model Models such as the four-layer model consist of a set of equations. The equations contain parameters which represent properties of the system.

202

N. J. BARROW

Examples are the binding constants for ions, the values for maximum adsorption, and the capacitance of the region between the planes. In order to evaluate the model, values have to be allocated to these parameters. This enables the equations to be solved and values to be allocated to the variables. Examples of variables are the charge on the several planes, the amount of absorbate in these planes, and the electrostatic potentials of the planes. The values of only a few of the variables will have been measurable, and it is the ability of the model to match these observations that is usually tested. Thus to evaluate a model we need a way to allocate values to the parameters and a way to solve the equations and thus find the values of the variables. In the four-layer model the values allocated to the parameters have been chosen on the basis of their ability to match the observations. Thus the aim has been to test whether the model can describe the observations rather than to find values of constants. Nevertheless, the more measured variables there are, and the wider the range of conditions of measurements, the more the possible values of the parameters are restricted. This is illustrated by Barrow et al. ( 198 1a). Because of the structure of this model, it is not possible to use general computer programs for chemical equilibria to solve the equations. Nor is it possible to rearrange the equations so that they can be solved analytically. Hence iterative methods have to be used. This involves, in essence, guessing a value for one of the variables and using this value to calculate the value of the other variables. To be solvable the number of equations must be at least equal to the number of variables and hence, if one variable is guessed, there must be an equation “left over.” If the correct value of the guessed variable had been chosen, this equation would be true. Iteration therefore involves repeated guessing of the chosen variable until the “left over” equation is true to within a required level of accuracy. A simpler, rapid, but approximate way of fitting the model has also been suggested (Posner and Barrow, 1982). This involves assuming that the change in potential with adsorption is linear. This is only an approximation, because the change in charge is not linearly related to the amount of adsorption (Fig. 5). Nevertheless, this method provides a simple way of testing the model against the data. It is also very useful when attempting to adapt models developed with simple systems to explain the complex behavior of soils (Section V,C,l). 5. Reproduction of Observations by the Four-Layer Model

The ability of the four-layer model to reproduce observations has been tested under a wide range of conditions. It has been shown to reproduce the observed charge on goethite suspended in a range of electrolytes at different

203


concentrations (Barrow et al., 1980a). It also closely reproduced both the effect of pH on phosphate adsorption and the effect on the charge conveyed to the surface (Figs. 7 and 8). That is, it was possible to select a position of the mean plane of phosphate adsorption such that the effect of pH on both adsorption and charge was reproduced. The charge conveyed to the surface by phosphate adsorption formed a U-shaped curve against pH (Fig. 8). The reason for this is analogous to that discussed in Section III,B,3,b. At low pH the surface was initially positive and adsorption displaced electrolyte anions; at high pH the surface was negative and adsorption was balanced by electrolyte cations. In both cases the charge conveyed to the surface was large. At intermediate pH, there were few adsorbed electrolyte ions and the charge was balanced by displacement of OH- ions and increased adsorption of H + ions. A similar close description of citrate adsorption was also obtained (Bowden et al., 1980b). When adsorption of selenite was plotted against pH, the model reproduced the data and the characteristic bend in the curve near the pK, of selenious acid. Figure 9 shows how this bend was associated with the interaction between the declining value of the potential $a and the increasing value of the concentration of SeOs-. The rapid, approxi-

b

f

ot Y

0

200 400 Phosphate concentration (pM)

600

FIG.7. Modeling of the observed effects of pH and of phosphate concentration on adsorption of phosphate by goethite using the four-plane model. (From Bowden et al., 1980b.)

204

N. J. BARROW -a,/S

*.Or

P 2

In

-3 -m0

1.5-

E

Y m

.c

B

r P

1.0.

L

P

I

I

01 t

4.0

8.0

6.0

,J

10.0

PH

FIG.8. Modeling the observed effects on the charge conveyed to goethite by adsorption of phosphate. The diagram indicates that the observed net charge is the sum of that due to adsorption of divalent phosphate ions (csJS) less that due to displacement of OH- ions and gain of H + ions (AcsJS), where cs indicates charge in the appropriate adsorption plane and S the amount of phosphate adsorption. (From Bowden et al., 1980b.)

mate way of fitting the model was also shown to give a good description of selenite adsorption (Posner and Barrow, 1982). McKenzie (1983) modeled the effect of pH on molybdate adsorption by goethite by postulating that both the HMoO; and the MOO:- species were adsorbed. However, he did not investigate the effects of changing the capacitance, and thus the position of the adsorption plane relative to the s plane. An observation that has been a source of controversy in soil science for some time is that anion adsorption is usually increased when there is an increase in the electrolyte concentration of the medium in which adsorption is measured. This is also observed for adsorption of anions on oxides, but only at medium and high pH values (Fig. 10). This effect is also closely described by the four-plane model and is produced in the model by effects of electrolyte concentration on the potential t,ha (Fig. 10). The pragmatic explanation is that, at high pH, the surface charge is negative and a high concentration of


205

01 PH

FIG. 9. Modeling the observed effect of pH on adsorption of selenite by goethite. (From Bowden et al., 1980b.)

0

I

4.0

6.0

8.0

10.0

FIG. 10. Modeling the observed effect of pH and of electrolyte concentration on the adsorption of phosphate by goethite. (From Barrow et al., 1980b.)

N. J. BARROW 1 .o

a

b

0

0.8

0.075M NaCl 0-0.075 M KN03 b---0.0075 M NaCl

0-

(13

0.6

0.4

0.2

(

4.0

4.5

5.0

4.5

5.5

5.0

Pl-

FIG.11. Modeling the observed effect of pH and of different electrolytes on the adsorption of (a) copper and (b) lead by goethite. (From Barrow et al., 1981a.)

electrolyte causes a high concentration of electrolyte cations in the outer planes of adsorption. This permits an increased adsorption of phosphate. At low pH, the surface charge is positive and a high concentration of electrolyte anions in the other planes decreases adsorption of phosphate. The four-plane model was able to closely describe adsorption of copper and lead from a range of electrolytes (Fig. 11). As was the case for anions, the model was able to reproduce the effects of adsorption on charge (Barrow et al., 1981a). Thus it was possible to choose a position for the mean plane of metal ions that reproduced both adsorption and charge observations. The data modeled in this case were unusual in that absorption of lead increased more slowly with pH than adsorption of copper (Fig. 11). This contrasts to the results of McKenzie (1980) (Fig. 1) and Benjamin and Leckie (1981). The low slope for lead could only be reproduced by assuming that Pb2 ions were absorbed rather than PbOH'. However, this conclusion should be treated with some caution because of uncertainties about the correct constants with which to calculate ion pairs with chloride and nitrate. The adsorption of zinc by goethite was also closely modeled, in this case using the assumption that ZnOH' ions were adsorbed (Barrow et al., 1981a). +


207

A simplified version of the four-layer model has also been used to describe “cooperation” between ions of opposite charge and competition between ions of like charge and to describe adsorption of phosphate on montmorillonite (Bowden et al., 1980a). 6. Overall Evaluation of the Four-Layer Model

Of the three models considered, the four-layer model is the most complex in its conceptual structure. Yet there are some conceptual advantages. There is a simple subdivision of the effects of pH into the effects on the species of ions present and effects on the electrostatic potential. It is a conceptual advantage to a soil scientist that the adsorption equation is similar to a Langmuir equation. It is also a conceptual advantage that the closeness of approach of adsorbed ions to the surface is an important distinction between different reactants. Although the structure is complex, the number of parameters needed to describe adsorption is not large-in most cases only three. In this respect it compares well with the other models. The model has been shown to be comprehensive in that it describes a wide range of observations. It is also precise in that it describes the observations closely. Specific surface complexes are not a part of the model. Nev,ertheless, the model is consistent with direct measurements of these complexes.

IV. RATES OF ADSORPTION AND DESORPTION In the presentation so far it has been assumed that equilibrium had been reached. Indeed, this idea is so deeply entrenched that the word “equilibrate” is almost always used to describe mixing of adsorbate and adsorbent. Before desorption of the adsorbed ions can be discussed, we need to discuss the rates of desorption and adsorption reactions. Suppose we imagine a sequence of reaction steps that might be written as U P V P W e X P Y P Z. Suppose that, in the sequence U -+Y, one of the steps is much slower than the rest. The rate of that step will effectively control the rate of conversion of U to Y.Let us further suppose that the step for Y to Z is very much slower still. Then Y will be produced fairly quickly and will slowly be converted to Z. It is in this sense that we can regard a reaction as being controlled by two processes yet within each process there could be a number of steps. This appears to be the case for adsorption in that there is a fairly rapid initial process which comprises a number of steps. Direct evidence for this was provided by Zasoski and Burau

208

N. J. BARROW

(1978) for cadmium adsorption on manganese oxide. They showed different rate curves for Cd adsorption and for H’, Na’, and K + displacement. This rapid initial process is followed by a slow process. These will be examined in turn. A. THERATEOF THE ADSORPTIONPROCESS The adsorption reaction involves the approach of charged particles to charged surfaces. There might therefore be electrostatic effects on the rate of the reaction. Unfortunately, this has not always been taken into account (Mikami et al., 1983). It may be shown that the position of the rate-limiting step relative to the electron-transfer steps determines the magnitude and direction of the electrostatic effects (Barrow et al., 198lb). For phosphate adsorption on goethite, Madrid and Posner (1979) observed that the adsorption of phosphate was fastest at low pH. It was shown that this meant that the rate-limiting step came before the electron-transfer steps and did not, of itself, involve electron transfer (Barrow et al., 1981b). The effect of this is that the electrostatic effect increases the rate of the forward (adsorption) reaction when phosphate reacts with a positively charge goethite surface but does not affect the rate of the backward (desorption) reaction. Further, the equilibrium favors the product of the reaction, and hence the rate constant of the forward reaction is larger than that of the backward reaction (Barrow et al., 1981b). Therefore, if we disturb an equilibrium by adding phosphate, the time required to reach a new equilibrium will be less than if we disturb the equilibrium by removing the phosphate. It should not be assumed that a period that suffices for the forward reaction to apparently reach equilibrium will also suffice for the backward reaction. Some of the reported difficultiesin inducing desorption may be due to this effect. In this context it is appropriate to discuss the use of the word “irreversible” to describe adsorption. This word is used in differing senses in the chemical literature. A common meaning is that the equilibrium for a chemical reaction is so far to the right that it goes virtually to completion. It therefore is practically impossible to drive it in the opposite direction, that is, it is practically irreversible. This sense is obviously inappropriate for adsorption for, if the reaction went to completion, there would be no adsorbate left in solution and adsorption-concentration curves could not be drawn. It is therefore a contradiction in terms to speak of irreversible adsorption. If desorption is found to differ from adsorption, there are only two possible interpretations. One is that desorption is slower and that insufficient time has been allowed. The other is that a further process has followed adsorption and some of the reactant is, as a result, no longer “adsorbed” and thus not in


209

equilibrium with the solution. The evidence for this further process is considered next. B. THESLOW PROCESS THAT FOLLOWS ADSORPTION There are three kinds of evidence that a molecular rearrangement may follow the adsorption process. One is direct observation of overall rate. The reaction often seems to consist of one process that appears to approach an end point after periods of a few minutes followed by a much slower process. This observation has been frequently made, and the following list of references is meant to be illustrative rather than exhaustive: for acid-base titration of goethite, see Madrid and Arambarri (1978); for reactions of copper, zinc, cadmium, and lead with amorphous iron hydroxide see Benjamin and Leckie (1981); for reaction of copper with goethite see Padmanabham (1983); and for reaction of phosphate with goethite see Madrid and Posner (1979) and Sigg and Stumm (1980). The second kind of evidence is that high temperatures are observed to increase the extent of the overall reaction [for phosphate on goethite, see Muljadi et al. (1966); for cations on silica, see Bye et al. (1983)l. In both of these cases, this observation was explained in terms of an effect of temperature on the position of an equilibrium and thus the value of an equilibrium constant. However, in both cases, it was also observed that the maximum amount of adsorption also increased. Muljadi et al. (1966) explained this in terms of the breaking of surface bonds at high temperatures, thus increasing the surface area. However, the temperatures used were not very high. Bye et al. (1983) did not bring this aspect of their results to attention. A much simpler explanation of these results is that a slow reaction followed adsorption and that the rate of this slow reaction was increased at high temperatures. The third kind of evidence is that the longer the period that has elapsed before desorption is started, the less complete the desorption. This has been reported for the reaction of copper, zinc, and cobalt with goethite (Padmanabham, 1983a,b). Again, this can be explained by a slow reaction that follows adsorption. Indeed, this is the only explanation that can explain observations that, after an initial desorption, a resumption of adsorption may occur (Hingston et al., 1974; Cabrera et al., 1981). It is curious that such an extensive body of evidence has often been ignored, especially when one considers that the slow and continuing process that follows adsorption is very important in the long-term reaction of plant nutrients and pollutants with soils. It is thus important both for the residual value of fertilizers and for the retention of pollutants. It is even more curious

210

N. J. BARROW

when we realize that study of the mechanism of this slow process preceded much of the work on the adsorption reaction. For example, Wei and Bernstein (1959) reacted deuterium oxide vapor with boehmite (an aluminium oxyhydroxide) and showed that there was a rapid initial surface exchange followed by a slower diffusion-controlled exchange within the crystal. Berube et al. (1967) analyzed titration data with crystalline ferric oxide and concluded that the slow step was controlled by diffusion of protons into, or out of, the surface. They obtained a value of 20 kcal/mole for the activation energy, that is, an increase in temperature would increase the rate of the process. A question that then arises is whether ignoring this diffusive process invalidates studies on the supposed equilibrium. One defense is that equilibrium models have usually been tested against data obtained from a fairly short period of reaction, so that the diffusive process is less important. Another is that the immediate source of the diffusive process is the surface concentration of adsorbate, not the solution concentration. If we assume that diffusion is into a plane and ignore, for the time being, changes in surface potential which the diffusion process will itself induce, then 2 M , = -r,@

&

where M , is the amount of substance i penetrating at time t, Ti is its surface concentration, and D is a solid-state diffusion coefficient. (The mathematics of. solid-state diffusion are more complex than indicated here, but this complexity will not affect the current argument.) If we accept that diffusive penetration of the surface occurs, then the amount of “adsorption” observed is really true adsorption plus penetration, that is, M , + Ti. It would therefore be better called “sorption” or “retention.” After a given period, the observed retention is proportionately higher than true adsorption. The parameters of “adsorption” models are therefore overestimated. This is especially true of the maximum adsorption. A further question is, By what molecular mechanism does diffusion occur? Diffusion in crystals is basically an atomic process, with the diffusing molecules undergoing a series of jumps through the crystal. Mechanisms include exchange, in which two atoms change place, and vacancy mechanisms, in which an atom moves in one direction and the vacancy, or hole, in the other (Manning, 1968). The vacancy mechanism is most likely to occur if the crystals are imperfect. Wei and Bernstein (1959) noted that hydrogen occurs in bonded sheets in boehmite crystals and suggested that diffusion of hydrogen is more rapid than diffusion of oxygen. Similarly, Padmanabham (1983a,b) suggested that copper, zinc, and cobalt atoms might occupy the


211

lattice positions normally occupied by iron. They could therefore diffuse into goethite by an exchange mechanism. Padmanabham (1983b) found that lead could be more easily desorbed than copper, zinc, or cobalt and suggested that the large lead atoms could not replace iron atoms in the crystal. For larger ions such as phosphate, a vacancy mechanism seems more likely, and imperfections in the crystal would therefore be important. Cabrera et al. (1981) found that reaction of phosphate with the iron lepidocrocite continued for a longer period than reactions with goethite. They noted that lepidocrocite appeared to consist of small crystals forming large aggregates with micropores between them. The resolution that can be obtained with electron microscopy is now so good that such boundaries between microcrystals can be examined in detail. Cornell et al. (1983) have shown that a sample of goethite was highly coherent across such boundaries, that is, there were not many vacancies or micropores. An interesting research project would be to test whether samples of iron oxides with less well-organized boundaries are more prone to continue to react with, say, phosphates. It is largely because of the diffusion process that follows adsorption that desorption does not seem to follow the same path as retention (hereafter the word “retention” is used to describe the sum of adsorption and penetration). This difference between retention and desorption is often more clearly expressed, and certainly more extensively documented, with soils and is discussed further in the content.

V. TRANSFERRING THE VARIABLE-CHARGE MODELS TO SOILS A. THEOBSERVATIONS To BE DESCRIBED A large proportion of all the studies on soil chemistry have involved reactions with the variable-charge components of soil. It is obviously impractical to deal with more than a small fraction of this work here, and hence only the main aspects will be considered.

1. Spec$city and Concentration It is a common experience that many ions react specifically, that is, when several species of ions are present in solution, some are retained by soil in much greater proportions than their proportion in solution. For example, phosphate is strongly preferred over, say, chloride, and zinc is strongly preferred over, say, sodium. Nevertheless, some of the specifically reacting

212

N. J. BARROW

-

10,000

.SI

2 1000rn

I

P

L

Q

8

L

n

10

I,

0.01

I

I

I

1.o 10 Phosphate concentration ( p g P/ml)

0.1

FIG.12. Relation between retention of phosphate and concentration over a wide range of concentrations for three soils of widely differing phosphate retention. The lines drawn are taken from a model of the effect of pH and concentration on retention. They are a section in the concentration dimension of Fig. 16.

ions remain in solution, and plots may be prepared of the amount retained against the concentration remaining in solution (Fig. 12). Such curves are often called “adsorption isotherms.” However, it is now becoming clear that the process involved is not simply adsorption, and the term “isotherm” implies that temperature is the only other variable affecting the curve. I have therefore argued that this impressive-sounding terminology be avoided (Barrow, 1978). A better terminology is Q/Z (quantity/intensity) plots or retention curves. Much effort has been expended in seeking a mathematical description of these Q/Z plots. The reasons for this effort have not always been clear. I have argued that the only good reason is to seek a way of summarizing behavior by a few numbers (Barrow, 1978). For this purpose, the fewer the numbers the better. A simple equation that often describes Q/Z plots is the Freundlich equation S

= acb

(16)

were S is the amount retained and should include any adsorbate initially present, c is the concentration remaining in solution, and a and b are constants. If this equation holds, logarithmic plots of S against c give a straight line. This is often a convenient approximation over important concentration ranges, but over larger ranges the nonlinearity becomes more important (Figs. 12, 13,14, and 15). The Langmuir equation, or versions of it, have been widely used to describe Q/Z curves on the grounds that they are

213

REACTION OF ANIONS AND CATIONS 1ooc

3oa

Days

v 91

--.-

I00

.

A

43

0

10

E

0 3

a.

0 1

m

m

B e

2 3

2

30

0.03

0.1 0.3 1.o Phosphate concentration ( p g Plml)

3.0

1000

z

a.

300

100 I

I

0.03

I

0.3 Phosphate concentration 0.1

I

1 .o

1

3.0

(Hg P/rnl)

FIG. 13. Effect of time and temperature (“C) of incubation on the relationship between phosphate retention and solution concentration. The data are from Barrow and Shaw (1975a) and the lines are for the mechanistic model of Barrow (1983b).

more mechanistic. However, the mechanism of ion reaction with soils does not meet the Langmuir criteria. Furthermore, a truly mechanistic approach should be capable of also describing the effects of variables other than concentration. This will be discussed further in Section V,C. 2. Period of Reaction It is widely observed that when phosphate and several other nutrients react with soil, the solution concentration continues to decline for a long period. The problems that this behavior produces are often circumvented by making measurements after some convenient, but constant, period. Again a semantic trap arises. Many workers use the verb “equilibrate” to describe the process of mixing the soil and nutrient solution, and then deal with the results as if equilibrium had been reached. Rather than avoiding the problems caused by

214

N. J. BARROW

looor

Fluoride concentration ( p g F/ml)

FIG.14. Effect of the time and temperature (“C)of incubation on the relationship between fluoride retention and solution concentration. The data are from Barrow and Shaw (1977a) and the lines were obtained by fitting the mechanistic model of Barrow (1983b). (From Barrow, 1985a.)

the slow rate of reaction, it would be better to measure the rate of reaction. It is, after all, an important property affecting both the long-term effectiveness of fertilizers and the long-term retention of pollutants. As for any other reaction, measuring the rate involves mixing the reactants and measuring either their rate of decrease or measuring the rate of increase of the product. In this case, the most practical measurement is the rate of decrease in nutrient concentration. However, there are difficulties both in making these measurements and in interpreting them [reviewed by Barrow (1983a)l. One problem is that, at the commonly used solution/soil ratios, concentration and retention both vary through time. If the effect of time is to be defined, we need to be able to see the effect on concentration at constant retention-or the effect on retention at constant concentration. Approaches that relate just two of the three important variables can give misleading results. It was shown (Barrow, 1983a) that the Elovich equation, which relates retention to time and which was advocated by Chien and Clayton (1980), could not be used to characterize soils. The rate measured using this equation varied with both the level of addition of nutrient and the solution/ soil ratio. One way of overcoming these difficulties is to incubate soil and nutrient solution at moisture contents near those occurring in the field. The solution/soil ratio is thus low (say, 0.2/1) and the amount of nutrient


215

100

/

40 0

25' 40'

rn

60'

ZnOH' concentration ( p g Zn/rnl)

FIG.15. Effect of time and temperature ("C) of incubation on the relationship between zinc retention by soil and calculated solution concentration of ZnOH+. Lines were obtained by fitting the mechanistic model of Barrow (1983b). Retention is plotted against ZnOH+. (From Barrow, 1985b.)

remaining in solution is low relative to the amount reacted with the soil. Changes in concentration can therefore be measured at near-constant values for retention. Another approach is to use an apparatus to hold concentration constant (van Riemsdijk and De Haan, 1981). However, it seems easier to measure both retention and concentration through time and to mathematically describe the surface which interrelates them. If desired, the changes in retention at constant concentration, and vice versa, could then be interpolated. Equations that describe the changes in retention and concentration of phosphate through time have been summarized and reviewed by Berkheiser et a!. (1980). They concluded that the most satisfactory equation for describing phosphate retention was

216

N. J. BARROW

where b , and b, are constants. They also concluded that such equations may describe the observations but they do not do so in a mechanistic sense. Equation (17) also applies to molybdate, fluoride, sulfate (Barrow and Shaw, 1975b, 1977a,b), arsenite (Elkhatib et al. 1984), and zinc retention by soil (Kuo and Mikkelsen, 1979). When plotted on a logarithmic scale, it gives a planar surface relating log S to log c and log t. However, when a wide range of concentrations are used, this is seen to be an approximation, and the relationship is curved in the log c direction (Kuo and Mikkelsen, 1979; and Figs. 13, 14, and 15). Although the rate of change in solution concentration is perhaps the best way of characterizing the rate of reaction, the rate of change of the effectiveness of a fertilizer is of such practical importance that it merits further discussion. The decrease, with time, in the effectiveness of phosphate fertilizer is well documented [reviewed by Barrow (1980a)l. The rate of change has been measured over a range of periods and has been shown to be consistent with Eq. (17) (Barrow, 1974). Furthermore, the coefficients of Eq. (17) have been shown to be of value in characterizing differences between soils and the residual effectiveness of phosphate fertilizers (Barrow, 1980b). Rather less work has been done on other nutrients. There is field evidence that the effectiveness of molybdenum fertilizer decreases with time (Smith and Leeper, 1969), and it was shown, in pots, that the rate of decrease was also consistent with Eq. (17) (Barrow and Shaw, 1975b). The effectiveness of copper fertilizers was found to decrease with increased period of incubation with moist soil (Brennan et al., 1980, 1984). A similar effect was also observed when sulfate was incubated with a soil which retained sulfate strongly (Barrow and Shaw, 1977b).

3. Effects of Temperature In general, temperature may have two distinct effects on a chemical reaction: it may affect the rate of approach to equilibrium, and it may affect the position of the equilibrium. Both these effects occur for the reaction of nutrients with soil [reviewed by Barrow (1979b)I. High temperatures of reaction of soil and phosphate decrease the effectiveness of phosphate fertilizer to plants and usually give rise to decreased concentrations in solution (Fig. 13). Both of these effects are due to an increased rate of reaction. Increasing the temperature of incubation also increases the rate of reaction of copper with soil and decreases the effectiveness of copper fertilizers (Brennen et al., 1984). Similarly, high temperatures increase the rate of desorption. Analogous effects occur for molybdate and sulfate (Barrow and Shaw, 1975b, 1977b) and for fluoride (Fig. 14) and zinc (Fig. 15). In ?neral, effects of temperature on the rate of a reaction occur because the


217

reaction involves an intermediate, high-energy state. Only those molecules with sufficient energy can make the transition over this stage, and the higher the temperature, the higher the proportion of sufficiently energetic molecules. The effect of temperature is conveniently described by the energy required to jump over the barrier, the activation energy. For phosphate, the effect of temperature on the forward (sorption) reaction is about the same as that on the backward (desorption) reaction, that is, the activation energy is about the same (Barrow, 1979b). This suggests that the rate-limiting step in sorption is similar to that in desorption, as it would be for a diffusion process. Further, the lack of difference indicates that temperature would have little effect on the position of the equilibrium eventually reached by the rate-limiting steps. Yet, if conditions are chosen such that the slow step has almost stopped, then high temperatures are found to increase the concentration of phosphate (Barrow and Shaw, 1975a; Barrow, 1979b) and of molybdate in solution (Barrow and Shaw, 1975b). This is strong evidence that there is an initial adsorption step, that the equilibrium for the step is reached quickly, and that it is exothermic and thus high temperatures decrease the amount of product. For agronomists, the effects of temperature in increasing the rate of reaction between nutrients and soil can provide a valuable experimental tool. It means that effects that might take years in the field can be produced in a few days in the laboratory.

4. Efects of pH Soil pH can be readily measured and fairly readily changed. It has therefore been widely investigated as a means of controlling the reaction with nutrients. Most work has been done on phosphate. Nevertheless, it remains a poorly understood subject. This is because seemingly contrasting effects have been observed by different workers. This has been noted in recent reviews: “Reports on the effect of liming on the sorption of phosphate are conflicting” (Probert, 1980); “Considerable controversy exists in the literature regarding whether or not liming decreases P fixation” (Sanchez and Uehara, 1980); “Liming has been reported to increase, decrease, or not affect the phosphate that can be extracted from soils” (Haynes, 1982). There are several reasons for these conflicting observations. Haynes (1982) suggested two: that changing the pH may affect mineralization of organic phosphate and that airdrying the soil after applying lime may cause an artifact leading to decreased retention. Further reasons are that the direction of the effect may differ in different parts of the pH range (Barrow, 1984; and Fig. 16), that the direction and magnitude will depend on the electrolyte in which retention is measured (Barrow, 1984), and that the effect of changing pH on the release of “native”

218

N. J. BARROW

FIG.16. Effect of pH on retention of phosphate from four widely differing soils. Points are interpolated values for retention at indicated concentrations and pH values. Lines are modeled values. (From Barrow, 1984.)

phosphate may differ from that on the retention of new phosphate (Barrow, 1984). Nevertheless, the main reason for inconsistent results is that the effect of pH on phosphate retention is fairly small. Ancillary effects can therefore have a relatively large effect and can change the direction of the net effect. The effect of pH on reactants other than phosphate is much greater. Retention of sulfate and especially of molybdate decreases as pH increases. Sulfate, molybdate, and phosphate retention were compared for a group of soils that ranged in pH (in calcium chloride) from about 4 to just above 6 (Barrow, 1970).The decrease in sulfate retention between pH 4 and pH 6 was about three times greater than for phosphate. For molybdate, the decrease was about 20 times greater than for phosphate. Such marked effects tend to dominate any other effects. Boron retention is also affected by pH, but increasing the pH increases retention (Hatcher et al., 1967). For most soils, the maximum pH attainable is usually limited by the equilibrium between calcium carbonate and carbon dioxide. This seems to be lower than the pH at which maximum retention of boron occurs. Sims and Bingham (1967, 1968a,b) showed that for soil constituents such as clay minerals and aluminium and iron oxides, in sodium or potassium systems, maximum retention was near pH 9. They concluded that retention by clay was mostly caused by iron and aluminium oxide impurities. A retention maximum near 9

219

REACTION OF ANIONS AND CATIONS Dekalb B Pb

-

0 I

.7.1

cu

:F8”

c

3

‘m

Y

‘5.3 4.8 4.3

- 1

P

41

p7.1

32

Ni

= 4.6 k

1

0 0

4.3 *

.; .

.

.

.

0.2 0.3 0.1 0:2 Equil. Solution Concentration (Mrnol /ml)

0.1

0.3

FIG. 17. Effect of pH on retention of lead, copper, zinc, and nickel by two soils. (From Harter, 1983.)

was also reported for clays by Keren and Mezuman (1981) and Keren et al. (1981). Retention of the hydrolyzable cations increases markedly with increasing pH. A comprehensive set of results for lead, copper, zinc, and nickel was provided by Harter (1983) (Fig. 17).

N. J. BARROW

220

Time (h)

0

I 0.5

I

1 .o Fluoride concentration ( p g F/rnli

I 1.5

FIG.18. Sorption and desorption of fluoride. The fluoride had been incubated with soil for 4 days at 80°C and desorption was then induced by mixing the soil with a range of solution/soil ratios. The lines indicate the modeled values for sorption and for desorption after 40 h, using the model fitted to Fig. 14. The inset shows the modeled and observed effects of time of desorption for two solution/soil ratios and the 700 pg F/g soil level of addition. (From Barrow, 1985a.)

5. Desorption

Suppose that a range of concentrations of a nutrient is mixed with samples of soil for a specified period and then a plot is prepared of retention against solution concentration. If the concentration in solution of one of these samples is then decreased, some desorption will occur but the original retention curve will not be retraced. Such observations have been made several times, and an example is illustrated in Fig. 18. Nevertheless, this description is inadequate inasmuch as desorption, like the forward retention reaction, is also a slow process. Rather than merely specifying the position of the desorption curve at one time, it is better to specify the change in position through time and thus the rate of desorption. This poses some problems because, in most practical systems, increasing desorption is accompanied by an increasing solution concentration. It is therefore difficult to measure desorption at a constant solution concentration and seemingly impossible to measure it at the ideal of zero solution concentration. It has been argued (Barrow, 1979a) that desorption can be specified by two


221

end points. One is at zero concentration in solution. Although this point cannot be achieved experimentally, it provides a conceptual limit and is the point at which maximum desorption would occur. Observed desorption can be described by assuming that desorption at this point is proportional to a fractional power of time. This was the case for phosphate (Barrow and Shaw, 197%; Barrow 1979a), fluoride (Barrow and Shaw, 1977a), and sulfate (Barrow and Shaw, 1977b). Sharpley et al. (1981), in a slightly different approach, also found that desorption was proportional to a fractional power of time. The other conceptual end point is the concentration in solution at which neither sorption nor desorption occurs. Clearly desorption can only occur if the concentration is lower than this null point. The value of the null point decreases with increasing period of reaction, that is, as the retention reaction proceeds. Consequently, the rate of desorption decreases (Barrow and Shaw, 1975~).These are not direct cause-and-effect relations but rather separate manifestations of the same process-the continuing reaction between the nutrient and the soil. OF THE REACTION WITH SOILS B. QUALITATIVE MODELING

The four-layer model is well able to describe qualitatively many of the observed effects with soils. Consider the contrasting effects of pH on molybdate, sulfate, and borate (Section V,A,4). The pK, for sulfuric acid is about 2 and hence, at soil pH values, only SO:- ions are present. Furthermore, there is little change in the proportion of the ions with changes in pH. The only effect that can occur is due to the change in electrostatic potential with change in pH. If the sulfate ion does not approach the s plane very closely, the change in potential will not be large, and hence the effect of pH on adsorption will not be large. Molybdate is more complex, as both pK, and pK, are near 4. Hence both H,MoO, and HMoO; are weak acids and both dissociation products (HMoO,, MOO:-) could be adsorbed (Section 111,A). The concentration of HMoO; decreases rapidly as the pH increases above 4, and this could be one reason for the marked effect of pH in retention. A further reason could be that the mean plane of molybdate adsorption is closer to the s plane than the mean plane of sulfate adsorption. This mechanism would also give a steeper effect of pH on retention (Fig. 4). The tendency for borate retention to increase up to pH 9, and to then decrease, is consistent with the adsorption of the monovalent B(0H); ion, for which the pK value is about 9 (Fig. 6). This seems to be a simpler assumption than to assume that boric acid, borate, and hydroxide all compete for the same adsorption site (Keren et al., 1981).

222

N. J. BARROW

c. QUANTITATIVE MODELING OF THE REACTION WITH SOILS 1. Description of a Quantitative Model

Although qualitative models can help one understand observations, they only become convincing and satisfying if they can be rigorously tested against data, that is, if they become quantitative models. The challenge is to adapt the detailed quantitative models used for simpler systems to the more complex problems of soils. In the development of models for the reaction with variable-charge surfaces, the pathway was to understand and to describe the charging process. Adsorption models were then added to the charge models. This does not seem to be a feasible pathway for soils. It is too daunting a task to describe the charge for all of the materials that might be present in a soil. If, however, we emphasize the importance of the electrostatic potential, then the charge models can be regarded simply as a way of estimating this potential. They show that the potential decreases as pH increases and that it also changes as adsorption proceeds. For soils we may not know a priari what these rates of change are, but we can investigate the rates of change that are needed to describe the observed behavior. This is the first step in adapting detailed models to soils. The second step is to take account of the heterogeneity of soils. This seems to show itself in the Freundlich equation that often approximately relates retention and concentration. A Freundlich equation can be generated if it is assumed that the adsorbing surface is nonuniform and that there is an appropriate distribution of values of the parameters of the adsorption equation. It has been shown (Sposito, 1980) that one appropriate distribution is a normal distribution of the logarithms of the binding constants of a Langmuir equation. This is similar to assuming a normal distribution of log ki of Eq. (13) or of assuming a normal distribution of $ a of Eq. (1 3). These two assumptions are mathematically indistinguishable. I have assumed that $, varies (Barrow, 1983b). The third and final step is to accept that the diffusive process that seems to occur in pure systems also occurs in soils but is even more marked. Many authors have suggested that diffusion mechanisms may be involved in the slow reaction between nutrients and soils. Often this has been a qualitative suggestion and the consequences have not been quantitatively explored. Further, the suggestion has often been that diffusion in the solution phase, from the bulk solution to the surface, is the rate-limiting process. It is important to emphasize that, in the present case, the postulated diffusion is solid-state diffusion within the adsorbing particle. The source of this diffusion is the surface concentration of adsorbate, that is, it is postulated that


223

movement from the solution to the surface is relatively quick-and is therefore not rate limiting. Before we consider the consequences of this assumption, let us consider its feasibility. The variable-charge materials of soil differ from those prepared in the laboratory in several ways. They are far from pure. Iron oxides may contain a considerable portion of aluminum atoms (Taylor et al., 1983); Schwertmann, 1984). Possibly this is one source of strain in the crystals and thus a source of the vacancies which would increase diffusion. Natural iron oxides may also contain appreciable silicon and phosphorus (Norrish and Rosser, 1983). They also have a much lower point of zero charge than pure oxides, and it seems likely that this is a consequence of the contained silicon and phosphorus. It is possible that the silicon and phosphorus atoms are scattered throughout the oxide. However, recent work using high-resolution electron microscopy suggests that it is more organized than this. Smith and Eggleton (1983) have shown that natural goethites are composed of very fine needles with a cross section averaging about 3 0 0 k Between these needles there appears to be a “grouting,” and it is consistent with their data that this grouting is a monolayer of silica. If this is the case for soil goethites, it would provide a feasible pathway for diffusion of anions. Smith and Eggleton (1983) also note that the fineness of the needles causes a broadening of the X-ray diffraction profiles. A similar broadening is observed for soil goethite. This has been taken to indicate that soil goethites occur in very small particles. If an adsorbed ion diffuses into the interior of the adsorbing material, a surface site will be vacated. However, the electrostatic potential of the surface will be less favorable for adsorption than it was originally. Hence, even if the solution concentration of the ion is kept constant, the surface concentration will decrease. There is thus the paradox that, while the total retention increases, the true adsorption actually decreases. This provides important feedback, because it is the surface concentration that is the source for the solid-state diffusion mechanism. If the surface concentration were constant, we would expect the amount diffusing to be proportional to the square root of time. Because it is not constant, the amount diffusing is found to be proportional to a smaller power of time. A solid-state diffusion mechanism was also used by van Riemsdijk et al. (1984) to describe the effects of time on the reaction of phosphate with soils. However, they did not consider the effectsof reaction on charge. Rather, they explained the observation that the rate was not proportional to the square root of time due to the presence of a range of different particles. In each particle the square root relationship held, but the observed rate was the sum of the individual sites. A mathematical specification of this soil model was given by Barrow (1983b). It suffices to note here that the electrostatic potential in the plane of adsorption plays a central role in this model. It is used to describe the

224

N. J. BARROW

heterogeneity of the soil, and it is the mechanism by which the feedback effects of adsorption and of diffusive penetration modify surface concentration via Eq. (13). The model takes the changing surface concentration of adsorbate into account in a noncontinuous way, that is, by breaking the process into a number of steps. This permits the operator to interrupt the process by changing the solution concentration and thus induce desorption. 2. Application of the Quantitative Model The quantitative model reproduces the effects of concentration on retention (Figs. 12, 13, 14, and 15). This is largely by appropriate choice of the parameters that describe the distribution of values of the potential $a. Examples are given in Barrow (1983b) to show the effects of changing the value of these parameters on the shape of the plots of retention against concentration. The effect of time and temperature of incubation are also closely reproduced (Figs. 13, 14, and 15). The effects of temperature arise because solidstate diffusion involves a series of jumps from a favored position in the crystal, over an energy barrier, to another favored position. Thus there is a substantial activation energy and the process is faster at high temperatures. This contrasts with diffusion in solution, for which the activation energy is low. The process limiting the rate of desorption is reverse diffusion in response to a change in gradient. The activation energy for desorption is therefore the same as that for sorption. However, it can be shown that the value of the binding constant ki of Eq. (13) decreases as temperature increases (Barrow, 1983b). Thus, if the adsorption reaction per se can be isolated, higher temperatures produce a lower value for the binding constant, and hence a higher solution concentration is needed to maintain the same adsorption. That is, for the adsorption reaction per se, high temperatures give high solution concentrations. The model thus encompasses the effects of temperature. For phosphate, the effects of pH on retention are closely described (Fig. 16). The important aspect of the model, in this case, is the assumed rate of change in the electrostatic potential $,with pH. It is necessary to assume that the decrease with increasing pH is less than that observed for pure goethite and that it is almost counterbalanced by the increasing concentration of HPOi- ions (Barrow, 1984). As a result, the overall effect is fairly small, and the direction of the effect can change through the pH range and can differ between soils. For hydrolyzable cations, the model postulates that there is a synergistic effect. With increasing pH, the increasing proportion of MeOH' ions and the

REACTION OF ANIONS A N D CATIONS

50 -

225

Pb

10 5 0.

I

10.~

10.2

I 10''

I 1.0

I 10

. -

3 101 0

5

4.8

1

5.3

0

6.2 A 6.7 v

2 10.~

I

I

I

10-2

10-1

1.o

I 10

Concentration of MeOH (prnol//)

FIG.19. Replotting the data of Fig. 17 against the calculated concentrations of PbOH', CuOH', ZnOH', and NiOH' ions.

decreasing value of $, both favor increased retention. The contribution for the $a term can be assessed by plotting retention against calculated values of the concentration of MeOH' ions. This requires values for the pH of each observation because reaction with cations releases protons. Values for pH will therefore vary not only with lime treatments but also with the level of addition of cation. No published data which gave values for the final pH could be located. The nearest approximation, therefore, was to use the initial pH. Details of the extensive observations of Harter (1983) were kindly provided by Professor

226

N. J. BARROW

Harter and are plotted against the concentration of MeOH' in Fig. 19. The figure shows that there was a strong tendency for the points to cluster about a common line. Most of the outlying points were either at low pH or were for the lowest concentration of a pH set. Both of these could be imprecise due to the difficulties of measuring small amounts of retention or low concentrations. Figure 19 seems to provide a useful way of investigating metal retention and one that could be more widely used. It suggests, for example, that the affinity of the soil for nickel is higher than for the other metals because nickel is retained at a lower concentration of MeOH' ions. It also suggests that, for this soil, there was little contribution to the effects of pH from the I),term, that is, that change in the electrostatic potential with pH had little effect on metal retention. This conclusion depends, of course, on the assumption that the reacting ion is indeed MeOH'. If we do accept this assumption, it follows that the metals were retained by surfaces that are not variable in charge. One possibility is that the reaction was entirely with the fixed-charge components of the soil. Another possibility is that the reaction was, at least partly, with the metal oxides in soil, but that these were closely associated with a large negative charge. This could be from close association with fixed-charge clays or from extensive penetration of the oxide with an anion such as phosphate. It can be shown that close association of an oxide with negative charge can result in all the surface sites being occupied by protons balancing this charge. Variation in the H f concentration in solution then has no effect on the surface charge. Cations would tend to react preferentially with such negative surfaces. This may seem speculative, but it is included here to show how consideration of a model can lead to further insights and to testable hypotheses. As an extension of this process, consider the persistence of metal oxide particles in soils. Over a very long time, the more favored particles will tend to grow at the expense of the less favored due to a greater ability to retain the appropriate metal ions. This depends not only on the chemical stability of the particular oxide but also on its charge. It is therefore advantageous for a particle to be negatively charged. The inclusion of anions in metal oxides in soil may not be merely accidental; it may be a necessary component in their survival. Desorption is also well described by the model (Fig. 18; Barrow, 1983~).It is important to note that desorption is modeled in a mechanistic sense and using parameters chosen to describe the sorption side of the model. Desorption is slow because it involves a reversal of the slow, solid-state diffusion. The longer the nutrient has had to react with a soil, the more deeply it will have penetrated and hence the more slowly it can be recovered by desorption.


227

V. CONCLUSIONS In this article I have outlined models that seem to provide a consistent explanation of many seemingly disparate observations. The models are attempts to describe bulk behavior from a consideration of molecular behavior. As such they involve two principles of modeling: one is the principle of sufficiency, and the other is the principle of hierarchies. Although the models may seem complex, they are, in fact, greatly simplified. For example, it is assumed that the surfaces with which ions react are so large relative to the ions that they can be regarded as planes. This is an example of the principle of sufficiency-models are made just sufficiently complex to describe the observations at hand. A valuable way of testing models, therefore, is to test whether they can describe further observations. There is an important difference between retrospective fitting of a model to existing observations and prospective testing of a model by making observations with the specific aim of testing whether the model is adequate or whether additional complexity needs to be included. Hopefully this will be an important future activity. The principle of hierarchies means that complex models may be used at one level of detail but these are then abstracted and simplified to describe behavior at a more general level. Thus the model to describe total soil behavior contains a simplified and abstracted version of the model that describes the behavior of soil constituents. This process may be continued. For example, Eq. (17) might be treated as a simplified description of part of the soil behavior and used as an input to a more general model of, say, leaching or uptake of nutrients. It is an extension of this hierarchical process that I shall now use in attempting to summarize the article by a few short statements. The reactions of a diverse range of ions with soil may be described by three rules. The first is that there is an initial reaction between a variable-charge surface and certain of the ions in solution, that is, reaction is between charged surfaces and charged particles. The second is that the soil surfaces are heterogeneous. And the third rule is that the initial surface reaction is followed by a diffusive penetration, the rate of which is modified by electrostatic effects. The consequences of the first rule are largely shown in the effects of pH on retention. They depend on the sign and magnitude of the charge on the ionic species that reacts with the surface, the pH at which this reacting ion is present, and the closeness of approach of the adsorbed ion species to the surface. As a result, the effect of pH differs between anions and cations and differs been monovalent adsorbing anions (e.g., F-, H3Si04) and divalent adsorbing anions (e.g., HPOi-, SeOZ-).

228

N. J. BARROW

The consequences of the heterogeneity are largely shown in the shape of the relation between concentration and retention, the Q/I relation. If the model is correct in this respect, it is futile to seek simpler “mechanistic” equations to describe the shape of such relations. Among the consequences of the diffusive penetration are the continuing reaction between soil and nutrient, the slow desorption of added nutrient, and, from the agronomic point of view, the need for repeated applications of fertilizers. The unusual kinetics of this process are explained by feedback via the electrostatic effects. Because of this, the rate is not proportional to the square root of time but rather to a small fractional power. In modern agriculture, most fertilizer is applied to previously fertilized land. The continuing reaction between the soil and the previously added nutrients is one of the main determinants of the need for further fertilization. It is ironic that this process has been ignored in so many studies on soil chemistry. It is to be hoped that this will be rectified in the future.

REFERENCES Baes, C. F., and Mesmer, R. E. 1976. “The Hydrolysis of Cations.” Wiley, New York. Barrow, N. J. 1970. Soil Sci. 109, 282-288. Barrow, N. J. 1974. S ~ i Sci. l 118, 380-386. Barrow, N. J. 1978. J. Soil Sci. 29, 447-462. Barrow, N. J. 1979a. J. Soil Sci. 30, 259-270. Barrow, N. J. 1979h. J. Soil Sci. 30,271-279. Barrow, N. J. 1980a. In “The Role of Phosphorus in Agriculture” (F. E. Khasawneh, E. C. Sample, and E. J. Kamprath, eds.), pp. 333-359. ASA-CSSA-SSSA, Madison, Wisconsin. Barrow, N. J. 1980b. Ausr. J. Soil Res. 18, 215-224. Barrow, N. J. 1983a. Fert. Res. 4, 51-61. Barrow, N. J. 1983b. J. Soil Sci. 34, 733-750. Barrow, N. J. 1983c. J. Soil Sci. 34, 751-758. Barrow, N. J. 1984. J. Soil Sci. 35, 283-297. Barrow, N. J. 1985a. J. Soil Sci. (submitted). Barrow, N. J. 1985b. J. Soil Sci. (submitted). Barrow, N. J., and Shaw, T. C. 1975a. Soil Sci. 119, 167-177. Barrow, N. J., and Shaw, T. C. 1975b. Soil Sci. 119,301-310. Barrow, N. J., and Shaw, T. C. 1975c. Soil Sci. 119, 311-320. Barrow, N. J., and Shaw, T. C. 1977a. Soil Sci. 124, 265-278. Barrow, N. J., and Shaw, T. C. 1977b. Soil Sci. 124, 347-354. Barrow, N. J., Bowden, J. W., Posner, A. M., and Quirk, J. P. 1980a. Ausi. J. Soil Res. 18,37-47. Barrow, N. J., Bowden, J. W., Posner, A. M., and Quirk, J. P. 1980b. Aust. J. Soil Res. 18, 395-404. Barrow, N. J., Bowden, J. W., Posner, A. M., and Quirk, J. P. 1981a. Aust. J. Soil Res. 19, 309-32 1. Barrow, N. J., Madrid, A. M., and Posner, A. M. 1981b. J. Soil Sci. 32, 399-407. Benjamin, M. M., and Leckie, J. 0. 1981. J. Colloid Interface Sci. 79,209-221.

REACTION O F ANIONS AND CATIONS

229

Berkheiser, V. E., Street, J. J., Rao, P. S. C., and Yuan, T. L. 1980. CRC Crit. Rev. Environ. Control 10, 179-224. BCrube, J. Y., Onada, G. Y., and de Bruyn, P. L. 1967. Surf. Sci. 8,448-461. Bolan, N. S., and Barrow, N. J. 1984. J. Soil Sci. 35, 273-281. Bolt, G. H., and van Riemsdijk, W. H. 1982. In “Soil Chemistry: B, Physico-chemical Models” (G. H. Bolt and M. G. M. Bruggenwert, eds.), pp. 459-504. Elsevier, Amsterdam. Bowden, J. W., Posner, A. M., and Quirk, J. P. 1977. Aust. J. Soil Res. 15, 121-136. Bowden, J. W., Posner, A. M., and Quirk, J. P. 1980a. In “Soils with Variable Charge” (B. K. G. Theng, ed.), pp. 147-166. New Zealand Society of Soil Science, Lower Hutt. Bowden, J. W., Nagarajah, S., Barrow, N. J., Posner, A. M., and Quirk, J. P. 1980b. Aust. J. Soil Res. 18,49-60. Brennan, R. F., Gartrell, J. W., and Robson, A. D. 1980. Aust. J. Soil Res. 18,447-459. Brennan, R. F., Gartrell, J. W., and Robson, A. D. 1984. Aust. J. Soil Res. 22, 165-172. Bye, G. C., McEvoy, M., and Malati, M. A. 1983. J. Chem. SOC.Faraday Trans. I 79,2311-2318. Cabrera, F., De Arambarri, P., Madrid, L., and Toca, C. G. 1981. Geoderma 26,203-216. Chien, S . H., and Clayton, W. R. 1980. Soil Sci. SOC.Am. J. 40, 265-268. Cornell, R. M., Mann, S., and Skarnulus, A. J. 1983. J. Chem. SOC.Faraday Trans.79,2679-2684. Davis, J. A,, and Leckie, J. 0. 1978. J. Colloid Interface Sci. 67, 90-107. Davis, J. A., and Leckie, J. 0. 1980. J. Colloid Interface Sci. 74, 32-43. Davis, J. A., James, R. O., and Leckie, J. 0. 1978. J. Colloid Interface Sci. 63, 480-499. Elkhatib, E. A., Bennett, E. A., and Wright, R. J. 1984. Soil Sci.SOC.Am. J. 48, 758-762. Fordham, A. W., and Norrish, K. 1983. Aust. J. Soil Res. 21,455-477. Goldberg, S., and Sposito, G. 1984. Soil Sci. SOC.Am. J . 48, 772-778. Harter, R. D. 1983. Soil Sci. SOC.Am. J. 47,47-51. Hatcher, J. T., Bower, C. A,, and Clark, M. 1967. Soil Sci. 104, 422-426. Haynes, R. J. 1982. Plant Soil 68, 289-308. Hingston, F. J. 1970. Ph.D. thesis, University of Western Australia. Hingston, F. J. 1981. In “Adsorption of Inorganics at Solid-liquid Interfaces” (M. A. Anderson and A. J. Rubin eds.), pp. 51-90. Ann Arbor Science Publ., Ann Arbor, Michigan. Hingston, F. J., Posner, A. M., and Quirk, J. P. 1972. J. Soil Sci. 23, 177-192. Hingston, F. J., Posner, A. M., and Quirk, J. P. 1974. J. Soil Sci. 25, 16-26. James, R. O., and Healy, T. W. 1972. J. Colloid Interface Sci. 40, 65-81. James, R. O., Stiglich, P. J., and Healy, T. W. 1980. Adsorption from aqueous solutions. A.C.S. Symp. Ser. Keren, R., and Mezuman, U. 1981. Clays Clay Miner. 29, 198-204. Keren, R., Cast, R. G., and Bar-Yosef, B. 1981. Soil Sci. SOC.Am. J. 45,45-48. Kuo, S., and Mikkelsen, D. S. 1979. Soil Sci. 128, 274-279. McKenzie, R. M. 1980. Aust. J. Soil Res. 18, 61-73. McKenzie, R. M. 1983. Aust. J. Soil Res. 21, 505-513. Madrid, L., and Arambarri, P. 1978. Geoderma 21, 189-208. Madrid, L., and Posner, A. M. 1979. J. Soil Sci. 30, 697-707. Manning, J. R. 1968. “Diffusion Kinetics for Atoms in Crystals,” 1st Ed. Van Nostrand, Princeton, New Jersey. Mikami, N., Sasaki, M., Hachiya, K., Astumian, R. D., Ikeda, T., and Yasunaga, T. 1983. J. Phys. Chem. 87, 1454-1458. Morel, F. M. M., Westall, J. C., and Yeasted, J. G. 1981. In “Adsorption of Inorganics at Solidliquid Interfaces” (M. A. Anderson and A. J. Rubin eds.), pp. 263-294. Ann Arbor Science Publ., Ann Arbor, Michigan. Mott, C. J. B. 1981. In “The Chemistry of Soil Processes” (D. J. Greenland and M. H. B. Haynes, eds.), pp. 179-220. Wiley, New York.

230

N. J. BARROW

Muljadi, D., Posner, A. M., and Quirk, J. P. 1966. J. Soil Sci. 17, 238-247. Nanzyo, M., and Watanabe, Y. 1982. Soil Sci. Plant Nu@. 28, 359-368. Norrish, K. L., and Rosser, H. 1983. In “Soils: An Australian Viewpoint” (Division of Soils, CSIRO, eds.), pp. 335-361. CSIRO Melbourne/Academic Press. Padmanabham, M. 1983a. Aust. J. Soil Res. 21,308-320. Padmanabham, M. 1983b. Aust. J. Soil Res. 21, 515-525. Parfitt, R. L., Russell, J. D., and Farmer, V. C. 1976. J. Chem. SOC.Faraday Trans. 1 72, 1082-1087. Posner, A. M., and Barrow, N. J. 1982. J. Soil Sci. 33,211-217. Probert, M. E. 1980. Aust. J. Exp. Agric. Anim. Husb. 20, 240-246. Pyman, M. A. F., Bowden, J. W., and Posner, A. M. 1979. Aust. J. Soil Res. 17, 191-195. Sanchez, 0. A,, and Uehara, G. 1980. In “The Role of Phosphorus in Agriculture” (F. E. Khasawneh, E. C. Sample, and E. J. Kamprath, eds.), pp. 471-514. ASA-CSSA-SSSA, Madison, Wisconsin. Schindler, P. W. 1981. In “Adsorption of Inorganics at Solid-liquid Interfaces” (M. A. Anderson and A. J. Rubin, eds.), pp. 1-49. Ann Arbor Science Publ., Ann Arbor, Michigan. Schwertmann, V. 1984. Z . Pflanzenernaehr. Bodenkd. 147, 385-399. Sharpley, A. N., Ahuja, L. R., Yamamoto, M., and Menzel, R. G . 1981. Soil Sci. SOC.Am. J. 45, 493-496. Sigg, L., and Stumm, W. 1980. Colloids & SurJ 2, 101-117. Sims, J. R., and Bingham, F. T. 1967. Soil Sci. SOC.Am. Proc. 31, 728-732. Sims, J. R., and Bingham, F. T. 1968a. Soil Sci. SOC.Am. Proc. 32, 364-369. Sims, J. R., and Bingham, F. T. 1968b. Soil Sci. SOC.Am. Proc. 32, 369-373. Smith, B. H., and Leeper, G. W. 1969. J. Soil Sci. 20, 246-254. Smith, K. L., and Eggleton, R. A. 1983. Clays & Clay Miner. 31, 392-396. Sposito, G. 1980. Soil Sci. SOC.Am. J. 44,652-654. Sposito, G. 1983. J. Colloid Interface Sci. 91, 329-340. Taylor, R., McKenzie, R. M., Fordham, A. W., and Gillman, G. P. 1983. In “Soils: An Australian Viewpoint” (Division of Soils, CSIRO, eds), pp. 307-334. CSIRO Melbourne/Academic Press. Van Olphen, H. 1977. “An Introduction to Clay Colloid Chemistry,” 2nd Ed. Wiley, New York. Van Raij, B., and Peech, M. 1972. Soil Sci. SOC.Am. Proc. 36, 587-593. Van Riemsdijk, W. M., and de Haan, F. A. M. 1981. Soil Sci. SOC.Am. J. 45,261-266. Van Riemsdijk, W. M., Boumans, L. J. M., and de Haan, F. A. M. 1984. Soil Sci. SOC.Am. J. 48, 537-541. Wei, Y.K., and Bernstein, R. B. 1959. J. Phys. Chem. 63, 738-741. Westall, J., and Hohl, H. 1980. Adu. Colloid Interface Sci. 12, 265-294. White, R. E. 1980. In “Critical Reports on Applied Chemistry, Vol. 2. Soils and Agriculture” (P, B. Tinker, ed.), pp. 71-114. Blackwell, Oxford. Yates, D. E., Levine, S., and Heally, T. W. 1974. Trans. Faraday SOC.70, 1807-1818. Zasoski, R. J., and Burau, R. G. 1978. Soil Sci. SOC.Am. J. 42, 373-374.


KINETICS OF IONIC REACTIONS IN CLAY MINERALS AND SOILS Donald L. Sparks

.

Department of Plant Science College of Agricultural Sciences. Universitv of Delaware. Newark . Delaware

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I1. Methodologies Used in Kinetic Studies . . . . . . . . . . . . . . . . . . . . . . . . . A . Batch Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Flow or Miscible Displacement Techniques . . . . . . . . . . . . . . . . . . . . 111. Application of Chemical Kinetics to Soil Solutions . . . . . . . . . . . . . . . . . . A. Theoretical Aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. BasicEquations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV. Rate-Determining Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Concurrent Processes Involved in Ion Exchange . . . . . . . . . . . . . . . . . B. Identifying Rate-Limiting Steps in Soil Solutions . . . . . . . . . . . . . . . . . V. Kinetics Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A . Use of Kinetic Data to Calculate Thermodynamic Parameters . . . . . . . . . B. Eyring’s Reaction Rate Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Comparison of Kinetic and Equilibria1 Models to Describe Thermodynamic Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI . Kinetics of Ionic Exchange in Clay Minerals . . . . . . . . . . . . . . . . . . . . . . A. Rate of Exchange on Kaolinite, Montmorillonite, and Illite. . . . . . . . . . . B. Rate of Exchange on Vermiculite and Micaceous Clays . . . . . . . . . . . . . C . Rate of Reaction on Other Pure Surfaces . . . . . . . . . . . . . . . . . . . . . VII. Kinetics of Ionic Reactions in Heterogeneous Soil Systems. . . . . . . . . . . . . . A. Potassium Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Phosphorus Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Nitrogen Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Sulfur Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E. Microelement and Heavy Metal Kinetics . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

231 233 234 235 238 238 240 251 251 252 256 256 257 258 258 258 259 260 261 261 262 262 263 263 264

I. INTRODUCTION Chemical kinetics as applied to the soil solution might be the least investigated and is certainly one of the most enigmatic areas in soil chemistry. The equilibrial. rather than kinetic. aspects of ionic reactions in clay and soil 23 1

Copyright 0 1985 by Academic Press. Inc. All rights of reproduction in any form reserved.

232

DONALD L. SPARKS

systems have received considerable attention over the years and are reviewed in detail by Sposito (1981a) and Goulding (1983). These thermodynamic investigations have proved valuable, but they do not assist one in understanding the mechanisms and rates of ionic reactions in clay minerals and soils, nor are they often applicable to field conditions. Agricultural soils are nearly always in a state of disequilibrium with regard to ion transformations. Soils that have been intensively cropped and fertilized with optimal fertilizer additions for many years belong in this group because equilibrium is precluded by periodic additions of fertilizers. To fully understand and model elemental reactions in soils, a knowledge of the kinetics is fundamental. Unfortunately, kinetic investigations of clay and soil systems have been sparse. This is particularly true for soils that contain complex mixtures of clay minerals, noncrystalline components, oxides, hydroxides, and organic matter. This complex array of inorganic and organic components exacerbates the difficulty in studying ion exchange kinetics. Perhaps our lack of knowledge in this area was most elegantly stated by Professor G. W. Thomas (1977) in an address entitled “Historical Developments in Soil Chemistry: Ion Exchange”: Our understanding of ion exchange in soils, where the exchangers are complex mixtures of clay minerals, oxides, and organic matter, has not kept pace with the work on purer exchange material. Further, the details of how and the rate at which the ion exchange reactions occur are not understood, even in ideal systems.

J. Thomas Way, whom many consider as the patriarch of soil chemistry, was the first to study ion exchange kinetics. In his classic and pioneering research, he showed that the rate of NH: adsorption on a British soil was rapid and concluded that ion exchange was an instantaneous process. Similar conclusions with other cations were found by Gedroiz (1914) in Russia and by Hissink (1924) in Holland. However, Kelley (1948), in his beautiful book “Cation Exchange in Soils,” questioned whether these investigators’ results would be valid in all soils. He speculated, and correctly so, that ion exchange would not be an instantaneous process in vermiculitic and micaceous systems. Other than Kelley, it seems that most researchers accepted the conclusions of previous workers and did not bother to corroborate or to refute their findings. Accordingly, little further work on ion exchange kinetics was conducted until the classic work of Boyd et al. (1947) appeared. Working in conjunction with the Manhattan project, they investigated ion exchange kinetics on organic zeolites and elucidated mechanisms and rate-determining steps in ion exchange kinetics. The purpose of this article is not to give in-depth discussions on pure chemical kinetics. For these, the reader should consult the excellent books of

KINETICS OF IONIC REACTIONS

233

Frost and Pearson (1961), Hammes (1978), and Laidler (1965). Rather, the intentions of this article are to apply the principles of kinetics to clay minerals and soils and to review recent literature dealing with the kinetics of elemental reactions in these systems.

II. METHODOLOGIES USED IN KINETIC STUDIES One of the most important aspects of any kinetic study is the method one employs to measure rate parameters. Unfortunately, many soil scientists have used techniques that are not suitable for measuring rapid ion exchange reactions and that sometimes alter the surface chemistry of the exchangers they are studying. The study of ion exchange kinetics has seen the development and use of several methodologies. H. S . Thompson (1850) and J. T. Way (1850) conducted cation exchange experiments using columns of soil through which adsorbate solutions were leached. As the adsorbate solution passed through the soil it miscibly displaced the existing solution, allowing the incoming cations to displace the existing cations on the colloid. The displaced cations were removed from the reaction site as the leachate was collected. However crude, the above experiments were likely the first examples of miscible displacement or flow techniques. Way’s work was criticized by the renowned chemist Liebig (Thomas, 1977). Liebig believed that Way’s “exchange” was simply due to cations being held within the capillaries of the soil column, much like water in a sponge. If true, this implied that the length and packing of the column would affect the exchange capacity of the adsorbent. The realization that columns of the same soil did not always yield similar results lead soil scientists, including Way, to use digestion procedures (Kelley, 1948). These procedures involved placing known quantities of adsorbent and adsorbate in a closed vessel, and after time (digestion), analyzing the liquid, viz., the beginning of the batch technique. The above techniques were quite cumbersome, and thus did not enable soil scientists to accurately study the rates of the exchange reactions. When expanded over time these methodologies lead to the general acceptance of Way’s conclusion: exchange reactions were instantaneous. Not until the work of Kelley (1948) was Way’s conclusion about the kinetics of ion exchange seriously questioned. The methodologies that have evolved from these early authors, and that are now used most often, are the flow or miscible displacement technique and the closed vessel or batch technique.

234

D O N A L D L. SPARKS

A. BATCHTECHNIQUES The bulk of kinetic studies with soil and clay minerals have employed batch techniques. These involve placing an exchanger and the adsorbate in a vessel such as a centrifuge tube. The suspension is then agitated using a reciprocal shaker or it is stirred. Then, the suspension is usually centrifuged to obtain a clear supernatant solution for subsequent analysis. The use of batch techniques to measure the rate of ionic reactions in clay minerals and soils has a number of disadvantages. Unless the batch technique of Zasoski and Burau (1978) is used, most batch techniques require centrifugation which requires about five minutes to separate the solid from the liquid phases. Many exchange reactions are complete by this time or less (Harter and Lehmann, 1983; Jardine and Sparks, 1984a). If the usual batch technique is employed, one will never observe these reactions. This is a particular problem in soils that contain organic matter, kaolinite, and montmorillonite, and other layer silicates that have readily accessible sites for exchange. In these systems, ion exchange kinetics are very rapid (Sparks and Jardine, 1984), and most batch techniques should not be used. However, if one uses the batch technique of Zasoski and Burau (1978) which is a filtration technique, one can observe rapid exchange reactions. Also, to properly measure the kinetics of a chemical reaction, the technique must not change the reactant concentration (Zasoski and Burau, 1978). Thus, the sample and the suspension should have a similar solid to solution ratio at all times. Unfortunately, this has not been true in most batch studies, as Barrow (1983) discusses in an excellent review article. Most studies have employed large solution: soil ratios where the concentration in the solution and the quantity of adsorption both vary simultaneously. Exceptions are the studies of Zasoski and Burau (1978), Van Riemsdijk (1970), and Van Riemsdijk and DeHaan (1981). In the latter two studies, the solution concentration was held constant. Thus, the relation between adsorption and time should only be treated as a simple, two-dimensional relation if adsorption is constant. Unless a batch technique similar to that of Van Riemsdijk and DeHaan (1981) is utilized, these conditions do not apply to experiments in which a wide so1ution:soil ratio is used. Another problem with the batch technique is mixing the solution and exchanger. If mixing is inadequate, the rate of reaction will be limited. However, vigorous mixing, which many investigators have used, can cause abrasion of the exchanger particles leading to high rates of reaction and alterations in the surface chemistry of the particles (Barrow and Shaw, 1979). Barrow and Shaw (1979) studied the rate of phosphate adsorption by soil and found that under prolonged agitation, breakdown of soil particles occurred, increasing surface area, and causing greater phosphate adsorption. The

235


breakdown of particles during shaking was less marked at high so1ution:soil ratios. The abrasion of colloidal particles could be a serious problem when studying most ionic reactions, but particularly with potassium since release could occur from potassium-bearing minerals. A type of batch technique that has been employed by a number of investigators is the isotopic exchange method. This has been particularly true with soil phosphorus. Many authors have suggested that the fraction of total soil phosphorus which is readily accessible to isotopic exchange may represent the phosphorus available to plants (White, 1976). Investigations using isotopic exchange methods are evaluated in several ways. The earliest approach was to use a linear combination of exponential terms analogous to a series of simultaneous first-order reactions as (Neuman and Neuman, 1958):

x = 1 - a, exp(-k,t)

- u2 exp(-k,t)

-

u3 exp(-k,t)

- a,exp(-k,t)

(1)

where x is the fraction of the tracer in the adsorbed state at time t ; k,, k,, and k, are the adsorption rate constants; and a,, a,, and u3 are constants such that the sum is equal to 1. However, the process of fitting experimental data to the above equation is quite cumbersome. Probert and Larsen (1972) adopted a two-constant formula to approximate the sum of exponential terms. Their formula was 1 - x = [(t

+

y)/y-b]

where y and b are constants. Probert and Larsen (1972) report that Eq. (2) adequately described 32Pexchange data in several soils. It must be pointed out that neither of the two equations above makes a reference to adsorption or desorption rates per se. The equations simply describe the rate of disappearance of the tracer from solution and its exchange with solid-phase phosphorus. Presumably, this involves both adsorption and desorption reactions, and the algebraic sum of both gives the net adsorption rate of the isotope. For both of these equations, equilibrium is asymptotically approached with increasing time. B. FLOW OR MISCIBLE DISPLACEMENT TECHNIQUES

Flow or miscible displacement techniques have been used to a lesser extent to investigate the kinetics of reactions in clay minerals and soils (Sivasubramaniam and Talibudeen, 1972; Sparks et ul., 1980b; Sparks and Jardine, 1981; Jardine and Sparks, 1984a). However, flow techniques are increasingly being recommended over batch studies to study sorption-desorption phenomena on colloids, particularly if one wishes to relate kinetic studies to solute transport under field conditions (Murali and Aylmore, 1981, 1983a,b).

236

DONALD L. SPARKS

Perhaps Murali and Aylmore (1983b) best stated this: It seems self-evident that adsorption studies should be performed under conditions close to those encounteredin the field, viz., realistic water contents with no shaking or agitation,if these results are to be related to solute transport models.

Most previous adsorption studies employing a batch technique were performed at high solution-to-solid ratios with continuous shaking or stirring. Such experiments usually yielded reaction rates that were instantaneous. Many investigators thus, incorrectly, concluded that all ion exchange processes were instantaneous. This conclusion was often reached because one could not measure short reaction times using the batch technique. However, flow studies performed at realistic solution-to-solid ratios (usually I1) clearly indicate that for many chemical species of interest, such as potassium, phosphate, and selenite, the solute-solid interactions are much slower than with batch techniques (Murali and Aylmore, 1983a,b; Sparks and Rechcigl, 1982). The amount of solution in contact with colloidal particles is also an important attribute of a flow technique. Supplied with a solution of the same concentration, soil particles with solution flowing past them will be exposed to a greater mass of ions (concentration x flow rate x time) than the soil particles in a static system (concentration x solution volume) by the time equilibrium is established. More importantly, with solution flowing through the soil system, the solution not only brings in more ions but also removes the desorbed ions of other species that were present originally at potential sorption sites (Akratanakul et al., 1983). This is particularly important in studying potassium reactions since small amounts of potassium in the equilibrium solution will prevent further release of adsorbed potassium which, consequently, results in marked hysteresis (Martin and Sparks, 1983). Additionally, the number of introduced ions that can be adsorbed also depends on how easily they can be exchanged with other ions of different kinds already adsorbed by the soil surfaces, thus affecting the magnitude of the heat of adsorption. With a closed system, exchange cannot be complete without increasing the concentration of the replaced ions in the bulk solution. This would, in turn, drive these ions back into the adsorbed phase. However, in an open flow system, the exchange can be complete, as the replaced ions are continuously carried out of the system while more of the introduced ions can take their place. Sparks et al. (1980b) developed a miscible displacement technique to study kinetics of potassium adsorption from soils. With this technique, a soil suspension is injected with a syringe into a 47-mm Nucleopore filter (Fig. 1). The filter is attached to a fraction collector, and the sample is leached with KCl and CaCl, for adsorption and desorption studies, respectively. The


237

FIG.1. Miscible displacement technique.

electrolyte is passed through the soil at a constant flow rate using a peristaltic pump, and aliquots are collected at various times. Later, Jardine and Sparks (1984a) modified the technique so that aliquots of leachate could be taken at 2-min increments. Thus, one could investigate very rapid ionic reactions which could not be measured with a batch technique. Using the miscible displacement technique of Jardine and Sparks (1984a), one can investigate adsorption and desorption on the same soil sample. This technique has proved extremely advantageous in studying the kinetics of potassium reactions in soil since solution phase potassium is constantly being removed and no inhibition of further potassium release occurs (Sparks et al., 1980b; Sparks and Jardine, 1981; Sparks and Rechcigl, 1982; Jardine and Sparks, 1984a). Thus Sparks and co-workers have obtained almost complete reversibility in potassium exchange. This would not have been possible using a batch technique. The miscible displacement technique developed by Sparks et al. (1980b) was recently modified by Carski and

238

DONALD L. SPARKS

Sparks (1985) to allow the study of systems which are known to adsorb very small quantities of ions. In summary, the flow technique has several advantages over the traditional batch technique for studying kinetics in clay minerals and soils: (1) it more closely simulates ionic reactions under field conditions, (2) one can measure short reaction times, (3) one avoids separation of liquid from solid phases by centrifugation; and (4) one can maintain a relatively constant solid: solution ratio.

111. APPLICATION

OF CHEMICAL KINETICS TO SOIL SOLUTIONS

A. THEORETICAL ASPECTS Chemical kinetics is one of the most fascinating, yet one of the most difficult areas in physical chemistry. Before applying chemical kinetics to soil solutions, we shall discuss some of the theoretical aspects of this topic. Chemical kinetics deals with chemical reaction rates and how these rates can be explained in terms of reaction mechanisms (Laidler, 1965). There are two salient reasons for studying the rates of chemical reactions: ( 1 ) to predict how quickly a reaction mixture will move to its equilibrium state, and (2) to reveal reaction mechanisms. Thus kinetics, unlike thermodynamics, provides information along each step of a reaction pathway. Unfortunately, due to theoretical and experimental difficulties, it is often arduous to apply pure chemical kinetics to even simple homogeneous solutions! When kinetic theories are applied to soil solutions, the problems are intensified. To fully comprehend the above ideas, a knowledge of the rate equations or rate law explaining the reaction system is required. Acquiring an empirical rate law necessitates knowledge of the concentration of the reactants and the stoichiometric equation, as well as the mechanism of product formation. One can express the rate equation as rate = - l/vi d[i]/dt

(3) where [ i ] is the concentration of reactant i, t is time, and v is a stoichiometric parameter. The dependence of rate on reactant concentrations is expressed by the law of mass action. Thus, for a given stoichiometric reaction, v,A

+ v,B

-

vtC

+ odD

rate = - 1/ui d [ i ] / d t = - I / u i k [ ~ ] " [ B I b -

(4)


239

where a and b indicate the reaction order for the individual constituents, [ A ] and [B] are the concentrations of the reactants A and B, respectively, and k is the rate constant. One should realize that the rate law is determined by experimentation and it cannot be inferred by simply examining the overall chemical reaction equation. The rate law serves three primary purposes: (1) it permits the prediction of the rate, given the composition of the mixture and the experimental value of the rate constant or coefficient; (2) it enables one to propose a mechanism for the reaction; and (3) it provides a means for classifying reactions into various orders. The order of a reaction is the summation of the powers to which the concentrations of the components are raised in the rate law. A number of equations have been employed to describe the kinetics of reactions in clay minerals and soils (Chien et al., 1980; Sparks and Jardine, 1981,1984; Martin and Sparks, 1983; Jardine and Sparks, 1984a). These have included the first-order, Elovich, parabolic diffusion, zero-order, secondorder, and two-constant rate equations. Since a comprehensive article on kinetics as applied to soil solutions has not been previously published, complete derivations of most of these equations will be given. The final forms of some of the above equations for adsorption kinetics using a miscible displacement technique are given in Table I.

Table I Equations Describing the Kinetics of Adsorption Reactions in Clay Minerals and Soils Using a Miscible Displacement Technique" 1. Elovich: C, = a + b In t

2. Parabolic diffusion law: C,/C,

=a

+ btIi2

3. First order:

log(1 - C,/C,) = a - bt 4. Zero order: (1 - CJC,) a

= a - bt

The terms in each equation are defined in the text.

240

DONALD L. SPARKS

B. BASICEQUATIONS I . First-Order Equations

For a batch technique, if the rate of adsorption of an ion onto a colloidal surface is proportional to the quantity of the ion remaining in solution, a firstorder equation is expressed as

d(C0 - C)/dt = k,C

(5)

where C is the concentration of the ion in solution at time t, C , the initial concentration of the ion added at time zero, and k, the adsorption rate coefficient. The integrated form of Eq. (5) is In C = In C , - kt

(6) Thus, plotting In C versus t will yield a straight line of slope - k and an intercept of In C o if the data conform to first-order kinetics. For a miscible displacement technique, if the rate of adsorption of an ion onto a colloidal surface follows first-order kinetics, then d(CJCm) = ka(Cm - Ct) dt (7) where C , is the amount of ion on the colloid at time t, C , is the amount of ion on the colloid at equilibrium, and k, the adsorption rate coefficient. Separating variables and integrating results in log( 1 - CJC,) = - k, C , t

(8)

Since C , is a constant, one can call the product k,C, a constant. Thus, log(1 - C,/C,) = -&t

(9)

where & = k,C,/2.303 and is an apparent adsorption rate coefficient. Using miscible displacement, adsorption will vary only slightly with flow rate, and since by definition the adsorption rate coefficient k, is constant, a new term, the apparent adsorption rate coefficient k,, is defined for each flow rate in the system (Sparks et al., 1980b). If the rate of desorption of an ion from a colloidal surface follows firstorder kinetics, then for a batch technique

dC/dt = kd(C0 - C )

(10)

where C is the amount of ion released at time t, Co the total amount of ion that could be released at equilibrium, and kd the desorption rate coefficient. For the initial condition of C = 0 at t = 0, the integrated form of Eq. (10) becomes ln(C, - C ) = In Co - kdt (11)


24 1

If the rate of release of an ion from a colloidal surface follows first-order kinetics, then for a miscible displacement technique (for the solid surface phase) d(CJC0) = -kdC, dt

(12)

where Co is the amount of ion on the exchange sites of the colloid at zero time of desorption, C, the amount of ion on the exchange sites of the colloid at time t, and kd the desorption rate coefficient. Integrating with appropriate boundary conditions, one obtains h(c,/co) = kdt

(13)

Expressed in terms of base 10 logarithms, we have log(C/C,)

= &t

(14)

where kd = kJ2.303 and is an apparent desorption rate coeliic;en:. 2. Application of First-Order Kinetics to Clay Minerals and Soils

a. Single First-Order Reaction. One should realize that the adsorption rate coefficients (k, and kdrrespectively) determined from the previous first-order equations are composed of numerous diffusional and chemical rate constants (Jardine and Sparks, 1984a). It is appropriate to suggest that for the adsorption process the rate of adsorption ra is ra cc k,

+ k2 + k,

(1 5 )

where k, is the rate constant associated with the film diffusion process, k2 the rate constant associated with intraparticle diffusion or, independently, the rate constant associated with surface diffusion, and k 3 the chemical rate constant. Depending on the type of systems being studied, one or more of the above rate constants (viz., k , , k2, k 3 ) may be negligible or absent. The rate constant which is lowest will be the rate-limiting parameter and will have the greatest impact on the observed constant k,. Analogous expressions can be obtained for the desorption process. First-order equations have been used by many soil chemists to describe the kinetics of reactions in clay minerals and in soils. Sawhney (1966) described the adsorption of cesium on vermiculite as a pseudo first-order reaction. Sparks and Jardine (1984) studied the kinetics of potassium adsorption on the standard clay minerals kaolinite, montmorillonite, and vermiculite. The first-order equation described potassium adsorption on the clays extremely well (Fig. 2).

TIME (rnin) 0

20 I

40 I

80

60

I

I

100

I

I

120 I

I

140 I

160 I

180 ~

200 I

220 ~

240 I

I

KAOLlNlTE 0 MONTMORILLONITE A VERMICULITE

0

m

1.2

-

1.4

-

FIG. 2. First-order plots of potassium adsorption on clay minerals. (After Sparks and Jardine, 1984.)

0

0.3 -

Y

0.6 -0.9 1.2 -

-?!

1.5

.

I

8

-F I

1.8 2.1 2.4

.-

I

20 I

40

I

60

I

I

I

80 I

I

100 1~ 1

120 I

l

140 160 l 1 1 1 MATAPEAKE A H MATAPEAKE B A KENNANSVILLE A A KENNANSVILLE B 0 DOWNERA

0

DOWNER B

I

~

I

~

243


Sparks and co-workers (Sparks et al., 1980b; Sparks and Rechcigl, 1982; Jardine and Sparks, 1984a; Sparks and Jardine, 1984) and Talibudeen and co-workers in Great Britain (Sivasubramaniam and Talibudeen, 1972) have also found that potassium reactions between solution and exchangeable phases in soil systems follow first-order kinetics. An illustration of potassium adsorption in soils conforming to first-order kinetics (Sparks and Jardine, 1984) is shown in Fig. 3. Sparks et al. (1980b) investigated potassium desorption kinetics in two Dothan soils from Virginia. The first-order rate equation described potassium desorption for an average of 165 and 173 min for the aluminum- and calcium-saturated samples, respectively, in the Ap, A2, and B21t horizons, and for an average of 439 and 505 min for the aluminumand calcium-saturated samples, respectively, in the B22t horizon (Table 11). These represented times when potassium desorption was virtually complete in the respective soil horizons. The first-order rate equation described potassium desorption well, with r values ranging from -0.993 to -0.998. b. Multiple First-Order Reactions. Many researchers have found that the kinetics of ionic adsorption and desorption conform to a single first-order reaction. However, a number of workers have found that the kinetics of reactions in clays and soils are characterized by multiple first-order reactions (Li et al., 1972; Griffin and Burau, 1974; Griffin and Jurinak, 1974; Jardine

Table I1 Values of K O and the Amount of Time the First-Order Rate Equation Described Potassium Desorption for Dothan Soil from Greensville and Nottoway Counties" ~~~

~

KOb

(mol/lO- kg soil) Saturation treatment

Soilhorizon

Time of first-order conformity' (min)

Greensville

Nottoway

Greensville

Nottoway

5.64 6.44 6.10 6.64 6.39 6.80 7.95 9.00

5.77 6.64 6.21 6.77 6.44 7.00 8.00 9.28

152 163 161 166 162 169 438 500

160 170 175 183 177 186

-

~

AP

Al Ca

A2

A1 Ca

B21t

Al

B22t

Ca A1 Ca

~

4-40 5 10

~~

' Data after Sparks et al. (1980b). Used by permission of the SoiL Science Society of America Journal. Represents quantity of potassium on exchange sites at zero time of potassium desorption. Represents time for which the first-order rate equation described potassium desorption.

Time (rnin)

0

50

100

150

200

250

300

350 1

A=283 K a = 298 K = . 313 K

-8

-0.6-

$

-0.8-

Y -c

-m

TIME b i n )

FIG.4. First-order kinetics for potassium adsorption at three temperatures on Evesboro soil; the inset shows the initial 50 min of the first-order plots at 298 and 313 K. (After Jardine and Sparks, 1984a.)


245

and Sparks, 1984a). Griffin and Jurinak (1974) investigated the adsorption of phosphate on calcite and noted two simultaneous first-order reactions. They assumed the rates of the two reactions were independent and that the faster reaction went to completion before the slower reaction began. The results indicated a linear relationship existed for reaction times of about 10 min to 4 hr. The data for reaction times less than 10 min were curvilinear when plotted according to first-order kinetics. The linear portion of the plot was then extrapolated back to zero time. The slope allowed the calculation of the phosphate concentration in solution at any previous time due to the slower first-order reaction. The authors then took the total phosphate concentration in solution at intervening reaction times ( t = 0-10 min) and corrected for the influence of the slow reactions. The corrected data were then plotted according to a second-order kinetic expression which successfully described the first 10 min of the reaction process. Griffin and Jurinak (1974) found that the bulk of the phosphate was desorbed during the more rapid reaction, which was dominant during the initial 400 min of the reaction. The authors attributed the faster ZnZ+> CdZ+> Pb2+ > Cu2+. Since the ions used had diffusion coefficients of similar value, the authors inferred that the ion with the highest selectivity for peat would be the ion which was most rapidly adsorbed. The quantity of metal ions desorbed by the hydronium ion (H,Q+) was too small to predict reliable rate processes. Cavallaro and McBride (1978), working with Cu2+ and CdZ+,showed rapid adsorption of these metals on selected acid and calcareous soils. Salim and Cooksey (1980), investigating the adsorption of PbZ+on river muds, found that the rate of exchange conformed to first-order kinetics. ACKNOWLEDGMENTS

I thank Mrs. Cyndi Timko for typing this manuscript. Appreciation is also expressed to Ted Carski, Jerry Hendricks, and Richard Ogwada for their helpful comments and suggestions. I wish to dedicate this article with much love to my wife, Joy. REFERENCES Aharoni, C., and Ungarish, M. 1976. J. Chem. Soc. Faraday Trans. 72,400-408. Akratanakul, S., Boersma, L., and Klock, G. 0. 1983. Soil Sci. 135, 331-341. Amer, F., Bouldin, D. R., Black, C. A,, and Duke, F. R. 1955. Plant & Soil 6, 391-394. Ardakani, M. S., and McLaren, A. D. 1977. Soil Sci. Soc. Am. J. 41,877-879. Atkinson, R. J., Hingston, F. J., Posner, A. M., and Quirk, J. P. 1970. Nature (London) 226, 148-149. Atkinson, R. J., Posner, A. M., and Quirk, J. P. 1972. J. Inorg. Nucl. Chem. 34, 2201-2211. Ayodele, O., and Agboola, A. A. 1981. Soil Sci. SOC.Am. J. 45, 462-464. Barrow, N. J. 1983. Fert. Res. 40, 41-59. Barrow, N. J., and Shaw, T. C. 1979. J. Soil Sci. 30,67-76. Bolt, G. A., Sumner, M. E., and Kamphort, A. 1963. Soil Sci. SOC.Am. Proc. 27, 294-299. Bowman, R. A., and Focht, D. D. 1974. Soil Biol. Biochem. 6,297-301. Boyd, G . E., Adamson, A. W., and Myers, L. S., Jr. 1947. J. Am. Chem. Soc. 69,2836-2848. Broadbent, F. E., and Clark, F. 1965. In “Soil Nitrogen” (W. V. Bartholomew and F. E. Clark, eds.), pp. 344359. Am. SOC.of Agron., Madison, WI. Brown, J. L. 1981. Soil Sci. SOC.Am. J. 45,475-477. Bunzel, K., Schmidt, W., and Sansoni, B. 1976. J. Soil Sci. 27, 32-41. Burns, A. F., and Barber, S. A. 1961. Soil Sci. SOC.Am. Proc. 25, 349-352. Carski, T. H., and Sparks, D. L. 1985. Appl. Clay Sci. 1, in press. Cavallaro, N., and McBride, M. B. 1978. Soil Sci. Soc. Am. J. 42, 550-556. Chang, M. L., and Thomas, G. W. 1963. Soil Sci. SOC.Am. Proc. 21, 281-283. Chien, S. H., Clayton, W. R., and McClellan, G. H. 1980. Soil Sci. SOC.Am. J. 44, 260-264. Chute, J. H., and Quirk, J. P. 1967. Nature (London) 213, 1156-1157. Crank, J. 1976. “The Mathematics of Diffusion.” Oxford Univ. Press, London and New York. Deist, J., and Talibudeen, 0. 1967. J. Soil Sci. 18, 125-137.


265

Evans, R. L., and Jurinak, J. J. 1976. Soil Sci.121, 205-21 1. Eyring, H., Lin, S. H., and Lin, S. H. 1980. “Basic Chemical Kinetics.” Wiley, New York, Feigenbaum, S., and Hadas, A. 1980. Soil Sci. SOC.Am. J. 44, 1006-1010. Feigenbaum, S., and Levy, R. 1977. Geoderma 19, 159-169. Fenn, L. B., Matocha, J. E., and Wu, E. 1981. SoilSci. Soc. Am. J. 45, 883-886. Fenn, L. B., Matocha, J. E., and Wu, E. 1982. SoilSci. Soc. Am. J. 46, 78-81. Fiskell, J. G. A., Mansell, R. S., Selim, H. M., and Martin, F. G. 1979. J. Environ. Qual. 8,579-584. Frost, A. A., and Pearson, R. G. 1961. “Kinetics and Mechanism.” Wiley, New York. Gedroiz, K. K. 1914. Zhur. Ophr. Agron. 15, 181-208. Glasstone, S., Laidler, K. J., and Eyring, H. 1941. “The Theory of Rate Processes.” McGraw-Hill, New York. Goulding, K. W. T. 1983. Adv. Agron. 36, 215-264. Griffin, R. A., and Burau, R. G. 1974. SoilSci. Soc. Am. Proc. 38,892-897. Griffin, R. A., and Jurinak, J. J. 1974. Soil Sci. SOC.Am. Proc. 38, 75-79. Hackerman, N., and Stephens, S. J. 1954. J. Phys. Chem. 58,904-908. Hague, R., and Sexton, R. 1968. J. Colloid Interface Sci. 27, 818-823. Hammes, G. G. 1978. “Principles of Chemical Kinetics.” Academic Press, New York. Harter, R. D., and Lehmann, R. G. 1983. Soil Sci. SOC.Am. J. 41,666-669. Helfferich, F. 1962. “Ion Exchange.” McGraw-Hill, New York. Hingston, F. J. 1981. In “Adsorption of Inorganics at Solid-Liquid Surfaces” (M. A. Anderson and A. J. Rubin, eds.), pp. 51-90. Ann Arbor Science, Ann Arbor, MI. Hissink, D. J. 1924. Trans. Faraday Soc. 20, 551 -566. Ismail, F. T., and Scott, A. D. 1972. SoilSci. Soc. Am. Proc. 36, 506-510. Jardine, P. M., and Sparks, D. L. 1984a. Soil Sci. Soc. Am. J. 48,39-45. Jardine, P. M., and Sparks, D. L. 1984b. Soil Sci. Soc. Am. J . 48,45-50. Keay. J., and Wild, A. 1961. Soil Sci. 92, 54-60. Keeney, D. R. 1973. J. Environm. Qual. 2, 15-29. Kelley, W. P. 1948. “Cation Exchange in Soils.” Van Nostrand-Reinhold, Princeton, NJ. Kohl, D. H., Vithayathil, F., Whitlow, R., Shearer, G., and Chien, S. H. 1976. SoilSci.SOC.Am. J. 40, 249-253. Komareni, S. 1978. Soil Sci. Soc. Am. J . 42, 531-532. Kuo, S., and Lotse, E. G. 1972. Soil Sci.Soc. Am. Proc. 36,725-729. Kuo, S., and Lotse, E. G. 1974. Soil Sci. Soc. Am. Proc. 38, 50-54. Kyle, J. H., Posner, A. M., and Quirk, J. P. 1975. J. Soil Sci. 26, 32-42. Laidler, K. J. 1965. “Chemical Kinetics.” McGraw-Hill, New York. Li, W. C., Armstrong, D. E., Williams, J. D. H., Harris, R. F., and Syers, J. K. 1972. Soil Sci. Soc. Am. Proc. 36, 279-285. Liu, M., and Thomas, G. W. 1961. Nature (London) 192, 384. Low, M. J. D. 1960. Chem. Rev. 60, 267-271. Malcom, R. L., and Kennedy, V. C. 1969. Soil Sci. Soc. Am. Proc. 33,245-253. Martin, H. W., and Sparks, D. L. 1983. Soil Sci. SOC.Am. J. 47,883-887. Mortland, M. M. 1958. Soil Sci. Soe. Am. Proc. 22, 503-508. Murali, V., and Aylmore, L. A. G. 1981. A w f .J. Soil Rex 19, 23-29. Murali, V., and Aylmore, L. A. G. 1983a. Soil Sci. 135, 143-150. Murali, V., and Aylmore, L. A. G. 1983b. Soil Sci. 136, 279-290. Neuman, W. F., and Neuman, M. W. 1958. “The Chemical Dynamics of Bone Mineral.” Univ. Chicago Press, Chicago. Olsen, S. R., and Khasawneh, F. E. 1980. In “The Role of Phosphorus in Agriculture” (F. E. Khasawneh, E. C. Sample, and E. J. Kamprath, eds.), pp. 361-410. Amer. SOC.of Agronomy, Madison, WI.

266

DONALD L. SPARKS

Onken, A. B., and Matheson, R. L. 1982. Soil Sci. Soc. Am. J . 46, 276-279. Parravano, G., and Boudart, M. 1955. In “Advances in Catalysis” (W. G. Frankenburg, V. I. Kamarewsky, and E. K. Rideal, eds.), pp. 47-62. Academic Press, New York. Patrick, W. H., Jr. 1960. Int. Congr. Soil Sci. Trans., 7th 2, 494-500. Probert, M. E., and Larsen, S. 1972. J. Soil Sci.23, 76-80. Quirk, J. P., and Chute, J. H. 1968. lnt. Congr. Soil Sci. Trans., 9th 2, 671-681. Rajan, S. S. S. 1978. Soil Sci. SOC.Am. J. 42, 39-44. Rao, P. S. C., and Davidson, J. M. 1978. Soil Sci. Soc. Am. J. 42, 668. Reddy, K. R., Patrick, W. H., Jr., and Phillips, R. E. 1978. Soil Sci. SOC.Am. J. 42, 268-272. Reed, M. G., and Scott, A. D. 1962. Soil Sci. Soc. Am. Proc. 26,437-440. Reichenberg, D. 1957. In “Ion Exchanges in Organic and Biochemistry”(C. Calmon and L. R. E. Kressman, eds.), pp. 66-85. Wiley (Interscience), New York. Salim, R., and Cooksey, B. G. 1980. Plant & Soil 54,399-417. Sawhney, B. L. 1966. Soil Sci. Soc. Am. Proc. 30, 565-569. Selim, H. M., Mansell, R. S., and Zelazny, L. W. 1976. Soil Sci. 122, 77-84. Sharpley, A. N. 1983. Soil Sci. Soc. Am. J. 47,462-467. Sharpley, A. N., Ahuja, L. R., and Menzel, R. G. 1981a. J. Enuiron. Quul. 10, 386-391. Sharpley, A. N., Ahuja, L. R., Yamamoto, M., and Menzel, R. G. 1981b. SoilSci. Soc. Am. J. 45, 493-500. Sivasubramaniam, S., and Talibudeen, 0. 1972. J. Soil Sci. 23, 163-176. Sparks, D. L. 1985. “Soil Physical Chemistry.” CRC Press, Boca Raton, FL, in press. Sparks, D. L., and Jardine, P. M. 1981. Soil Sci. Soc. Am. J. 45, 1094-1099. Sparks, D. L., and Jardine, P. M. 1984. Soil Sci. 138, 115-122. Sparks, D. L., and Rechcigl, J. E. 1982. Soil Sci. Soc. Am. J. 46, 875-877. Sparks, D. L., Zelazny, L. W., and Martens, D. C. 1980a. Soil Sci. SOC.Am. J. 44, 37-40. Sparks, D. L., Zelazny, L. W., and Martens, D. C . 1980b. Soil Sci. Soc. Am. J. 44, 1205-1208. Sposito, G. 1981a. “The Thermodynamics of Soil Solutions.” Oxford Univ. Press, London and New York. Sposito, G. 1981b. In “Chemistry in Soil Environment” (R. H. Dowdy, J. A. Ryan, V. V. Volk, and D. E. Baker, eds.), pp. 13-30. (Amer. SOC.Agron. Spec. Publ. No. 40). Stanford, G., Vanderpol, R. A., and Dzienia, S. 1975a. Soil Sci. SOC.Am. Proc. 39, 284-289. Stanford, G., Dzienia, S., and Vanderpol, R. A. 1975b. Soil Sci. Soc. Am. Proc. 39, 867-870. Stevens, R. G., and Reuss, J. 0. 1975. SoilSci. Soc. Am. Proc. 39, 787-793. Tabatabai, M. A., and Al-Khafaji, A. A. 1980. Soil Sci. SOC.Am. J. 44, 1000-1006. Talibudeen, O., and Dey, S. K. 1968. J. Agric. Sci. (Camb.) 71, 95-104. Tamers, M., and Thomas, H. C. 1960. J. Phys. Chem. 64,29-32. Thomas, G . W. 1977. Soil Sci. SOC.Am. Meet., Houston. Thompson, H. S. 1850. J . R. Agric. SOC. Engl. 11, 68-74. Tripathi, P. S. M., Tripathi, R., and Prasad, B. B. 1975. Proc. Indian Natl. Sci. Acad. 41, 156-159. Ungarish, M., and Aharoni, C. 1981. J. Chem. Soc. Faraday Trans. 1,975-979. Van Riemsdijk, W. H. 1970. Ph.D. thesis, Agricultural University, Wageningen. Van Riemsdijk, W. H., and DeHaan, F. A. M. 1981. Soil. Sci. SOC.Am. J. 45,261-266. Way, J. T. 1850. J . R. Agric. Soc. Engl. 11, 313-379. White, R. E. 1976. Phosphorus Agric. 67, 9-14. Zasoski, K. J., and Burau, R. G. 1978. Soil Sci. Soc. Am. J. 42, 372-374.


ENHANCING NITROGEN FIXATION BY USE OF PESTICIDES: A REVIEW Martin Alexander Department of Agronomy Cornell University Ithaca. N e w York

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. Rhizobium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111. Free-Living Heterotrophs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1v. Blue-Green Algae in Flooded Soils . . . . . . . . . . . . . . . . . . . . . . . . . . . V. Resistant Isolates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI. Choice of Pesticides and Inocula . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII. Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII. Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

267 269 273 274 276 277 278 280 281

I. INTRODUCTION Under natural conditions, the presence of active nitrogen-fixing microorganisms is a necessary but often not a sufficient condition for significant nitrogen fixation to occur. This is true of reactions brought about by Rhizobium in association with leguminous plants, by a variety of bacterial genera in the rhizosphere of cereals or other nonlegumes, or by blue-green algae (cyanobacteria), which often proliferate in waterlogged soils planted with rice. The lack of appreciable activity in regions where the microorganisms are naturally present and also in areas where they are deliberately added by inoculation is often attributable to abiotic stresses that affect the survival, proliferation, or metabolism of these organisms. However, recent evidence suggests that the low activity may frequently be a result of biotic stresses, specifically a consequence of interactions between the nitrogen fixers and other microorganisms or lower animals. These harmful organisms prey or graze upon or compete with the active species and thus potentially reduce the amount of nitrogen introduced into agricultural or natural ecosystems. The purpose of this review is to present evidence that these harmful predators, grazers, or competitors can be controlled by pesticides and to suggest that these chemicals may thus function not only to control pests but also, in an indirect fashion, to effect a nitrogen gain in agricultural practice. 267

Copyright 0 1985 by Academic Press, Inc. All rights of reproduction in any form reserved.

268

MARTIN ALEXANDER

Inoculation with nitrogen-fixing microorganisms is known to cause nitrogen gains associated with the cultivation of legumes and sometimes cereals. However, such inoculation often has only a small effect, at least compared with the potential nitrogen gains that would occur if the microorganisms were added to plants grown in sterile soil or to unplanted sterile flooded or nonflooded soil. Often, the inoculum has no effect at all. In the case of legumes, for example, inoculation with Rhizobium sometimes does not lead to the appearance of nodules formed by the inoculum (Obaton, 1977), nodules being required for nitrogen fixation by this symbiosis. In some cases, only about 20% (Kuykendall and Weber, 1978) or as few as 5 % (Dobereiner, 1978) of the nodules are derived from the inoculum strain of Rhizobium, the use of which is recommended because of the low nitrogen-fixation effectiveness of the indigenous rhizobia in the soil. Similarly, the poor ability of Rhizobium to colonize the roots even of its host legumes is evident in the benefit arising from large as compared to small inocula applied to clover (Holland, 1970) and soybeans (Kapusta and Rouwenhorst, 1973). Extensive colonization of the rhizosphere of legumes by effective strains of Rhizobium, of the root zone of nonlegumes by free-living nitrogen-fixing bacteria, and of flooded soils and the overlying water is necessary for nitrogen gains to be extensive. This is true for both indigenous nitrogen fixers and organisms used as inocula. In the case of Rhizobium, the degree of colonization when roots are susceptible to infection is reflected in the extent of nodulation. Thus, a threshold number of the root-nodule bacteria appears to be necessary for nodulation to occur (Bohlool and Schmidt, 1973; Purchase and Nutman, 1957), and the abundance of nodules is directly related to the number of rhizobia in the root zone (Lim, 1963). For the free-living bacteria and blue-green algae, the extent of nitrogen accretion is likely a direct function of the extent of growth because a small biomass probably brings about little nitrogen input and growth is required for the initially small population in the soil or the inoculum to give the large cell mass necessary for significant activity. This colonization requires that the active species attain or maintain large populations in environments with numerous competitors, many of which grow faster, as well as protozoa that prey on them and invertebrates that graze on them. Despite the abundance and activity of competitors, predators, and other organisms that may be inimical to the nitrogen fixers, few attempts have been made to favor the inoculum organism in some selective way. The inoculum strain is added to the seed, the floodwater, or sometimes the soil, and in the absence of a means of selective stimulation, large inocula are used. Even many of these large inocula are of little value. In contrast, the approach suggested here is designed to overcome the natural biotic stresses or biological controls that hold the nitrogen fixers in check. The approach involves (1) first establishing the importance of predation, competition, or

ENHANCING NITROGEN FIXATION

269

other biotic stresses that suppress the desired organism and (2) then seeking chemicals that inhibit the predators, competitors, or other antagonists. Should the nitrogen-fixing organism whose establishment is desired be naturally susceptible to the compounds, mutants or other cultures are obtained that are resistant to the inhibitors. In this way the biotic stress is reduced or overcome, and the introduced nitrogen fixer has a selective advantage in the heterogeneous community in which its establishment is desired.

II. RHIZOBIUM Considerable attention has been given to explaining why populations of Rhizobium are rarely large in soil, why they do not reach as large numbers in the rhizosphere as species of other gram-negative bacteria, and why large numbers experimentally added to soil or the high cell densities that briefly appear in the rhizosphere decline markedly. The likely biological explanations include the deleterious actions of bacteriophages, lytic organisms, antibiotic producers, competitors, Bdellovibrio, myxobacteria, or protozoa. However, the density in soil of Bdellovibrio and bacteriophages able to destroy Rhizobium appears to be too small to have a significant effect on the population of root-nodule bacteria (Keya and Alexander, 1975; Ramirez and Alexander, 1980). The numbers of myxobacteria and of Bdellovibrio, bacteriophages, and lytic microorganisms able to destroy Rhizobium juponicum have been found to be fewer than 300 per gram of rhizosphere soil (Hossain and Alexander, unpublished data), numbers that are probably too small to significantly affect the R. juponicum population. Similarly, small numbers of lytic microorganisms, Bdellovibrio, and bacteriophages were found in the rhizosphere of Phaseolus vulgaris at the time that the population size of Rhizobium phaseoli was declining by about three orders of magnitude (Lennox and Alexander, 1981). Hence, the cause of the biologically induced decline or biological control of the rhizobia must be sought elsewhere. It has long been known that protozoa are more abundant in the root zone than in adjacent nonrhizosphere soil (Katznelson, 1946). Bacteria flourish at the expense of the array of readily metabolizable organic compounds excreted by roots, and the resulting large cell densities are highly suitable as prey for the protozoa that presumably are active when they have many cells on which to feed. Direct evidence for the role of protozoa in reducing high cell densities of Rhizobium was obtained in studies of nonrhizosphere soil to which large rhizobial populations were deliberately added (Danso and Alexander, 1975; Habte and Alexander, 1977). Subsequent research suggested that the same relationship held in the rhizosphere of P . vulgaris. Thus, R. phuseoli inoculated onto bean seeds proliferated rapidly shortly after

270

MARTIN ALEXANDER

planting the seeds, and the rhizobium soon became quite abundant. As the rhizobia as well as other bacteria replicated, their high densities led to a marked rise in the abundance of protozoa. The protozoa, in turn, brought about a drastic reduction in the abundance of all bacteria, including R. phaseoli (Lennox and Alexander, 1981). Protozoa not only may markedly reduce the size of large populations but they also may prevent the development of such large populations when the cell density initially is low. This is suggested by observations that small numbers of R. phaseoli, Rhizobium meliloti, R. japonicum, and a strain nodulating cowpeas failed to increase in abundance in natural soil even when a carbohydrate they can use was added, but they increased from 6- to 170fold in 4 days when inhibitors of eukaryotic microorganisms were added (Pena-Cabriales and Alexander, 1983).The suppression by protozoa of small numbers of rhizobia and the reduction in abundance of rhizobia to values below about lo6 per gram of soil, which were sometimes observed (Chao and Alexander, 198 l), were unexpected because it was generally believed that protozoa could not actively graze on small populations of bacteria, that is, populations below about lo6 per gram (Alexander, 198 1). However, although the need for a threshold density of prey does indeed appear to be necessary for such predation, not all prey cells need to be of the same species, and the protozoa are able to feed on alternative species in addition to the bacteria of interest. In this way, the protozoa graze actively and replicate at the expense of the alternative prey, the density of which is above the threshold, but in their relatively indiscriminate grazing, the protozoa also consume and thereby affect a particular species existing at a value below the threshold (Mallory et al., 1983). Such an effect of alternative prey is particularly likely in the rhizosphere, in which enormous numbers of gram-negative bacteria are present and are growing by using the supply of organic nutrients released by the roots. Hence, a suppression by protozoa of the slow-growing rhizobia, few of which can reach the population sizes of the many fast-growing bacterial species in the root zone, is probably widespread. The results of a test of the possible significance of alternative prey to the suppression of Rhizobium provide further evidence for the ecological role of protozoa. For example, when simple organic nutrients are added to soil to simulate the products released by plant roots, bacteria able to use these nutrients multiply. This leads to a dramatic rise in the population of protozoa. At the same time, R. phaseoii declines to values at day 7 of less than 0.1% of the initial population size. Similarly, the deliberate addition to soil of cells of an alternative prey for protozoa results in a suppression of R. phaseoli (Ramirez and Alexander, 1980). Granting that high rates of nitrogen fixation require the presence of large numbers of an effective strain of Rhizobium when legume roots are susceptible to infection, and granting further that protozoa are important in


271

suppressing the rhizobia or holding them in check, then a chemical that is toxic to protozoa but has no direct effect on rhizobia or their leguminous host should enhance colonization by the root-nodule bacteria, promote nodulation, and hopefully bring about greater nitrogen fixation. Initial investigations to test these possibilities involved the use of mannitol-amended soil to simulate root excretions and additions of cycloheximideand Triton X100, a mixture that is highly detrimental to protozoa. Under these conditions, R. phaseoli proliferated extensively, rather than becoming more sparse (Ramirez and Alexander, 1980). In subsequent studies, thiram [bis(dimethylthiocarbamoyl)disulfide], which is highly toxic to protozoa, was coated on the seeds. The host plant was P. oulgaris, and a thiram-resistant strain of R. phaseoli was inoculated directly onto seeds, as is the common field practice. The rise in abundance of R. phaseoli was identical in the presence or absence of the pesticide, but the subsequent fall in their abundance was appreciably slowed by the chemical. Of special importance, the population of R. phaseoli was higher on roots derived from thiram-treated seeds at the time that rhizobial infections were being initiated on young roots and nodules were being formed on the bean seedlings. Under the test conditions, thiram killed many of the protozoa, and for several days their numbers remained smaller than in the rhizosphere of bean seedlings derived from unamended seeds. With time, however, the protozoan population increased. It was not clear whether the return of the protozoa resulted from the breakdown of the pesticide, the appearance of resistant protozoa, or protozoan colonization of roots at a site distant from the point of placement of the treated seeds. That the beneficial effect of thiram resulted from its inhibition of protozoa was confirmed by a study in which it was found that stimulation of R. phaseoli took place in sterile soil inoculated with a protozoa-containing mixture of soil microorganisms, but not if the mixture was protozoa free (Lennox and Alexander, 1981). Similar effects have been noted in the rhizosphere of soybeans derived from seeds coated with benomyl [methyl 1-(butylcarbamoyl)-2-benzimidazolecarbamate] and inoculated with a benomyl-resistant strain of R. japonicum. The fungicide reduced the density of protozoa in the rhizosphere and slowed the decline of R. japonicum after its initial flush of growth. The suppression of protozoa by benomyl also disappeared, as occurred with thiram applied to beans, as the soybeans developed further (Hossain and Alexander, unpublished data). By maintaining larger populations of rhizobia, the pesticides presumably increased the opportunities for the bacteria to infect root hairs as the root system developed, but the effect was limited to young plants. From a practical viewpoint, the critical issue is not the inhibition of the potential predator or the enhanced colonization of the rhizobia, but rather the frequency of nodulation, the relative abundance of effective nodules, and particularly the response of the above-ground portions of the plant. This

212

MARTIN ALEXANDER

issue was first addressed in investigations of P . vulgaris. Application of thiram and a thiram-resistant R. phuseoli strain to seeds increased the number of nodules on 21-day-old plants as well as the percentage of nodules formed by the resistant strain in a soil containing other R. phuseoli strains. The differences in nodule numbers on beans derived from treated and untreated seeds disappeared as the plants matured. Nevertheless, adding thiram to the inoculated seeds increased the nitrogen content and yield of the tops of the beans and also increased the number and weight of the pods as compared to plants derived from inoculated seeds not treated with the fungicide. Furthermore, when the resistant rhizobia were present in the soil, as would occur if these bacteria were introduced as a consequence of inoculation during a preceding cropping season, pesticide treatment of the seed still increased the yield and nitrogen content of the plants. The suppression of protozoa and the consequent improved colonization by the effective, pesticide-resistant rhizobia appear to explain the beneficial responses of the plant (Lennox and Alexander, 1981). In trials with soybean, the seeds received benomyl or oxamyl (N',N'dimethyl-l\r-[(methylcarbamoyl)oxy]-1-thiooxamimidic acid methyl ester) and R. juponicum strains resistant to one or the other of these pesticides. The combination of benomyl and the bacteria increased the relative frequency of nodules formed by the inoculated rhizobia and also increased the percentage nitrogen, the total nitrogen content, the yield at one of the planting densities tested, and the pod weight in one of the soils examined in the greenhouse. The combination of resistant R. japonicum and oxamyl-which was applied to the seeds, foliage, or both-increased the relative abundance of nodules produced by the inoculum organism and also increased the yield, percentage nitrogen, total nitrogen content, and weight of pods and grain. The beneficial effects on soybeans grown in the greenhouse could have resulted from a favoring of the resistant strain of R.japonicum, a suppression of pathogens, or both (Hossain and Alexander, unpublished data). Predation appears to be the major, and possibly the sole, reason for the decline of the large rhizobium populations that appear as seeds germinate and seedlings develop. However, predation is not the sole mechanism involved in preventing the increase in rhizobial numbers. The most numerous microorganisms in the rhizosphere are bacteria, many of which undoubtedly compete with rhizobia for the supply of organic compounds needed for growth. Moreover, because many of these bacteria grow rapidly and some require few or no growth factors, they probably are better competitors than the slower growing and often more fastidious rhizobia. Such organisms probably use up nearly all of the organic carbon that is excreted from the roots, leaving little for the root-nodule bacteria. Although competition by fast-growing species is a likely mechanism to prevent or reduce the extent of


273

growth of slow-growing organisms, it is not easy to demonstrate unequivocally its significance in natural ecosystems. Possibly the best way to establish its importance is to eliminate all other likely mechanisms for the biological control of a test species and then to show that suppression of the likely competitors relieves the check maintained on growth of the population of interest. This is the approach that has been used in an investigation of R. phaseoli, R. rneliloti, R.japonicurn, and an isolate that nodulates cowpeas. The soil contained too few bacteriophages or cells of Bdellovibrio or myxobacteria to have an appreciable impact. Other studies failed to reveal significant numbers of cells capable of producing lytic enzymes or antibiotics active against rhizobia. In these circumstances, the addition to soil of streptomycin and erythromycin, antibiotics that have little or no effect on protozoa but that are toxic to bacteria (which are the most likely competitors with rhizobia), resulted in an enormous stimulation of the four rhizobia. For the purposes of the study, the rhizobia were made resistant to the two antibiotics. The density of the rhizobia increased by 400- to 3700-fold, a stimulation far greater than that observed when protozoa and other eukaryotes were suppressed. In line with the suggestion that competition limits the growth of small populations of Rhizobium in soil was the finding that their numbers were not detectably increased by additions of 0.01 or 0.10% mannitol, but that the counts rose upon the addition of 0.50% mannitol; that is, the competitors used up the lower concentrations before sufficient time had elapsed for the rhizobia to replicate to a significant degree, but enough of the carbon source was available at the higher concentration for even the slowgrowing species (Pena-Cabriales and Alexander, 1983). Earlier studies also showed that Rhizobium trifolii and Rhizobiurn lupini only increased in number in soil if large amounts of carbohydrates were added (Chowdhury, 1977). If the explanation of the effects of the antibiotics is correct, then the use of antibiotics, synthetic chemicals, or other inhibitors that are detrimental to the competitors should result in markedly enhanced growth of mutants of Rhizobium that are tolerant of the inhibitors. If the chemicals have a broad spectrum of action and are injurious to both competitors and predators, the should be still more marked. stimulation of Rh~zob~um

II I. FREE- LlVl NG H ETEROTRO PHS The approach of using inhibitors to control predators and competitors together with the inhibitor-tolerant inocula has not been pursued with the nitrogen-fixing bacteria that may enhance the growth of cereals or other

274

MARTIN ALEXANDER

nonlegumes. Such bacteria sometimes bring about small nitrogen gains under natural conditions (Balandreau and Villemin, 1973; Steyn and Delwiche, 1970; Vlassak et al., 1973), but their abundance is rarely great enough to exploit fully the organic excretions of the root. To help overcome the scarcity of nitrogen fixers, the practical use of inoculation has been investigated, e.g., Azospirillum sp. on millet (Barber et al., 1979) and Azotobacter chroococcum (Thompson, 1974a) and Beijerinckia derxii (Thompson, 1974b) on wheat. Although the amount of carbon excreted by the root may rarely be great enough to support the fixation brought about within the nodule by Rhizobium, allowing the free-living nitrogen fixer to use more of the available carbon than it otherwise would, but for the competition with other organisms or the impact of predation, should result in greater nitrogen gains. That the approach is feasible is evident from studies designed to promote colonization of a strain of Pseudomonas in the corn rhizosphere (MendezCastro and Alexander, 1983). The fungicide used was mancozeb [Mn and Zn salt of ethylenebis(dithiocarbamate)], and the isolate was a strain resistant to this pesticide. Following inoculation of the pseudomonad onto corn seeds, the bacterium became established on the roots, but its abundance was never great. In contrast, when the seeds were coated with mancozeb at the time they were inoculated, the roots that appeared had populations of the inoculum strain that were 100-fold or more larger. The types of organisms suppressed by mancozeb were not established.

IV. BLUE-GREEN ALGAE IN FLOODED SOILS In fields of rice planted in flooded soil, blue-green algae (cyanobacteria) may add appreciable nitrogen by fixation (App et al., 1980; Roger and Kulasooriya, 1980), and much or nearly all of the nitrogen removed from the soil with the harvested grain may be returned by their activity. The quantity of nitrogen fixed undoubtedly varies appreciably with the type of algae that dominates in the flooded field and with the prevailing environmental circumstances, but a figure of 30 kg/ha/yr seems reasonable under natural conditions (Watanabe et a]., 1977). Because the indigenous algae may not be active in fixation, inocula are sometimes added to the paddy water to increase the microbial fixation of nitrogen, and these inoculations may sometimes increase the yield of rice (Roger and Kulasooriya, 1980). Several factors affect the rate and extent of development of these algae and the amount of nitrogen they fix. Among the more significant of these factors are grazing and competition by other organisms that proliferate in the


215

floodwater. As the algae multiply, the large biomass they create serves as a food source for invertebrates, and the populations of these grazers often become quite large. The grazers may build up on either the blue-green algae or on the diatoms or unicellular green algae that precede them in the succession of organisms in the water. At times, the unicellular green algae that flourish support large numbers of waterfleas (cladocera), and these animals may be so dense that they eliminate a blue-green alga that is introduced as an inoculum (Watanabe et al., 1955a). Ostracods may sometimes be very abundant in the paddy field, and numbers of 5,000 to 15,000 per square meter have been recorded. An active population of such ostracods could potentially consume 12 tons of algae per crop of rice (Grant and Alexander, 1981). These microcrustaceans are often common in rice fields of Southeast Asia (Fernando, 1977). Grazing by Cypris sp. can prevent the development of or suppress inocula of blue-green algae tested under laboratory conditions, and these effects are evident in terms of the biomass or nitrogen-fixing activity of the algae. The extent of grazing and the influence on nitrogen fixation vary with the alga available to Cypris sp., the ostracod exhibiting a marked preference in its feeding habits (Osa-Afiana and Alexander, 1981; Wilson et al., 1980). Both cladocerans and ostracods are susceptible to an array of insecticides, and grazing by these invertebrates can be markedly reduced or totally abolished by the use of low levels of several insecticides. As early as 1967, Raghu and MacRae reported that the application of lindane (y-isomer of hexachlorocyclohexane) in rice fields killed indigenous Cypris, thereby allowing the flourishing of native blue-green algae. As little as 0.1 pg/ml of lindane totally prevented the feeding by Cypris sp. on Tolypothrix tenuis in laboratory tests (Grant and Alexander, 1981), and 0.1 pg of parathion per milliliter of floodwater completely abolished the feeding on Aulosira by the same ostracod (Osa-Afiana and Alexander, 1981). Low concentrations of methyllindane and carbofuran (2,3-dihydro-2,2-dimethyl-7-benzofuranol carbamate) brought about the cessation of grazing by the ostracods Cyprinotus carolinensis and Heterocypris luzonensis (Grant et al., 1983a). The cladoceran Daphnia is also easy to control, and both lindane (Maity and Saxena, 1979) and parathion (Watanabe et al., 1955b) are effective in this regard. Field tests in the Philippines demonstrated that Perthane [1,ldichloro-2,2-bis(p-ethylphenyl)ethane] suppressed ostracods and resulted not only in a 10-fold increase in biomass of blue-green algae and nitrogen fixation but an increase in rice yield (Grant et al., 1983b). Hence, a significant factor that often appears to reduce nitrogen fixation by blue-green algae can be overcome by the use of insecticides. The insecticides themselves, at the low concentrations needed to suppress grazing by the invertebrates, usually have no significant effect on algae (Roger and Kulasooriya, 1980).

276

MARTIN ALEXANDER

In the floodwater, algae other than the blue-green algae appear, sometimes in thick growths and often before the onset of significant development of the nitrogen-fixing blue-green algae. Because growth of the nonfixing algae is often limited by the same inorganic nutrients that limit growth of the blue-green algae, it is likely that algae early in succession may preempt much of the supply of a common nutrient element, leaving little for the blue-green algae that appear later in the succession. Alternatively, there may be a direct competition between nitrogen-fixing and nonfixing algae for a common limiting nutrient. Support for the view that nonfixing species affect bluegreen algae in such ways comes from a report that suppression of indigenous algae was necessary to permit the colonization of flooded rice fields by Tolypothrix tenuis (Hirano et al., 1955; Watanabe, 1962). The finding that the growth, the nitrogen-fixing activity, or both of T. tenuis or Aulosira sp. is reduced by indigenous algae or an experimentally introduced green alga adds weight to the view that indigenous algae may significantly interfere with the fixation of nitrogen by photosynthetic microorganisms in lowland rice production (Wilson et al., 1979). Control of indigenous algae is not difficult because they are quite susceptible to a variety of herbicides (McCann and Cullimore, 1979; Wright, 1978). Hence, a potentially useful approach to increasing nitrogen fixation in lowland rice production is the control of indigenous algae, which often have low or no capacity for nitrogen fixation, with a suitable herbicide and inoculation of the field with a herbicideresistant alga that has high nitrogenase activity and a rapid growth rate. Repeated reinoculation would not be necessary if the alga is also one that can survive the period when the soil is dried following harvest of the rice. Obviously, the pesticide must be one that does not injure the rice plant. This approach has been tested in the laboratory. The herbicide used was simetryne [2,4-bis(ethylamino)-6-(methylthio)-1,3,5-triazine], and the alga was a variant of Aulosira sp. that was resistant to the pesticide at levels that suppressed proliferation of indigenous algae. When added to flooded soil, simetryne controlled the indigenous algae, and this suppression of potential competitors allowed the herbicide-tolerant Aulosira sp. to proliferate and reach high levels of nitrogen fixation (Wilson et al., 1979). Strains of blue-green algae resistant to other pesticides have also been obtained (Sharma and Gaur, 1981 ; Vaishampayan, 1984).

V. RESISTANT ISOLATES Several methods have been used to obtain isolates of microorganisms that are resistant to the chemicals that are introduced for the control of predators or possible competitors that affect the nitrogen fixers. To obtain a fungicide-


277

resistant strain of Rhizobium, the original sensitive culture may be serially transferred in media with increasing concentrations of the chemical until a suitably resistant organism is obtained. This approach has been used to obtain rhizobia resistant to thiram, Spergon (2,3,5,6-tetrachloro-p-benzoquinone), Phygon (2,3-dichloro- 1 ,Qnaphthoquinone) (Odeyemi and Alexander, 1977), benomyl, streptomycin, and erythromycin (Hossain and Alexander, unpublished data), and the identical procedure has been used to obtain an isolate of Pseudomonus sp. resistant to mancozeb (Mendez-Castro and Alexander, 1983). In contrast, isolates of Rhizobium resistant to streptomycin and erythromycin have been obtained in a single step by plating a large number of cells on an agar medium containing the toxicant and selecting the colonies that appeared on the agar (Pena-Cabriales and Alexander, 1983). On the other hand, the herbicide-resistant alga used in such studies was obtained by exposing cells to a mutagen and then allowing the phenotypic expression of the resistance trait to become evident following introduction of the cells into a herbicide-free medium (Wilson et al., 1979). No special procedures are needed to obtain insecticide-tolerant blue-green algae, because they are resistant to the pesticides used to date for the control of invertebrates. Each isolate should be tested to be sure that the cultures finally evaluated for usefulness as inoculants retain their resistance when grown in the absence of the pesticide and also maintain their ability to fix nitrogen as rapidly as the wild-type culture. For Rhizobium, the cultures also must be tested to determine that they still nodulate their leguminous hosts. Fungicide-resistant rhizobia generally nodulate the same hosts as the wild type and fix nitrogen in symbiosis with the host as well as the parent culture; indeed, these strains can be used for inoculation when there is concern that fungicides needed for the control of pathogens attacking seeds or seedlings may prevent nodulation (Odeyemi and Alexander, 1977). The herbicide-resistant alga that has been evaluated retained its nitrogenase activity (Wilson et al., 1979). On occasion, however, rhizobia that gain the ability to grow in the presence of an inhibitor simultaneously lose their capacity to fix nitrogen (Schwinghamer, 1964).

VI. CHOICE OF PESTICIDES AND INOCULA In some instances, the enhanced nitrogen fixation arising from the joint use of pesticides and pesticide-resistant inocula will be appreciable. As pointed out above, these inocula need not be used often if they become established in the soil. Indeed, their establishment will benefit from the repeated use of the chemical so that persistence in soil of the introduced resistant organism may be greater than has been observed with nonresistant organisms, for which the

278

MARTIN ALEXANDER

selective advantage of the toxicant is not present. Whether the fixation is markedly or only modestly enhanced, a savings in cost can be achieved if the pesticide used to promote nitrogen fixation is also part of the farmer’s system of pest control. From the viewpoint of the marketing of pesticides, a chemical that has two functions would have special attraction. Many of the compounds now widely sold for pest control can be the basis for the selective enhancement of nitrogen fixation. For example, legume seeds are often treated with antifungal agents to prevent or minimize such fungal diseases as damping-off and seed and seedling rots, and these protective compounds applied to the seeds are often ideal choices for the inhibition of protozoa preying on rhizobia and bacteria competing with the root-nodule microsymbionts. The very fact that many fungicides at recommended rates are deleterious to rhizobia (Fisher, 1976; Fisher and Hayes, 1981; Tu, 1980) indicates they are good antibacterial as well as antifungal agents, and the data cited above also show that widely used fungicides have antiprotozoan activity. Similarly, the insecticide used to control ostracods, cladocerans, or other invertebrate grazers could be one already being applied to control pests of rice. Thus, compounds such as lindane or carbofuran would have a dual purpose. Not only would they control grazing on the algae, but they would also have the already desired function, lindane being important for protection of rice against stem borers and carbofuran being widely used to control a variety of insects, mites, and nematodes. Similarly, in regions where herbicides are already used to control weeds in rice fields, the herbicide might also be selected for its toxicity to the indigenous algae that otherwise would reduce nitrogen fixation. Such dual functions might also make pesticide use more attractive to farmers not already applying them for the control of particular groups of pests. The pesticide-resistant inoculum must be carefully chosen, however. Not only must it be active in nitrogen fixation but it should tolerate stresses of the habitat, and it should survive well when the crop is no longer being grown or, for lowland rice, during drying of the soil. Furthermore, the mechanism of its resistance to the pesticide must be checked to be sure the organism does not owe its tolerance to its ability to degrade the toxicant, because then the desired control of both the pests and the species affecting the nitrogen-fixing inoculum would be lost.

VII. LIMITATIONS One of the major limitations of the proposed approach is the cost of the pesticide. This expense, although real, can be absorbed as part of the cost of


279

pest control if the chemical is already used for the suppression of pathogens, insects, or weeds. A potential problem is the lack of movement of many chemicals from the site of their first introduction into the soil. A fungicide applied to the seed may inhibit the predatory protozoa and competing bacteria and thus promote colonization by the inoculum strain around the germinating seed and the nearby roots, but the pesticide will have no such effects on roots at some distance away unless the chemical moves through or with the developing roots or is translocated through the soil. Even where such movement is not appreciable, nitrogen fixation may be stimulated, as reported above, possibly because of the initial benefits to the plant and the continuing activity of the organisms so favored. Nevertheless, the availability of chemicals that are translocated would provide a more widespread influence in soil, and thus presumably a greater benefit, than those compounds that have a restricted zone of influence. Chemicals that move from seeds to roots or that move downward following foliar application would overcome such problems. Evidence that basipetally translocated pesticides may be effective in favoring nitrogen fixation comes from studies of oxamyl. This compound is translocated downward through leguminous plants (Martin and Edgington, 1981), and after its application to the foliage of soybeans, it increased the yield, nitrogen content, and percentage of nitrogen in soybeans inoculated with an oxamyl-resistant strain of R.japonicum (Hossain and Alexander, unpublished data). Several other fungicides applied to the above-ground portions of plants alter the composition of the rhizosphere microflora (Halleck and Cochrane, 1950; Rao and Sharma, 1978), although such changes may also arise from alterations in the composition of the root excretions. A significant limitation is the poor mobility of bacteria along roots and through soil. This is true of many and possibly of all bacteria, and only a small percentage of rhizobial cells have been found to be transported for distances as short as 2.7cm (Madsen and Alexander, 1982), although movement of even a few cells could be a prelude to their establishment if many of the organisms grazing or competing with them are suppressed by the pesticide. Possibly the best means for overcoming this limitation is to use, as inocula, bacteria that persist in soil; in this way, the pesticide-resistant bacterium that was inoculated onto seeds in earlier years would endure and become better dispersed in soil with root growth, cultivation, and water movement, and although present at distant sites from newly planted seeds, it would be available to colonize roots that grow to microsites that contain the persistent microorganism. The persistence of Rhizobium in soil is thoroughly documented (Lowendorf, 1980). As with other organisms whose control is sought by pesticides, the predators and competitors may become resistant to the chemicals. Acquired

280

MARTIN ALEXANDER

resistance because of frequent exposure to insecticides, for example, may explain the tolerance of the ostracod Heterocypris luzonensis to carbofuran and lindane (Grant et al., 1983a). The solution to the problem of acquired resistance may be the introduction of new pesticides or the rotation of chemicals to prevent the acquired resistance from eliminating the benefit from enhanced nitrogen fixation. Chemicals that are readily degraded also may pose difficulties. The compound must persist long enough to control the predators and competitors. For example, the effectiveness of carbofuran in controlling ostracods diminished as the insecticide was degraded, although lindane, which was more persistent under the test conditions, continued to suppress invertebrates (Grant et al., 1983a). This is not a limitation unique to the control of species harmful to nitrogen fixers, however.

VIII. SUMMARY AND CONCLUSIONS For appreciable fixation of nitrogen, Rhizohium must colonize the rhizosphere of its host legume; free-living nitrogen-fixing bacteria must grow around the roots of nonlegumes; and blue-green algae (cyanobacteria) must multiply in fields of lowland rice. Inoculation may lead to little or no nitrogen fixation, because the introduced microorganism may not become established or may not reach large numbers. Rhizobium populations may not become large or may be reduced in size because of predation by protozoa or competition with bacteria in the rhizosphere. Similar biological controls may affect free-living nitrogen-fixing bacteria in the root zone of nonlegumes. Grazing by ostracods and cladocerans reduces the extent of colonization of flooded soils by blue-green algae. The effects of predation by protozoa and competition with other bacteria in soil and grazing by invertebrates and competition with algae in paddy fields may be overcome by use of pesticides that suppress the predators and competitors together with inocula of nitrogen-fixing species resistant to these pesticides. The proposed method represents a novel approach to enhancing nitrogen fixation. Some of the initial studies needed to establish the feasibility of the approach have been conducted, although several issues still need to be resolved. Additional investigations are required to find better chemicals, to obtain resistant bacteria active on various legumes as well as cereals and other nonlegumes, to develop resistant algae, and to test these procedures under various field conditions. It is likely that future study will show other organisms to be important in suppressing or reducing the rate of growth and colonization of rhizobia,


28 1

bacteria around the roots of nonlegumes, and blue-green algae in rice fields. Moreover, no attention has yet been given to other nitrogen-fixing microorganisms or symbiotic associations, for example, the Azolla-Anabaena association. Nevertheless, the approach of using pesticides to control species that are harmful to the nitrogen fixers and nitrogen fixers resistant to these compounds should serve as a new means to bring about greater nitrogen gains in agriculture. REFERENCES Alexander, M. 1981. Annu. Rev. Microbiol. 35, 113-133. App, A. A., Watanabe, I., Alexander, M., Ventura, W., Daez, C., Santiago, T., and DeDatta, S. K. 1980. Soil Sci. 130, 283-289. Barber, L. E., Russell, S. A., and Evans, H. J. 1979. Plant Soil 52,49-57. Balandreau, J., and Villemin, G. 1973. Reo. Ecol. Biol. Sol 10, 25-33. Bohlool, B. B., and Schmidt, E. L. 1973. Soil Sci. Soc. Am. J. 37, 561-564. Chao, W.-L., and Alexander, M. 1981. Soil Sci. SOC.Am. J. 45, 48-50. Chowdhury, M. S. 1977. In “Exploiting the Legume-Rhizobium Symbiosis in Tropical Agriculture” (J. M. Vincent, ed.). Misc. Publ. (145), College of Agriculture, University of Hawaii, Honolulu. Danso, S. K. A., and Alexander, M. 1975. Appl. Microbiol. 29, 515-521. Dobereiner, J. 1978. In “Limitations and Potentials for Biological Nitrogen Fixation in the Tropics” (J. Dobereiner, R. H. Burris, and A. Hollaender, eds.), pp. 13-24. Plenum, New York. Fernando, C. H. 1977. Geo-Eco-Trop. 3, 169-188. Fisher, D. J. 1976. Pestic. Sci. 7, 10-18. Fisher, D. J., and Hayes, A. L. 1981. Ann. Appl. Biol. 98, 101-107. Grant, I. F., and Alexander, M. 1981. Soil Sci. SOC.Am. J. 45, 773-777. Grant, I. F., Egan, E. A., and Alexander, M. 1983a. Soil Biol. Biochem. 15, 193-197. Grant, I. F., Tirol, A. C., Aziz, T., and Watanabe, I. 1983b. Soil Sci. SOC.Am. J. 47, 669-675. Habte, M., and Alexander, M. 1977. Arch. Microbiol. 113, 181-183. Halleck, F. E., and Cochrane, V. W. 1950. Phyroputhology 40,715-718. Hirano, T., Shiraishi, K., and Nakano, K. 1955. Shikoku Nogyo Shikenjo Hokoku 2, 121-137. Holland, A. A. 1970. Plant Soil 32, 293-302. Kapusta, G., and Rouwenhorst, D. L. 1973. Agron. J. 65,916-919. Katznelson, H. 1946. Soil Sci. 62, 343-354. Keya, S. O., and Alexander, M. 1975. Soil Biol. Biochem. 7, 231-237. Kuykendall, L. D., and Weber, D. F. 1978. Appl. Environ. Microbiol. 36,915-919. Lennox, L. B., and Alexander, M. 1981. Appl. Environ. Microbiol. 41, 404-411. Lim, G. 1963. Ann. Bot. (London) 27,55-67. Lowendorf, H. J. 1980. Adv. Microb. Ecol. 4, 87-124. McCann, A. E., and Cullimore, D. R. 1979. Residue Rev. 72, 1-31. Madsen, E. L., and Alexander, M. 1982. Soil Sci. SOC.Am. J. 46,557-560. Maity, A. K., and Saxena, J. 1979. Nutl. Acud. Sci. Lett. (India) 2, 113-114. Mallory, L. M., Yuk, C.-S., Liang, L.-N., and Alexander, M. 1983. Appl. Environ. Microbiol. 46, 1073-1079. Martin, R. A., and Edgington, L. V. 1981. Pestic. Biochem. Physiol. 16,87-96. Mendez-Castro, F. A., and Alexander, M. 1983. Appl. Emiron. Microbiol. 45, 248-254.

282

MARTIN ALEXANDER

Obaton, M. 1977. In “Biological Nitrogen Fixation in Farming Systems of the Tropics” (A. Ayanaba, and P. J. Dart, eds.), pp. 127-133. Wiley, New York. Odeyemi, O., and Alexander, M. 1977. Soil Biol. Biochem. 9,247-251. Osa-Afiana, L. O., and Alexander, M. 1981. Soil Bid. Biochem. 13.27-32. Pena-Cabriales, J. J., and Alexander, M. 1983. Soil Sci. SOC.Am. J. 47, 241-245. Purchase, H. F., and Nutman, P. S. 1957. Ann. Bot. (London) 21,439-454. Raghu, K., and MacRae, I. C. 1967. Can. J. Microbiol. 13, 173-180. Ramirez, C., and Alexander, M. 1980. Appl. Environ. Microbiol. 40, 492-499. Rao, A. V., and Sharma, R. L. 1978. Acra Bor. Indica 6, 71-74. Roger, P. A., and Kulasooriya, S. A. 1980. “Blue-green Algae and Rice.” International Rice Research Institute, Los Banos, Philippines. Schwinghamer, E. A. 1964. Can. J. Microbiol. 10, 221-233. Sharma, V. K., and Gaur, Y. S. 1981. Int. J. Ecol. Environ. Sci. 7, 117-122. Steyn, P. L., and Delwiche, C. C. 1970. Environ. Sci. Technol. 4, 1122-1128. Thompson, J. P. 1974a. Queensl. J. Agric. Anim. Sci. 31, 129-138. Thompson, J. P. 1974b. Queensl. J. Agric. Anim. Sci. 31, 139-144. Tu, C. M. 1980. Bull. Environ. Conram. Toxicol. 25, 364-368. Vaishampayan, A. 1984. New Phyrol. 96, 7-11. Vlassak, K., Paul, E. A., and Harris, R. E. 1973. Plant Soil 38, 637-649. Watanabe, A. 1962. J. Gen. Appl. Microbiol. 8, 85-91. Watanabe, A., Ito, R., and Sasa, T. 1955a. J. Gen. Appl. Microbiol. 1, 137-141. Watanabe, A., Ito, R., and Sasa, T. 1955b. J. Gen. Appl. Microbiol. 1, 190-193. Watanabe, I., Lee, K. K., Alimagno, B. V., Sato, M., Del Rosario, D. C., and DeGuzman, M. R. 1977. IRRI Res. Pap. Ser. 3, 1-16. Wilson, J. T., Greene, S., and Alexander, M. 1979. Appl. Environ. Microbiol. 38, 916-921. Wilson, J. T., Greene, S., and Alexander, M. 1980. Soil Biol. Biochem. 12, 237-240. Wright, S. J. L. 1978. In “Pesticide Microbiology” (I. R. Hill, and S. J. L. Wright, eds.), pp. 535-602. Academic Press, New York.

ADVANCES IN AGRONOMY . VOL . 38

WEEDS AND WEED MANAGEMENT IN UPLAND R I C E S . Sankaran and S . K. De Datta Department of Agronomy International Rice Research Institute Manila. Philippines

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I1. Weed Flora of Upland Rice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A . Weeds in Upland Rice in Seven Asian Countries . . . . . . . . . . . . . . . . B. Weed Composition in Upland Rice in Africa. . . . . . . . . . . . . . . . . . . C. Weed Composition in Upland Rice in Latin America . . . . . . . . . . . . . D. Distribution Pattern of Weeds in Upland Rice . . . . . . . . . . . . . . . . . 111. Ecology of Upland Rice Weeds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A . Factors Influencing the Weed Flora in Upland Rice Fields . . . . . . . . . . B. Shifts in Weed Flora Due to Weed Control Methods . . . . . . . . . . . . . . C. Adaptation and Growth of Weeds in Upland Rice . . . . . . . . . . . . . . . IV . Weed Competition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A . The Upland Rice Ecosystem . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Critical Period of Weed Competition . . . . . . . . . . . . . . . . . . . . . . . C. Nature and Effect of Competition . . . . . . . . . . . . . . . . . . . . . . . . . D. Factors Influencing Competition . . . . . . . . . . . . . . . . . . . . . . . . . . E. Yield Losses Due to Weeds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V . Land Preparation and Crop Establishment Techniques . . . . . . . . . . . . . . . A . Land Management before and after Upland Rice . . . . . . . . . . . . . . . . B. Methods of Seedbed Preparation in Different Regions . . . . . . . . . . . . . C. Reduced Tillage and Weed Growth . . . . . . . . . . . . . . . . . . . . . . . . D. The Effect of Time of Land Preparation on Weed Emergence. . . . . . . . . VI . Fertilizer Application and Weed Management . . . . . . . . . . . . . . . . . . . . VII . Soil Moisture-Herbicide Relationships in Upland Rice . . . . . . . . . . . . . . . A . Rainfall Distribution and Weed Emergence. . . . . . . . . . . . . . . . . . . . B. Soil Moisture Content and Herbicide Activity . . . . . . . . . . . . . . . . . . C. Leaf Water Potential of Upland Rice and Weeds . . . . . . . . . . . . . . . . VIII . Weed Control Methods in Upland Rice . . . . . . . . . . . . . . . . . . . . . . . . A . Cultural Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Mechanical Weed Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Chemical Weed Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Weed Control in Upland Rice-Based Cropping Systems . . . . . . . . . . . . E. Biological Weed Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . F. Integrated Weed Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX. Yield Response of Rice to Herbicides and Herbicide Combinations . . . . . . . . 283

284 285 285 286 288 288 294 294 294 295 295 295 297 299 300 303 303 303 304 304 305 306 310 310 310 312 313 313 315 316 319 321 321 323

Copyright 0 I985 by Academic Press,Inc.

All rights of reproduction in my form reserved.

284

S. SANKARAN AND S. K. DE DATTA

X. Phytotoxicity of Herbicides and Residues . . . . . . . . . . . . . . . . . . . . . . . XI. Economics of Weed Control in Upland Rice . . . . . . . . . . . . . . . . . . . . . XII. Critical Research Needs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix: Common Names and Chemical Formulas of Herbicides . . . . . . References.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

..

323 327 328 330 330

I. INTRODUCTION Upland rice (Oryza satiua L.) is grown in Asia, Africa, and Latin America, mostly by subsistence farmers. In Africa and Latin America, the predominant cultural practice used for rice production is upland culture, in which rice is produced in very much the same way as other cereals. In Asia most rice is produced under lowland conditions, in which the crop is flooded for most of the growth period. Nevertheless, 11 million ha of upland rice is planted each year in Asia. With 2 million ha in Africa and 6 million in Latin America, this makes about 19 million ha used for upland rice worldwide each year, or about 12% of the total rice area. Upland rice is grown under a wide range of management intensities, varying from shifting cultivation, as in Malaysia, the Philippines, West Africa, and Peru, to highly mechanized systems, as in Brazil (De Datta, 1981). In Asia, upland rice is important in India, Bangladesh, Sri Lanka, Indonesia, the Philippines, Thailand, and Vietnam. In West Africa, 75% of the rice-growing area (1.87 million ha) is upland rice. Key upland rice-growing countries are Sierra Leone, Guinea, Nigeria, Ivory Coast, and Liberia. Latin America has 6.5 million ha of rice, 6 5 7 0 % of which is upland. Brazil has 5.4 million ha of rice, of which 4.2 million are upland. Most upland rice in Brazil is grown on small- to medium-sized farms with rolling topography. Colombia, Guyana, Panama, Ecuador, Peru, Venezuela, and several Central American countries also grow upland rice. Rice yields under upland conditions are as low as 1 ton/ha (De Datta and Ross, 1975) because of poor cultivars, irregular and inadequate rainfall, and weed competition. Weeds and soil moisture are the greatest limitations to higher yields. Weeds can cause complete crop failure. Weed competition varies with culture type, seeding method, cultivar, and production practices. This variation presents opportunities to develop combinations of practices to minimize weed problems.

WEEDS AND WEED MANAGEMENT

285

II. WEED FLORA OF UPLAND RICE Information on the weed flora of the world’s upland rice is useful when describing the ecology and habitat of weeds and for planning appropriate weed control. A. WEEDS IN UPLAND RICE IN SEVEN ASIANCOUNTRIES India. Upland rice grows under varying edaphic and climatic situations in India. Pande et al. (1967), Patro and Misra (1969), Chatterjee et al. (1971), Khan (1971), Misra and Roy (1971), and Mukhopadhyay et al. (1972) reported that Echinochloa colona (L.) Link, Echinochloa crus-galli (L.) Beauv., Cynodon dactylon (L.) Pers., Eleusine indica (L.) Gaertn., Zpomoea sp., Fimbristylis sp., Commelina benghalensis L., Phyllanthus niruri L., and Amaranthus spp. are the weeds in upland rice irrespective of edaphic differences. Misra and Roy (1971) showed that 80-90% of the weeds belonged to the families Commelinaceae, Gramineae, and Cyperaceae. Nine families of dicots represent only l0-12% of the weed population. The major graminaceous weeds were E. indica, E. colona, and Digitaria sanguinalis (L.) Scop. The dominant dicots were Celosia argentea L., Amaranthus viridis L., Amaranthus spinosus L., Acanthospermum hispidum DC., and Croton sparszjiorus Morong. The common Cyperaceae were Cyperus rotundus L. and Fimbristylis miliacea (L.) Vahl. Bangladesh. Gaffer (1983) reported that C. rotundus, Aneilema vaginatum R. Br., and Dopatrium junceum (Roxb.) Buch.-Ham. ex. Benth. were the most widely distributed weeds in upland rice in Bangladesh. An ecological survey identified 88 weeds from 30 families (Mian, 1971). Sri Lanka. Gunasena (1974) reported that 8 1 % of the weeds in upland rice in Sri Lanka were monocots similar to those found in India. Cyperus iria L., E. colona, E. crus-galli, and C. dactylon were present within 25 days after rice emergence.F. miliacea, E. indica, C. rotundus, and Panicum spp. were found at the later growth stages of the rice crop. Indonesia. The predominant weed species in high-rainfall, brown latosol areas in Indonesia were Digitaria sp., E. indica, C. rotundus, Ageratum conyzoides, Phyllanthus sp., and Amaranthus lividus L. (Sutidjo, 1969). Subsequent studies by Mangoensoekardjo and Kadnan (1971a) and Soerjani and Tirtarahardja (1971) showed that grasses were the dominant weeds in upland rice, followed by the perennial sedge, C. rotundus. C. rotundus is the most important of the 85 weed species associated with upland food crops (Ronoprawiro, 1975). The most reported weeds in upland

286


rice were the grasses E. colona, E. indica, Paspalum conjugatum Berg., C. dactylon, and Digitaria ciliaris (Retz.) Koel. and the broadleaf weeds Physalis angulata L., Amaranthus spp., Cleome aspera Koen. ex. DC., and A. conyzoides. The area planted to upland rice might be expanded if Imperata cylindrica (L.) Beauv. could be controlled. Philippines. De Datta (1979), Lopez et al. (1980), Tasic et al. (1980), Castin et al. (1983a,b),and Sankaran and De Datta (1984) listed the dominant weeds in experimental plots and in farmers’ fields in the Phillippines. Moody (1983) named 50 species of 21 families growing in upland rice. As in other Asian countries, monocots constituted 8045% of the weed flora, followed by a wide variety of broadleaf weeds. The 10 most common weed species were the monocots E. colona, C. rotundus, E. indica, Digitaria spp., Rottboellia exaltata L. f., Dacty loctenium aegyptium (L.) Willd., and C. benghalensis and the dicots Portulaca oleracea L., A. spinosus, and Ipomoea triloba L. Thailand. Earlier reports on the weed flora in upland rice in Thailand have identified 21 families and 74 species (Kanchanomai, 1975; Schiller and Indhaphun, 1979; Teerawatsakul, 1981). The predominant weeds were C. benghalensis, the grasses E. colona, E. indica, C. dactylon, D. ciliaris, and R. exaltata, and the sedges C. rotundus and F. miliacea. Vietnam. Vuong (1973) observed that the weed flora in Vietnam is determined by soil texture and soil moisture content. He identified E. colona, E. indica, C. iria, C. dactylon, A. conyzoides L., Ipomoea sp., and Alternanthera sessilis (L.) R. Br. ex. R.&S. as the important weeds.

B. WEED COMPOSITION IN UPLAND RICEIN AFRICA The upland rice weed spectrum in Africa is less diverse than in Asia. Shifting cultivation, the common cultural practice in several African regions, limits weed problems and diversity (Moody, 1975). There also is little documentation of the weed problem in the African upland rice-growing regions. Ivory Coast. Merlier (1974, 1983) reported more than 100 weed species in upland rice in Ivory Coast. Six accounted for 75% of the weed biomass in rice fields: Digitaria setigera Roth ex. R.&S., Brachiaria lata (Schumach.) C. E. Hubb., E. indica, A. hispidum, C. benghalensis, and Trianthemaportulacastrum L. R. exaltata is a noxious weed in the Philippines and Thailand but not a serious problem in Ivory Coast. Sierra Leone. Jones and Tucker (1978) found that Pennisetum subangustum (Schum.) Stapf and Hubb., Sida corymbosa R. E. Fries, Cyperus sp.,

287


Chlorisfasciculata (L.) Thell, and A. conyzoides were the predominant weed species in upland rice in Sierra Leone. Tanzania. The weed flora recorded by Ghosh (1976) in the upland rice fields of Tanzania varied little from that in neighbouring countries. R. exaltata, E. crus-galli, and Oryza punctata Steud. were the predominant grass weeds. The major broadleaf weeds were C. benghalensis, Melochia concatenata L., P. oleracea, Phyllanthus odontadenius Muel1.-Arg., and Amaranthus spp.; Cyperus compressus L. and Cyperus diflormis L. were the common sedges. Nigeria. At Ibadan, Nigeria, Fagade (1976) found that Chloris pilosa Schumach. was dominant (60%) in upland rice fields, followed by A. hispidum (20%) and A. conyzoides (5%). Malagasy. Falais (1978) reported that Cyperus spp., R. exaltata, and A. conyzoides were major upland weeds in Malagasy. The most cited upland rice weeds in Africa were the annual broadleaf weeds A. conyzoides, C . benghalensis, P. oleracea, Ipomoea spp., Euphorbia Table I The Most Common Weeds in Upland Rice in Latin America"

Grasses Cynodon dactylon (L.) Pers. Digitarin sanguinalis (L.) Scop. x Echinochloa colona (L.) Link Echinochloa crus-galli (L.) Beauv. Eleusine indica (L.) Gaertn. Leptochloa panicea (Retz.) Ohwi x Oryza satiua L. (red rice) Rottboellia exaltata L. f. Sedges Cyperus ferax L. x Cyperus rotundus L. Broadleaves X Amaranthus sp. Amaranthus spinosus L. X Commelina diffusa Burm. f. lpomoea sp. Physalis angulata L. Portulaca sp. (I

x

x

x

x

X

x

x

x

x

x

x x

x

X

x x

x x

X

x

x

X

x x

x x

x

X

X

x

X

x

x

x

x x x

x x

x x

x x

x x x x x

x x x x x

X X

x

x x

x x x

x x x

x x x

x x

x

x

X

X

x

x

x

x

X

x

x x x X

x

X

X

X

x

x

X

x X X

X

X

x

X

x

Adapted from Gonzalez et al. (1983). Other weeds in Latin America are Eclipta alba (L.) Hassk., C. odoratus, and Paspalum spp.

X

288


hirta L., and Amaranthus spp.; the annual grasses E. indica, R. exaltata, C. pilosa, E. colona, E. crus-galli, and Pennisetum spp.; the annual sedge Cyperus odoratus L.; the perennial grass I. cylindrica; and the perennial sedge C. rotundus.

c. WEED COMPOSITION IN UPLAND RICE IN LATINAMERICA Gonzales et al. (1983) collected information on weed composition in upland rice from 13 Latin American countries (Table I). E. colona was the most serious grass weed.

D. DISTRIBUTION PATTERN OF WEEDSIN UPLAND RICE There are 49 families of weeds in 30 upland rice-growing countries of the world. Asia has 248 species from 44 families, Africa 163 species from 27 families, and Latin America 65 species from 15 families. Among the families, Poaceae (gramineae) encompasses nearly 28 % of all weed species found in upland rice. Cyperaceae and Asteraceae each represent 10%;Amaranthaceae, Euphorbiaceae, and Papilionaceae each represent 5 %; and Commelinaceae, Malvaceae, Rubiaceae, and Convolvulaceae each represent 3% of the total weed species present. These 10 families account for more than 74% of all the weed species reported in upland rice. The number of times the weed species was reported, the spread of the weed in different countries, and the extent of yield reduction caused in upland rice were considered to determine the order of importance of weeds for each upland rice-growing continent. The 25 most commonly mentioned weeds in each of the upland rices of Asia, Africa, and Latin America are listed in Tables II-IV. C. rotundus is the most noxious weed in upland rice regions and occurs in almost all rain-fed rice-growing countries. Holm and Herberger (1969) and Ronoprawiro (1975) reported that C. rotundus was a problem wherever rain-fed rice grew. C. rotundus is formidable because it has an extensive underground root and tuber system (Holm and Herberger, 1969) and apical dominance (Muzik and Cruzado, 1952; Smith and Fick, 1973). It also is a problem because it germinates and grows with upland rice (De Datta, 1974a). E. colona is the second most serious weed in upland culture, probably because it needs less soil moisture for growth than E. crus-galli (Noda, 1977). In order of importance, other problem weeds are E. indica, C.dactylon, E. crus-galli, A . conyzoides, R. exaltata, C.benghalensis, P. oleracea, and C. iria.

Table I1

A Ranking of the 25 Most Important Upland Rice Weeds in Asia" Botanical name Cyperus rotundus L. Echinochloa colona (L.) Link Eleusine indica (L.) Gaertn. Cyperus iria L. Cynodon dactylon (L.) Pers. Echinochloa crus-galli (L.) Beauv. Portulaca oleracea L. Ageratum conyzoides L. Amaranthus spinosus L. Digitaria ciliaris (Retz.) Koel. Digitaria sanguinalis (L.) Scop. Amaranthus viridis L. Commelina benghalensis (L.) Rottboellia exaltata L.t Celosia argentea L. Dactyloctenium aegyptium (L.) Willd. Ipomoea triloba L. Phyllanthus niruri L. Cyperus compressus L. Fimbristylis miliacea (L.) Vahl Euphorbia hirta L. Lepthochloa chinensis (L.) Nees Commelina diffusa Burm. f. Eclipta alba (L.) Hassk. Trianthema portulacastrum L.

Common nameb Purple nutsedge Jungle rice Goose grass Rice flatsedge Bermuda grass Barnyard grass Common purslane Tropic ageratum Spiny amaranth

Life form'

PS AG AG AS PG AG AB AB AB AG Large crabgrass AB Slender amaranth AB AB Itch grass AG AB Crowfoot grass AG 3-lobe morning glory AB Niruri AB AS AS AB Garden spurge AG Spreading day flower AB Yerba-de-tag0 AB Horse purslane AB

Family

Countries where reportedd

Cyperaceae Poaceae Poaceae Cyperaceae Poaceae Poaceae Portulacaceae Asteraceae Amaranthaceae Poaceae Poaceae Amaranthaceae Commelinaceae Poaceae Amaranthaceae Poaceae Convolvulaceae Euphorbiaceae Cyperaceae Cyperaceae Euphorbiaceae Poaceae Commelinaceae Asteraceae Aizoaceae

BND, IND, IDO, JPN, PHI, SLK, THI BND, IND, IDO, SLK, VIE, PHI, THI BND, IND, IDO, JPN, PHI, SLK, VIE BND, IND, IDO, JPN, KOR, PHI, VIE,THI BND, IND, IDO, PHI, SLK, VIE, THI BND, IND, JPN, KOR, SLK, THI IND, IDO, JPN, KOR, PHI, THI IND, IDO, JPN, PHI, VIE, THI BND, IND, IDO, PHI, THI IND, IDO, JPN, KOR, THI BND, IND, IDO, PHI IND, IDO, JPN, THI IND, PHI, THI PHI, THI BND, IND, PHI, SLK IND, PHI, THI BND, PHI, VIE IND, IDO, PHI IND, IDO, PHI, THI IND, SLK, THI IND, PHI, THI JPN, PHI, THI PHI, SLK, THI IND, IDO, THI IND, PHI, SLK

Number of references 41

36 29 19 24 14 12 10 13 9 10

8 15

5 5 10 10 11

5 10 6 6 3 6 4

Based on their occurrence and/or yield reductions reported in rice. Standard common names of weeds. -, Not available. 'AG, Annual grass; AB, annual broadleaf; AS, annual sedge; PG, perennial grass; PS, perennial sedge. BND, Bangladesh; IND, India; IDO, Indonesia; JPN, Japan; KOR, Korea; PHI, Philippines; SLK, Sri Lanka; VIE, Vietnam; THI, Thailand. a

Table m N

A Ranking of the 25 Most Important Upland Rice Weeds io Africa"

B

Botanical name Cyperus rotundus L. Eleusine indica (L.) Gaertn. Ageratum conyzoides L. Rottboellia exaltata L.f. Commelina benghalensis L. Portulaca oleracea L. lpomoea spp. Cyperus odoratus L. Chloris pilosa (L.) Schumach. Acanthospermum hispidum DC. Echinochloa colona (L.) Link Euphorbia hirta L.

Common nameb Purple nutsedge Goose grass Tropic ageratum Itch grass -

Common purslane -

-

Bristly starbur Jungle rice Garden spurge

Life form' PS AG

AB AG AB AB AB AS AG AB AG AB

Family


Cyperaceae Poaceae Asteraceae Poaceae Commelinaceae Port ulacaceae Convolvulaceae Cyperaceae Poaceae Asteraceae Poaceae Euphorbiaceae

GHA, MAL, NIG, TNZ, SLE, IVO, SEG MAL, NIG, SLE, IVO, SEG MAL, NIG, SLE, IVO MAL, NIG, SLE, IVO GHA, TNZ, IVO MAL, TNZ, IVO NIG, IVO NIG, IVO NIG, IVO NIG, IVO IVO, GUI, SEG MAL, SLE, IVO

Number of references 12 7 6 6 5 4

3 3 3 3 3 3

Echinochloa crus-galli (L.) Beauv. Digitaria spp. Amaranthus spinosus L. Amaranthus viridis L. Solanum nigrum L. Stachytarpheta spp. Bidens pilosa L. Aspilia spp. Tridax procumbens L. Pennisetum spp. Dactylocteniurn aegyptium (L.) WillId. Boerhavia diffusa L. Imperata cylindrica (L.) Beauv. a

2

Barnyard grass Barnyard grass Spiny amaranth Slender amaranth Black nightshade Nettle leaf vervain Hairy beggarsticks -

Crowfoot grass Spiderling Cogon grass

AG AG AB AB AB AB AB AB AB AG AG AB PG

Poaceae Poaceae Amaran thaceae Amaran thaceae Solanaceae Verbenaceae As teraceae Asteraceae As teraceae Poaceae Poaceae Nyctaginaceae Poaceae

NIG, TNZ IVO IVO IVO IVO IVO IVO IVO IVO IVO IVO IVO IVO

Based on their occurrence and/or yield reduction reported in rice. Standard names of weeds. -, Not available. AG, Annual grass; AB, annual broadleaf; AS, annual sedge; PG, perennial grass; PS, perennial sedge. GHA, Ghana; MAL, Malagasy; NIG, Nigeria; TNZ, Tanzania; SLE, Sierra Leone; IVO, Ivory Coast; GUI, Guinea; SEG, Senegal.

2 3 3 3 3 3 2 2 2

2 2 2 2

Table IV A Ranking of the 25 Mast Important Upland Rice Weeds in Lath America'

W N

Botanical name

Common nameb

Life form'

Family


N

Echinochloa colona (L.) Link

Jungle rice

AG

Poaceae

Cyperus rotundus L.

Purple nutsedge

PS

Cyperaceae

AS

Cyperaceae

Cyperus odoratus L.

-

Cynodon dactylon (L.) Pers

Bermuda grass

PG

Poaceae

Oryza sativa L.

Red rice

AG

Poaceae

Rottboellia exaltata L.f. Ipomoea spp.

Itch grass Morning glory

AG AB

Poaceae Convolvulaceae

Eleusine indica (L.) Gaertn. Digitaria sanguinalis (L.) Scop.

Goose grass Large crabgrass

AG AG

Poaceae Poaceae

MEX, GUA, HON, SAL, NIC, CR, PAN, VEN, COL, ECU, BOL, BRA MEX, GUA, HON, SAL, NIC, CR, PAN, PER, BRA GUA, SAL,CR, PAN,VEN,COL, ECU, PER, BRA GUA, HON, SAL, NIC, VEN, PER, BOL, BRA MEX, GUA, HON, SAL, CR, PAN, ECU, BRA HON, CR, PAN, VEN, COL, PER, BOL GUA, HON, NIC, PAN, COL, PER, BRA SAL, NIC, ECU, PER, BOL, BRA SAL, CR, PAN, COL, BOL, BRA

Number of references

.Id

Leptochloa panicea (Retz.) Ohwi Amaranthus spinosus L. Echinochloa crus-galli (L.) Beauv. Commelina diflusa (Burm. f.) Portulaca sp. Physalis angulata L. Ischaemum rugosum Salisb. Sorghum halepense (L.) Pers. Eclipta alba (L.) Hassk. Paspalum spp. Ixophorus unisetus. (Pers.) Schlecht. Malachra spp. Bidens pilosa L. Cassia tora L. Cenchrus echinatus L. Panicum fasciculatum Sw.

Red sprangletop Spiny amaranth Barnyard grass Spreading daytlowei Purslane Ground cherries -

Johnson grass Yerba-de-tag0 -

Hairy beggarsticks Sicklepod Southern sandbur Brown top panicum

AG AB AG AB AB AB AG AG AB PG AG AB AB AB AG AG

Poaceae GUA, HON, NIC, CR, PER, BOL Amaranthaceae SAL, CR, PAN, PER, BRA CR, ECU, PER, BRA Poaceae Commelinaceae MEX, CR, PAN, BRA Portulaceae NIC, CR, PAN, BRA CR, PAN, PER, BOL Solanaceae Poaceae PAN, COL, PER MEX, NIC, PER Poaceae Asteraceae PAN, COL, ECU PER, BOL,BRA Poaceae Poaceae SAL, NIC, CR Malvaceae NIC, CR, PER Asteraceae PAN, BRA Caesalpiniaceae COL, PER Poaceae NIC, BRA Poaceae HON, PER

6 5 4 4 4 4 3 3 3 3 3 2

2 2 2 2

W rD

Based on their frequency of occurrence and/or yield reductions reported in rice. Standard common names of weeds. -, Not available. ' AG, Annual grass; AB, annual broadleaf; AS, annual sedge; PG, perennial grass; PS, perennial sedge. MEX, Mexico; GUA, Guatemala; HON, Honduras; SAL, El Salvador; NIC, Nicaragua; CR, Costa Rica; PAN, Panama; VEN, Venezuela; COL, Colombia; ECU, Ecuador; PER, Peru; BOL, Bolivia.

294


111. ECOLOGY OF UPLAND RICE WEEDS A. FACTORS INFLUENCING THE WEED FLORA IN UPLAND RICE FIELDS Weed distribution in upland rice is influenced by several environmental and management factors. Janiya et al. (1983) reported that the emergence pattern of weed species was governed mainly by soil moisture from 0 to 15 cm as influenced by rainfall. The type and number of weeds growing with preceding crops, weeding treatments, and soil moisture content before and after crop establishment (Janiya et al., 1983) determined the weed flora. Studies in West Africa showed that weed infestation in rice grown on newly cleared forest lands is much less than on land cultivated for several years (Merlier, 1978). Savanna regions and short-term fallows in forest zones have heavy weed infestations that increase risk and labor requirements beyond economic levels (USDA/USAID, 1968; Brown, 1969; Cates, 1969; Aryeetey, 1970; Moody, 1973). The weed flora of upland rice do not vary between climatic zones, although rice is cultivated from forests through savannas to mid-altitudes (Akobundu and Fagade, 1978). However, population intensity and dominant species vary. R. exaltata may be the major weed limiting rice production in one area, while E. colona may predominate in another area of the same climatic zone. A. conyzoides and Tridax procumbens L. are broadleaved annuals found in mature rice. They continue to germinate following dissipation of herbicides or after the last hand weeding. They have little effect on yield but interfere with harvest and harbor rodents.

B. SHIFTSIN WEED FLORA DUETO WEED CONTROL METHODS One study suggests that Compositae and Commelinaceae species become dominant in fields where other weeds have been controlled with preemergence herbicides (Akobundu and Fagade, 1978). Bhandari and Moody (1981) reported that C. rotundus populations increased in plots where pendimethalin' was used to control R. exaftata in rain-fed, rice-based cropping systems. A similar shift favoring C. rotundus was reported by Navarez et al. (1983) when preemergence pendimethalin was followed by postemergence 2,4-D in rice and when pendimethalin was applied preemergence in the succeeding mungbean [Vigna radiata (L.) Wilczek] crop to control R. exaftata in a rice-mungbean sequence. Munroe et al. (1981) 'See Appendix for complete chemical names.


295

reported that continued application of butachlor caused a shift from monocots to dicots. Sankaran and De Datta (1984) reported that C. benghalensis, an annual broadleaf weed, became dominant in upland rice when R. exaltata was completely controlled by preemergence applications of pendimethalin. A shift favoring slow-growing broadleaf weeds over fast-growing annual grasses is desirable because it significantly reduces weeding time. Unfortunately, the shift sometimes is from a moderately easy-to-control weed to one that is difficult to control (Mercado, 1983).

c. ADAPTATIONAND GROWTHOF WEEDS IN UPLAND RICE Iwata and Takayanagi (1974a) reported that D. ciliaris seeds were more adaptable to low soil moisture content (40% of field capacity) than upland rice and other weeds such as P . oleracea, Amaranthus retro$exus L., and E. indica. E. indica was more adaptable to high soil moisture content, even waterlogging, than other weeds. Iwata and Takayanagi (1974b) studied the growth rate of D. ciliaris and upland rice sown at the same time in monoculture and mixed culture. The optimum growth period of Digitaria was shorter, but the growth increments during the period were larger than those of rice. In mixed cultures, Digitaria reduced the number of tillers, the plant height, and the dry weight of rice. In West Africa, perennial weeds of cleared, drier forest and derived savanna zones include Cyperus spp. and I. cylindrica. Their rhizomes make them more difficult and expensive to control than annual weeds. Infested areas are generally allowed to lay fallow for several years before rice is planted again (Akobundu and Fagade, 1978).

IV. WEED COMPETITION A. THEUPLAND RICE ECOSYSTEM Weed competition largely governs the development of upland rice. The usual low yield of upland rice has been attributed mainly to inadequate and irregular moisture supply, heavy weed infestations, lack of suitable cultivars, nutritional imbalance, and inadequate cultural practices, including inefficient control of disease and insect pests (De Datta and Beachell, 1972). Among these limiting factors (except poor water supply), inadequate weed control is, perhaps, the most difficult constraint to increasing upland rice production (De Datta, 1972; Madrid et al., 1972).

296


..

ENVIRONMENT Rainfall Solar radiation EdaDhic conditions

m

.

YIELD

+I

MANAGEMENT

WEED MANAGEMENT

..

Primary Herbicide Hand hoeing and weeding

.. .. ...

Secondary Cultivation Cropcompetition Seeding method Variety Seeding rate Fertilizer Preventive

-

I 1-

1

k* WEED PRESSURE

FIG.1. Schematic diagram of an upland rice ecosystem showing the relationship between weed management, weed pressure, and crop yield. (Adapted from OBrien, 1981.)

Figure 1 depicts the crop-weed balance in the upland rice ecosystem and the various factors that influence it. Primary weed control methods affect rice yield by reducing weed pressure. Secondary weed control methods, such as seedbed preparation, moisture conservation, and crop rotation, directly reduce weed pressure and increase rice yields. Other secondary methods, like seeding method and density, increase the competitive ability of rice. Fertilizer application increases weed and rice growth. In the upland rice ecosystem, seed is sown when soil moisture is adequate for germination. Rice and weed seedlings compete for moisture, nutrients, and light during germination and growth (Akobundu and Fagade, 1978; Borgohain and Upadhyay, 1980). The rice crop and weeds use the same environmental factors for their growth needs. Competition begins as soon as a factor cannot support the normal growth needs of the crop and the weeds (Utomo, 1981). Weeds in upland rice can withstand drought better than rice because they have deeper roots and high root-length density to tap moisture from deeper soil layers (Sankaran and De Datta, 1984). High-cost inputs such as seeds and fertilizers are useless under upland conditions without efficient weed control. In shifting cultivation, the land is abandoned when weeds seriously reduce yields. After reviewing current practices in shifting cultivation, Moody (1975) reported that existing weed control methods are unsuitable for continuous or large-scale farming. He emphasized the need for research to find economical, effective control measures.

297


B. CRITICAL PERIOD OF WEEDCOMPETITION The critical period of weed-crop competition is between early growth, during which weeds can grow without affecting crop yield, and the point after which weed growth does not affect yield (Zimdahl, 1980). Establishing the critical period of competition is essential to develop effective, economical weed control measures (Sharma et al., 1977). In upland rice, researchers (Kawatei et al., 1966; Park and Kim, 1971; Mercado, 1979; Schiller and Indhaphun, 1979; De Datta, 1980; Kolhe and Mittra, 1981) pinpointed the optimum weed-free period (Table V). Table V The Critical Period of Weed-Crop Competition in Upland Rice

Country

No. of days weeds can be left in competition after rice emerges

Days after which emerging weeds can be left to compete with the crop

Asia India

10

20

21

20 15

42 30 45 42 50 25 40 63 50

Ghosh et al. (1977); Sharma et al. (1977) Upadhyay and Choudhary (1979) Kolhe and Mittra (1981) Sahai et al. (1983) Kawatei et al. (1966) Park and Kim (1971) Akhanda (1966) Mercado (1979) Wells and Cabradilla (1981) Schiller and Indhaphun (1979)

15 42

45 56

Morales and Vargas (1976) Carson (1975)

India India India Japan Korea Philippines Philippines Philippines Thailand Africa and Latin America Colombia Ghana

~

15 -

-

Source

The first 15 days after seeding (DAS) rice seem to be the maximum period during which weeds can be tolerated without affecting the final crop yield (Morales and Vargas, 1976; Ghosh et al., 1977; Sharma et al., 1977; Schiller and Indhaphun, 1979; Upadhyay and Choudhary, 1979; De Datta, 1980; Wells and Cabradilla, 1981).The weed-free requirement for upland rice varies from the first 30 to 60 DAS, depending on edaphic and climatic conditions and weed flora. In their studies on weed-crop competition, Wells and Cabradilla (1981) found that competition began during the first 3 weeks (20 days) of the crop.

298


Grain yield decreased as weed growth duration increased. Weed growth increased exponentially during the first 60 days, reaching a maximum dry weight of 6.6 ton/ha. Removing weeds at this stage gave little yield increase. The critical period of weed competition was between 2 and 9 weeks after sowing. Moody (1982) wrote that if weeds mature rapidly, shorter-duration rice cultivars would compete for a proportionately longer time than longerduration cultivars. Because the late-maturing cultivar grows longer, it has more ability to compensate for weed competition. Yields from plots not weeded for 10 days after germination were equal to those of the weed-free check (Ghosh et al., 1977). Weed presence for 20 days or more reduced yield significantly. O n the other hand, keeping the plots weed-free for the first 20 days was as good as keeping them weed-free for 50 days or throughout the growing period. No weed competition occurred and yield was unaffected if weeds grew at 10 DAS. George et al. (1968), Morales and Vargas (1976), Sharma et al. (1977), Schiller and Indhaphun (1979), and Wells and Cabradilla (1981) had similar findings that weeds do not appreciably reduce yield if they begin to grow 15 DAS rice. Merlier (1978) observed that weed dry weight and yield of the rice cultivar Iguape cateto were negatively correlated (Fig. 2) and that increasing the duration of grass weed competition progressively reduced panicle weight and number. I

L

- - - Dry weight of weeds 10-

o A

Yield 1974

'

,'

1975

0c

I

I

I

I I

/'

?'

I

I

P'

0 \ 1

P

'P

/' I

I I

I

FIG.2. The effect of dry weight of weeds after different periods of infestation and yield of variety Iguupe cateto. (Adapted from Merlier, 1978.)

299


In Ghana, Carson (1975) reported that weeding upland rice could be delayed until 6 weeks after seeding without adversely affecting yield. Beyond 6 weeks, however, yield losses increased with a corresponding delay in weed removal. Eight weeks of early weed-free conditions prevented substantial yield reduction. The critical stage of weed competition was between 6 and 8 weeks after seeding. The weed-free requirement of upland rice emphasizes the need for selecting preemergence herbicides that are residually active and effective for the first 50 DAS to ensure optimum grain yield. Weed control after the critical period could prove wasteful.

c. NATURE AND EFFECTOF COMPETITION Blackman and Templeman (1938) reported that, in a normal rainfall year, competition is principally for nitrogen. Chakraborty (1973) reported higher nitrogen content in weed species at the vegetative, flowering, and postflowering stages, indicating severe competition for nitrogen throughout the upland rice growing season. At IRRI,De Datta (1974a,b) and Okafor and De Datta (1976b) studied the effect of nitrogen level on varying densities of C. rotundus in upland rice. De Datta (1974b) reported a significant negative correlation between C. rotundus dry weight and rice grain yield as influenced by nitrogen application (Fig. 3).

C

0

120 kg N/ha 318138-000369X r = -09822**

5-

r = -0

n- L., '

200

I

9339* I

I

400 500 Dry weight of perennial nutsedge (g/rn21

300

600

FIG.3. Correlation and linear regression between weed weight and grain yield of upland rice as influenced by nitrogen level. IRRI, 1972 wet season. *,Significant at 5% level; **, significant at 1% level. (From De Datta, 1974b.)

300

S. SANKARAN A N D S. K. DE DATTA

Severe weed infestation in upland rice depressed the height (Wells and Cabradilla, 198 l), dry matter (Chakraborty, 1973), tiller and panicle numbers (Okafor and De Datta, 1974; Sharma et a!., 1977; Kolhe and Mittra, 1981), leaf area index, and light transmission ratio (Okafor and De Datta, 1974) of rice. Utomo (198 1) observed that rice plants and weeds compete in two ways. The tops of the plants compete for light and the root zone competes for nutrition (nitrogen and humus). In these competitions, rice plants develop less chlorophyll and contain less nitrogen, which causes fewer panicles and spikelets per plant.

D. FACTORS INFLUENCING COMPETITION Many factors influence competition. Weeds and rice modify the environment by influencing the growth of constituent plants. Plant species, plant density, and distribution and duration of plants in the ecosystem directly affect competition (Bleasdale, 1960, cited by Zimdahl, 1980). Climatic and edaphic conditions modify rice-weed competition (Fig. 4). Upland rice yield is influenced by weed density per unit area, especially during early growth. Weed density and rice yield have a sigmoid relationship (Utomo, 1981). Eussen (1981) reported that A . conyzoides densities from 32 to 1024 plants/m2 reduced grain yield of an upland rice cultivar in Bicol,

- species - densities -distribution

- duration

Weeds

-7

(weeding)

edaphic and climati

- densities - othercrops - distribution

- duration

Upland r i c e 1 production

(crop)

FIG.4. Schematic outline of the factors influencing the upland rice-weed complex. (Adapted from Bleasdale, 1960.)

301


Philippines, by 19 to 72%. Grain yields declined progressively as weed density increased. Eussen and Hadi (1981) observed a similar effect of D. ciliaris on grain yield. Takayanagi and Iwata (1978) reported that D. ciliaris had a higher dry matter production at early growth stages. C. rotundus, generally considered the world's worst weed (Holm et al., 1977), reduced the rice yield 33 and 93% at 32 and 1024 plants/m2, respectively. Grain yield reduction was mainly caused by reduced spikelet and panicle number. Eussen and Martoyo (1981) found that Porophyllum ruderale (Jacq.) Cass. at 64 plants/m2 present for 60 DAS retarded tillering and reduced grain yield by 50%. At 1024 plants/m2, the effect appeared 30 DAS and reduced yield by 91%. D. aegyptium populations reduced yield similarly (Utomo, 1981). The level of competition varies with weed species. The extent of competition can be determined by calculating the relative space occupied by each species following the De Witt (1960) method. Utomo (1981) reported that D. ciliaris occupied more space than Digitaria fuscencens (Presl) Henr. and Digitaria ternata (A. Rich.) Stapf. In a greenhouse experiment, Soetrisno et al. (1981) showed that D. ciliaris retarded rice growth and decreased yield. They found the relative yield value of rice in mixed culture to be smaller than that of D. ciliaris. Weeds occupied more space than rice. Okafor and De Datta (1974) compared the effects of three groups of weeds (annuals, C. rotundus, and annuals plus C. rotundus) on the growth and yield of drilled and broadcast IR5 (Table VI). A combination of annual weeds and Table VI Effect of Weed Composition on Leaf Area Index (LAI), Light Transmission Ratio (LTR), and Grain Yield of IR5 Rice under Two Methods of Seeding" Drilled

Weed community Weed free Annuals Perennial sedge (C. rotundus) Annuals plus perennial sedge

Broadcast

LTR at panicle initiation stage

LA1 at flowering stage

Grain yield (tonha)

LTR at panicle initiation stage

LA1 at flowering stage

Grain yield (tonha)

7.1 a 3.4 b 3.4 b

4.9 a 2.7 b 2.8 b

4.7 a 1.2 c 2.7 b

5.6 a 3.1 ab 4.1 a

5.6 a 2.2 b 3.2 b

4.9 a 1.6 c 2.4 b

3.1 b

1.3 b

0.8 c

2.5 b

1.4 bc

0.9 d

IRRI, 1972 wet season. In a column, values followed by the same letter are not significantly different at the 5 % level as determined by DMRT. From Okafor and De Datta (1974).

302


C. rotundus caused the greatest reduction in leaf area index (LAI), light transmission ratio (LTR), and grain yield. Many soil and environmental factors modify weed competition. Soil nutrient level influences the nature and duration of competition in upland rice ecosystems (De Datta and Malabuyoc, 1976). Increasing nitrogen fertilizer application increased nitrogen uptake of C. rotundus (Okafor, 1973). Weeds grow better under adequate levels of nutrients, thus making them Table VII Yield Reduction in Upland Rice Due to Uncontrolled Weed Growth in Asia, Africa, and Latin America Country Asia Bangladesh

India

Indonesia Philippines

Sri Lanka Thailand Africa Ghana

Liberia Nigeria Senegal Sierra Leone Gambia Upper Volta Latin America Brazil Panama

Yield reduction (%)

90 (80- 100)” 59 (32-86)

75 (70-80) 68 (35-100) 50 62 (38-86)

71 (58-84) 63 (39-87) 90 (80-100) 48 58 (25-90) 100 62 95 100

Source

BRRI (1981) Mani et al. (1968); Manna et al. (1971); Mukhopadhyay et al. (1971); Spratt and Chowdhury (1978); Singh and Sharma (1981); Moorthy and Dubey (1981) Syam and Effendi (1977) IRRI (1967); Vega et al. (1967); De Datta (1972); Okafor and De Datta (1974); Mercado (1979); Sarkar and Moody (1981)* Jayasekera and Velmurugu (1964) Schiller and Indhaphun (1979); Teerawatsakul (1981) Aryeetey (1970); Carson (1975) WARDA (1981) Moody (1973); Fagade (1976) WARDA (1976) Jones and Tucker (1978) WARDA (1976) WARDA (1976) Burga and Tozani (1980) Smith (1973)

‘Numbers in parentheses denote the range of grain yield reductions. Mean of 10 wet seasons (1971-1980) yield data at IRRI and mean range of loss in yield at 9 locations in the Philippines.

WEEDS A N D WEED MANAGEMENT

303

more competitive. Eussen and Zulfaldi (1981) reported that at 100-300 kg N/ ha, weeds reduced grain yields by 25 %. Grain yield reduction was independent of nitrogen level. Therefore, weed control becomes more imperative as fertilizer applications increase.

E. YIELD LOSSESDUETO WEEDS Weeds cause two types of crop losses. The most important is the direct yield loss resulting from competition, followed by indirect loss from reduced crop quality (De Datta, 1980). All cereal crops yield less under severe weed stress (Table VII). Data from the upland rice experiments from 1971 to 1980 at the International Rice Research Institute (IRRI) show weeds reduce grain yield by 2 ton/ ha with a mean of 89% (Sarkar and Moody, 1981). Yield reduction fluctuated from 0.9 to 3.3 ton/ha, with a mean loss of 2.3 tons/ha. When weed pressure is extremely severe at early rice growth, the yield loss can be 100% (IRRI, 1967, 1968; Vega et al., 1967; Manna et al., 1971; De Datta, 1972; Madrid et al., 1972; Moody, 1973; Smith, 1973; Williams, 1975; Fagade, 1976; WARDA, 1976). Dubey and Thomas (1977) reported 4 5 8 0 % yield loss in upland rice, 42% of which occurred in the first 3 weeks of crop growth.

V. LAND PREPARATION AND CROP ESTABLISHMENT TECH NlQU ES Land preparation reduces the weed problem in the following crop. The rain-fed environment is so heterogenous that there is not one set of land preparation methods for all situations (Pillai, 1981).

A. LANDMANAGEMENT BEFORE

AND AFTER

UPLANDRICE

Most countries of Asia, Africa, and Latin America produce upland rice during the rainy season, which normally lasts for 3 to 7 months. Generally, a single rice crop is grown in uplands, after which the field remains fallow. This practice facilitates weed growth and weeds complete their life cycle with adequate soil moisture, which increases weed problems for the next crop. Weekly harrowing in the dry season gives good soil tilth, which favors rice seed emergence and weed growth. For example, Castin et al. (1983b) found

304 2.0

-

b

1.5 b

Jz

\ C

c

0 1.0

.-

) I

._C

-

a a

FIG.5. Effect of previous land usage on grain yield of upland rice. (Adapted from Castin et al., 1983b.)

that planting maize (Zea mays L.) or mungbean in the dry season reduced weed growth and weeding time and increased herbicide performance and rice yields in the following season as compared with plots kept weed-free by using paraquat or being maintained as weedy fallow (Fig. 5). In plots where maize was planted during the preceding dry season, yields were 1.6 tons/ha compared to 1.3 and 0.7 ton/ha, respectively, in the mungbean and weedy fallow plots. OF SEEDBED PREPARATION IN DIFFERENT REGIONS B. METHODS

In Asia, little mechanization is used to prepare seedbeds. Fields are plowed and harrowed with animal-drawn implements when rain falls (De Datta and Ross, 1975). In Africa, upland rice fields are tilled manually. Deep plowing is also a weed control method in upland rice. Curfs (1975) found fewer weeds on soils that were deep plowed and rototilled with a powerful tractor than on those that were shallow plowed or zero tilled. Pande and Bhan (1966a) also showed the effectiveness of deep (28 cm) tillage in reducing weed biomass and increasing panicle number and grain yield as compared to tillage at 7, 14, and 21 cm (Fig. 6). At IRRI, Lopez et al. (1980) and De Datta and Llagas (1984) reported that increasing the number of rototillings reduced weed population, weed biomass, and weeding time (Lopez et al., 1980) and increased grain yields. c . REDUCEDTILLAGE AND WEED GROWTH Crops can be grown with minimum soil disturbance and reduced energy inputs for cultural operations if herbicides are used to control weeds.

305


;j 1.5

150 Nc.-45N-

L \

Advances in Agronomy, Volume 38

Advances in Agronomy, Volume 60

Advances in Agronomy, Volume 50

Advances in Agronomy, Volume 68

Advances in Agronomy, Volume 54

Advances in Agronomy, Volume 11

Advances in Agronomy, Volume 31

Advances in Agronomy, Volume 33

Advances in Agronomy, Volume 65

Advances in Agronomy, Volume 21

Advances in Agronomy, Volume 64

Advances in Agronomy, Volume 105

Advances in Agronomy, Volume 20

Advances in Agronomy, Volume 100

Advances in Agronomy, Volume 108

Advances in Agronomy, Volume 69

Advances in Agronomy, Volume 83

Advances in Agronomy, Volume 91

Advances in Agronomy, Volume 46

Advances in Agronomy, Volume 111

Advances in Agronomy, Volume 62

Advances in Agronomy, Volume 56

Advances in Agronomy, Volume 28

Advances in Agronomy, Volume 53

Advances in Agronomy, Volume 14

Advances in Agronomy, Volume 26

Advances in Agronomy, Volume 103

Advances in Agronomy, Volume 81

Advances in Agronomy, Volume 40

Advances in Agronomy, Volume 45

Advances in Agronomy, Volume 101

Advances in Agronomy, Volume 38

Advances in Agronomy, Volume 60

Advances in Agronomy, Volume 50

Advances in Agronomy, Volume 68

Advances in Agronomy, Volume 54

Advances in Agronomy, Volume 11

Advances in Agronomy, Volume 31

Advances in Agronomy, Volume 33

Advances in Agronomy, Volume 65

Advances in Agronomy, Volume 21

Advances in Agronomy, Volume 64

Advances in Agronomy, Volume 105

Advances in Agronomy, Volume 20

Advances in Agronomy, Volume 100

Advances in Agronomy, Volume 108

Advances in Agronomy, Volume 69

Advances in Agronomy, Volume 83

Advances in Agronomy, Volume 91

Advances in Agronomy, Volume 46

Advances in Agronomy, Volume 111

Advances in Agronomy, Volume 62

Advances in Agronomy, Volume 56

Advances in Agronomy, Volume 28

Advances in Agronomy, Volume 53

Advances in Agronomy, Volume 14

Advances in Agronomy, Volume 26

Advances in Agronomy, Volume 103

Advances in Agronomy, Volume 81

Advances in Agronomy, Volume 40

Advances in Agronomy, Volume 45

Advances in Agronomy, Volume 101

Recommend Documents