WATER RESEARCH A Journal of the International Water Association
Editor-in-Chief Mogens Henze Institute of Environment & Resources Technical University of Denmark Bygningstorvet DK-2800 KGS Lyngby Denmark Tel: 45 4525 1477 Fax: 45 4593 2850 E-mail:
[email protected] Editors J. Block Université H. Poincaré, Nancy I France David Dixon University of Melbourne Australia Hiroaki Furumai The University of Tokyo Japan Eberhard Morgenroth University of Illinois Urbana-Champaign USA W. Rauch University Innsbruck Austria Maria Reis Universidade Nova de Lisboa/FCT Portugal Hang-Shik Shin Korea Advanced Institute of Science and Technology Korea Mark van Loosdrecht Delft University of Technology The Netherlands
Damien Batstone The University of Queensland Australia G-H. Chen The Hong Kong University of Science & Technology Hong Kong China Tom Curtis Univ. of Newcastle upon Tyne UK Ana Deletic Monash University USA Francis de los Reyes III North Carolina State University USA Rob Eldridge The University of Melbourne Australia Rosina Girones University of Barcelona Spain Stephen Gray Victoria University Australia Kate Grudpan Chiang Mai University Thailand Xiaodi Hao Beijing University of Civil Engineering and Architecture China E.E. Herricks University of Illinois - Urbana USA H-Y. Hu Tsinghua University China P.M. Huck University of Waterloo Canada Bruce Jefferson Cranfield University UK
Thomas Ternes Bundesanstalt für Gewässerkunde Germany
Sergey Kalyuzhnyi Moscow State University Russian Federation
Stefan Wuertz Univ. of California, Davis USA
Jaehong Kim Georgia Institute of Technology USA
Hanqing Yu University of Science & Technology of China China
Associate Editors Andrew Baker The University of Birmingham UK
Gregory Korshin University of Washington USA Jes La Cour Jansen Lund Institute of Technology Sweden G. Langergraber BOKU/Univ. of Natural Res. and Applied Life Scs. Austria
Anna Ledin Technical University of Denmark Denmark S-L. Lo National Taiwan University Taiwan Y. Matsui Hokkaido University Japan Max Maurer EAWAG Switzerland How Yong Ng National University of Singapore Singapore Satoshi Okabe Hokkaido University Japan S.L. Ong National University of Singapore Singapore Jong M. Park Pohang University of Science & Technology Korea Miguel Salgot University of Barcelona Spain David Sedlak University of California, Berkeley USA Jean-Philippe Steyer LBE-INRA France M. Takahashi Hokkaido University Japan Kai Udert EAWAG Switzerland V.P. Venugopalan BARC Facilities India E. von Sperling Federal University of Minas Gerais Brazil A.I. Zouboulis Aristotle University Greece
Editorial Office Mailing address Water Research Radarweg 29 1043 NX Amsterdam The Netherlands Tel: 31 20 485 2701 Fax: 31 20 485 2521; 31 20 485 3264 E-mail:
[email protected] Publication information: Water Research (ISSN 0043-1354). For 2010, volume 44 is scheduled for publication. Subscription prices are available upon request from the publisher or from the Elsevier Customer Service Department nearest you or from this journal’s website (http://www.elsevier.com/locate/watres). Further information is available on this journal and other Elsevier products through Elsevier’s website (http://www.elsevier.com). Subscriptions are accepted on a prepaid basis only and are entered on a calendar year basis. Issues are sent by standard mail (surface within Europe, air delivery outside Europe). Priority rates are available upon request. Claims for missing issues should be made within six months of the date of dispatch. Author Enquiries: Please visit this journal’s homepage at http://www.elsevier.com/locate/watres. You can track accepted articles at http://www.elsevier.com/ trackarticle and set up e-mail alerts to inform you of when an article’s status has changed. Also accessible from here is information on copyright, frequently asked questions and more. Contact details for questions arising after acceptance of an article, especially those relating to proofs, will be provided by the publisher. Orders, claims, and journal enquiries: Please contact the Elsevier Customer Service Department nearest you: St. Louis: Elsevier Customer Service Department, 11830 Westline Industrial Drive, St. Louis, MO 63146, USA; phone: (877) 8397126 [toll free within the USA]; (+1) (314) 4537076 [outside the USA]; fax: (+1) (314) 5235153; e-mail:
[email protected] Oxford: Elsevier Customer Service Department, The Boulevard, Langford Lane, Kidlington OX5 1GB, UK; phone: (+44) (1865) 843434; fax: (+44) (1865) 843970; e-mail:
[email protected] Tokyo: Elsevier Customer Service Department, 4F Higashi-Azabu, 1-Chome Bldg, 1-9-15 Higashi-Azabu, Minato-ku, Tokyo 106-0044, Japan; phone: (+81) (3) 5561 5037; fax: (+81) (3) 5561 5047; e-mail:
[email protected] Singapore: Elsevier Customer Service Department, 3 Killiney Road, #08-01 Winsland House I, Singapore 239519; phone: (+65) 63490222; fax: (+65) 67331510; e-mail:
[email protected] Application for membership of International Water Association should be made to: Executive Director, IWA, Alliance House, 12 Caxton Street, London SW1H 0QS, U.K. (Tel.: +44 207 654 5500; Fax: +44 207 654 5555; e-mail:
[email protected]; website: http://www.IWAhq.org.uk). Registered Charity (England) No. 289269. Individual membership is available from £30 upwards. For details contact IWA.
water research 44 (2010) 1–19
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Review
Nitrification and me – A subjective review Willi Gujer*,a,b a b
Swiss Federal Institute of Aquatic Science and Technology, Eawag, 8600 Dubendorf Institute of Environmental Engineering, ETH, 8093 Zurich, Switzerland
article info
abstract
Article history:
Based on the subjective experience of the author it is discussed how the nitrification
Received 22 April 2009
processes served as an important basis for the development of today’s understanding and
Accepted 25 August 2009
mathematical models for many wastewater treatment processes (activated sludge, biofilm
Published online 1 September 2009
reactors) and self-purification processes in rivers. Besides being an important process for the protection of receiving waters, nitrification served as a proxy for the understanding of
Keywords:
the behavior of a narrowly defined group of microorganisms growing on known substrates
Nitrification
under environmental conditions. Until the upcoming of readily available microbial genetic
Activated sludge process
techniques, nitrification was the single most studied microbial process in environmental
Biofilm
engineering.
Self-purification
ª 2009 Elsevier Ltd. All rights reserved.
Modeling
Contents 1. 2. 3. 4. 5.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Nitrification – catalyst for the change of the paradigm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Nitrification before my time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Nitrification becomes a task for water pollution control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Nitrification in the activated sludge process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 5.1. Sampling frequency makes the difference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 5.2. A ‘‘safety factor’’ controls design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 5.3. Long term and short term temperature effects are comparable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 5.4. Inhibition of nitrification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 5.5. Process control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 5.6. Peak shaving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 5.7. Geography affects the performance of wastewater treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 5.8. Design concepts must be revisited . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 5.9. Nitrification kinetics depends on many environmental factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 5.10. Detailed understanding of ammonium oxidation requires enzyme kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 5.11. Interaction of nitrification and denitrification may cause loss of nitrous oxide (N2O) . . . . . . . . . . . . . . . . . . . . . . . . . 8
* Swiss Federal Institute of Aquatic Science and Technology, Eawag, 8600 Dubendorf, Switzerland. Tel.: þ41 44 823 5036; fax: þ41 44 823 5389. E-mail address:
[email protected] 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.08.038
2
water research 44 (2010) 1–19
6. 7. 8.
9. 10. 11.
12. 13. 14.
1.
Dynamic activated sludge models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Biofilm models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Experiments with biofilm systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 8.1. Laboratory systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 8.2. Rotating biological contactors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 8.3. Tertiary trickling filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 8.4. Dual media sand filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 8.5. Hybrid systems outcompete two stage processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 8.6. Summary on tertiary nitrification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Nitrification in receiving waters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Immission standards for nitrogen species . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Nitrification as a case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 11.1. From Nitrobacter to nitrite-oxidizing bacteria (NOB) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 11.2. Nitrification as an indicator for micropollutant degradation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 11.3. The case of bioaugmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 11.4. The case of conventional activated sludge versus membrane bioreactors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 11.5. Model structure uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 11.6. Kinetic parameters are stochastic variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 11.7. Batch tests may not yield reliable kinetic information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 11.8. Chemical nitrification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 11.9. Nitrification provides evidence for the anammox process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 11.10. Ammonium as a reactive tracer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Open questions and outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Do we stand on the brink of a new paradigm again? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Introduction
Nitrification was the single most important process in our development of today’s theoretical understanding of biological wastewater treatment processes. It is an important process in wastewater treatment plants however research on this process has two entirely different aspects: 1. In order to protect receiving waters we elaborate the engineering information required for establishing reliable nitrification performance of biological wastewater treatment systems and an understanding of nitrogen transformation in receiving waters. Here the goal is making full scale use of the nitrification process in order to achieve improved water pollution control. 2. We use the nitrification processes to follow the behavior and performance of specific, narrowly defined groups of microorganisms in an otherwise ill-defined mixed population. Alternatively we follow the transformation of a specific compound (ammonium) in a complex chemical matrix (wastewater). Here nitrification is just a proxy for learning more about the detailed behavior of microorganisms and pollutants in general. In this report I am trying to analyze how these two aspects evolved over the last decades and specifically how they have affected my own research and perception of these aspects of engineering biological processes in the environment and in wastewater treatment systems.
Do not expect a careful, detailed, broad and well balanced review of nitrification. Rather accept this report to bear to a large extent the color of my subjective analysis and experience. I think that stepping back and trying to understand how science in this area evolved over the decades may teach us how to proceed successfully into the future. If I can contribute to this end I have more than reached my goal.
2. Nitrification – catalyst for the change of the paradigm For many decades empirical ratios such as mean hydraulic residence time, volumetric loading, food to microorganism ratio (F/M), etc. served as the basis for sizing the reactors of biological treatment plants (trickling filters, activated sludge tanks). Beginning in the mid 1950s and heavily influenced by chemical engineers, sanitary engineers began to analyze their systems based on systems analytical methods, mass balances, transformation processes and transformation rates, kinetics and stoichiometry, reactor hydraulics etc. This change of the paradigm allowed or improved productive communication between engineers and natural scientists. Nitrification as a transformation process which is easily identified and perfectly serves water pollution control was a very welcome example to demonstrate the advantages of the new tools. For over 30 years many new concepts were introduced and first demonstrated with the aid of nitrification.
water research 44 (2010) 1–19
Thus we ‘‘owe’’ a lot to this process which supported to a large extent the development of today’s understanding and technological know how in biological wastewater treatment.
3.
Nitrification before my time
‘‘Before parameters such as BOD, COD, and organic carbon were used to judge the efficiency of a wastewater treatment system, a high degree of nitrification in a secondary effluent was assumed to be an indicator of a well-treated sewage’’ (Gujer, 1974). Nitrification was initially not seen as a necessity from the point of view of the receiving water but was rather experienced as a cause of eutrophication and nuisance (Fair and Geyer, 1954). Obviously nitrification activity was here used in lieu of better alternatives to judge the progress of the treatment process and not in order to judge the load of reduced or oxidized nitrogen on receiving waters. The introduction of the chemostat by Monod (1950) and Novick and Szilard (1950) laid ground for the understanding and mathematical modeling of continuous microbial culture systems. Garrett (1958) seems to be the first author who related microbial growth to the wasting rate of activated sludge; he realized the direct relationship between wasting rate and washout of a group of microorganisms. In his report he writes: ‘‘The monthly averages of the total nitrite-plus nitrate– nitrogen ranged from 0.2 to 0.7 ppm. This is not a significant amount of oxidized nitrogen, and is probably a result of wasting solids at a rate more rapid than the maximum rate of growth of the nitrifying organisms under the conditions in the aeration tanks.’’ A substantial step in understanding nitrification in the activated sludge process is due to a research group at the British Water Pollution Research Laboratory (Water Pollution Research, 1964). Here Downing et al. (1964) developed a comprehensive theoretical concept for the design of nitrifying activated sludge plants based on kinetic concepts and reactor technology. Wuhrmann (1964) substantiated this concept and introduced the German term ‘‘Schlammalter’’ (sludge age) in the context of washout of nitrifiers. Other authors report on more empiric studies of nitrification and did not yet integrate the upcoming theoretical approaches (see e.g. Balakrishnan and Eckenfelder, 1969). By 1970 the use of the synonymous terms Solids Retention Time (SRT), Mean Cell Residence Time (MCRT) or Sludge Age (SA, its meaning was revised after its first definition based on incoming solids) in modeling and design of nitrifying activated sludge plants have been firmly established (see e.g. Lawrence and McCarty, 1970) and became part of modern sanitary engineering education. By 1975 the first fully dynamic models of nitrification became available (Lijklema, 1973 or Poduska and Andrews, 1975), with supporting data based on experimental work with artificial sewage. Physical/chemical treatment options for nitrogen removal were studied and realized in a few full scale plants in the early 1970s. Breakpoint chlorination, ion-exchange (on clinoptilolite) and air-stripping of ammonia (NH3) were considered to be competitive. The fact that processes, specific for ammonium removal, were studied extensively indicates that ammonium (and nitrate) started to be recognized as a problem in water pollution control.
3
In 1975 USEPA produced a then rather advanced design manual for nitrogen control which describes the state of the art at this time: Complex treatment schemes such as three sludge systems (high loaded activated sludge for COD removal, separate system for nitrification and a third system fed with methanol for denitrification) and rather involved physical– chemical processes are documented in this manual. The introduction to this manual states: ‘‘This manual could not have been produced five years ago (1970) because of the state of nitrogen control technology at that time.’’ It is interesting to follow up on this manual. USEPA (1993) published ‘‘an update and a revision of the original 1975 edition’’. It states: ‘‘Since the first manual’s publication, the trend in nitrogen control technology applications has been overwhelmingly in favor of biological processes, with only a few instances in which physical/chemical processes have been implemented.’’ Thus when I started my career in process engineering of wastewater treatment in 1971 as a young PhD candidate, secured design information for biological nitrification was still lacking. It was however rapidly developed throughout the 1970s. My own first contribution was a steady state model for nitrification in the contact stabilization activated sludge process (Gujer and Jenkins, 1975).
4. Nitrification becomes a task for water pollution control Discharge requirements for ammonium, nitrite and nitrate started to be enacted in the 1970s. In Switzerland the first wastewater treatment plant for which nitrification became required was the Werdho¨lzli plant of the city of Zurich, where less than 2 gNH4–N m3 in more than 80% of the 24 h flow proportional composite samples had to be reached above 10 C. In 1973 the city of Zurich announced an international competition for the design of the extension of its wastewater treatment plant. From this competition with world wide participation it became clear that secured design information for many of the proposed process alternatives was lacking and full scale experience was still rather scarce (Wiesmann, 1982). As a young sanitary engineer, I was assigned the task to develop the design criteria for the extension of the Werdho¨lzli plant with the aid of pilot plants that were available at the Swiss Federal Institute for Water Resources and Water Pollution Control (Eawag) directly on the main sewer feeding into the treatment plant. The challenge of this project was a major factor in the future development of my career.
5. Nitrification in the activated sludge process 5.1.
Sampling frequency makes the difference
The performance of pilot plants as well as full scale plants is typically monitored based on 24 h composite samples. It is only recently that reliable on-line sensors became available which provide much higher time (and possibly space) resolution. In the 1970s, when automatic sampling was hardly available and all monitoring was based on wet-chemical
4
water research 44 (2010) 1–19
analysis, data with high temporal resolution were scarce and expensive to obtain. In the context of the pilot tests for the extension of the wastewater treatment plant Werdho¨lzli in Zurich we performed a detailed and then very costly sampling procedure with 2 h composite samples in the influent and grab samples in the effluent of the biological reactor of a single CSTR type activated sludge process with partial nitrification (Fig. 1). The fact that we sampled the effluent of the biological reactor rather than the effluent of the plant (secondary clarifier), where concentration fluctuation would be hydraulically attenuated, proved to be very rewarding. The immediate breakthrough of ammonium upon the increased loading in the morning hours (urine) clearly revealed that nitrification is a highly dynamic phenomenon which cannot easily be described based on static models. These results lead me into dynamic simulation which remained a central topic of my research for years to come. My first dynamic model (Gujer, 1977) concentrated on the prediction of the nitrification activity of activated sludge. Receiving waters were included with some simple complete mixing models and allowed to predict diurnal variation of ammonium (NHþ 4 ), ammonia (NH3) and nitrate (NO3 ) throughout the year. In 1975 it was necessary to develop a FORTRAN code, specific for this case, and implement it on the high capacity main frame of the Swiss Federal Institute of Technology. Altogether this was an effort which required several weeks. Today, using advanced simulation tools, a similar model and program would be available within hours. In addition, systems analysis tools (sensitivity, parameter estimation, etc.) would be available to support and simplify model development and identification (Gujer, 2006).
5.2.
A ‘‘safety factor’’ controls design
Lawrence and McCarty (1970) introduced the concept of a safety factor (SF) in the design of activated sludge plants which relates the solids retention time chosen for the design to the solids retention time at which complete failure of the plant (complete washout of the relevant organisms) would result. Later it was shown (USEPA, 1975) that it is reasonable to choose SF in excess of the ratio of the daily peak ammonium
Fig. 1 – Diurnal variation of ammonium concentration in the influent and effluent of the aeration tank of a pilot plant operated for the design of the biological wastewater treatment plant Werdho¨lzli in Zurich. The plant was operated at a sludge age of 5.4 d, the samples were collected at temperatures around 13 8C, the aeration tank was completely mixed. Adapted from Gujer and Erni (1978).
load divided by the daily average ammonium load (Lmax/Lavg). This became an important relationship in many design procedures for nitrifying activated sludge systems. It allowed including many specific local conditions into the design: Diurnal load variations, design temperature, expected inhibition, sludge production, etc. For the design of the Werdho¨lzli treatment plant a combination of experimental and modeling results yielded Fig. 2. Here SF is defined as: SF ¼ mmax $SRT where mmax is the maximum specific growth rate of ammonium oxidizers under design operating condition (temperature, dissolved oxygen, pH, inhibitors) which stands for the maximum activity of the nitrifying population. SRT is the expected aerobic solids retention time for the design loads of COD, TSS and P, it is related to the size of the nitrifier population in the system.
5.3. Long term and short term temperature effects are comparable An important aspect of long term dynamic simulation of activated sludge plants is the question how microorganisms respond to short term (diurnal variation) and long term (seasonal) temperature change. We answered this question with the aid of pilot plants that we operated at different temperatures close to the washout of nitrifying organisms. Nitrification efficiency was maintained at about 50%, excess sludge removal was increased or decreased based on daily analytical result. In addition the nitrification activity of biomass grown at different temperatures (6 and 14 C) was evaluated after a rapid change of the temperature (hours). As indicated in Fig. 3, it turned out that the long term maximum growth rate and the short term activity of the biomass both increased by a factor of 0.11 C1. In mathematical modeling this allows using just one temperature dependency for nitrifying biomass, independent of the time frame of the temperature change.
Fig. 2 – Nitrification efficiency as a function of the safety factor for design of the Werdho¨lzli treatment plant. Basis is a diurnal ammonium load variation by a factor of 2 (diurnal peak to average load). Developed for winter conditions, 8–12 8C. Adapted from Gujer (1977).
water research 44 (2010) 1–19
5
Fig. 3 – Long term (left) and short term (right) effects of temperature on ammonium oxidizing organisms. Growth rates are based on 12 weeks of operation of 3 pilot plants close to washout of the organisms at 6, 10 and 14 8C. Short term effects are based on batch results with activated sludge grown at 6 and 14 8C. Adapted from Gujer (1977).
5.4.
Inhibition of nitrification
Early reports on nitrification in activated sludge processes typically included the remark that industrial wastewater may have inhibited nitrification if not fully then at least partially. Even though heavy metals and some organics are known to inhibit the growth of nitrifying organisms (Tomlinson et al., 1966), my personal experience (without scientific proof) deviates from the above remarks. This was the time when oxygen electrodes were not available and dynamic behavior of biological systems was poorly understood and resulted in poor operating strategies frequently far from ‘‘steady state’’. Thus frequent periods of lack of oxygen, poor control of SRT and time dependant ammonium loads may have been the dominant cause of reports on the inhibition of nitrification. Today, with more reliable control of oxygen concentration, reports on inhibition are less frequent. In addition raw wastewater in industrialized countries is under permanent control such that toxic compounds must not be expected with high frequency. Personally I have yet to find a case where an industrial effluent can be demonstrated to be the cause of reduced nitrification. Gujer and Boller (1978) report on the effect of different chemicals for the precipitation of phosphorus in activated sludge plants (pre-precipitation and simultaneous precipitation) on nitrification. We found a reduced maximum growth rate of nitrifying organism when Ferrous Sulfate (FeSO4) was used as a precipitant but we could not identify any mechanism which caused this apparent inhibition. In addition we found a weak effect of digester supernatant on the maximum growth rate of ammonium oxidizing organisms. Digester supernatant is an important recycle stream which may contain reducing, inhibitory (sulfur) compounds when directed back to the wastewater treatment plant. We tested the effect of such recycling with a pilot plant which was operated close to washout of nitrifying organisms (see above). As indicated in Fig. 4 the effect of digester supernatant is only small but statistically significant. A typical NH4-load in the supernatant is in the order of 10–20%, depending on the sludge thickening and dewatering processes applied.
5.5.
Process control
Equipped with a calibrated and field tested dynamic model for nitrification in activated sludge systems Gujer and Erni (1978)
could simulate the effect of hydraulic flow scheme, ammonium load balancing, limitation of nitrification by oxygen and some process control strategies. Given the process is supported by sufficient dissolved oxygen, ammonium load balancing was proven to be by far the most efficient means for improving nitrification performance. An example of such load balancing by digester supernatant is given in Fig. 5: Growing more nitrifiers during the night prepares the activated sludge to better deal with high loads during peak loading situations. Since digester supernatant is rich in ammonium (roughly 700 gN m3) it proves to be very efficient to store this liquid and recycle it in the best possible moment. Bringing it back to the treatment plant when it is generated (typically during working hours when raw sludge is fed to the digesters or when digested sludge is dewatered) would only add to the peak load and would thus bleed through the plant. In addition any inhibitory effects of digester supernatant (Fig. 4) would have fewer consequences during low load rather than high load periods. Load balancing is an early version of what was later termed waste design, a term which stands for the generation of wastewater amenable to improved or optimal treatment (Larsen and Gujer, 2001).
Fig. 4 – Effect of digester supernatant on maximum growth rate of ammonium oxidizing organisms. Each point is the weekly average growth rate of activated sludge operated close to washout. Ammonium is used as the tracer for an unknown possible toxic compound. Adapted from Gujer (1976a).
6
water research 44 (2010) 1–19
Fig. 5 – Balancing of ammonium load with digester supernatant in order to enhance the nitrifier population (Gujer and Erni, 1978).
Today the IWA Task Group on Benchmarking of Control Strategies for WWTPs (http://www.benchmarkwwtp.org/, see also Copp, 2002) provides a fully developed framework for testing alternative control strategies for nitrification/denitrification in an activated sludge plant. In the latest version of this framework load balancing is included in the defined options however the flow scheme itself cannot be optimized.
5.6.
Peak shaving
Since nitrifying biological treatment must typically be designed for peak ammonium loads, it is advantageous trying to even out the ammonium load throughout the day not only by using ammonium-rich recycle streams from sludge handling but potentially directly at the source. Urine separation toilets were introduced in Sweden in the early 1990s in order to recycle valuable nutrients in a concentrated separate stream (Kirchmann and Pettersson, 1995; Hanæus et al., 1997). The concept and the consequences of urine separation for improved water pollution control were introduced and discussed by Larsen and Gujer (1996). Rauch et al. (2003) suggest separating urine at the source, storing it and randomly discharging it to the sewer. This will result in an even ammonium load on the treatment plant (peak shaving) which will improve nitrification performance and could even enhance denitrification. Combined with a strategy to withhold urine during rain events, ammonium in combined sewer overflow (CSO) could be reduced to further enhance water pollution control. Such activities which optimize the composition of wastewater in view of efficient wastewater treatment are summarized today under the terms waste design and source control (Larsen and Gujer, 2001).
5.7. Geography affects the performance of wastewater treatment The load variation in the influent to the treatment plant is the result of the convolution of time dependent input of pollutants into the sewer and the residence time distribution of the sewage in the sewer (Fig. 6). Whereas concentric catchments lead to large load variation, linear and very large catchments
lead to load equalization. Since diurnal load variation controls to a large extent the nitrification performance of biological treatment processes, resulting load equalization in linear or large catchments is advantageous. Fig. 7 summarizes the extreme 2 h ammonium load relative to the daily average from a variety of catchments. Since the safety factor (SF) for the design of nitrifying treatment plants is typically chosen in the order of the ratio of the maximal to average Load (Lmax/Lavg) this figure provides important design information. The choice of SF according to Fig. 7 has the interesting and desired feature that SF and thus the solids retention time becomes larger the smaller the treatment plant and thus the less professional and more difficult the operation.
5.8.
Design concepts must be revisited
Dominguez and Gujer (2006) discuss the evolution of the wastewater treatment plant Werdho¨lzli in the period of 1985– 2003. The plant was initially designed for nitrification (see above) and included simultaneous precipitation of phosphorus. The design loads chosen exceeded observed loads by about 15%. The treatment concept relied on the idea that an old existing activated sludge plant could be used to pretreat about 50% of the primary effluent in order to facilitate nitrification in a new second stage activated sludge process. Over the 18 years in question, the population of the city of Zurich decreased rather than increased by about 20%. Many large, wastewater producing industries (brewery, milk processing plant, slaughterhouse, etc.) left the city. Phosphate was banned in textile detergents which resulted in less sludge production from precipitation. Groundwater infiltration into the sewers was drastically reduced and drinking water consumption decreased by 33% which allowed reducing the maximum hydraulic load of the treatment plant and thus resulted in an increase of the allowable activated sludge concentration. The old activated sludge process was taken out of operation, pre-denitrification with 28% of the volume was introduced in the new plant without extending reactor volume. A second wastewater treatment plant of the city of Zurich was taken out of operation and the wastewater was fed into the Werdho¨lzli plant, this added an extra 20% to the load. Temporarily deicing fluids from Zurich airport were treated as well. And so on. We realized from this analysis that a wastewater treatment plant is a ‘‘living organism’’ and will hardly ever be operated in the way and with the performance it was designed for. Over the short period of 18 years the boundary conditions as well as the flow scheme of the Werdho¨lzli plant changed dramatically. The future is difficult to predict and our design concepts should consider this uncertainty.
5.9. Nitrification kinetics depends on many environmental factors Holiencˇin (1996) developed a kinetic model for the production of nitrite in the context of nitrification in the activated sludge process. Due to the many environmental factors involved, this model is complex and cannot be presented here. A summary
7
water research 44 (2010) 1–19
Concentric catchment
Linear catchment
A
B B A
C C
WWTP
WWTP
Relative load 10 Concentric catchment
8 6
Relative Load 6
4
4
2
2
0
0
6
12 18 Time of day, hrs
24
0
0
6
Linear catchment
12 18 Time of day, hrs
24
Fig. 6 – Concentric or linear arrangement of urban areas affects diurnal variation of pollutant loads in wastewater treatment plants.
of the trends is given in Tables 1 and 2. They relate to the following simple model for two step nitrification: r ¼ rmax $
S Ks þ S
with
Most of today’s dynamic models for the activated sludge process do not include enzyme dynamics. For most practical purposes Manser et al. (2006) conclude that the approach chosen in ASM3 (Gujer et al., 1999; see below), which differentiates between biomass decay rates under aerobic and anoxic conditions, is sufficient.
r ¼ rate of oxidation of ammonia or nitrite [gN m3 d1] rmax ¼ maximum value of r under operating conditions 3 S ¼ concentration of substrate (NH3 or NO 2 ) [gN m ] KS ¼ saturation coefficient for true substrate.
5.10. Detailed understanding of ammonium oxidation requires enzyme kinetics Manser et al. (2006) analyzed decay processes of nitrifying bacteria under aerobic and anoxic conditions. They found large differences in apparent decay rates under aerobic conditions (large rates) and anoxic conditions (rates close to zero). However in order to interpret their batch results in detail, they had to introduce the dynamics of enzyme saturation of the organisms. Under aerobic conditions at 20 C they found a decay rate of enzymatic activity of ammonium oxidation of kdecay ¼ 3 d1 and a regeneration of this activity in the presence of ammonium of ksynthesis ¼ 30 d1. Under anoxic conditions the decay of enzymatic activity was negligible.
Fig. 7 – Extremes of diurnal ammonium load variation in the influent to wastewater treatment plants as a function of average load or size of the treatment plant. Assuming 10 gN capL1 dL1 the X-coordinate covers the range of 1000– 1,000,000 population equivalents. Adapted from Gujer and Erni (1978).
8
water research 44 (2010) 1–19
Table 1 – Effects of environmental conditions on max. activity (rmax) and saturation coefficient KS of ammonium oxidation (nitritation). Parameter
rmax KS
NH3 (ammonia) NHþ 4 (ammonium) HNO2 NO 2 (nitrite) HNO3 NO 3 (nitrate) pH value (Hþ activity) Temperature O2 (dissolved oxygen) Alkalinity Organic substrate Hydraulic retention time Solids retention time
Remark
Table 2 – Effects of environmental conditions on max. activity (rmax) and saturation coefficient KS of nitrite oxidation (nitratation). Parameter
rmax KS
Yes No True substrate No No In equilibrium with NH3, depending on pH and temperature Yes No Non-competitive inhibition No No In equilibrium with HNO2, depending on pH and temperature No No No No Yes Yes pH range 6.2–8.0: non-competitive inhibition by Hþ, KS increases with increasing pH Yes No Yes No We do not have our own results for KS
NO-2 (Nitrite) HNO2 NO-3 (Nitrate) HNO3 NH3 (Ammonia) NHþ 4 (Ammonium) pH value (Hþ activity) Temperature O2 (dissolved Oxygen) Alkalinity
Yes Yes No Yes Yes No
Yes No Affects pH inside flocs (diffusion limitation) Yes No Indirect effect, reduced O2 available inside flocs Yes No Affects biomass concentration
Organic Substrate Hydraulic retention time Solids retention time
Yes No Activity decrease due to decay at elevated SRT
No No Yes No No No
Remark Most probably true substrate Non-competitive inhibition Competitive inhibition Non-competitive inhibition Very weak dependency
Yes No pH range 6.2 to 8.0: Only weak at elevated pH values Yes Yes Yes Yes We do not have our own results for KS Yes No Weak indirect effect from nitritation due to pH change inside flocs Yes No Indirect effect, reduced O2 available inside flocs Yes No Affects biomass concentration
Yes No Activity decrease due to decay at elevated SRT
Adapted from Holiencˇin and Gujer (1996).
Adapted from Holiencˇin and Gujer (1996).
5.11. Interaction of nitrification and denitrification may cause loss of nitrous oxide (N2O) In the process of heterotrophic denitrification some N2O is produced as an intermediary product. Since separate denitrification reactors are not aerated this N2O is only stripped to a very small degree in a following nitrification reactor. More critical is the situation in plants operated with simultaneous or alternating nitrification/denitrification. These systems combine elevated nitrite and low oxygen concentrations with gas stripping, a situation which was identified as critical (von Schulthess et al., 1994). Later von Schulthess and Gujer (1996) measured N2O production in a full scale activated sludge process under different operating conditions. We concluded that not more than 0.072% of the incoming nitrogen is released to the atmosphere as N2O if nitrification (2 gO2 m3) and denitrification (0 gO2 m3) are optimized separately. In a national balance, this would be a negligible amount. However Poth and Focht (1985) demonstrated that Nitrosomonas europaea which is present in activated sludge, especially under ammonium-rich conditions (Manser et al., 2005a; Manser, 2005), is able to denitrify nitrite to N2O under low oxygen, high nitrite conditions. Several researchers have demonstrated that this mechanism contributes substantial amounts of N2O to off gas and dissolved nitrogen in effluents of biological treatment. Thus today’s trend towards simultaneous nitrification/denitrification may start to emit large amounts of N2O, a very undesirable greenhouse gas, even considered in the Kyoto agreement. N2O emission from nutrient removal plants is presently
under scrutiny by many researchers and may prove to be rather more complex than identified in the limited studies of a single PhD student.
6.
Dynamic activated sludge models
A first generation of dynamic models for nitrification in the activated sludge process was developed in the 1970s (see above). An important input came from the research group around Gerrit v. R. Marais at the University of Cape Town (UCT). This group started to develop models with a broad scope, integrating degradation of soluble, colloidal and particulate organics, nitrification, denitrification as well as oxygen consumption and sludge production in cascades of mixed reactors, first for steady state (Marais and Ekama, 1976) and later for dynamic behavior (Dold et al., 1980). In 1982 Poul Harremoe¨s, then Vice-President of IAWPR, initiated the IAWPR Task Group on Mathematical Modeling for Design and Operation of Biological Wastewater Treatment. Based on the advanced work of the group from UCT this task group developed the family of activated sludge models known today as ASM1 to ASM3 (see Henze et al., 2000, for the documentation of the models and Gujer, 2006, for an appreciation of their development). One of the major contributions of this task group was the so called matrix notation which allows communicating rather complicated integrated mathematical models in a well organized and condensed format which was first developed by Gujer (1985). Today it appears that this family of ASMs is broadly accepted as state of the art models for the activated sludge process. Initial acceptance of these models related to a large extent to the success of these models in predicting
water research 44 (2010) 1–19
nitrogen transformations (nitrification, denitrification). In the meantime these models (especially ASM2d) have reached a level of complexity which is difficult to handle routinely by consulting engineers. Their responsible application and adaptation is still the task of highly experienced engineers. But applied by specialists they truly help to improve plant design. Calibration of ASMs is tedious and often done by ad hoc tuning procedures. Brun et al. (2002) developed a systematic approach which allows identifying the most important model parameters and their interdependencies. Such a procedure is however at this time not readily available for practical engineers; the resources of time, required software as well as the theoretical background for well founded interpretation of the results are barely available. Today such techniques are primarily applied in research and development environments. In addition it appears that many experimental results are heavily influenced by uncharacterized hydraulic deficiencies of reactors. Calibration concentrates however on the adaptation of the biokinetic models, thus, hydraulic deficiencies may be mimicked by adapted biological parameters, not exactly a productive procedure in dynamic modeling. Definitely it was the positive experience with dynamic models of nitrification in activated sludge processes which provided ample motivation to step into the development of the more comprehensive and integrated models. Presently a further valuable contribution is developed by the IWA Task Group on Good Modeling Practice which is working on ‘‘Guidelines for Use of Activated Sludge Models’’. Whether this will facilitate and improve the application of the rather detailed and complex models remains to be seen.
7.
Biofilm models
First models which explicitly considered diffusion of pollutants in the depth of biofilms appeared in the mid 1970s (Williamson and McCarty, 1976; Harremoe¨s, 1976). The model by Williamson and McCarty was able to deal with electron acceptor as well as electron donor. These first models could not deal with the competition of different groups of microorganisms. Thus an a priori prediction of nitrification performance in the presence of organic substrates was not possible. Whereas in suspended growth reactors the competition between different groups of organisms (say nitrifiers and heterotrophs) is rather easy to model and to understand, this competition becomes more involved when organisms grow attached inside a biofilm. What is the activity of organisms buried in the depth of a biofilm? How can slow growing autotrophic organisms be enriched when rapidly proliferating heterotrophic organisms grow close to the surface of a biofilm? Biofilm models must combine transformation and transport processes whereas suspended growth models are typically based on the assumption of complete mixing, which is a very simple description of complicated transport processes. Mueller et al. (1978) provide an early report on the performance of a rotating biological contactor (RBC) with simultaneous degradation of BOD and nitrification (Fig. 8). Clearly
9
heterotrophic activity is located in early, upstream reactors whereas nitrification sets in once soluble BOD is degraded. The distribution of the relative biomass depends on the composition of the external wastewater. Based on this observation Wanner and Gujer (1984) developed a steady state model which successfully described the competition between autotrophic and heterotrophic organisms within a biofilm. The model was qualitatively validated with the data of Mueller et al. (1978) and later expanded into a fully dynamic model describing species competition in biofilms (Wanner and Gujer, 1986). These models predict the distribution of different particulate fractions of biomass as well as pollutant concentrations over the entire biofilm (Fig. 9). Fruhen et al. (1991) worked with a highly controlled system and obtained experimental evidence that this mixed culture biofilm model allows to describe the competition of nitrifiers and heterotrophic organisms in a biofilm rather well. Changes in the external substrate composition had dramatic effects on the nitrification performance of a biofilm and on species distribution within the biofilm. An application of this model to a rotating biological contactor (RBC) is provided by Gujer and Boller (1990). It is based on a model similar to ASM1 but includes nitrite from nitrification. It is used to discuss the consequences of different operating strategies and possible problems of continuous operation. The model is generally applicable for the description of competing microorganisms in fixed biomass (biofilms). It teaches us the controlling factors which affect the relative abundance of organisms competing for space and substrate within biofilms.
8.
Experiments with biofilm systems
8.1.
Laboratory systems
Siegrist and Gujer (1987) used a laboratory scale biofilm reactor to simulate a trickling filter. The process of nitrification which is heavily pH dependent was used to demonstrate mass transfer effects within biofilms. A closed chamber allowed exposing the biofilm to different atmospheres (O2, N2, CO2) and together with the choice of alkalinity (buffer capacity, HCO 3 ) in the influent the drop of pH across the biofilm could be controlled. Model predictions closely matched biofilm behavior. We learnt how to combine diffusion, reaction and pH equilibrium models.
8.2.
Rotating biological contactors
From our modeling efforts (see above) we derived that high nitrification rates in trickling filters or rotating biological contactors (RBC) could be achieved if heterotrophic organisms would be excluded from biofilm reactors as much as possible, thereby high nitrifier biomass density within the biofilm could be reached. We compared the nitrification performance of an RBC after high rate activated sludge treatment without nitrification but with and without tertiary filtration to remove residual TSS. The idea was that TSS in the effluent of the
10
water research 44 (2010) 1–19
Fig. 8 – Experimental results obtained from a nitrifying rotating biological contactor operated with six stages in series. Data from Mueller et al. (1978), adapted from Wanner and Gujer (1984).
secondary clarifier would only dilute the biomass in the biofilm and would thereby reduce nitrification performance. The results of pilot tests are given in Fig. 10. Clearly filtration held back the heterotrophic organisms and resulted in the expected positive effect. Unfortunately this effect was not sustained in full scale equipment: the production of nitrifying biomass was smaller than its consumption by higher organisms (worms, snails, fly larvae) and nitrification activity was periodically lost due to predation. In addition the secondary, non-nitrifying treatment step contains a large volume of ammonium-rich wastewater. During storm events, this water is rapidly flushed into the tertiary nitrification plant which has only a small water volume. This instantaneous increase in load cannot typically be handled by such a tertiary system and results in massive bleeding of ammonium. Thus full scale performance did not match our expectations. We learnt from this experiment that on the one hand models are useful to develop new technology. However on the other hand we had to realize that not all aspects of full scale operation can successfully be piloted at reduced scale and in limited time periods.
8.3.
Tertiary trickling filters
In trickling filters, the biomass has a fixed position within the reactor whereas the wastewater passes by. Nitrifying organisms can only grow when their substrate is available and since diurnal load variation and temperature may result in a lack of substrate in the lower part of the filter, biomass development is not usually distributed evenly over the depth. In pilot experiments with a plastic media tertiary trickling filter for nitrification Boller and Gujer (1986) found the situation indicated in Fig. 11. The biomass in the lower part of the trickling filter is exposed to ammonium for about 1 h d1. Under these conditions biomass predation and decay are important relative to biomass growth. Thus, nitrification
Nitrification rate at 10°C, gNH4-N m-2 d-1 3 after filtration: jNH4 = 3.1⋅
2
SNH4 1.8 + SNH4
without filtration: jNH4 = 2.0 ⋅
1
SNH4 2.3 + SNH4
0 0
Fig. 9 – Relative biomass distribution over the depth of a biofilm and oxygen concentration profile. Adapted from Wanner and Gujer (1984).
5 10 15 Ammonium concentration, gNH4-N m-3
20
Fig. 10 – Nitrification rate of a tertiary RBC after high rate, non-nitrifying biological pretreatment. Suspended solids in the secondary effluent were either left in the effluent or removed by filtration. Rates are adjusted to 10 8C. Figure adapted from Boller et al. (1990).
11
water research 44 (2010) 1–19
activity in the lower parts of the filter is reduced. This is especially critical in autumn, when temperatures decrease. At elevated temperatures biofilm regions at the low end have hardly been exposed to ammonium, thus no biofilm activity could develop. With decreasing temperature this reactor region is now required in order to reach full performance, however it will take weeks until an active biofilm is available. The suggested strategy to deal with this problem is to cut the filter in two sections, what is the first section one week then becomes the second the other week and vice versa. Such a strategy allows developing a substantial biofilm throughout the entire reactor system even under summer conditions, when only half the reactor volume would be sufficient. However, a first full scale design which relied on this strategy, was abandoned due to high cost. The strategy was transferred to a tertiary RBC plant, where the reversal of flow direction was implemented rather than providing a two stage process. In tertiary trickling filters biomass production and thus biomass accumulation is rather small. Gujer and Boller (1984) report on massive invasions of higher organisms (trickling filter fly larvae, Psychodidae, and some worms, Naididae) which were grazing on this biomass and virtually wiped out the nitrification performance of tertiary trickling filters for extended periods of time. This problem was overcome by increasing the hydraulic load of the trickling filter, a strategy causing extra operating costs (pumping energy). The pilot experience with tertiary trickling filters led to the development of a rather simple but efficient mathematical model for the design of this technology (Gujer and Boller, 1986). The model deals with the competition of oxygen and ammonium and readily allows adjusting pilot experience to different temperatures. Even though this technology has never found broad application I still use this model in class to teach simple but meaningful biofilm models.
8.4.
Dual media sand filters
Sand filters after nitrifying biological treatment units accumulate nitrifiers and may thus be prepared for additional nitrification. The advantage of using a tertiary trickling filter for nitrification is twofold, (i) the biomass is highly enriched in nitrifiers and (ii) the effluent contains rather elevated concentrations of oxygen. Pilot tests revealed that a sand filter operated after a tertiary trickling filter with an effluent rich in oxygen could nitrify up to 1.7 gN m3 (see Fig. 12, Boller and Gujer, 1986), a substantial amount considering that discharge requirements in Switzerland typically are 2 gNH4–N m3. In addition this amount of nitrification has a very positive effect on nitrite too, especially in summer, when ammonium is low and nitrite may be elevated (s.a. Fig. 15). These results were obtained even directly after backwashing, indicating that some nitrifying biomass adheres to the filter material.
8.5.
Hybrid systems outcompete two stage processes
Today, two stage biological systems rely on optimized management of biomass and substrates. Matsche´ and Moser (1993) report on the performance of a two stage hybrid activated process which combines the biomass of the second, nitrifying activated sludge system with the sludge in the first system in order to improve nitrogen control (nitrification and denitrification). Later this concept has been implemented successfully in an adapted version in the main wastewater treatment plant of Vienna (Wandl et al., 2006). Here an optimal management of primary effluent (denitrification in the second stage), final effluent (nitrate, denitrification in the first stage), activated sludge from the second biological treatment step (nitrifiers, nitrification in the first step) and sludge from the first activated sludge process (sorbed organics for denitrification in the second step) leads to substantial improvements of nutrient removal performance. In our own research we followed the concept of separating the individual functions (organics removal by activated sludge and nitrification in fixed biomass reactors). This concept stems from a period when denitrification was
Effluent NH4 in gN m-3
6
4
2
Fig. 11 – Diurnal variation of ammonium concentration profiles over the depth of a tertiary nitrifying plastic media trickling filter (from top to bottom). Different concentration profiles are exceeded for the indicated time during the day. Example: At a depth of 1.2 m 6 gN mL3 are exceeded during 13 h dL1 when the influent varied between 6.5 and 21 gN mL3 (Boller and Gujer, 1986).
0
0
2
4
6
8
Influent NH4 in gN m-3 Fig. 12 – Correlation between influent and effluent ammonium concentration in a nitrifying dual media sand filter after a nitrifying tertiary trickling filter (Boller and Gujer, 1986).
12
water research 44 (2010) 1–19
hardly considered in Switzerland. Clearly its performance comes nowhere near the potential of the highly integrated hybrid systems. However optimized control of such a system might only be available on large, professionally operated plants.
8.6.
Summary on tertiary nitrification
Initially many biological wastewater treatment plants were designed for BOD removal only. First generation activated sludge processes in Switzerland typically did not nitrify even in summer. It was tempting to develop end of pipe type technology which could nitrify the effluent of such a plant. Thus, we developed design information for several alternative technologies for tertiary nitrification (see Boller et al., 1994). It turned out that none of these technologies were ever applied in more than a handful of applications. The rapid development of nutrient removal technology in the 1990s made tertiary nitrification soon obsolete. We learnt that predicting the future is difficult at best.
9.
Nitrification in receiving waters
It is not sufficient to predict nitrification performance of wastewater treatment but it is rather essential to understand the fate of nitrogen in the receiving waters as well. Fig. 13 shows the length profile of some nitrogen species along a small creek close to steady state. The anaerobic influent must first be aerated and after 200 m the process of nitrification becomes clearly visible. Nitrite is first produced and accumulates. As ammonium is degraded to sufficiently low levels nitrite follows suit and is itself degraded again to low residual levels. Here only high spatial resolution of the data can reveal these details. Fig. 14 shows the diurnal variation of the ammonia (NH3) concentration in a river about 1.5 km below the discharge of the effluent of a partially nitrifying treatment plant. The synchronization of temperature, pH and ammonium load by sunlight results here in extreme variations of the toxic compound.
Concentration in gN m-3 1.00 0.80 NH4+
0.60 18°C
0.40
NO2-
0.32
0.20 0
Ntot NO3-
0.16
0
200
400
600 800 Flowdistance in m
1000
1200
Fig. 13 – Length profile of mineral nitrogen species in a small creek. The creek drains the anaerobic hypolimnion of an eutrophic small lake, becomes reaerated and then nitrification sets in (unpublished, provided by M. Koch). Dashed lines relate to the example in Fig. 15.
Fig. 14 – Diurnal variation of ammonia concentration in the river Birs, 1.5 km below a partially nitrifying wastewater treatment plant. Variation is affected by temperature, photosynthesis (pH), and ammonium load all synchronized by sunshine as is the oxygen concentration (unpublished, provided by B. Hurni).
Good interpretation of water quality in receiving waters thus requires a detailed understanding of the processes not only in the treatment plants but also by self-purification (and self-polluting as in the case of nitrite) processes within the receiving waters themselves. Two aspects complicate this endeavor: - In creeks and small rivers the biomass responsible for selfpurification processes is concentrated in fixed biomass (biofilms) on the surfaces of the sediment and the leaves of macrophytes. Thus modeling its behavior requires developing some biofilm models for nitrification, subject to the extra complication of competition of abundant hetero- and phototrophic bacteria and algae. In addition growth surfaces vary enormously over the seasons because leaf surface of macrophytes depends heavily on sunshine. - Whereas activated sludge reactors typically are modeled as a series of completely mixed compartments a river resembles more a plug flow type reactor. Thus non-stationary mathematical models of rivers typically result in partial differential equations. A first simple model (Gujer, 1976b) allowed quantifying ammonium oxidation in small rivers as a function of growth surfaces, temperature and competing organisms. Later this model was extended to predict the maximum nitrite concentration that is reached in such rivers due to the oxidation of ammonium (Gujer, 1978). Since nitrite is toxic for fish we must understand the dynamics of this compound or else our investment into nitrification of wastewater might not be successful in restoring natural fish populations. Based on the model of competition of ammonium and nitrite-oxidizers Fig. 15 indicates the maximum nitrite concentration that will be reached in the context of selfpurification processes. Dashed lines in Fig. 13 at 650 m are repeated in Fig. 15, thus it becomes possible to estimate the maximum nitrite concentration that may be reached. Fig. 16 shows the results from a continuous monitoring exercise of ammonium and nitrite in the river Glatt in Switzerland (Berg, 1991). This river was at the time heavily loaded with non-nitrified secondary effluent. Clearly nitrite becomes a significant problem with increasing temperature
water research 44 (2010) 1–19
13
endeavor. In view of the costs of nitrogen control it became however a necessity to have arguments for or against nitrification at hand – and immission standards are strong arguments. Table 3 summarizes possible limits based on the following arguments for the choice of maximum allowable ammonium concentration in receiving waters under Swiss conditions (Gujer, 1978):
Fig. 15 – Equilibrium nitrite concentration in small rivers derived from a model suggested by Gujer (1978). Dashed lines relate to the example in Fig. 13.
and in summer may potentially be more of an ecological threat than ammonia. Jancarkova et al. (1997) quantified the distribution of nitrifying biomass in a shallow stream. Depending on the local hydraulic situation, they found very significant amounts of active biomass deep in the loose sediment of the river. In addition a very significant fraction of the biomass was eroded and self-purification capacity was lost during a storm event. Understanding self-purification thus requires us to know the ‘‘hydraulic history’’ of the river, to consider the exchange of river water between bulk and sediment and to include erosion and regrowth processes of biomass. All together a formidable task at least. I learnt from these examples that only a holistic understanding of technical (wastewater treatment) and ecological systems (self-purification, toxicity) can be the basis in generating successful proposals for environmental protection.
10.
Immission standards for nitrogen species
It is not the task of environmental engineers alone to suggest immission standards for receiving waters but rather do we expect ecologists and ecotoxicologists to support us in this
- Allowing 20% of the oxygen saturation concentration to be consumed for the nitrification of river water after infiltration into groundwater; - Limiting the ammonium concentration to 0.5 gNH4 m3 to protect possible surface water use for water supply (equal to drinking water tolerance values in EU and Switzerland); - Accepting the limiting value of 0.02 gNH3–N m3 as suggested by the European Inland Fisheries Advisory Commission (EIFAC, 1970) for the protection of freshwater fish and applied in the EU; - Choosing a temperature and pH value typically observed on sunny afternoons, when pH is the highest due to photosynthesis; - Considering toxicity of nitrite which might arise from nitrification (Fig. 15); - Considering the effect of chloride ion (Cl) on nitrite toxicity for fish as suggested by Mu¨ller (1990) and derived from EIFAC (1984): The higher the chloride concentration, the lower the toxicity of nitrite. Based on the arguments in Table 3 the Swiss ordinance on water pollution control (GSchV, 1998) prescribes that the ammonium concentration (NH4–N plus NH3–N) should not exceed 0.4 gN m3 in the receiving water below 10 C and should be below 0.2 gN m3 above 10 C. These values are a compromise between ecological requirements and cost of wastewater treatment. In many situations they are difficult to reach.
11.
Nitrification as a case
11.1. (NOB)
From Nitrobacter to nitrite-oxidizing bacteria
Wagner et al. (1996) demonstrated with the aid of molecular techniques (FISH) that Nitrobacter spp. cannot be the main
Fig. 16 – Ammonium and nitrite concentration over 3 full days in January, May and June in the river Glatt. Data provided by M. Berg (s.a. Berg, 1991).
14
water research 44 (2010) 1–19
Table 3 – Basis for immission standards for ammonium in receiving waters in Switzerland (s.a. Gujer, 1978; Mu¨ller, 1990; EIFAC, 1984). Reasoning
Allowable ammonium concentration in gN m3 River water temperature
Tolerance
10 C
20 C
Protection of groundwater
20% of O2 saturation reserved for nitrification (4.57 gO2/gNH4–N)
0.49
0.40
Protection of water supply
Tolerance concentration for drinking water in Switzerland and EU 0.5 gNH4 m3
0.39
0.39
Limiting ammonia concentration for protection of local fish 0.02 gNH3–N m3 (average)
0.02 gNH3–N m3
pH ¼ 8.0 pH ¼ 8.5 pH ¼ 8.75
1.0 0.33 0.20
0.49 0.17 0.10
Limiting the nitrite concentration depending on local chloride concentration for the protection of fish, considering nitrification
20 gCl1 m1
0.02 gNO2–N m3 0.05 gNO2–N m3 0.10 gNO2–N m3
0.1 0.2 0.5
0.05 0.1 0.2
organisms responsible for the oxidation of nitrite to nitrate in typical biological wastewater treatment plants. Until then it was assumed by environmental engineers that Nitrosomonas and Nitrobacter are the responsible organisms for nitrification. I suspect many engineers assumed these organisms to be well defined entities growing on well defined and easily accessible substrates. These assumptions made nitrification the ideal process for following the behavior of a specific organism within the mixed population that makes up activated sludge. In these studies frequently it is not nitrification that is of interest but nitrification is only a proxy for the analysis of species behavior in mixed cultures. Today careful engineers use the term ammonium oxidizing organisms (AOB) and nitrite-oxidizing organisms (NOB) rather then giving specific names to the catalysts of these processes. Nevertheless experience with these processes remained valid, independent of the more advanced microbiological findings. The unique situation that simple experiments (respiration in batch tests, nitrite and nitrate production rates, etc.) allowed to establish kinetic and stoichiometric information, develop mathematical models for activated sludge population dynamics and dynamic system behavior, biofilm models, etc. was extremely helpful in the development of models for activated sludge processes and attached growth systems. Today many models of mixed culture mixed substrate interactions actually follow the lines first calibrated and validated with the processes of nitrification. The lack of quantitative microbial techniques to follow different groups of organisms and their activity in the activated sludge made it necessary for engineers to use nitrification as a readily and easily quantifiable process for the development of many models. I am convinced that nitrification was a blessing for engineers involved in the development of mathematical models. I foresee that in the future when molecular microbial techniques become more and more quantitative and readily available, these techniques will partially replace the use of nitrification as a source for further understanding of the
interplay of substrate and microbial populations. But there are still questions to be answered with the aid of domesticated nitrifying organisms. In addition nitrification is such an easy to understand and well behaved system that it will remain important in the education of generations of engineers.
11.2. Nitrification as an indicator for micropollutant degradation Nitrification performance of a biological wastewater treatment plant can easily be followed and is today a frequent requirement in many industrialized countries. In contrast, degradation of micropollutants is difficult and costly to follow and is not routinely included in plant performance control. Clara et al. (2005) demonstrate that at solids retention times typically used in nitrogen removal plants many micropollutants are efficiently degraded. Thus nitrification efficiency is a valid indicator for micropollutant removal. For some micropollutants like EE2 it may even be the nitrifiers themselves which are responsible for their degradation (Forrez et al., 2009). As mentioned early in this review, nitrification has for a long time served as an indicator for good secondary treatment. With its importance related to micropollutants this indicator function may get value again.
11.3.
The case of bioaugmentation
It is tempting (and is even patented, US Patent 5811009) to pregrow nitrifying organisms with warm ammonium-rich liquids originating e.g. from sludge handling (supernatants from digesters operated at 35 C) and then to add these organisms to an activated sludge reactor in order to augment the nitrifier concentration and thus to obtain better performance from a given, highly loaded reactor. With the aid of FISH (Fluorescent In Situ Hybridization) Manser (2005) demonstrated that different organisms are enriched in ammonium-rich liquids (R-strategists: Nitrosomonas europaea and Nitrobacter) and in domestic wastewater
water research 44 (2010) 1–19
with lower ammonium concentrations (K-strategists: Nitrosomonas oligotropha and Nitrospira). With experiments and based on simulations Manser concluded that the R-strategists are rapidly washed out from the activated sludge and the use of the extra nitrogen for load balancing (Fig. 5) would actually result in better plant performance, however at the cost of extra oxygen input into the reactor. Simply working with oxygen electrodes and ammonium, nitrite and nitrate analysis would not necessarily lead to this conclusion (see patent application). In this case molecular microbiological techniques lead directly to an explanation of a not a priori expected result. Using the activated sludge with nitrifiers grown in the activated sludge process itself as an influent into a reactor where a warm, concentrated ammonium solution is nitrified, appears to be a more successful strategy. Salem et al. (2003) simulated this strategy based on ASM1 and predicted a very positive effect which was later validated in full scale (Salem et al., 2004; Krhutkova et al., 2006). The initial simulations did however not differentiate between alternative possible groups of nitrifiers. It is only with the full scale validation that these predictions became valuable.
11.4. The case of conventional activated sludge versus membrane bioreactors It is not a priori clear that the experience with conventional activated sludge systems (CAS), where biomass is retained based on sedimentation, can directly be transferred to membrane bioreactors (MBR). Here biomass is quantitatively retained and does not have to settle. It is well possible that under these differing operating conditions different organisms with different properties are enriched. Manser et al. (2005a) used FISH and found only minor differences between the two systems for both ammonia-oxidizing and nitriteoxidizing bacteria. Kinetic parameters differed between the two systems. Apparent Monod saturation coefficients for nitrifiers are larger in CAS than in MBR systems. Manser et al. (2005b) explain these differences with mass transfer effects. In CAS the flocs are larger than in MBR systems. The longer diffusion paths result in a larger apparent saturation value. Thus some kinetic parameters are system specific.
11.5.
15
since frequently their models are only crude approximations of the fine details of reality. Today we do not have a scientific strategy to deal with model structure uncertainty. Pragmatic approaches are to add extra noise to the data until structural problems are masked or to thin out data until structural problems cannot be identified any more.
11.6.
Kinetic parameters are stochastic variables
Mathematical models for biological wastewater treatment such as the family of Activated Sludge Models No. 1–3 (ASM1, ASM2, ASM3) typically are assumed to be deterministic and based on fixed parameter values (which may however have to be calibrated for a specific situation and system). Using nitrification we could argue that different groups of nitrifiers exist under the operating conditions of an activated sludge system, which is genetically open to the environment. Thus, it is well possible that over time different groups of organisms may be enriched in the activated sludge. This would then result in apparent time dependent kinetic parameters if nitrification is modeled with only one ‘‘species’’ of organisms. In addition varying activated sludge floc size could lead to variable diffusion limitations inside the flocs which from a macroscopic point of view would be identified as a variable value of Monod saturation coefficients. Daebel et al. (2007) identified the saturation coefficient for ammonium oxidizing (AOB) and nitrite-oxidizing (NOB) bacteria in activated sludge from a conventional activated sludge plant with sedimentation and a membrane bioreactor (MBR). The kinetic parameters varied over time (Fig. 17) and since flocs were smaller in the MBR than in the CAS the parameters were also different for the two systems. Since the variation of the kinetic parameters is statistically highly significant, we must assume that such parameters are not constant in time but may be subject to time dependent stochastic processes (effects of processes not captured with today’s models). At this moment it is unclear what the consequence of such results is, but we might have to accept that our nice deterministic models capture only a small fraction of the complex behavior of activated sludge.
Model structure uncertainty
In calibrating our mathematical models we frequently assume that the mathematical structure of our models provides a good image of reality. Daebel et al. (2007) analyzed the residuals (deviations) between experimental observation and model prediction for some respiratory tests with nitrifiers. We found systematic deviations (autocorrelated residuals) which go back to structural deficiencies of our models. We concluded that by using standard least square procedures for parameter identification, parameter uncertainty is underestimated. Neumann and Gujer (2008) follow up on this problem based on artificial data and conclude that we do not yet have the techniques to deal with parameter uncertainty in view of structural problems in our mathematical models. Especially environmental engineers are suffering from this situation,
Fig. 17 – Temporal variation of the oxygen saturation coefficient (Monod model) of ammonium (AOB) and nitrite (NOB) oxidizing bacteria in a continuously operated membrane bioreactor plant (MBR). Expected value and 95% confidence region of a lognormally distributed parameter value. Adapted from Daebel et al. (2007).
16
water research 44 (2010) 1–19
11.7. Batch tests may not yield reliable kinetic information Mass transfer and utilization of oxygen strongly interfere. What we observe on a macroscopic level (using an oxygen electrode) is quite different from what microorganisms inside flocs experience. Results obtained from a batch test in which both AOB and NOB are active (test spiked with ammonium) and a test in which NOB activity dominates (spiking nitrite only) may result in different apparent Monod saturation coefficients for oxygen. Since AOB consume much more oxygen than NOB, mass-transfer may result in different oxygen concentration profiles inside the flocs (Manser et al., 2005b). The situation gets even more complicated when heterotrophic activity must be considered. The same applies to saturation coefficients of nitrite: If nitrite is supplied from the outside of the flocs by spiking, we must expect a larger apparent saturation coefficient than when nitrite is produced from ammonia by AOBs, when nitrite might actually diffuse out of the flocs.
11.8.
Chemical nitrification
In special cases microbial and chemical nitrification processes are strongly interlinked. Trying to nitrify a highly concentrated ammonium-nitrite solution Udert et al. (2005) found the oxidation of ammonium to nitrite to be catalyzed by microorganisms which reduced the pH far below 5.5. In the resulting solution conditions developed which induced chemical oxidation of nitrite to nitrate and a final pH below 3. It is presently not known where such processes are of importance; nitrogen emission from acid soils is a candidate. The interesting aspect of these experiments is the simultaneous activity and interaction of significant microbial and chemical processes.
11.9. Nitrification provides evidence for the anammox process Nitrification has the favorable property that substrates (educts) and products of the process can quite easily be followed and analyzed with the aid of mass balances. This is not the case with the degradation of organic compounds where carbon dioxide (CO2) may easily be lost to the atmosphere and may interact with the carbonate buffer system. Siegrist et al. (1998) operated a tertiary rotating biological contactor (RBC) for the nitrification of the pretreated effluent from a hazardous-waste landfill. The influent contained a minimum of organic substrate nevertheless up to 70% of the ammonium which was nitrified was lost in the process. This unexpected result led these authors to the discovery that anammox bacteria have developed in their system. Clearly this observation would have been difficult without the possibility to follow substrate and product of the expected processes.
11.10. Ammonium as a reactive tracer Engineers frequently use inert tracer compounds in order to obtain experimental information on reactor hydraulics
(hydraulic residence time distribution). It is rather difficult to derive information on reactor internal mixing from such experiments since time constants for internal mixing are considerably shorter than mean hydraulic residence times, thus internal mixing is masked by the averaging process of mixing of the tracer. If reactive tracers are used their time constants (mean life expectancy) may be much shorter, thus following their concentration inside the reactor may yield more information on mixing processes. Using ammonium and dissolved oxygen as reactive tracers Braun and Gujer (2008) used on-line electrodes inside the reactor and found oscillations of these two compounds with different frequencies. A low frequency (1 h1) related to problems in the aeration control, a high frequency (9 h1, Fig. 18) with a period of approximately 7 min originated from problems with macroscopic internal mixing. These oscillations affect the performance of the biological reactor. They are not typically contained in our models but might actually be quite common in biological wastewater treatment, where mixing energy is costly and thus kept to a minimum. The time constants of nitrification are such that ammonium and related also oxygen can be used to identify such problems with reasonable effort. With non-reactive tracers we could not identify these problems with the same resolution and reproducibility. When we use data on treatment performance from pilot or full scale plants, such mixing effects may affect the results but may not be realized because typically we do not measure state variables routinely within our reactors. In the process of calibration of our models for dynamic simulation we primarily adjust kinetic and stoichiometric parameters but seldom improve the hydraulic reactor models. Thus defects of reactors are copied onto kinetic parameters. The value of these parameters for the simulation of another system is then questionable. The question arises: to what degree can models calibrated with results from pilot plants be extrapolated to full scale plants under design?
12.
Open questions and outlook
Nitrite is a known toxic compound for fish and thus is an important aspect of water pollution control. The reliable
Fig. 18 – Oscillations of ammonium and dissolved oxygen concentrations in a non-optimized activated sludge reactor caused by internal, macroscopic mixing processes. Adapted from Braun and Gujer (2008).
water research 44 (2010) 1–19
prediction of nitrite in the effluent of biological treatment systems is however still an open problem: Unexpected spills of high concentrations of nitrite are frequently observed. The problem is not to develop a model structure but rather to understand the variation of the kinetic parameters. It might well be that only stochastic models for parameters or alternatively structured biomass (microorganisms with cell internal structure) will be successful in capturing some of these phenomena. Reaching very low (.2 gN m3) residual ammonium concentrations in the effluent of activated sludge plants is sometimes difficult, especially when diurnal load variations are high. We do not yet have a full understanding of these problems. Here too it may be necessary to include cell internal structure (organism activity) to explain our experience. In addition a further understanding of predation, decay and lysis processes under different redox conditions might be required. Sometimes the cause is related to poorly characterized hydraulic mixing conditions in the aeration tank. The economics of wastewater treatment could be improved if nitrification were stopped with the production of nitrite only. Nitrification would be cheaper (oxygen supply) and denitrification would be more efficient. At the low temperatures of urban wastewater we do however not yet have the technology to stop nitrification at the level of nitrite. In addition nitrous oxide production is related to high nitrite concentration. Sensor technology has made significant progress in the last decade. Broad application of such technology combined with advanced control strategies has the potential to provide us with vastly different treatment technologies. The full potential of this development has not yet been developed. Increasingly anaerobic ammonium oxidation (anammox) is recognized to be an important process of the nitrogen cycle in dilute natural systems. Kuypers et al. (2003) state: In fact, the widespread occurrence of ammonium consumption in suboxic marine waters as well as in sediments suggests that anammox bacteria could have an important but as yet neglected role in the oceanic loss of fixed nitrogen. Whereas anammox is increasingly used for the removal of nitrogen from highly concentrated ammonium solutions (Kuenen, 2008), we have yet to see a substantial application of this process for the treatment of dilute wastewater. Anaerobic treatment with production of methane followed by nitritation combined with anammox would be an interesting combination for urban wastewater in many situations, especially in warm climates.
13. Do we stand on the brink of a new paradigm again? Today we have a solid understanding of what I would call the backbone of biological wastewater treatment which is responsible for the removal of the macro-nutrients from urban wastewater: TSS, COD, nitrogen and phosphorus. There will still be further valuable developments but compared to the second half of the 20th century I do not expect advances in in-depth understanding at the same rate. What lies ahead of us is more the development of detailed understanding of the behavior of specific chemicals (micropollutants, .), specific groups of organisms (filaments, anammox, .), novel
17
treatment technologies (membranes, anaerobic processes, granular biomass, .) and of detailed engineering methods (computational fluid dynamics, CFD, .). At the same time there is a trend away from an interest for highly integrated, large, centralized sewer and prototype treatment systems to the development of more decentralized and smaller, potentially even industrially produced units. In addition we learnt to admit that there will remain some uncertainty in engineering design and we rapidly find techniques to quantify this uncertainty and to integrate it into our decision processes. The transition from general understanding of bulk performance of publicly owned wastewater treatment plants towards the specific behavior of individual entities in small, possibly industrially produced and privately owned treatment units requires a new approach with new tools. On the one hand new stakeholders will be involved and their interest and potential must become understood and considered. On the other hand new specific techniques (microbial, chemical and engineering) will become available which will allow for very specific and detailed results however at considerable cost. Combining these two aspects requires the consideration of societal relevance and thus more transdisciplinary work. The glory of time of gaining generally valid information from analyzing nitrification as a proxy for many important processes slowly vanishes, the new paradigm however still waits to be explicitly defined.
14.
Conclusion
There is no doubt, ammonium is today recognized as an important quality parameter in receiving waters and nitrification is the dominant process to rid wastewater of ammonium. Thus nitrification is here to stay and a detailed understanding of this process is key to modern wastewater treatment. By now the organisms responsible for nitrification of urban wastewater are domesticated and a broad suite of technologies is available for their productive application. Mathematical models, which are an important tool for design and optimization of biological treatment units have been developed and are used on a broad scale. The future will result in some refinement but not necessarily in an entirely new structure. Nitrification can easily be quantified thus it has served as a proxy to learn about many problems of biological wastewater treatment processes. Today we increasingly get more specific chemical and microbiological techniques which reduce the importance of working with nitrification. Definitely open questions remain in the context of nitrification of dilute and concentrated nitrogen solutions. I am convinced that research and development will continue in this exiting field of environmental engineering.
references
Balakrishnan, S., Eckenfelder, W.W., 1969. Nitrogen relationships in biological treatment processes. 1. Nitrification in the activated sludge process. Water Res. 3 (1), 73–81.
18
water research 44 (2010) 1–19
Berg, M., 1991. Stickstoffdynamik in der Glatt. GWA 71 (12), 822–830. Boller, M., Gujer, W., 1986. Nitrification in tertiary trickling filters followed by deep-bed filters. Water Res. 20 (11), 1363–1373. Boller, M., Gujer, W., Tschui, M., 1994. Parameters affecting nitrifying biofilm reactors. Water Sci. Technol. 29 (10–11), 1–11. Boller, M., Gujer, W., Nyhuis, G., 1990. Tertiary rotating biological contactors for nitrification. Water Sci. Technol. 22 (1–2), 89–100. Braun, D., Gujer, W., 2008. Reactive tracers reveal hydraulic and control instabilities in full-scale activated sludge plant. Water Sci. Technol. 57 (7), 1001–1007. Brun, R., Ku¨hni, M., Siegrist, H., Gujer, W., Reichert, P., 2002. Practical identifiability of ASM2d parameters – systematic selection and tuning of parameter subsets. Water Res. 36, 4113–4127. Clara, M., Kreuzinger, N., Strenn, B., Gans, O., Kroiss, H., 2005. The solids retention time – a suitable design parameter to evaluate the capacity of wastewater treatment plants to remove micropollutants. Water Res. 39, 97–106. Copp, J.B., 2002. The COST Simulation Benchmark: Description and Simulator Manual. Office for Official Publications of the European Community, Luxembourg. Daebel, H., Manser, R., Gujer, W., 2007. Exploring temporal variations of oxygen saturation constants of nitrifying bacteria. Water Res. 41, 1094–1102. Dold, P.L., Ekama, G.A., Marais, GvR, 1980. A general model for the activated sludge process. Prog. Water Technol. 12 (6), 47–77. Dominguez, D., Gujer, W., 2006. Evolution of a wastewater treatment plant challenges traditional design concepts. Water Res. 40 (7), 1389–1396. Downing, A.L., Painter, H.A., Knowles, G., 1964. Nitrification in the activated sludge process. J. Inst. Sew. Purif., 130–158. EIFAC, 1970. Water Quality Criteria for European Freshwater Fish. Report on Ammonia and Inland Fisheries. EIFAC Technical Papers T11. European Inland Fisheries Advisory Commission, FAO, Rome. EIFAC, 1984. Water Quality Criteria for European Freshwater Fish. Report on Nitrite and Freshwater Fish. European Inland Fisheries Advisory Commission, FAO, Rome, ISBN 9251021775. Fair, G.M., Geyer, J.C., 1954. Water Supply and Waste-Water Disposal. John Wiley & Sons, Inc., New York. Forrez, I., Carballa, M., Boon, N., Verstraete, W., 2009. Biological removal of 17a-ethinylestradiol (EE2) in an aerated nitrifying fixed bed reactor during ammonium starvation. J. Chem. Technol. Biotechnol. 84, 119–125. Fruhen, M., Christan, E., Gujer, W., Wanner, O., 1991. Significance of spatial-distribution of microbial species in mixed culture biofilms. Water Sci. Technol. 23 (7–9), 1365–1374. Garrett Jr., M.T., 1958. Hydraulic control of activated sludge growth rate. Sew. Ind. Wastes 30 (3), 253–261. GSchV, 1998. Gewa¨sserschutzverordnung vom 28. October 1998. Swiss Law, SR 814.201. Gujer, W., 1974. Oxygen and Nitrogen Mass Balances for the Contact Stabilization Activated Sludge Process. Dissertation, University of California, Berkeley, USA, p. 155. Gujer, W., 1976a. Auswirkungen von abgesetztem Faulwasser auf die Nitrifikation. Eawag Report to Tiefbauamt der Stadt Zu¨rich, Contract Nr. 4489. Gujer, W., 1976b. Nitrifikation in Fliessgewa¨ssern – Fallstudie Glatt. Schweiz. Z. Hydrol. 38 (2), 171–189. Gujer, W., 1977. Design of a nitrifying activated sludge process with the aid of dynamic simulation. Prog. Water Technol. 9, 323–336. Gujer, W., 1978. Nitrit in Fliessgewa¨ssern – Ein erweitertes Nitrifikationsmodell’’ Schweiz. Z. Hydrol. 40 (2), 211–230. Gujer, W., 1985. Ein dynamisches Modell fu¨r die Simulation von komplexen Belebtschlammverfahren. Habilitationsschrift, ETH Zurich. Gujer, W., 2006. Activated sludge modeling: past, present and future. Water Sci. Technol. 53 (3), 111–119.
Gujer, W., Boller, M., 1978. Basis for the design of alternative chemical–biological waste water treatment processes. Prog. Water Technol. 10, 741–758. Gujer, W., Boller, M., 1984. Operating experience with plastic media tertiary trickling filters for nitrification. Water Sci. Technol. 16 (10–11), 201–213. Gujer, W., Boller, M., 1986. Design of a nitrifying tertiary trickling filter based on theoretical concepts. Water Res. 20 (11), 1353–1362. Gujer, W., Boller, M., 1990. A mathematical-model for rotating biological contactors. Water Sci. Technol. 22 (1–2), 53–73. Gujer, W., Erni, P., 1978. The effect of diurnal ammonium load variation on the performance of nitrifying activated sludge processes. Prog. Water Technol. 10 (5/6), 391–407. Gujer, W., Jenkins, D., 1975. Nitrification model for the contact stabilization activated sludge process. Water Res. 9 (5–6), 561–566. Gujer, W., Henze, M., Mino, T., van Loosdrecht, M.C.M., 1999. Activated Sludge Model No. 3. Water Sci. Technol. 39 (1), 183–193. Hanæus, J., Hellstro¨m, D., Johansson, E., 1997. A study of a urine separation system in an ecological village in Northern Sweden. Water Sci. Technol. 35 (9), 153–160. Harremoe¨s, P., 1976. The significance of pore diffusion to filter denitrification. J. WPCF 46 (2), 377–388. Henze, M., et al., 2000. Activated Sludge Models ASM1, ASM2, ASM2D and ASM3. Scientific and Technical Report No. 9. IWA Publishing, London, England. Holiencˇin, A., 1996. Nitritumsatz in nitrifiziererenden Belebungsanlagen. Dissertation ETH Nr. 11875. Holiencˇin, A., Gujer, W., 1996. Produktion von Nitrit in der Nitrifikation. GWA 76 (11), 841–846. Jancarkova, I., Larsen, T.A., Gujer, W., 1997. Distribution of nitrifying bacteria in a shallow stream. Water Sci. Technol. 36 (8–9), 161–166. Kirchmann, H., Pettersson, S., 1995. Human urine – chemical composition and fertilizer use efficiency. Fert. Res. 40, 149–154. Krhutkova, O., Novak, L., Pachmanova, L., Benakova, A., Wanner, J., Kos, M., 2006. In situ bioaugmentation of nitrification in the regeneration zone: practical application and experiences at full-scale plants. Water Sci. Technol. 53 (12), 39–46. Kuenen, J.G., 2008. Anammox bacteria: from discovery to application. Nat. Rev. – Microbiol. 6, 320–326. Kuypers, M.M.M., Sliekers, A.O., Lavik, G., Schmid, M., Jørgensen, B.B., Kuenen, J.G., Damste´, J.S.S., Strous, M., Jetten, M.S.M., 2003. Anaerobic ammonium oxidation by anammox bacteria in the Black Sea. Nature 422 (6932), 608–611. Larsen, T.A., Gujer, W., 1996. Separate management of anthropogenic nutrient solutions (human urine). Water Sci. Technol. 34 (3–4), 87–94. Larsen, T.A., Gujer, W., 2001. Waste design and source control lead to flexibility in wastewater management. Water Sci. Technol. 43 (5), 309–317. Lawrence, A.W., McCarty, P.L., 1970. Unified basis for biological treatment design and operation. J. Sanitary Eng. Div., ASCE 96 (3SA), 757–778. Lijklema, L., 1973. Model for nitrification in activated sludge process. Environ. Sci. Technol. 7 (5), 428–433. Manser, R., 2005. Population Dynamics and Kinetics of Nitrifying Bacteria in Membrane and Conventional Activated Sludge Plants. Dissertation, ETH No. 16217, ETH Zurich, http://ecollection.ethbib.ethz.ch/view/eth:28260?q=(author:manser) Manser, R., Gujer, W., Siegrist, H., 2006. Decay processes of nitrifying bacteria in biological wastewater treatment systems. Water Res. 40 (12), 2416–2426.
water research 44 (2010) 1–19
Manser, R., Gujer, W., Siegrist, H., 2005a. Membrane bioreactor versus conventional activated sludge system: population dynamics of nitrifiers. Water Sci. Technol. 52 (10–11), 417–425. Manser, R., Gujer, W., Siegrist, H., 2005b. Consequences of mass transfer effects on the kinetics of nitrifiers. Water Res. 39 (19), 4633–4642. Marais, GvR., Ekama, G.A., 1976. The activated sludge process. Part 1 – steady state behaviour. Water SA 2 (4), 163. Matsche´, N., Moser, D., 1993. Operation of a two-stage activated sludge package plant for high efficiency treatment. Water Sci. Technol. 28 (10), 299–307. Monod, J., 1950. La technique de culture continue: the´orie et applications. Ann. Inst. Pasteur 79, 390–410. Mueller, J.A., Paquin, P., Famularo, J., 1978. Nitrification in rotating biological contactors. In: 51st Annual Conference WPCF, Anaheim, CA. Mu¨ller, 1990. Stickstofftoxizita¨t fu¨r Fische und herzuleitende Grenzwerte. Eawag News No. 30D, 33–36. http://library.eawagempa.ch/eawagnewsarchive.htm. Neumann, M.B., Gujer, W., 2008. Underestimation of uncertainty in statistical regression of environmental models: influence of model structure uncertainty. Environ. Sci. Technol. 42 (11), 4037–4043. Novick, A., Szilard, L., 1950. Description of the chemostat. Science 112 (2920), 715–716. Poduska, R.A., Andrews, J.F., 1975. Dynamics of nitrification in the activated sludge process. J. WPCF 47 (11), 2599–2619. Poth, M., Focht, D.D., 1985. 14N kinetic analysis of N2O production by Nitrosomonas europaea: an examination of nitrifier denitrification. Appl. Environ. Microbiol. 49, 1134–1141. Rauch, W., Brockmann, D., Peters, I., Larsen, T.A., Gujer, W., 2003. Combining urine separation with waste design: an analysis using a stochastic model for urine production. Water Res. 37 (3), 681–689. Salem, S., Berends, D.H.J.G., Heijnen, J.J., Van Loosdrecht, M.C.M., 2003. Bio-augmentation by nitrification with return sludge. Water Res. 35, 1794–1804. Salem, S., Berends, D.H.J.G., van der Roest, H.F., van der Kuij, R.J., van Loosdrecht, M.C.M., 2004. Full-scale application of the BABE((R)) technology. Water Sci. Technol. 50 (7), 87–96. Siegrist, H., Gujer, W., 1987. Demonstration of mass transfer and pH effects in a nitrifying biofilm. Water Res. 21 (12), 1481–1487.
19
Siegrist, H., Reithaar, S., Koch, G., Lais, P., 1998. Nitrogen loss in a nitrifying rotating contactor treating ammonium-rich wastewater without organic carbon. Water Sci. Technol. 38 (8–9), 241–248. Tomlinson, T.G., Boon, A.G., Trotman, G.N.A., 1966. Inhibition of nitrification in the activated sludge process of sewage disposal. J. Appl. Bacteriol. 29 (2), 266–291. Udert, K.M., Larsen, T.A., Gujer, W., 2005. Chemical nitrite oxidation in acid solutions as a consequence of microbial ammonium oxidation. Environ. Sci. Technol. 39 (11), 4066–4075. USEPA, 1975. Process Design Manual for Nitrogen Control. U.S. Environmental Protection Agency, Technology Transfer. October 1975. USEPA, 1993. Manual: Nitrogen Control. U.S. Environmental Protection Agency Office of Research and Development, Office of Water, EPA/625/R-93/010, p. 311, September 1993. von Schulthess, R., Gujer, W., 1996. Release of nitrous oxide (N2O) from denitrifying activated sludge: verification and application of a mathematical model. Water Res. 30 (3), 521–530. von Schulthess, R., Wild, D., Gujer, W., 1994. Nitric and nitrous oxides from denitrifying activated sludge at low oxygen concentration. Water Sci. Technol. 30 (6), 123–132. Wagner, M., Rath, G., Koops, H.P., Flood, J., Amann, R., 1996. In situ analysis of nitrifying bacteria in sewage treatment plants. Water Sci. Technol. 34 (1–2), 237–244. Wandl, G., Kroiss, H., Svardal, K., 2006. The main wastewater treatment plant of Vienna: an example of cost effective wastewater treatment for large cities. Water Sci. Technol. 54 (10), 79–86. Wanner, O., Gujer, W., 1984. Competition in biofilms. Water Sci. Technol. 17, 27–44. Wanner, O., Gujer, W., 1986. A multispecies biofilm model. Biotechnol. Bioeng. 28, 314–328. Water Pollution Research, 1964. Yearly Report. Ministry of Technology, Her Majesty’s Stationary Office, London. Wiesmann, J., 1982. Ideenwettbewerb. Schweiz. Ing. und Arch. 100 (13), 222–226. Williamson, K., McCarty, P.L., 1976. A model of substrate utilization by bacterial films. J. WPCF 48 (1), 9–24. Wuhrmann, K., 1964. Grundlagen fu¨r die Dimensionierung der Belu¨ftung bei Belebungsanlagen. Schweiz. Z. Hydrol. 26 (2), 310–337.
water research 44 (2010) 21–40
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Review
Impacts of salinity on the performance of high retention membrane bioreactors for water reclamation: A review Winson C.L. Lay a,b, Yu Liu a,b, Anthony G. Fane a,b,* a
Singapore Membrane Technology Centre, Nanyang Technological University, Singapore 637723, Singapore Division of Environmental and Water Resources Engineering, School of Civil and Environmental Engineering, Nanyang Technological University, Singapore 639798, Singapore b
article info
abstract
Article history:
Recent efforts in the field of used water treatment and water reclamation have led to the
Received 13 April 2009
development of a number of innovative high retention membrane bioreactor (HRMBR)
Received in revised form
systems. These systems invariably combine a high rejection membrane separation with
2 July 2009
a biological treatment. A common positive outcome of these systems is that smaller size
Accepted 10 September 2009
organic contaminants are effectively retained, which facilitates their biodegradation and
Available online 17 September 2009
thus produces high quality product water. This provides the desired high level of separation, but also leads to salt accumulation with potentially adverse effects on the operations.
Keywords:
The effects of elevated salt condition are complex, and impact on aspects covering phys-
Membrane bioreactor
icochemical parameters, microbiology and membrane performance. The salt concentra-
High retention
tion factor is an important operating parameter to be optimised in the HRMBR systems.
Salt
This paper aims to elucidate the important issues associated with the use of HRMBR
Concentration factor
systems under elevated salt conditions up to 50 g L1.
Water reclamation
ª 2009 Elsevier Ltd. All rights reserved.
Contents 1. 2. 3.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Problem definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Issues associated with elevated salt condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 3.1. Effect on physicochemical aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 3.1.1. Oxygen transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 3.1.2. Density, turbidity and viscosity of suspension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 3.1.3. Salt precipitation, solute interactions and colloid chemistry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.2. Microbiological aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 3.2.1. Microbiology in elevated salt environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 3.2.2. Biological carbon removal in elevated salt environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 3.2.3. Biological nutrient removal in elevated salt environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 3.2.4. Biomass characteristics and biological operating conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
* Corresponding author. Tel.: þ65 67905272; fax: þ65 67910676. E-mail address:
[email protected] (A.G. Fane). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.026
22
water research 44 (2010) 21–40
Membrane aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 3.3.1. Driving force, concentration polarisation, flux and product quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 3.3.2. Membrane fouling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 3.4. Concentration factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 3.3.
4.
1.
Introduction
Membrane bioreactors (MBRs) are finding increasing application for industrial and municipal used water treatment. The conventional MBR uses microporous microfiltration (MF) or ultrafiltration (UF) membranes to retain the mixed liquor of the bioreactor, and delivers particle-free treated effluent. However, the molecular weight cut-off (MWCO) of the MF/UF membranes means that a portion of the organic species is not retained. For these species, the organic retention time (ORT) is the same as the hydraulic retention time (HRT). The effect of this is that recalcitrant organics may not be well degraded, and the direct reuse potential of the permeate may be limited. In an earlier attempt to overcome this, (Rautenbach and Mellis, 1994) combined MF/UF followed by nanofiltration (NF) with an activated sludge biosystem, where the NF reject stream is recycled back to the bioreactor. More recent efforts in the field of used water treatment and water reclamation have led to the development of a number of innovative high retention membrane bioreactors (HRMBRs). One example is the nanofiltration MBR (NFMBR) (Choi et al., 2002, 2007), where the NF membrane is used in place of MF/UF. Other recent developments include the membrane distillation bioreactor (MDBR) (Fane et al., 2005; Phattaranawik et al., 2008), and the osmotic MBR (OMBR) (Cornelissen et al., 2008; Oo et al., 2008). The underlying motivation for developing HRMBR systems is to explore the feasibility of combining the conventional MBR and high rejection membrane separation in a single step. The HRMBR systems will be able to retain effectively smaller size and persistent contaminants, which facilitates their biodegradation in the bioreactor, thereby producing higher quality product water. This attribute is important for large-scale water reclamation intended for augmenting water supply where high water quality is required (PUB, 2002). Under favourable circumstances, such as the availability of a waste heat source for the MDBR, some HRMBR systems may also achieve comparative economical advantage by being able to remove pollutants from the used water with lower primary energy demand (Phattaranawik et al., 2008). In the conventional MBR processes, the mixed liquor suspended solids are retained in the bioreactor, but a significant fraction of the dissolved solids will pass through the membrane into the effluent. Therefore, no salt accumulation takes place in conventional MBR processes. The HRMBR systems, however, will retain most of the colloidal and dissolved solids. This provides the desired high level of separation, but also leads to salt accumulation with potentially adverse impacts on the biological treatment and the operation
of the MBR. The elevated salt condition is therefore an inherent issue that HRMBR systems will need to grapple with. High salinity (halophilic condition) is known to affect the physical and biochemical properties of the microorganisms necessary for the biological treatment, and it will also affect the membrane performance (Reid et al., 2006). For the MDBR, there is the additional impact of operating at higher temperature (thermophilic condition), beyond the mesophilic range of conventional MBR systems. This paper aims to elucidate the important issues associated with the use of HRMBR systems for treating used water under elevated salt conditions, with the focus on the treatment and reclamation of used water of domestic origin. It is envisaged, however, that the HRMBR systems would also be applicable in many other water reuse scenarios including used water of industrial origin, when high quality product water is required.
2.
Problem definition
The biological treatment of saline used water have been reported. Examples include the reviews by Lefebvre and Moletta (2006) and McAdam and Judd (2008). However, the subject becomes more complex when the condition of elevated salt occurs within an HRMBR system. The elevated salt concentrations can impact on physicochemical, microbiological and membrane performance of the system, which in turn are mutually engaged in a dynamic interaction. For definition within the scope of this paper, an HRMBR system refers to a non-conventional combination of an activated sludge bioreactor and a high rejection membrane separation for treatment of used water. An HRMBR system produces high quality product water favourable for water reclamation purpose, but it also invariably leads to the retention and accumulation of inorganic species in the bioreactor. As noted above, examples of HRMBR systems include NFMBR, OMBR and MDBR. For the MDBR, higher temperature effects will also need to be considered. The schematic of an HRMBR system is illustrated in Fig. 1. The only way for salt removal in an HRMBR system is with the waste sludge. A salt mass balance shows that the concentration factor (CF) is related to the recovery of the system (4), and can be calculated as follows (for complete retention): CF ¼
SRT Q 1 ¼ ¼ HRT Qw ð1 4Þ
(1)
where SRT is the solids retention time (V/Qw); HRT, hydraulic retention time (V/Q); Q, flow rate of the influent used water;
water research 44 (2010) 21–40
List of symbols A a B BOD BNR c c0 CECP CF COD CP D DO E EPS F FISH F/M GC/MS HRMBR HRT IAP J k kd KL KLa Ksp Mw MBR MDBR MF ML MLSS MLVSS MWCO NFMBR OLR OMBR ORT OTR
Water permeability constant Air–liquid interfacial area for mass transfer Salt permeability constant Biochemical oxygen demand Biological nutrient removal Concentration in g L1 Concentration in mol L1 Cake enhanced concentration polarisation Concentration factor Chemical oxygen demand Concentration polarisation modulus Diffusion coefficient Dissolved oxygen Removal efficiency, E ¼ (S0 S )/S0 100 [%] Extracellular polymeric substances Faraday’s constant Fluorescence in situ hybridisation Food to microorganisms ratio Gas chromatography/ mass spectrometry High retention membrane bioreactor Hydraulic retention time Ion activity product Flux Mass transfer coefficient Endogenous decay or death rate Liquid mass transfer coefficient Overall oxygen mass transfer coefficient Solubility product constant Molecular weight Membrane bioreactor Membrane distillation bioreactor Microfiltration Mixed liquor Mixed liquor suspended solids Mixed liquor volatile suspended solids Molecular weight cut-off Nanofiltration membrane bioreactor Organic loading rate Osmotic membrane bioreactor Organic retention time Oxygen transfer rate
QW, waste sludge flow rate; V, volume of the MBR system (see list of symbols). Thus, the concentration of salt in the HRMBR will increase by a factor determined by the ratio of SRT to HRT. Table 1 shows the possible accumulation of some selected chemical constituents typical in domestic used water under various values of recovery (f) and the concentration factor (CF) in an HRMBR system with complete retention of the constituents. The table is computed from Eq. (1) based on reference concentrations adapted from the literature (Tchobanoglous et al., 2004) for untreated medium-strength domestic used water. It is to be pointed out that while Table 1 provides useful reference for discussion, the presented concentrations could vary considerably from one source to the other. Considering that the recovery of an HRMBR system could be between 75%
23
p pressure PAC Powdered activated carbon PCR–DGGE Polymerase chain reaction followed by denaturing gradient gel electrophoresis Q Flow R Resistance RO Reverse Osmosis S Substrate for the microorganisms in the form of COD or BOD SEM Scanning electron microscopy SI Supersaturation index SMP Soluble microbial products SOTR Standard oxygen transfer rate SRT Solids retention time T Temperature TDS Total dissolved solids TOC Total organic carbon UF Ultrafiltration V Volume w Weight fraction X Microorganisms expressed in MLSS or MLVSS Y Growth yield coefficient z Charge of particle a Oxygen transfer correction factor b Salinity-surface tension correction factor d Boundary layer thickness Thickness of the fouling cake dcake Permittivity of the solution 3perm 3 Porosity of the fouling cake f Volume fraction of solids h Viscosity 4 Recovery 1/k Debye-length m Specific growth rate of the microorganisms Chemical potential mchem n Factor for mole increase due to dissociation of the dissolved salts p Osmotic pressure q Temperature correction factor r Density s Tortuosity of the fouling cake
(typical RO recovery used in water reclamation (Thompson and Powell, 2003; Coˆte´ et al., 2005)) and 99% (typical MBR recovery with long SRT and short HRT (Judd, 2006), CF would be between 4 and 100. On the one hand, at the lower CF values (CF < 5), the system volume would need to be excessive in order to allow for adequate SRT for large-scale application. On the other hand, at the higher CF values (e.g. CF ¼ 100), the TDS concentration could reach up to 50 g L1 (¼100 500 mg L1) in an HRMBR system, which is more saline than seawater. Furthermore, at the higher CF values, there are other effects, which could be detrimental to the system. These effects are discussed later in this paper. Higher TDS concentrations would also bring about considerable osmotic pressures that are normally not experienced in conventional used water treatment systems. The osmotic
24
water research 44 (2010) 21–40
High quality permeate (To post-treatment where applicable)
Influent used water, Q
High retention MBR system, V
Salt accumulation
Waste sludge, Qw
Air
Fig. 1 – Schematic of an HRMBR system (the figure shows an immersed MBR configuration as a representation; an HRMBR system may also be in sidestream configuration and with other system components; e.g. the draw solution line for the OMBR, or the heating element for the MDBR.).
pressure (p) of the water under elevated salt condition can be estimated by the van’t Hoff equation (Belfort et al., 1994; Schaefer et al., 2005b; Melin and Rautenbach, 2007) as follows: p¼
X
above 40 bar, and would exert considerable impact on the performance of an MBR system in terms of physicochemistry, microbiology and membrane. This level of salt and osmotic pressure would be a problem in the OMBR, which relies on an osmotic pressure driving force. The optimum salt level and concentration factor will be a trade-off in terms of driving force and applied SRT and HRT. This is less of an issue for the MDBR as increased salt concentration has only a modest effect on the vapour pressure of water. Ideally, it is a design goal to achieve a recovery (f) as high as possible. However, as predicted by Eq. (1) and discussed above, high recovery equates to high CF, which in turn, results in high TDS. The corresponding increase in osmotic pressure and other treatment issues could then pose considerable operational challenges on the system.
ni ci RT=Mw;i
(2)
where the subscript i refers to the various salt components present in the used water. The osmotic pressure of the water increases about 8 bar for every 10 g L1 of NaCl, and at NaCl concentration of 50 g L1, the osmotic pressure would be
3. Issues associated with elevated salt condition Table 2 summarises some recent examples of MBR systems operated under elevated salinity. The salt level is presented in g L1 with reference to the concentration of sodium chloride (NaCl). The listed MBR systems use porous MF/UF membranes in treating saline used water of various origins, mostly under the conventional immersed set-up. They are not HRMBR systems, and do not exhibit the salt accumulation effect. Nevertheless, the higher salt concentrations encountered in these studies still provide useful and relevant information for our objective. In general, it appears that MBR systems are feasible for treating saline used water satisfactorily up to the required salt content.
Table 1 – Possible accumulation of selected chemical constituents under various values of the recovery (4) and the concentration factor (CF) in an HRMBR system Typical domestic used watera
f CF
0.5
0.75
0.9
0.95
0.967
0.99
2
4
10
20
30
100
Total dissolved solids [mg L1] TDSb
500
1000
2000
5000
10,000
15,000
50,000
Anions HCO3c CO3c Cl SO4
[mg L1] [mg L1] [mg L1] [mg L1]
100 10 50 30
200 20 100 60
400 40 200 120
1000 100 500 300
2000 200 1000 600
3000 300 1500 900
10,000 1000 5000 3000
Cations Ca Mg K Na
[mg L1] [mg L1] [mg L1] [mg L1]
16 10 15 70
32 20 30 140
64 40 60 280
160 100 150 700
320 200 300 1400
480 300 450 2100
1600 1000 1500 7000
Other constituents [mg L1] SiO2 [mg L1] CODd
10 430
20 86
40 172
100 430
200 860
300 1290
1000 4300
a Reference concentrations are adapted from Tchobanoglous et al. (2004) for medium strength untreated domestic used water. b The TDS comprises fixed and volatile dissolved solids. c The actual concentration of HCO3 and CO3 ions would depend on the carbonate equilibrium. d Assuming up to 10% of the influent biodegradable COD is converted to non-readily-biodegradable soluble microbial products, and accumulates in the system.
25
Dan et al. (2002) 0.30 0.19 Yeast MBR Bacterial MBR (5) Synthetic UW (similar to tuna fish processing UW)
32 32
85 91
– –
3.7 1.9
15 15
36 13.7
3.4 2.1
Tam et al. (2006) 0.23 0.14 1.93 1.72 Conventional (immersed) (4) Municipal sewage (seawater toilet flushing)
7.9 7.9
90–93 90–93
Nitrification; denitrification
– –
19 38
6.8 6.8
0.021 0.042 Conventional (immersed) (3) Municipal sewage
5 5
88 88
Nitrification; denitrification
8 16
64 64
72 36
0.36 0.72
Reid et al. (2006)
Sharrer et al. (2007) 0.029 0.55 64 7.1 Nitrification; denitrification; phosphorus >99 32 Conventional (immersed) (2) Backwash from aquaculture system
99 15
40.8
0.48 27.8
–
120
4
Artiga et al. (2008) 0.35 1.4 92
Biofilm-suspended (immersed) Biofilm-suspended (sidestream) (1) Fish canning factory UW
84
Nitrification; inhibited Nitrification; denitrification
4.6
73
120
F/M [kg COD kg1 ML(V)SS d1] COD removal [%] Salt conc. [g L1] Type of MBR process Type of used water (UW)
Table 2 – Description of MBR systems operated under elevated salt level
Nutrients removal
J [L m2 h1]
SRT [d]
HRT [h]
OLR [kg COD m3 d1]
Source
water research 44 (2010) 21–40
3.1.
Effect on physicochemical aspects
3.1.1.
Oxygen transfer
Adequate oxygen transfer is of fundamental importance to the aerobic activated sludge process, and aeration is essential for bacterial metabolism and contaminant oxidation, as well as for the mixing and distribution of the contents in the bioreactor. Aeration also has key influence on the economics of the treatment, as it typically provides the largest component of the process operating cost (Judd, 2006). Aeration is carried out by either diffused aeration system or mechanical aeration systems. For application to biological treatment of used water, the aeration equipment is commonly specified to maintain minimum dissolved oxygen (DO) of 2 mg L1 in the bioreactor (Tchobanoglous et al., 2004). The oxygen transfer rate (OTR) in an HRMBR system can be expressed as follows (Lange et al., 1972; Colt, 1984; Tchobanoglous et al., 2004): OTR ¼ SOTR a
b cS;clean cL qðT20Þ cS;20
(3)
Here, SOTR refers to the clean water test parameter, and assumes the standard conditions of tap water, temperature T ¼ 20 C, atmospheric pressure, and at initial DO ¼ 0 mg L1. The correction factors a, b, and q are defined in Table 3, and cS,clean is the dissolved oxygen saturation concentration in clean water under the operating condition, cL is the actual operating dissolved oxygen concentration, and cS,20 is the dissolved oxygen saturation concentration in clean water at 20 C and atmospheric pressure. It has been pointed out that the mixed liquor suspended solids (MLSS) concentration has controlling influence on the factor a and oxygen transfer in MBR systems (Krampe and Krauth, 2003; Germain et al., 2007), but the presence of salt at high concentrations could still exert significant impacts. Salinity influences the factor a indirectly by affecting the viscosity and the coalescence of air bubbles (Section 3.1.2); it has direct impact on the oxygen solubility in the form of the factor b. Fig. 2 (adapted from (Colt, 1984)) shows the dependence of DO on the salinity of the water around the mesophilic temperature range. A few observations can be made. Firstly, the dependence appears to be linear in the observed range of salinity and all are able to satisfy the minimum DO requirement of 2 mg L1. Secondly, by observing the gap between the isotherms, the dependence is larger at lower temperature than at higher temperature. Thirdly, by using the relationship given in Table 3 and the values from Fig. 2, b can be estimated. For instance, at the temperature of 20 C and lower salt level (TDS < 5 g L1), the effect is slight: b is 0.97 and bigger than 0.95, which is the value commonly adopted for used water applications (Stephenson et al., 2000; Tchobanoglous et al., 2004). However, b reduces with increasing salt level. At the same temperature, b is still somewhat modest at 0.94 for TDS ¼ 10 g L1 and 0.92 for TDS ¼ 15 g L1, but reduces to 0.74 for TDS ¼ 50 g L1. When the salt content is expected to be significantly higher than conventional biological used water treatment, the b value can vary significantly. In view of the operational and economical importance of aeration, it is necessary to consider the impact of salt elevation on the oxygen transfer for an
26
water research 44 (2010) 21–40
Table 3 – Correction factors for oxygen transfer rate Correction factor a b q
Describes effects of
Relationship
Typical range (Tchobanoglous et al., 2004; Judd, 2006)
Difference in oxygen mass transfer coefficient Salinity-surface tension Temperature
aðused waterÞ a ¼ KKLLaðclean waterÞ cS ðused waterÞ b ¼ cS ðclean waterÞ -
0.3–1.2a 0.7–0.98 1.024
a The correction factor a considers the complex inter-relationships between mixing intensity, type of aeration device, tank geometry and used water characteristics, and is influenced by a wide range of factors. Typical range of a for diffused aeration devices and mechanical aeration devices are 0.4–0.8 and 0.6–1.2, respectively (Tchobanoglous et al., 2004). a > 1 is possible under high degrees of turbulence or in the presence of certain surfactants that result in smaller specific surface areas for mass transfer (Hwang, 1979; Gillot et al., 2000).
3.1.2.
Density, turbidity and viscosity of suspension
An elevated salt environment could impact other physical parameters of a used water treatment system, such as density, turbidity and viscosity of suspension. The density of the mixed liquor (ML) of an HRMBR system may be expressed in relationship to its constituents as follows (Rautenbach, 1993):
rML
,01 P wi 1 Xwi i Aþ ¼1 @ rL ri i
(4)
where rML is the density of the mixed liquor; rL, density of water; ri, density of the individual constituents of the used water; wi, weight fraction of the individual constituents of the used water. When the salt content (wi) in the water increases, typically ri > rL, the density of the mixed liquor (rML) increases. Assuming NaCl is the only varying constituent in the water and referring to the literature at the temperature of 25 C (Lide, 2008), rML at the salt concentration of 10 and 15 g L1 would be about 0.7% and 1.1%, respectively, denser than its freshwater counterpart. However, this would increase to 3.4% at 50 g L1. The higher density of the suspension will mean that the liquid exerts a greater buoyant force, which retards sedimentation. This effect, couples with the other effects of higher salt level on the microbiology, result in weaker biological flocs’ structure and hence higher turbidity (Woolard and Irvine, 1995; Ng et al., 2005; Lefebvre and Moletta, 2006). The observation may appear counter-intuitive at first sight from the perspective of colloidal chemistry, where it is expected that greater salinity would decrease double-layers effects and encourage coagulation (Section 3.1.3). The explanation is that greater salinity could induce cell plasmolysis due to the increase in osmotic pressure of the environment and adversely impacting non-salt tolerant microorganisms including filamentous bacteria, and higher microorganisms such as protozoa and rotifiers (Section 3.2). This means that there would be a larger number of smaller-size particles dispersed in the system, which are not consumed by the higher microorganisms. Concomitantly, the deficiency of the filamentous bacteria weakens the mechanical integrity and
structure of the biological flocs, and results in the susceptibility of the flocs to break up into smaller particles. Furthermore, it is possible that microbial cells that do not lyse under the salt stress could develop surfaces that are less inclined to flocculate (Ng et al., 2005). This could result in poor effluent quality in a conventional biological treatment system that uses sedimentation, but not so for an MBR system, as the suspended solids will be effectively retained by the membrane. However, the issue here relates to the higher amount of the smaller particles that may occur in an HRMBR system and affect membrane fouling. Due to the non-Newtonian pseudo-plastic nature of the mixed liquor, the viscosity effect is complicated (Judd, 2006). In general, it may be anticipated that increasing salt concentrations will result in higher viscosity of a liquid, due to the increase in the dissolved solids content, as predicted by the Einstein’s equation (Thomas, 1965; Bird et al., 2007). Furthermore, increasing viscosity reduces the diffusivity and hence the mass transfer of oxygen in the mixed liquor (Section 3.1.1) The viscosity of the mixed liquor can be linked to the a factor ¨ zbek and Gayik, 2001; Krampe and in the following form (O Krauth, 2003; Judd, 2006): awhx
(5)
where x is an exponent to be experimentally determined. The higher viscosity due to greater salinity could therefore negatively affect oxygen transfer. It is to be noted, however, that the oxygen transfer may be positively affected by an elevated salt environment in another
12 11
Dissolved Oxygen [mgL-1]
HRMBR system. In some instances, the level of salt may need to be limited, which is determined by the CF defined by Eq. (1). If this is the case, it involves adjusting either or both SRT (decrease) and HRT (increase) away from more conventional values.
10 9
10 ºC
8
20 ºC
7
30 ºC
6 5
40 ºC 4 0
5
10
15
20
25
30
35
40
45
50
-1
Salinity [gL ]
Fig. 2 – DO in relationship with the salinity of the solution.
water research 44 (2010) 21–40
way. It was demonstrated (Zlokarnik, 1979) that the presence of salt could enhance oxygen mass transfer by promoting non-coalescence of the air bubbles in a water solution system, possibly due to changes in the water structure. The higher salt concentration favours the preservation of smaller primary bubbles and therefore increases the air–liquid interfacial area, and hence the a factor. It was reported that the enhancement for a slot injector system increased with increasing NaCl concentration in the experimental range between 0 and 15 g L1 of NaCl, though the enhancement effect was more pronounced between 3 and 5 g L1 of NaCl and less effective beyond 5 g L1 of NaCl. The enhancement effect also depended to a large extent on the interaction of material and process-related parameters and on the type of the gas dispersing device (Zlokarnik, 1979). The aforementioned discussion highlights the complicated subject matter of aeration and oxygen transfer. Many effects are involved, sometimes counteracting among themselves, such that a quantitative assessment of the overall impact of elevated salt environment on the oxygen transfer is difficult. Higher viscosity would also reduce the effectiveness of applying air bubbling to alleviate membrane fouling on a submerged hollow-fibre MBR system. A more viscous liquid would dampen membrane fibre movement and favour larger air bubbles with slower rise velocity (Wicaksana et al., 2006).
3.1.3. Salt precipitation, solute interactions and colloid chemistry An elevated salt environment with high TDS can lead to supersaturation condition that causes scaling on the membrane, and destabilises colloidal system in water that aggravates colloidal fouling. Both effects exert detrimental influence on membrane performance (Baker, 2004; Le-Clech et al., 2006), and when coupled together, it has been reported that they caused greater membrane flux decline than simple summation of the individual effects (Tarabara, 2007; Wang and Tarabara, 2007). It is therefore important to consider the physicochemical effects of precipitation (associated with scaling), solute interactions and colloid chemistry (associated with fouling) for salt accumulating systems. Scaling is essentially a crystallisation process (Gloede and Melin, 2008). It occurs when the saturation limits of the sparingly soluble salts are exceeded (Melin and Rautenbach, 2007). The common scalants include salts such as: calcium sulphate (CaSO4), calcium carbonate (CaCO3), silica (SiO2), and increasingly being observed also calcium phosphate (Ca3(PO4)2) in the application of the used water, or when phosphorus containing antiscalants are used (Schaefer et al., 2005a). The salts precipitate out of the water and form inorganic deposits on the membrane. A method to characterise the propensity of scaling that is often used is the supersaturation index (SI), defined as: SI ¼
IAP Ksp
(6)
where IAP is the ion activity product and Ksp is the solubility product constant for the mineral salt of interest. SI greater than one (SI > 1) implies that the solubility limit is exceeded, and scaling may occur.
27
As shown in Table 1, due to the salt accumulation effect of the HRMBR systems, the IAP of the potential scalants present in the feed water will necessarily be raised. Under the influence of concentration polarisation, this effect will be further aggravated, as the salt concentration at the membrane surface will be considerably higher than in the bulk feed water (Section 3.3.1). From conventional dense membrane processes used for water reclamation, the risk of scaling is often observed at 75% recovery (CF ¼ 4). For HRMBR systems, this would mean that SI > 1 for some potential scalants is likely to occur when CF > 4. The supersaturation condition serves as the thermodynamic driving force for the crystallisation process (Green and Perry, 2007), but it alone does not necessarily lead to scaling due to kinetic considerations (Gloede and Melin, 2008). This means that scaling may be controlled to a certain extent by the use of antiscalants, which are used in the practice in RO desalination. However, in the situation where there could be high salt accumulation, the use of antiscalants can be costly, and may not totally prevent scaling for high supersaturation conditions (Rahardianto et al., 2007). Furthermore, as particulate matter could adsorb antiscalants, they may not function well in an environment with high concentration of suspended solids, as is the case in a MBR system (Tanninen et al., 2005). Some antiscalants could also aggravate biofouling on the membrane (Vrouwenvelder et al., 2000). It is of interest to note that some form of scaling may be retarded in the presence of the other water constituents. For instance, it was reported that the presence of bicarbonate, magnesium ions and humic acid retarded the onset of gypsum (CaSO4$H2O) scaling (Le Gouellec and Elimelech, 2002). This might be a positive effect for used water application as such compounds are available, and if gypsum were to be the only scalant present. However, considering that scaling and fouling are a complex phenomenon (Schaefer et al., 2005a) that can be caused by a variety of scalants/ foulants and their interactions, and there are scalants such as silica and apatitie whose effects are aggravated in the presence of the other salts, it is likely that the elevated salt environment would intensify scaling and fouling. However, it is recognised that some organics such as humics retard the crystallisation process. Thus, the organics in the mixed liquor of an HRMBR system could partially alleviate the scaling problem. The effects will be system and feed specific. The MDBR presents a further complication. In addition to concentration polarisation, MD experiences temperature polarisation where the liquid temperature at the membrane surface is lower than in the bulk liquid (Schofield et al., 1987). For sparingly soluble salts, this increases the potential for scaling if solubility increases with temperature, and vice versa for salts with solubility decreasing with temperature (Tun et al., 2005). An elevated salt environment can also aggravate colloidal fouling by increasing the ionic strength of the water (Chong, 2007). This can be understood from the Deryagin–Landau– Verwey–Overbeek (DLVO) theory, which assumes the interaction between the particles is balanced by the van der Waals’ attractive force and the electrostatic repulsive force, and can
28
water research 44 (2010) 21–40
be characterised by the parameter k which is the reciprocal of the Debye length as follows (Gregory, 2006):
k¼
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1000F2 X 0 2 ci zi 3perm RT
(7)
where c0i and zi are the concentration and charge of the particles, F is the Faraday’s constant, 3perm is the permittivity of the solution, R is the universal gas constant, and T is the absolute temperature. With increasing salt concentration, the Debye length will therefore be reduced. This effect is known as double layer compression, and the particles become destabilised as they come closer towards one another resulting in agglomeration, though the effect on biofloc is more complex as discussed in Section 3.1.2. The effect is more pronounced at higher salt levels. For illustration, using Eq. (7), and taking NaCl as the reference TDS, the value of k at the salt levels of 10, 15 and 50 g L1 would be 4.5, 5.5 and 10 times, respectively, higher than a reference value at 500 mg L1 TDS. In this way, increasing salt concentration can aggravate the fouling on the membrane with the formation of a more densely packed cake layer (Faibish et al., 1998). This effect has been referred to as the salinitypromoted fouling in the literature (Reid et al., 2006), and has also been observed in real studies treating saline sewage (Tam et al., 2006). The positive side of the matter, however, is that the greater tendency of precipitation and coagulation in an elevated salt environment also presents itself as an opportunity for process enhancement. This is discussed further in Section 3.3.2. In summary, it is important to consider the impact of an elevated salt level on the physicochemical parameters, because these parameters could present the physical limits of the treatment process, which are not easily overcome by technical measures. The impact typically intensifies with increasing salt concentration. Considerable physicochemical challenges would be expected at the higher salt concentrations (w50 g L1), but may be tolerable at lower salt concentrations. In the context of this study, this would imply that it may be necessary to operate the treatment process at the lower salt levels (presumably around or less than 15 g L1 salt) due to physicochemical constraints. Consequently, the CF would need to be adjusted accordingly as predicted by Eq. (1), which in turn would result in lower recovery and could also affect other aspects of the system (Section 3.4).
3.2.
Microbiological aspects
3.2.1.
Microbiology in elevated salt environment
In the field of environmental biotechnology, the success of treatment depends on how well the microorganisms, often in mixed cultures, can survive and carry out the desired functions in complex ecosystems (Rittmann and McCarty, 2001). Within the context of this paper, the success of treatment essentially depends on the ability of the microorganisms to maintain growth and perform their function of biodegrading pollutants present in the used water under elevated salt condition. This is not a trivial matter, because microorganisms have specific growth requirements and those normally involved in conventional used water treatment are not habituated to higher salt level (Woolard and Irvine, 1995).
The growth range of the microorganisms in terms of salt concentration is therefore an important criterion in assessing the viability of the biological treatment under elevated salt conditions. Microorganisms with growth range that does not cover the actual salt concentration in the operating environment would not be capable of performing the treatment. In the literature, there is a distinction between halophilic microorganisms and halotolerant microorganisms, but this distinction is not clear-cut. True halophilic microorganisms or halophiles are those that grow in saline environment and require a certain minimum level of salt for survival. Halotolerant microorganisms, on the other hand, are those that grow better in freshwater environment, but can tolerate higher salt concentrations and can be found in saline environment too (Rodriguez-Valera et al., 1981). A more detailed classification is to categorise microorganisms according to the salt concentration that is optimal for growth. Under this classification, there are four main categories of microorganisms as shown in Table 4 (adapted from Ventosa and Nieto, 1995; Woolard and Irvine, 1995). The majority of microorganisms involved in conventional used water treatment, such as the activated sludge system, are non-halophilic (Woolard and Irvine, 1995). These microorganisms do not possess the mechanisms to cope with the osmotic stress exerted by an elevated salt environment. They are normally able to tolerate lower salt concentration up to 10 g L1 without acclimation. Below this concentration, salt may even result in a stimulatory effect with enhanced organic carbon removal (Ng et al., 2005). However, above 10 g L1, higher salt concentrations would bring about considerable osmotic stress on the microorganisms generated by the osmotic pressure of the environment (Eq. (2)). The osmotic stress would cause an outward flow of intracellular water, resulting in cell dehydration and eventually, plasmolysis and loss of activity of the cells for these microorganisms (Peyton et al., 2002; Uygur, 2006). The ability to cope with osmotic stress by maintaining osmotic balance between the intracellular environment of the cytoplasm and the elevated salt environment is therefore an essential attribute of the halophilic and halotolerant microorganisms. One strategy, known as the ‘‘salt-in’’ strategy, involves the accumulation of potassium (Kþ) and chloride (Cl) ions within the cytoplasm for osmotic balance (Oren, 1999). This mode of osmotic adaptation is found to be bioenergetically less expensive, but it requires the intracellular enzymatic systems to remain functional at high concentrations of inorganic salts. The use of this strategy is therefore confined only to a few specialised groups of extreme halophilic microorganisms such as those within the archaea order of Halobacteriales and the bacteria order of Halanaerobiales (Oren, 2007).
Table 4 – Categories of microorganisms according to the optimal growth range in NaCl Category Non-halophilic Marine or slightly halophilic Moderately halophilic Extremely halophilic
NaCl range for optimal growth [g L1] 150
water research 44 (2010) 21–40
Another strategy, known as the ‘‘compatible solute’’ strategy and which is more widely used among the larger group of moderately halophilic and halotolerant microorganisms across all three domains of Archaea, Bacteria and Eucarya, involves the accumulation of compatible organic osmotic solutes such as glycerol, glycine betaine, ecotine and various sugar alcohols and amino acids within the cytoplasm for osmotic balance. This mode of osmotic adaptation is found to be bioenergetically more expensive, but it does not require the intracellular enzymatic systems to adapt to high concentrations of inorganic salts. Microorganisms that adopt the ‘‘salt-in’’ strategy normally are able to grow in extremely high salt environment up to saturation level, but may suffer lysis of the cells when exposed to lower salt concentrations. Microorganisms that adopt the ‘‘compatible solute’’ strategy, on the other hand, do not normally thrive in extremely high salt environment, but are able to grow over a wide range of salt concentrations (Ventosa et al., 1998; Oren, 2002a, b). The foregoing discussion sheds light on some microbiological fundamentals observed under high salt concentrations. Due to the higher energetic cost necessary for osmotic adaptation, biological reactions need to provide sufficient energy to the microorganisms for their survival in elevated salt environment (Gerday and Glansdorff, 2007). On the basis of bioenergetics consideration, it becomes understandable why processes such as aerobic respiration and denitrification continue to occur at higher salt concentrations because of the larger amount of energy available from the reactions, whereas process such as nitrification, because of the smaller amount of energy available, is observed to occur at relatively lower salt concentrations (see also Section 3.2.3). On the same basis, it may be expected that the true growth yield (Y ) of microorganisms would reduce when adapting to increasing salt concentration, because more energy derived from substrate utilisation would be channelled to osmotic maintenance and less to growth. However, the observed yield (Yobs), measured experimentally by the mass of the dry cell over the mass of the utilised substrate, may reveal no difference or even increase with increasing salt concentration. This is because the organic osmotic solutes used for osmotic adaptation would contribute to the cell mass, but not to growth (Oren, 1999). Another point to note is the great microbial diversity of the halophilic and halotolerant microorganisms. These microorganisms can be found ubiquitously in nature ranging from salt lakes and saline soils to salted food and unusual habitats (Oren et al., 1992; Ventosa et al., 1998; Oren, 2002a). Some of these microorganisms are subjected to other extreme conditions such as extreme temperatures and extreme pH, and possess tolerance for such conditions. Consequently, there are halophilic or halotolerant microorganisms, which are thermophilic or thermotolerant (Madigan and Oren, 1999). Similarly, there are also halophilic or halotolerant microorganisms, which are alkaliphilic or alkalitolerant (Horikoshi, 1999). There are even the so-called polyextremophiles (Rothschild and Mancinelli, 2001), such as the haloalkalithermophiles, which can withstand simultaneously the elevated conditions of salt, pH and temperature. Within the scope of this study, the above knowledge is useful, as it allows an HRMBR system to be operated and optimised under different environment conditions.
3.2.2.
29
Biological carbon removal in elevated salt environment
In general, there is no difficulty in achieving biological carbon removal in elevated salt environment. Within the large and diverse group of halophilic and halotolerant microorganisms, there are a great number of aerobic heterotrophs that are able to biodegrade the organic carbon matter present in used water (Oren et al., 1992; Ventosa et al., 1998; Oren, 2002a). The issue here rather is to obtain the right type of microorganisms according to the salt range as discussed above. There are plentiful studies in the literature that confirm using halophilic or halotolerant microorganisms can achieve effective COD removal in used water with salt concentration as high as 150 g L1 NaCl and with substrates that are even considered bactericidal (Woolard and Irvine, 1995; Peyton et al., 2002). Some studies found that the addition of halophilic or halotolerant microorganisms enhanced COD removal in used water over the salt range from 0 to 60 g L1, but there could be a minimum at the concentration of around 30 g L1 and the enhancement effect was more pronounced only at higher salt concentration above 30 g L1. The explanation offered was that the minimum COD removal occurred at the salt concentration where it was high enough to slow the activity of the non-halophilic microorganisms, but too low for the effective operating range of the particular added halophilic microorganism, Halobacter halobium (Kargi and Dincer, 1996a, b). Acclimation or adaptation of microorganisms found in conventional systems to an increasing saline environment is another strategy adopted for treating used water. In fact, this strategy appears to work for salt concentration up to about 30 g L1. It is to be pointed out that, with the exception of one example which used osmo-tolerant yeast sludge (Dan et al., 2002), all four other examples given in Table 2 made use of acclimated cultures from existing systems in treating saline used water. In one example (Sharrer et al., 2007), high BOD removal exceeding 99.8% was obtained at all the tested salinity levels of 0, 8, 16 and 32 g L1 after adequate time had been allowed for acclimation. Operating at the SRT of 64 days, the time duration to reach stabilised acclimation or quasisteady-state conditions varied, and ranged between 6 and 117 days. However, this variation in the acclimation could be attributed to the adaptation of the autotrophic nitrifying microorganisms (see below), rather than to the adaptation of the heterotrophic microorganisms. Nevertheless, there could be an upper-bound salt limit for the acclimation strategy to work, and from the surveyed literature, this limit appeared to be between 30 and 50 g L1 NaCl. The acclimation achieved would also be non-permanent, and would be lost when the salinity of the environment changes. To summarise, for the treatment goal of carbon removal, two strategies may be considered to treat used water under elevated salt environment. For salt concentration up to 30 g L1, the acclimation strategy may work. For salt concentration higher than 30 g L1, addition of halophilic or halotolerant microorganisms should be considered. Furthermore, from a practical view point, the conditions at ‘start-up’ of the HRMBR systems will differ from the ‘steady state’, as the salt level gradually increases from the feed concentration to the value determined by the CF (SRT/HRT). Careful acclimation of the biomass may be necessary under these dynamic conditions.
30
water research 44 (2010) 21–40
3.2.3. Biological nutrient removal in elevated salt environment Biological nutrient removal (BNR) refers to the removal of nitrogen and phosphorus using microorganisms. Within BNR, there are two processes: nitrification to convert reduced nitrogen compounds to its oxidised nitrite (NO 2 ) or nitrate (NO 3 ) form, and denitrification to convert these to the final desired nitrogen gas (N2) form. The microorganisms involve in the two processes are broadly known as nitrifiers and denitrifers (Seviour and Blackall, 1999). Biological phosphorus removal utilises specific heterotrophic microorganisms known as the phosphorus accumulating organisms (PAOs) (Tchobanoglous et al., 2004). While it seems possible to find a range of microorganisms to carry out the task of carbon removal under different salt conditions as outlined in Section 3.2.2, there are more constraints for biological nutrient removal. All three types of nutrient removal microorganisms are affected by increasing salt content, with PAOs being the least salt tolerant and denitrifers being the most salt tolerant (Panswad and Anan, 1999; Uygur and Kargi, 2004).
3.2.3.1. Phosphorus removal. It is postulated that the sensitivity of phosphorus removal towards saline used water could be attributed to the accumulation of salt in PAO’s cells. This increases the osmotic pressure within the microbial cells, and thereby diminishes the phosphate accumulating capability of the microorganisms. As a result, phosphorus removal is inhibited, which leads to lower removal efficiency (Panswad and Anan, 1999). Of the studies surveyed, most show a continuous decline of phosphorus removal efficiency with increasing salt content (Uygur and Kargi, 2004; Sharrer et al., 2007). It was observed that satisfactory phosphorus removal was achieved up to around 2 g L1 NaCl, subject to the availability of sufficient soluble COD in the form of acetate (Intrasungkha et al., 1999). In another study, however, salt inhibition effects became pronounced and phosphorus removal efficiency decreased sharply at about 5 g L1 salt content. In this study, the phosphorus removal efficiency decreased from 84% to 22% when the salt content increased from 0 to 60 g L1 (Uygur and Kargi, 2004). Biological phosphorus removal may be enhanced by incorporating an anaerobic reactor before an anoxic-aerobic treatment process. This would provide an environment that is favourable to the PAOs by being rich in carbon and low in NO 2 and NO3 . Other possible strategies to enhance phosphorus removal include favourable carbon substrate, higher pH (w8), lower temperature (0.98 inhibition was observed resulting in 71–97% of the available NO 3 being converted to NO 2 for C:N values up to 8.5. Yoshie et al. (2006) also reported nitrite accumulation in concentrated brines indicating reductase activity may be very different at high salinity. Protein and polysaccharide transmission through the membrane at steady state were 27.3% 8.0% and 81.5% 10.5% respectively. Fawehinmi (2006) observed similar transmission rates for proteins and polysaccharides, recording 49% and 80% respectively, for operation of an anaerobic immersed hollow fibre (0.1 mm) MBR. In this study, SMP exhibited a principal protein peak of 55.1% between 70% (Low) 30–70% (Median) 700 mm) operated with low bulk liquid nitrate concentration, nitrate diffusion is expected to limit performance.
4.3.
Biofilm structure
Although methanotrophs in both reactor A and B were concentrated near the membrane, which provided their source of methane, the biofilm structures appeared significantly different. The images suggest that the direction of the oxygen supply affects biofilm structure resulting in a more porous structure in reactor B, in which oxygen was supplied from the membrane, and a very dense structure in reactor A, in which oxygen was supplied from the bulk liquid. In reactor A, the DO concentration in the methanotrophic zone is most likely low, so the methanotrophic activity is limited by oxygen availability. In reactor B, both methane and oxygen are abundant in the active methanotrophic zone. The porous cluster-shaped structure may be a way of enhancing oxygen transfer to the outer portions of the biofilm. The activities of non-methanotrophic microorganisms in the biofilms are unclear. The FISH images show that microorganisms not hybridizing with the Type I methanotrophic probes are indeed present (compare red clusters with green and blue clusters in Figs. 6 and 7). The non-methanotrophs in
these images typically have lower fluorescence intensity than the methanotrophs. The reason for this may be a lower rRNA content due to a lower growth rate (Wallner et al., 1993; Amann et al., 1995). Since non-methanotrophs in the biofilms are dependent on hydrolyzed biomass components or metabolites from methane oxidation for carbon and energy, they are likely living under substrate-limitation in the biofilms. The PCR amplification of nitrate reductase gene fragments from the biofilms in Experiment 2 shows that the biofilms indeed have genetic potential for denitrification. Whether this potential is expressed in Experiment 2 is unclear; however, the high nitrate removal efficiencies observed in Experiment 1 indicate denitrification took place. LaPara et al. (2006) analyzed the location of nirK, nirS, and ammonia monooxygenase (amoA) in a MBfR biofilm cultivated on oxygen, acetate, and ammonium. They found that nirK and nirS genes were mainly located a distance from the membrane surface, which provided oxygen (LaPara et al., 2006). In our study, we did not determine the spatial location of the nitrate reductase genes in the biofilms. The electron donor used by denitrifiers present in the biofilms is also not clear. Meschner and Hamer (1985) showed that methanol-utilizing denitrifiers could coexist with aerobic methanotrophs, Costa et al. (2000) suggested acetate was a more important electron shuttle between methanotrophs and denitrifiers. Eisentraeger et al. (2001) showed that methanol, acetate, and proteins could be utilized by coexisting denitrifiers whereas Rhee and Fuhs (1978) suggested citrate was used. Most likely a mixture of organic compounds are made available by methanotrophs and utilized by coexisting bacteria.
5.
Conclusions
The goals of the two experiments were to investigate whether MBfRs with varying gas supply regimes for methane and oxygen could achieve high nitrate removal efficiencies compared to previously known values for suspended cultures and to investigate the biofilm characteristics. Experiment 1 showed that methanotrophic biofilms supplied with both methane and oxygen from the membrane, supplied with the gases through separate intertwined membrane fibers, and supplied with only methane from the membrane and oxygen from the bulk liquid could all achieve high nitrate removal efficiencies of around 0.3–0.4 mol TIN mol1 CH4. The gas supply regime did not seem to be an important factor; instead it was the biofilm-mode of growth that was superior to suspended growth in terms of nitrate removal efficiency. Thus, our study suggests that in the future development of bioreactors for denitrification with methane, focus should be placed on biofilm reactors. Experiment 2 showed that anoxic zones tended to develop within the biofilms, particularly when both oxygen and methane were supplied from the membrane. A significant nitrite production was also observed with this gas supply regime at high influent nitrate concentrations. The biofilm in both reactors had methanotrophs primarily growing near the membrane, which in both cases supplied their source of methane. With mixed gases the biofilm structure near the membrane was porous and cluster-shaped whereas with only
water research 44 (2010) 85–96
methane from the membrane a very dense methanotrophic biofilm structure was observed near the membrane surface. The biofilms in this experiment had the genetic potential for denitrification as evidenced by PCR amplification of the nitrite reductase genes, nirS and nirK.
Acknowledgements A part of this research was financially supported by MEXT through Special Coordination Funds for Promoting Science and Technology (Project name: IR3S). O.M. was supported by MEXT through the Monbukagakusho scholarship.
references
Amann, R.I., Binder, B.J., Olson, R.J., Chisholm, S.W., Devereux, R., Stahl, D.A., 1990. Combination of 16S-ribosomal-RNA-targeted oligonucleotide probes with flow-cytometry for analyzing mixed microbial populations. Applied and Environmental Microbiology 56 (6), 1919–1925. Amann, R.I., Ludwig, W., Schleifer, K.-H., 1995. Phylogenetic identification and in-situ detection of individual microbial cells without cultivation. Microbiological Reviews 59 (1), 143–169. Amaral, J.A., Achambault, C., Richards, S.R., Knowles, R., 1995. Denitrification associated with groups I and II methanotrophs in a gradient enrichment system. FEMS Microbiology Ecology 18, 289–298. APHA, AWWA, WEF, 1998. Standard Methods for the Examination of Water and Wastewater, twentieth ed. American Public Health Association/American Water Works Association/Water Environment Federation, Washington DC, USA. Brindle, K., Stephenson, T., Semmens, M.J., 1998. Nitrification and oxygen utilization in a membrane aeration bioreactor. Journal of Membrane Science 144 (1–2), 197–209. Casey, E., Glennon, B., Hamer, G., 2000. Biofilm development in a membrane-aerated biofilm reactor: effect of flow velocity on performance. Biotechnology and Bioengineering 67 (4), 476–486. Clapp, L.W., Regan, J.M., Ali, F., Newman, J.D., Park, J.K., Noguera, D.R., 1999. Activity, structure, and stratification of membrane-attached methanotrophic biofilms cometabolically degrading trichloroethylene. Water Science and Technology 39 (7), 153–161. Costa, C., Dijkema, C., Freidrich, M., Garcia-Encina, P., FernandezPolanco, F., Stams, A.J.M., 2000. Denitrification with methane as electron donor in oxygen-limited bioreactors. Applied Microbiology and Biotechnology 53, 754–762. Daims, H., Bruhl, A., Amann, R., Schleifer, K.H., Wagner, M., 1999. The domain-specific probe EUB338 is insufficient for the detection of all bacteria: development and evaluation of a more comprehensive probe set. Systematic and Applied Microbiology 22 (3), 434–444. Downing, L.S., Nerenberg, R., 2008. Effect of oxygen gradients on the activity and microbial community structure of a nitrifying, membrane-aerated biofilm. Biotechnology and Bioengineering 101 (6), 1193–1204. Eisentraeger, A., Klag, P., Vansbotter, B., Heymann, E., Dott, W., 2001. Denitrification of groundwater with methane as sole hydrogen donor. Water Research 35 (9), 2261–2267. Eller, G., Stubner, S., Frenzel, P., 2001. Group-specific 16S rRNA probes for the detection of type 1 and type 2 methanotrophs by fluorescence in situ hybridization. FEMS Microbiology Letters 198, 91–97. Ettwig, K.F., Shima, S., van de Pas-Shoonen, K.P., Kahnt, J., Medema, M. H., op den Camp, H.J.M., Jetten, M.S.M., Strous, M., 2008.
95
Denitrifying bacteria anaerobically oxidize methane in the absence of Archaea. Environmental Microbiology 10 (11), 3164–3173. Freitas dos Santos, L.M., Livingston, A.G., 1995. Membraneattached biofilm for VOC wastewater treatment I: novel in situ biofilm thickness measurement technique. Biotechnology and Bioengineering 47, 82–89. Hallin, S., Lindgren, P.-E., 1999. PCR detection of genes encoding nitrite reductase in denitrifying bacteria. Applied and Environmental Microbiology 65 (4), 1652–1657. Hibiya, K., Terada, A., Tsuneda, S., Hirata, A., 2003. Simultaneous nitrification and denitrification by controlling vertical and horizontal microenvironment in a membrane-aerated biofilm reactor. Journal of Biotechnology 100, 23–32. Knowles, R., 2005. Denitrifiers associated with methanotrophs and their potential impact on the nitrogen cycle. Ecological Engineering 24, 441–446. LaPara, T.M., Cole, A.C., Shanahan, J.W., Semmens, M.J., 2006. The effects of organic carbon, ammoniacal-nitrogen, and oxygen partial pressure on the stratification of membrane-aerated biofilms. Journal of Industrial Microbiology and Biotechnology 33, 315–323. Leak, D.J., Dalton, H., 1986a. Growth yields of methanotrophs. Applied Microbiology and Biotechnology 23, 470–476. Leak, D.J., Dalton, H., 1986b. Growth yields of methanotrophs. Applied Microbiology and Biotechnology 23, 477–481. Lee, K.-C., Rittmann, B.E., 2000. A novel hollow-fiber membrane biofilm reactor for autohydrogenotrophic denitrification of drinking water. Water Science and Technology 41 (4–5), 219–226. Lide, D.R., 2008. CRC Handbook of Chemistry and Physics, eighty eighth ed. CRC Press, Boca Raton, USA. Meschner, K.L., Hamer, G., 1985. Denitrification by methanotrophic/metylotrophic bacterial associations in aquatic environments. In: Golterman, H.L. (Ed.), Denitrification in the Nitrogen Cycle. Plenum Press, New York, pp. 257–271. Modin, O., Fukushi, K., Yamamoto, K., 2007. Denitrification with methane as external carbon source. Water Research 41, 2726–2738. Modin, O., Fukushi, K., Nakajima, F., Yamamoto, K., 2008. A membrane biofilm reactor achieves aerobic methane oxidation coupled to denitrification (AME-D) with high efficiency. Water Science and Technology 58 (1), 83–87. Nerenberg, R., Rittmann, B., 2004. Hydrogen-based, hollow-fiber membrane biofilm reactor for reduction of perchlorate and other oxidized contaminants. Water Science and Technology 49 (11–12), 223–230. Okabe, S., Satoh, H., Watanabe, Y., 1999. In situ analysis of nitrifying biofilms as determined by in situ hybridization and the use of microelectrodes. Applied and Environmental Microbiology 66 (7), 3182–3191. Raghoebarsing, A.A., Pol, A., van de Pas-Schoonen, K.T., Smolders, A.J.P., Ettwig, K.F., Rijpstra, I.C., Schouten, S., Sinninghe Damste, J.S., Op den Camp, H.J.M., Jetten, M.S.M., Strous, M., 2006. A microbial consortium couples anaerobic methane oxidation to denitrification. Nature 440, 918–921. Ren, T., Roy, R., Knowles, R., 2000. Production and consumption of nitric oxide by three methanotrophic bacteria. Applied and Environmental Microbiology 66 (9), 3891–3897. Rhee, G.Y., Fuhs, G.W., 1978. Wastewater denitrification with onecarbon compounds as energy source. Journal WPCF 50 (9), 2111–2119. Rishell, S., Casey, E., Glennon, B., Hamer, G., 2004. Characteristics of a methanotrophic culture in a membrane-aerated biofilm reactor. Biotechnology Progress 20, 1082–1090. Rittmann, B.E., McCarty, P.L., 2001. Environmental Biotechnology: Principles and Applications. McGraw-Hill Book Co. Semmens, M.J., Dahm, K., Shanahan, J., Christianson, A., 2003. COD and nitrogen removal by biofilms growing on gas permeable membranes. Water Research 37, 4343–4350.
96
water research 44 (2010) 85–96
Shanahan, J.W., Semmens, M.J., 2006. Influence of a nitrifying biofilm on local oxygen fluxes across a micro-porous flat sheet membrane. Journal of Membrane Science 277, 65–74. Stewart, P.S., 2003. Diffusion in biofilms. Journal of Bacteriology 185 (5), 1485–1491. Throback, I.N., Enwall, K., Jarvis, A., Hallin, S., 2004. Reassessing PCR primers targeting nirS, nirK and nosZ genes for community surveys of denitrifying bacteria with DGGE. FEMS Microbiology Ecology 49, 401–417. Timberlake, D.L., Strand, S.E., Williamson, K.J., 1988. Combined aerobic heterotrophic oxidation, nitrification and denitrification
in a permeable support biofilm. Water Research 22 (12), 1513–1517. Waki, M., Tanaka, Y., Osada, T., Suzuki, K., 2002. Effects of nitrite and ammonium of methane-dependent denitrification. Applied Microbiology and Biotechnology 59, 338–343. Wallner, G., Amann, R., Beisker, W., 1993. Optimizing fluorescent in situ hybridization with rRNA-targeted oligonucleotide probes for flow cytometric identification of microorganisms. Cytometry 14, 136–143. Werner, M., Kayser, R., 1991. Denitrification with biogas as external carbon source. Water Science and Technology 23, 701–708.
water research 44 (2010) 97–104
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Influence of operating parameters on the arsenic removal by nanofiltration Alberto Figoli a,**, Alfredo Cassano a,*, Alessandra Criscuoli a, M. Salatul Islam Mozumder b, M. Tamez Uddin b, M. Akhtarul Islam b, Enrico Drioli a a b
Institute on Membrane Technology, ITM-CNR, c/o University of Calabria via P. Bucci 17/C, I-87030 Rende, Cosenza, Italy Department of Chemical Engineering and Polymer Science, Shahjalal University of Science and Technology, Sylhet, Bangladesh
article info
abstract
Article history:
Arsenic contamination of surface and groundwater is a worldwide problem in a large
Received 16 April 2009
number of Countries (Bangladesh, Argentina, Italy, USA, New Zealand, etc.). In many
Received in revised form
contaminated areas a continuous investigation of the available arsenic removal technol-
31 August 2009
ogies is essential to develop economical and effective methods for removing arsenic in
Accepted 2 September 2009
order to meet the new Maximum Contaminant Level (MCL) standard (10 mg/l) recom-
Published online 8 September 2009
mended by the World Health Organization (WHO).
Keywords:
laboratory scale by using two commercial nanofiltration (NF) spiral-wound membrane
Arsenic removal
modules (N30F by Microdyn-Nadir and NF90 by Dow Chemical). The influence of main
Nanofiltration
operating parameters such as feed concentration, pH, pressure and temperature on the As
Membrane technology
rejection and permeate flux of both membranes, was investigated. An increase of pH and
Drinking water
a decrease of operating temperature and As feed concentration led to higher As removal for
In this work the removal of pentavalent arsenic from synthetic water was studied on
both membranes, whereas higher transmembrane pressure (TMP) values slightly reduced the removal achievable with the N30F membrane. In both cases, the permeate flux increased with temperature and pressure and reached its maximum value at a pH of around 8. Among the parameters affecting the As rejection, feed concentration plays a key role for the production of a permeate stream respecting the limits imposed by WHO. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Arsenic is a natural tasteless and odourless element existing in the earth’s crust at average levels in the range of 2000–5000 mg/l (Mandal and Suzuki, 2002). Groundwater and surface contamination by arsenic is one of the major environmental problems in the present millennium, as many people are exposed to excessive arsenic amounts through contaminated drinking water (Smedley and Kinniburgh, 2002; Caceres et al., 2005). Acute and chronic exposure via drinking water has been reported in many Countries, especially Bangladesh, Argentina,
India, Mexico, Mongolia, Thailand and Taiwan, where a large proportion of groundwater is contaminated with arsenic at levels from 100 to 2000 ppb. Moreover, serious problems in terms of toxicity due to arsenic contamination are dominant in some Countries of South Asia such as West Bengal, India and Bangladesh (Bhattacharya et al., 1997; Bhattacharyya et al., 2003) where groundwater arsenic content (50–3200 ppb) can reach values above the national drinking water standards (50 ppb). In nature arsenic occurs in several chemical forms and oxidation states. The two states prevalent in water environment are trivalent (As (III)) and pentavalent (As(V)).
* Corresponding author. Tel.: þ39 0984 492067; fax: þ39 0984 402103. ** Corresponding author. E-mail addresses:
[email protected] (A. Figoli),
[email protected] (A. Cassano). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.007
98
water research 44 (2010) 97–104
Drinking water, after food, represents a secondary source of inorganic arsenic in the human system. Long term exposition to inorganic arsenic may cause a wide range of health effects including skin lesions such as hyperkeratosis and pigmentation changes, circulatory disorders, diabetes and cancers of bladder, lung, kidney and skin (National Research Council, 2001). Consequently, in recent years, authorities have taken a more stringent attitude to arsenic in the environment; in particular, World Health Organization (WHO) and US Environmental Protection Agency (USEPA) guidelines fixed the new standard limit for arsenic in drinking water to 10 ppb (WHO, 1998; USEPA, 2001). These new regulations impose the development of efficient methods for arsenic removal from drinking waters. Conventional arsenic removal technologies include adsorption and coagulation/precipitation processes. Conventional adsorbents such as activated carbon, activated alumina and ion exchanger resins have been used. New adsorbents (kaolinite–humic acid complexes, activated red mud, ferruginous manganese ore, porous resin loaded with crystalline hydrous zirconium oxide, zeolites, etc.) are currently being developed and evaluated (Mohan and Pittman, 2007). The main drawback in using adsorption processes for drinking water is the disposal of both spent media and the wastewater produced during regeneration/cleaning of the column. Different chemicals can be used for arsenic removal by coagulation/precipitation (Ng et al., 2004). They include: aluminium, ferric sulphate, ferric chloride, slaked or hydrated lime, ferric hydroxide, polyaluminium chloride (Meng et al., 2001). Disadvantages of this technology are: production of byproducts, release of taste and odour compounds due to chlorination, floc disposal and post-treatment. Membrane processes can be considered a promising technology for removing arsenic from water. The application of membrane techniques in the environmental protection involves a number of advantages in terms of: low consumption of energy, no requirement of chemical substances to be added, easy way to increase the capacity (modular system), separation in the continuous mode, possibility to easily join membrane processes with other unit processes (hybrid processes), separation carried out in mild environment conditions (Drioli et al., 2002; Drioli et al., 1999). The arsenic removal by pressure-driven membrane processes, including reverse osmosis (RO), nanofiltration (NF), ultrafiltration (UF) and microfiltration (MF), was recently reviewed by Shih (2005) and Uddin et al. (2008). Negatively charged UF membranes were studied by Brandhuber and Amy (2001) to evaluate the influence of membrane operating conditions and water composition on arsenic rejection. The presence of co-occurring divalent ions was shown to be sensitive to membrane operating conditions through the concentration polarization phenomena. A coupled process flocculation/MF for arsenic removal from drinking water was investigated by Han et al. (2002). MF of the flocculated water resulted in rejection of the flocs thus leading to low turbidity and arsenic removal in the permeate. RO and NF are capable of removing all kinds of dissolved solids including arsenic from water. The water, for treatment by membrane techniques, shall be free from suspended solids and the arsenic in water preferably in pentavalent form. In
fact, at neutral pH the predominant species for As(V) are 2 which means that As(V) exists as an H2AsO 4 and HAsO4 anion at a typical pH in natural water (pH 5–8), whereas in this range of pH As(III) is mainly present as uncharged species (H3AsO3) and, therefore, is less efficiently rejected. Amy et al. (1998) performed bench-scale RO testing by using Desal DK2540F obtaining a removal efficiency for arsenate (As (V)) and arsenite (As(III)) of 96% and 5%, respectively. Similarly, Geucke et al. (2009) obtained an As rejection significantly higher for As(V) than for As(III) by using marine RO desalinator with three different membrane modules (XLE2521, TW30-2521 and SW30-2521, all by Filmtec, Dow) made of thin-film polyamide composite membranes. NF membranes are usually asymmetric and negatively charged at neutral and alkaline drinking water pH. Therefore, separation of ions is based both on steric (sieving) and charge (Donnan exclusion) effects. These membranes are mainly used for the separation of multivalent ions from monovalent ones; however, it is also possible to achieve a certain separation of ions of the same valence by selecting the proper membrane and operating conditions (Lhassani et al., 2001). Since operating pressures in NF are lower than RO, separation occurs at low energy consumption (21% less than RO) and higher water fluxes can be achieved at lower transmembrane pressures. Additionally, the NF process is much more sensitive than RO to the ionic strength and pH of source water. The membrane surface charge is mainly due to anion adsorption from water rather than to fixed charged groups (as in the case of ion exchange membranes), therefore it depends strongly on bulk anion concentration (Velizarov et al., 2004). Different studies concerning the removal of arsenic from drinking water by NF are reported in literature. Sato et al. (2002) studied the performance of three types of NF membranes, ES-10 (polyamide), NTR-7250 (polyvinyl alcohol) and NTR-729HF (polyvinyl alcohol), supplied by Nitto Electric Industrial Co. (Japan), for arsenic removal. All membranes removed over 95% of pentavalent arsenic. Removal efficiencies of As(III) by NTR-7250 and NTR-729HF were lower than 22% due to the relatively larger pore size of these membrane. ES-10 showed removal efficiency of As(III) higher than 75%. Different rejection characteristics of arsenite (55%) and arsenate (99%) for ES-10 membrane were also found by Oh et al. (2000) in the low-operational-pressure range 0.2–0.6 MPa. NF-45, a fully aromatic polyamide thin-film composite NF membrane (Filmtec, Minneapolis, MN), removed 60–90% of arsenic from synthetic feed waters containing up to 316 ppb As(V) resulting in permeate arsenic concentrations up to 25 ppb (Vrijenhoek and Waypa, 2000). For this ‘‘loose’’ NF membrane the As rejection increased with increasing NaCl concentration; on the contrary, Sato et al. (2002) observed that the removal of As(V) for ‘‘tight’’ membranes was not affected by the ions concentration in groundwater. Saitu´a et al. (2005) studied the effect of operating conditions in removal of As(V) from water by using a spiral-wound thin-film composite polyamide membrane (192-NF300) supplied by Osmonics Inc. They found that arsenic rejection is independent of transmembrane pressure, cross-flow velocity and temperature. Moreover, arsenic rejection increased with arsenic retentate concentration and removals ranging from 93
water research 44 (2010) 97–104
to 99% and of 95% were obtained for synthetic feed waters and surface waters, respectively. Urase et al. (1998) investigated on the effect of pH on rejection of different species of arsenic by using a flat sheet aromatic polyamide NF membrane supplied by Nitto-Denko Co. Ltd. Arsenate rejection was almost constant (around 90%), while rejection of arsenite increased with pH. Recently, Uddin et al. (2007) studied the removal efficiency of two commercial polyamide NF membranes (NF-90 and NF-200) for As(III) and As(V), by analyzing the effect of the operating conditions on the rejections achievable. The feed stream consisted of tap water to which arsenate and arsenite were added. In all tests, As(V) was better rejected than As(III) and the highest removals obtained were above 98% for As(V) and around 65% for As(III). The removal of arsenic from natural groundwater was also investigated by Kosˇutic´ et al. (2005) by using thin-film polyamide NF membranes, NF270 and NFc (Filmtec Corporation, Dow Chemical Comp., Midland, MI). Rejection factor values of both NF membrane types for the sodium dibasic arsenate were higher (0.8–0.9) than those for the sodium chloride (0.53–0.65) and lower than those for sodium sulphate (>0.99). From the state of the art on the application of NF membranes for the arsenate removal from water it results, therefore, that the efficiency of the system can vary, depending on the membrane module properties and the feed water composition. Further investigations on other membrane modules and different As(V) feeds are, thus, of interest for the development of an arsenic removal strategy by NF. In this work, the effect of the membrane material on the NF performance was investigated by treating synthetic water containing pentavalent As. In particular, two commercial NF membranes (NF-90 and N30F) made of polyamide and polyethersulfone were chosen. The performance of each NF membrane was evaluated with relation to both permeate flux and As(V) rejection as a function of transmembrane pressure, temperature, As feed concentration and pH. The experimental work was performed by using only the As(V) species, which is the easier removable form of arsenic, considering that As(III) can be oxidised to As(V) by using a conventional pre-oxidation step with chemical oxidants (such as potassium permanganate or chlorine compounds) (Floch and Hideg, 2004; Zaw and Emett, 2002).
2.
Materials and methods
2.1.
Standards and reagents
The solution of pentavalent arsenic was prepared by dissolving analytical grade As2O53H2O (Sigma–Aldrich, Milan, Italy) in distilled water. Arsenic standard solutions, with concentration ranging from 100 to 1000 ppb, were prepared, immediately before use, by appropriate dilutions of a 1000 ppm stock solution. The pH of the solution was adjusted by either HCl or NaOH at 3, 6, 8 and 10, respectively.
2.2.
99
NF unit and procedures
NF experiments were carried out by using a bench-plant (Matrix Desalination Inc., USA) equipped with a feed and permeate container, a pressure vessel for 2.4 40 inches spiral-wound membrane modules, a pressurization pump, two pressure gauges, a thermometer for temperature measurement in the feed tank, a tap water heat exchanger for temperature control and a flow meter on permeate exit pipe. The effect of transmembrane pressure, pH, As feed concentration and temperature on the performance of the NF process, in terms of permeate flux and As rejection, was studied in experimental trials in which one of the variables was changed while the other ones were kept at a constant value. NF experiments were performed according to the total recycle configuration in which both permeate and retentate streams were recycled in the feed tank of the plant. Permeate fluxes and As rejection were analysed by changing the operating parameters in the range reported as follows: a) transmembrane pressure (TMP): 2–12 bar; b) pH: 3.5–10; c) temperature: 15–40 C; d) As feed concentration: 100– 1000 ppb. Each experimental run was repeated at least three times for verifying the reproducibility of results. A maximum error of 2% was registered.
2.3.
NF membrane modules
NF experiments were performed by using two types of commercial spiral-wound membrane modules named NF902540 and N30F-2440, supplied by Dow-Filmtec and MicrodynNadir GmbH, respectively. The characteristics of the membrane modules are summarised in Table 1.
2.4.
Sample analyses
The content of arsenic in the solution was determined by inductively coupled plasma-optical emission spectroscopy (ICP-OES) (Optima 2100 DV-Perkin Elmer) operating in the axial viewing mode. Argon, air and nitrogen were the used gases. The blank for the analysis was prepared by adding nitric acid to distilled water up to a HNO3 concentration of 2% v/v. Similarly, before measurements, samples and standard As solutions were acidified with nitric acid in order to obtain a final solution containing HNO3 at 2% v/v. The emission wavelength for arsenic was 193.696 nm. The system was equipped with an autosampler which automatically sent to the torch chamber the solution to be analysed. The deviation of each measurement was of 2% from the average value. Rejection factor R, defined as: R % ¼ 1 cp =cf 100
(1)
with cp and cf as permeate and feed concentration (ppb), respectively, was determined in each experiment. pH was measured by an Orion Expandable ion analyzer EA 920 pH meter (Allometrics, Inc. LA, USA) with automatic temperature compensation.
100
water research 44 (2010) 97–104
Membrane module
Active surface area (m2) MWCO (Da) MgSO4 rejection (%) NaCl rejection (%) Na2SO4 rejection (%) Maximum operating temperature ( C) Maximum operating pressure (bar) Maximum feed flow rate (m3/h) pH range, continuous operation pH range, cleaning Permeate flow rate (l/h) Water permeability (l/m2h bar)
NF90-2540
Polyamide thin-film composite 2.6
Hydrophilized polyethersulfone 1.6
w200a >97c 85–95 – 40
400b – 25–35d 80–95d 50
41
40
1.4
–
2–11
2–11
1–12 94.6c
1–12
–
>1e
a Estimated by Krieg et al. (2004). b Estimated from manufacturer’s data. c Salt rejection and permeate flow based on the following test conditions: 2000 ppm MgSO4, 25 C, 5 bar, 15% recovery. d Test conditions: 0.5%, 40 bar, 20 C, stirred cell (700 rpm). e At 20 C.
70 60 50 40 30 20 10
Results and discussion
3.1. Effect of transmembrane pressure (TMP) on permeate flux and arsenic removal In Fig. 1 the effect of the TMP on the permeate flux at different concentration of arsenic, for the NF90 and N30F membranes, respectively, is reported. Operating temperature and pH were fixed at 25 C and 8.0, respectively. The permeate flux at steady-state increased, as expected, with the applied pressure for all the As concentrations investigated. The increase of the As concentration in the feed solution did not influence the flux across the membrane for the NF-90 membrane (Fig. 1a) while it determined a slight decrease of the permeate flux for the N30F membrane (Fig. 1b). These results suggested that the water flux through the N30F membrane should be affected by the solute permeate flux at different As feed concentration. Fig. 2 shows the rejection of arsenic as a function of TMP for both membranes at a feed As concentration of 100 ppb, a pH of 8 and a temperature of 25 C. The removal of As by NF-90 membrane is higher than the value observed with the N30F membrane over the pressure range investigated. In particular, the removal of As with the NF-90 membrane is higher than 94%, while for the N30F membrane the rejection towards As was higher than 78%. This phenomenon can be explained assuming a lower molecular weight cut-off for the NF-90 membrane (w200 Da) in comparison with the N30F membrane (400 Da).
0
2
4
6
8
10
12
14
10
12
14
TMP (bar) 70
b
distilled water 100 ppb 500 ppb 1000 ppb
60 50 40 30 20 10 0
3.
distilled water 100 ppb 500 ppb 1000 ppb
80
0
Steady-state permeate flux (l/m2h)
Membrane material
NF90-2540
90
a Steady-state permeate flux (l/m2h)
Table 1 – Characteristics of the NF membrane modules.
0
2
4
6
8
TMP (bar) Fig. 1 – Effect of TMP on permeate flux at different As concentrations for the: a) NF-90 membrane and b) N30F membrane (temperature [ 25 8C, pH [ 8).
The increasing in operating pressure did not improve the As rejection in the range of pressure investigated, in particular for the NF-90 membrane. Saitu´a et al. (2005) obtained similar results in the NF of synthetic solutions with spiral-wound NF polyamide membranes (192-NF 300, Osmonics, Inc.) characterised by a molecular weight cut-off of 180 Da. Similarly, Sato et al. (2002) found that the As(V) removal was practically unrelated to the applied pressure in the NF of synthetic water with ES-10, NTR-729HF and NTR-7250 (all by Nitto Electric Industrial Co., Japan) membranes. Finally, the arsenic concentration in the permeate of the NF-90 membrane was lower than the EPA recommended MCL (10 ppb) and both membranes met the Bangladesh standard MCL (50 ppb) as showed in Fig. 2.
3.2. Effect of operating temperature on permeate flux and arsenic removal Fig. 3 shows the effect of the operating temperature on the steady-state permeate flux at different concentration of arsenic for the NF90 and N30F membranes, respectively. NF experiments were performed at a TMP of 6 bar and at a pH of
101
Bangladesh MCL
50
80
40
permeate As NF90 permeate As N30F rejection NF90 rejection N30F
30
60 40
20
0
20
EPA MCL
10
0
2
4
6
8
10
12
0 14
TMP (bar) Fig. 2 – Effect of TMP on the removal of As(V) (feed concentration [ 100 ppb, pH [ 8, temperature [ 25 8C).
8.0. Also in this case, the flux linearly increased with the operating temperature for the all the As concentrations investigated. The N30F membrane showed a decrease in permeate flux by increasing the As feed concentration (Fig. 3b) as already observed for Fig. 1.
a Steady-state permeate flux (l/m2h)
80 Water 100 ppb 500 ppb 1000 ppb
70
60
50
40
30
10
15
20
25
30
35
40
45
Temperature (°C)
b Steady-state permeate flux (l/m2h)
60 Water 100 ppb 500 ppb 1000 ppb
55 50 45 40 35 30 25 20
10
15
20
25
30
35
40
45
Temperature (°C) Fig. 3 – Effect of the operating temperature on permeate flux at different As concentrations for the a) NF-90 membrane and b) N30F membrane (TMP [ 6 bar, pH [ 8).
Fig. 4 shows the As rejection for both membranes as a function of temperature. The NF-90 membrane showed a higher As rejection if compared with the N30F membrane in all the range of temperature investigated. In particular, for the NF-90 membrane, the arsenic removal was 95.4% and 93.1% at temperatures of 15 C and 40 C, respectively. For both membranes a decrease in As rejection was observed by increasing temperature: this phenomenon can be explained assuming an increase in the diffusivity of arsenic with temperature which determined consequently an increase of the diffusive transport of arsenic across the membrane. Brandhuber and Amy (2001) observed a similar behavior in the treatment of drinking water by charged UF membranes. For the NF-90 membrane the As concentration in the permeate was lower than both EPA and Bangladesh MCL in all the investigated range of temperature.
3.3. Effect of arsenic concentration on permeate flux and arsenic removal Fig. 5 shows the effect of the As feed concentration on the steady-state permeate flux for both investigated membranes at a TMP of 6 bar and a temperature of 25 C. Basically, the permeate flux was not influenced by the arsenic concentration in the feed solution. However, a slight decrease of permeate flux, by increasing the As feed concentration, was observed for the N30F membrane. In Fig. 6 the influence of the feed concentration on the As rejection for both membrane modules is shown. The NF-90 membrane showed an higher As rejection, if compared with the N30F membrane, in all the range of the investigated As feed concentration. In particular, the As rejection for the NF90 membrane was higher than 97% independently of the As feed concentration, while it was between 74 and 79% for the N30F membrane. For both membranes the As concentration in the permeate increased in the range of feed As concentration investigated. In the case of the NF-90 membrane, the As detected in the permeate was lower than the EPA recommended MCL up to a feed As concentration of about 600 ppb; the arsenic concentration in the permeate of the N30F membrane was 60
100 Bangladesh MCL
50
80
40
As permeate NF90 As permeate N30F Rejection N30F Rejection NF90
30
60 40
20 EPA MCL
10 0
10
15
20
25
30
35
40
45
Rejection (%)
100
Permeate As concentration (ppb)
60
Rejection (%)
Permeate As concentration (ppb)
water research 44 (2010) 97–104
20 0
Temperature (°C) Fig. 4 – Effect of temperature on the removal of As(V) (feed concentration [ 100 ppb, pH [ 8, TMP [ 6 bar).
102
water research 44 (2010) 97–104
100 NF90 N30F
60
40
20
0
0
200
400
600
800
1000
50 40 30 20
0
1200
As feed concentration (ppb)
Effect of pH on permeate flux and arsenic removal
100
400 350
80
300 permeate As NF90 permeate As N30F Rejection NF90 Rejection N30F
60 40
150 100 Bangladesh MCL
50 0
20
EPA MCL
0
200
400
600
800
1000
0 1200
As feed concentration (ppb) Fig. 6 – Effect of As feed concentration on the removal of As(V) (temperature [ 25 8C, pH [ 8, TMP [ 6 bar).
Rejection (%)
Permeate As concentration (ppb)
In Fig. 7 the effect of pH on the permeate flux for both membranes, at a feed As concentration of 500 ppb, a temperature of 25 C and a TMP of 6 bar, is depicted. The permeate flux, at steady-state, reaches the highest value at pH of about 8 for both membranes. At a higher and lower pH values the permeate flux decreases more significantly for the N30F than NF-90 membrane. Fig. 8 shows the effect of pH on the As rejection for both membranes. For the N30F membrane the As(V) rejection increased significantly by increasing pH: in particular, the As rejection increased from 74% to 88% with increasing pH; this phenomenon can be explained assuming that the monovalent ion H2AsO 4 is dominant in the range of pH 4–6 while the divalent ion HAsO2 4 is dominant above pH 7. Divalent ions are
200
2
4
6
8
10
12
14
Fig. 7 – Effect of pH on the permeate flux for NF90 and N30F membrane modules (feed concentration [ 500 ppb, temperature [ 25 8C, TMP [ 6 bar).
rejected by the N30F membrane at a much higher rate compared to monovalent ions due to large hydrated radii of divalent ions compared to monovalent ions (Vrijenhoek and Waypa, 2000; Brandhuber and Amy, 2001). Moreover PES membranes, such as N30F, are negatively charged at high pH and the effective charge density decreases at lower pH. The iso-electric point, defined as the pH for which the net charge of the membrane is equal to zero, is located around 3 (Weis et al., 2003). For the NF-90 membrane the As(V) rejection increased from 94% to 98.4% in the range of pH investigated (3.4–10). This membrane is negatively charged in the neutral pH region and, similar to the N30F membrane, it becomes more negative as the pH value increases: therefore charge exclusion strongly effects the rejection. Urase et al. (1998) observed a similar trend by using the ES-10 NF membrane (Nitto-Denko Co., Ltd.) also made of aromatic polyamide. The concentration of As(V) in the permeate of the NF-90 membrane was lower than the Bangladesh MCL in all the range of the investigated pH and lower than the EPA MCL at pH 10. For
Permeate As concentration (ppb)
higher than the EPA recommended MCL independently of the concentration of As in the feed. Consequently, the feed concentration has to be strongly considered when treating contaminated arsenic water.
250
0
pH
Fig. 5 – Effect of feed concentration on permeate flux for NF90 and N30F membrane modules (temperature [ 25 8C, pH [ 8, TMP [ 6 bar).
3.4.
NF90 N30F
10
100
200
80
150
60 100
permeate As N30F permeate As NF90 Rejection N30F Rejection NF90
40 Bangladesh MCL
50
20 EPA MCL
0
0
2
4
6
8
10
12
14
0
pH Fig. 8 – Effect of the pH on the removal of As(V) (feed concentration [ 500 ppb, temperature [ 25 8C, TMP [ 6 bar).
Rejection (%)
80
Jperm (l/m2h)
Steady-state permeate flux (l/m2h)
60
water research 44 (2010) 97–104
both membranes the highest arsenic removal, which occurred at high pH, corresponds to the lowest arsenic concentration in the permeate as also observed by Uddin et al. (2007).
4.
Conclusions
Arsenic removal from synthetic water, prepared starting from arsenic pentaoxide, was studied by using two commercial nanofiltration membranes (NF90 and N30F). For both membranes the removal efficiency for As(V) was influenced by the operating conditions such as temperature, transmembrane pressure, pH and feed water concentration. Particularly, the As rejection of the NF-90 membrane was higher if compared with the N30F membrane (above 91%) for all the operating conditions investigated. The As concentration in the permeate of the NF-90 membrane resulted always lower than the Bangladesh MCL (50 mg/l) while the EPA MCL (10 mg/l) was reached for initial feed As concentration in the range 100–600 ppb. As a common trend, it was observed that an increase of pH and a decrease of operating temperature and As feed concentration determined a higher efficiency of As removal for both membranes, whereas the TMP slightly affected the As rejection of the N30F membrane (it reduced at higher TMP). In both cases, the permeate flux increased with temperature and pressure and it had a maximum value at a pH of about 8. On the basis of the experimental results, NF can be considered a viable approach to remove As(V) from contaminated water. However, the As feed concentration has to be strongly considered in order to produce a permeate stream containing an As concentration within the allowed limits.
Acknowledgements This work was carried out within the Asia Pro Eco Program ‘‘Technology partnership for innovative treatment of drinking and industrial water’’ (INNOWA) (BD Asia Pro Eco/07/96638) supported by the European Commission in the 6th Framework Programme.
references
Amy, G.L., Edwards, M., Benjamin, M., Carlson K., Chwirka, J., Bradhuber, P., McNeill, L., Vagliasindi, F., 1998. Draft report, AWWARF. Bhattacharya, P., Chattargee, D., Jacks, G., 1997. Occurrence of Arsenic-contaminated groundwater in alluvial aquifers from delta plains, Eastern India: options for safe drinking water supply. International Journal of Water Resources Development 13, 79–92. Bhattacharyya, R., Chatterjee, D., Nath, B., Jana, J., Jacks, G., Vahter, M., 2003. High arsenic groundwater: mobilization, metabolism, and mitigation – an overview in the Bengal delta plain. Molecular and Cellular Biochemistry 253, 347–355. Brandhuber, P., Amy, G., 2001. Arsenic removal by charged ultrafiltration membrane – influences of membrane operating conditions and water quality on arsenic rejection. Desalination 140, 1–14.
103
Caceres, D.D., Pino, P., Montesinos, N., Atalah, E., Amigo, H., Loomis, D., 2005. Exposure to inorganic arsenic in drinking water and total urinary arsenic concentration in Chilean population. Environmental Research 98, 151–159. Drioli, E., Criscuoli, A., Curcio, E., 2002. Integrated membrane operations for seawater. Desalination 147, 77–81. Drioli, E., Lagana`, F., Criscuoli, A., Barbieri, G., 1999. Integrated membrane operations in desalination processes. Desalination 122, 141–145. Floch, J., Hideg, M., 2004. Application of ZW-1000 membranes for arsenic removal from water sources. Desalination 162, 75–83. Geucke, T., Deowan, S.A., Hoinkis, J., Pa¨tzold, Ch, 2009. Performance of a small-scale RO desalinator for arsenic removal. Desalination 239, 198–206. Han, B., Runnels, T., Zimbron, J., Wickramasinghe, 2002. Arsenic removal from drinking water by flocculation and microfiltration. Desalination 145, 293–298. Kosˇutic´, K., Furacˇ, L., Sipos, L., Kunst, B., 2005. Removal of arsenic and pesticides from drinking water by nanofiltration membranes. Separation and Purification Technology 42, 137–144. Krieg, H.M., Modise, S.J., Keizei, K., Neomagus, H.W.J.P., 2004. Salt rejection in nanofiltration for single and binary salt mixtures in view of sulphate removal. Desalination 171, 205–215. Lhassani, A., Rumeau, M., Benjelloun, D., Pontie, M., 2001. Selective demineralization of water by nanofiltration. Application to the defluorination of brackish water. Water Research 35, 3260–3264. Mandal, B.K., Suzuki, K.T., 2002. Arsenic round the world: a review. Talanta 58, 201–235. Meng, X., Korfiatis, P.G., Christodoulatos, C., Bang, S., 2001. Treatment of arsenic in Bangladesh well water using a household co-precipitation and filtration system. Water Research 35, 2805–2810. Mohan, D., Pittman Jr., C.U., 2007. Arsenic removal from water/ wastewater using adsorbents-a critical review. Journal of Hazardous Materials 142, 1–53. National Research Council, 2001. Arsenic in Drinking Water. National Academy of Sciences, Washington, DC, USA. Ng, K.S., Ujang, Z., Le-Clech, P., 2004. Arsenic removal technologies for drinking water treatment. Reviews in Environmental Science and Biotechnology 3, 43–53. Oh, J.I., Yamamoto, K., Kitawaki, H., Nakao, S., Sugawara, T., Rahman, M.M., Rahman, M.H., 2000. Application of lowpressure nanofiltration coupled with a bicycle pump for the treatment of arsenic-contaminated groundwater. Desalination 132, 307–314. Saitu´a, H., Campderro´s, M., Cerutti, S., Padilla, A.P., 2005. Effect of operating conditions in removal of arsenic from water by nanofiltration membrane. Desalination 172, 173–180. Sato, Y., Kang, M., Kamei, T., Magara, Y., 2002. Performance of nanofiltration for arsenic removal. Water Research 36, 3371–3377. Shih, M.C., 2005. An overview of arsenic removal by pressuredriven membrane processes. Desalination 172, 85–97. Smedley, P.L., Kinniburgh, D.G., 2002. A review of the source, behaviour and distribution of arsenic in natural waters. Applied Geochemistry 17, 517–568. Uddin, M.T., Mozumder, M.S.I., Islam, M.A., Deowan, S.A., Hoinkis, J., 2007. Nanofiltration membrane process for the removal of arsenic from drinking water. Chemical Engineering Technology 30, 1248–1254. Uddin, M.T., Mozumder, S.I., Figoli, A., Drioli, E., Islam, M.A., 2008. Arsenic removal by conventional and membrane technology: an overview. Indian Journal of Chemical Technology 15, 441–450.
104
water research 44 (2010) 97–104
Urase, T., Oh, J.I., Yamamoto, K., 1998. Effect of pH on rejection of different species of arsenic by nanofiltration. Desalination 117, 11–18. US Environmental Protection Agency, 2001. Washington DC, USA. Panel 14: National Primary Drinking Water Regulations: Arsenic and Clarifications to Compliance and New Source Contaminants Monitoring, vol. 66 194. Velizarov, S., Crespo, J., Reis, M., 2004. Removal of inorganic anions from drinking water supplies by membrane bio/ processes. Reviews in Environmental Science and Biotechnology 3, 361–380.
Vrijenhoek, E.M., Waypa, J.J., 2000. Arsenic removal from drinking water by a ‘‘loose’’ nanofiltration membrane. Desalination 130, 265–277. Weis, A., Bird, M.R., Nystrom, M., 2003. The chemical cleaning of polymeric UF membranes fouled with spent liquor over multiple operational cycles. Journal of Membrane Science 216, 67–79. WHO, 1998. Guidelines for Drinking-Water Quality, Addendum to vol. 1, Recommendations Geneva. Zaw, M., Emett, M., 2002. Arsenic removal from water using advanced oxidation processes. Toxicology Letters 133, 113–118.
water research 44 (2010) 105–114
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Halonitromethane formation potentials in drinking waters Jia Hu a, Hocheol Song b,*, Jesse W. Addison a, Tanju Karanfil a,** a b
Department of Environmental Engineering and Earth Sciences, Clemson University, Anderson, SC 29625, USA Korea Institute of Geoscience and Mineral Resources, Daejeon 305-350, Korea
article info
abstract
Article history:
Halonitromethanes (HNMs) are highly cyto- and genotoxic nitrogenous disinfection by-
Received 16 May 2009
products (DBPs) that have been detected in some water distribution systems. In this study,
Received in revised form
a systematic investigation was conducted to examine the formation potential of HNMs in
1 September 2009
drinking waters under different oxidation conditions. Formation potential tests of samples
Accepted 2 September 2009
obtained from various drinking water sources showed that ozonation–chlorination
Published online 8 September 2009
produced the highest HNM yields followed by in the order of chlorination, ozonation– chloramination, and chloramination. Similar or higher HNM yields were observed in the
Keywords:
treated waters (i.e., after conventional water treatment) than in the raw waters, indicating
Halonitromethanes
that hydrophilic natural organic matter (NOM) components that are not effectively
Emerging DBPs
removed by conventional treatment processes are likely the main precursors of HNMs.
Nitrogenous DBPs
This was further confirmed by examining HNM formation potentials of NOM fractions
Formation potential
obtained with resin fractionation. Hydrophilic NOM fractions (HPI) showed significantly
Ozone disinfection
higher HNM yields than hydrophobic (HPO) and transphilic (TPH) fractions. The correlation
Chlorination
analysis of HNM formation potentials during ozonation–chlorination with various water
Natural organic matter
quality parameters showed the best correlation between the HNM yields and the ratio of
Drinking water
dissolved organic carbon to dissolved organic nitrogen concentrations in the water samples tested. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Over the last 30 years, significant amount of research efforts have been directed towards improving our understanding of disinfection by-products (DBPs), and to date more than 600 DBPs have been identified in drinking waters (Richardson, 2003). However, only a small fraction of those DBPs are currently regulated. The regulated organic DBPs in the United States (US) constitute about 30–60% of the overall total organic halogens formed in water (Richardson et al., 2002; Karanfil et al., 2008). In a nationwide occurrence study funded by the United States Environmental Protection Agency (USEPA), approximately 50
unregulated DBPs that have the potential to cause high human health risks (i.e. high priority DBPs) were selected and monitored in drinking waters across the US in 2002 (Krasner et al., 2006). These high priority DBPs included halonitromethanes (HNMs), iodo-trihalomethanes (I-THMs), haloacetonitriles, haloketones, haloamines and analogues of 3-chloro-4-(dichloromethyl)-5hydroxyl-2(5H)-furanone (MX). Among these DBPs, HNMs received special attention because of their potential high toxicity and their occurrence in finished waters at some treatment facilities. Although HNM concentrations were orders of magnitude lower than those of regulated trihalomethanes (THMs) and haloacetic acids (HAAs) in the US, the recent toxicology studies conducted on emerging DBPs showed that HNMs are one of
* Corresponding author. Tel.: þ82 42 868 3373; fax: þ82 42 868 3414. ** Corresponding author. Tel.: þ1 864 656 1005; fax: þ1 864 656 0672. E-mail addresses:
[email protected] (H. Song),
[email protected] (T. Karanfil). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.006
106
water research 44 (2010) 105–114
the most cyto- and genotoxic classes among the emerging DBPs, having orders of magnitude higher toxicity than THMs and HAAs (Plewa et al., 2004, 2008). There are a total of 9 species of chorine and bromine substituted HNMs including chloro- (CNM), dichloro- (DCNM), trichloro- (TCNM), bromo- (BNM), dibromo- (DBNM), tribromo- (TBNM), bromochloro- (BCNM), bromodichloro(BDCNM), and dibromochloronitromethane (DBCNM). The presence of TCNM in drinking water was first realized in late 1970s and early 1980s (Coleman et al., 1976; Becke et al., 1984; Merlet et al., 1985; Hoigne and Bader, 1988). Later, the other eight remaining species including mono- and di- chlorine and/or bromine substituted HNMs were detected in waters treated with ozone-chlorine, chlorine, and chloramines (Thibaud et al., 1988; Krasner et al., 1989, 1991, 2006; Richardson et al, 1999; Plewa et al., 2004). One of the most notable findings in those studies was that HNM formation substantially increased when ozonation was used prior to chlorination. For example, Hoigne and Bader (1988) reported that TCNM formation increased by 4–5 times when a lake water was pre-ozonated before chlorination. Various hypotheses have been proposed regarding the role of ozone (Merlet et al., 1985; von Gunten, 2003; Choi and Richardson, 2004), but none of them have been experimentally verified and the exact role of ozone in the formation of HNMs still remains unresolved. Despite the increasing body of literature on HNMs, systematic investigations with a whole array of HNM species have been rarely reported due to the lack of commercial chemical standards of all the species, which became available in early 2000’s. Therefore, formation and speciation characteristics of HNMs at varying environmental conditions, information about their precursors, and factors controlling their formation are mostly unknown. For example, the dissolved organic carbon (DOC) concentrations and the characteristics of DOC (e.g., hydrophobic/ hydrophilic nature of organic matter) have been linked to some extent to the formation and speciation of regulated THMs and HAAs (Reckhow et al., 1990; Croue´ et al., 2000; Kitis et al., 2002), whereas formation of nitrogenous DBPs such as N-nitrosodimethylamine (NDMA) and dichloroacetonitrile (DCAN) has been linked to the concentrations and composition of dissolved organic nitrogen (DON) in water (Lee et al., 2007). For HNMs, there is only one study that assessed the role of organic nitrogen in TCNM formation (Lee et al., 2007). The objectives of this study were to (i) examine the formation potentials (i.e., total amount of HNM precursors available) of HNMs in drinking waters with varying characteristics under different oxidation conditions commonly used for water treatment, (ii) identify the potential precursors of HNMs in natural waters by running experiments with fractionated natural organic matter (NOM), and (iii) examine the capabilities of various water quality parameters in predicting HNM formation potential in a water sample. Unlike the most of the previous researches, the samples were analyzed for all nine HNM species, thus formation potential of nine HNM species was investigated in this study.
2.
Experimental section
2.1.
Water samples
Water samples were collected from five drinking water treatment plants (DWTPs), Greenvillle (GV), Spartanburg (SP), Startex-Jackson-Wellford-Duncan (SJWD), Charleston (CH), and Myrtle Beach (MB), in South Carolina in the US. The sampling was performed three times between February and July 2007 for all the DWTPs except MB which was sampled twice. Samples were collected from the influents of the plants (i.e., raw water) and after conventional treatment processes before any disinfectant addition (i.e., treated water). Samples were filtered with pre-washed 0.2 mm Supor membrane filters to eliminate the particles and biological activity immediately after arrival at the laboratory, and stored in a dark constant temperature room (4 C) until the experiments were performed. Formation potentials of individual NOM fractions were examined for two sets of NOM fractions. The first set consisted of hydrophobic (HPO) and transphilic (TPH) fractions obtained from MB, CH, and SP waters using XAD-8 (superlite, DAX-8, Supelco) and XAD-4 (amberlite, Supelco) resin columns in sequence as described in a previous study (Karanfil et al., 2007; Song et al, 2009). Second set of NOM fractionations was obtained by using batch reactors and excess resin dose, determined through preliminary experiments, of 10 g/L at pH 2 for MB and CH raw waters. The water sample was mixed with the resin in a bottle on a shaker table for 7 days. After the XAD-8 fractionation, a portion of the supernatant was collected (TPH þ HPI) and the remaining NOM was further fractionated for another week in the same batch mode using XAD-4 to obtain HPI fraction. The resins used in this study were extensively cleaned prior to use. The DOC leaching from resins (i.e., controls containing deionized distilled water (DDW) at pH 2 and XAD-4 or XAD-8 resins) during batch fractionation period was about 0.3 mg/L. HNM formation tests were also conducted with these control solutions during chlorination and ozonation–chlorination. There was no HNM formation during chlorination, whereas 3 mg/L of TCNM formation was observed for ozonation–chlorination, indicating that small amount of HNM precursors has been leaching from resins. The HNM yields of HPI and TPI þ HPI fractions during ozonation–chlorination were corrected to account for this small amount of leaching effect. The pHs of the fractions were readjusted to 7 immediately after the fractionation.
2.2.
HNM formation potential (FP) tests
FP tests were designed to determine the extent of HNM formation, which is also a measure of the amount of HNM precursors in a sample, under the excess amount of oxidant for five different scenarios: ozonation, chlorination, ozonation–chlorination, chloramination, and ozonation–chloramination. The dosage of chlorine (Cl2) was determined using the formula approach developed by Krasner and coworkers studying the presence of DBP precursors in treated
water research 44 (2010) 105–114
wastewaters, reclaimed water and drinking waters from various sources with different compositions (Krasner et al., 2008, 2009). The following formula was used for chlorination:
Cl2 (mg/L) ¼ 3*[mg/L dissolved organic carbon (DOC)] þ 8*[mg/L NH3-N] þ 5*[mg/L NO 2 -N] þ 10 mg/L which expresses chlorine demand to oxidize organic carbon, ammonia (2NH3 þ 3 Cl2 / N2 þ 6Hþ þ 6Cl, 7.6 mg/L Cl2 per 1 mg/L NH3-N), and nitrite (NO 2 þ HOCl / þ NO 3 þ H þ Cl , 5 mg/L Cl2 per 1 mg/L NO2 -N). For chloramination FP tests, a monochloramine (NH2Cl) stock solution was prepared by mixing sodium hypochlorite (5–6% available free chlorine) in an ammonium sulfate solution at a Cl2/N mass ratio of 3.5:1 (0.69:1 molar ratio) and pH 9. Preformed NH2Cl dose used in the experiments was determined with the following formula:
NH2Cl (mg/L) ¼ 3*[mg/L DOC] þ 5*[mg/L NO2-N] which expresses chlorine demand to oxidize organic carbon þ and nitrite (NO 2 þ NH2Cl þ H2O / NH4 þ NO3 þ Cl , 5 mg -N). Ammonia is not included in NH2Cl/L as Cl2 per 1 mg/L NO 2 the formula since NH2Cl does not oxidize ammonia. Ozone dose was equal to DOC of the samples (i.e., 1:1 ratio) because this is a typical ratio used in ozonation during water treatment. This formula based approach allowed a consistent oxidant dosing strategy for waters with varying DOC, ammonia and nitrite concentrations, and always resulted in a positive residual at the end of the FP tests. Each reactor initially received a stir bar and was completely filled with the test water. Then, a pre-calculated volume of the water was removed from each reactor, with the volume removed being equal to the volume of the ozone stock solution to be added for ozonation. For the reactors involving only chlorination or chloramination, the removed volume was filled with DDW to yield the same DOC concentration as in the ozonated reactors. Ozonation of the sample was achieved by adding varying amount of ozone stock solution (z30 mg/L) to produce the desired ozone concentration. Ozone was produced using a GTC-1B ozone generator (Griffin Technics Inc.) fed with ultra-high purity oxygen. After the application of ozone, the reactors were mixed for 5 min prior to the addition of chlorine or chloramine. Ozone concentrations were measured after 5 min contact time to assure that there is ozone residual and ozone was not a limiting factor during the pre-oxidation period. Chlorination and chloramination of samples were accomplished by spiking varying amount of Cl2 (z1600 mg/L) and preformed NH2Cl (z800 mg/L) stock solutions, respectively, to achieve the desired concentration. The bottles were incubated in a water bath (22 C) and the reactions were allowed to occur for 24 h except those involved in chloramination, which were reacted for 72 h. For each oxidation scenario, duplicate reactors were prepared. The pH during the experiments remained in the range of 7–8. Additional details about the experimental procedures used in the study are reported elsewhere (Hu, 2009).
2.3.
107
Analytical methods
HNMs were measured using USEPA Method 551.1 with minor modifications. A 10 mL sample was extracted using 10 mL of methyl tert-butyl ether (MTBE, Sigma), 3 g of sodium sulfate and 1 g of cupric sulfate. The samples were then placed on a shaker table at 300 rpm for 30 min. The MTBE extract was analyzed with a HP 6850 gas chromatograph (GC) equipped with a DB-5 column (J&W Scientific, 30 m, 0.25 mm, 1.8 mm) and an electron capture detector (ECD). DB-5 column was used as the primary column, while DB-1 column was employed when necessary to resolve co-elution of target HNM with other peaks that may occur in DB-5 column. HNM standards were obtained from Orchid Cellmark (New Westminster, Canada, CNM 93.6%, DCNM 99þ%, BCNM 91.9%, BDCNM 93.9%, DBNM 91.4%, DBCNM 94.1%, TBNM 99þ%) and Sigma (TCNM 99þ%, BNM 99þ%). The GC temperature program used was 35 C for 6 min, 30 C/min to 190 C and hold for 1.5 min. The sample (2 mL) was injected in splitless mode. The carrier and make-up gases were ultra-high purity (UHP) helium at 2.3 mL/min and UHP nitrogen at 60 mL/min, respectively. The injector temperature was set at 117 C in order to minimize the thermal decomposition of HNM species (Chen et al., 2002), and the detector temperature was set at 297 C. Minimum reporting levels (MRLs) for HNMs were determined to be 0.7 mg/L. Bromide, nitrite and nitrate were measured using a Dionex DX-600 ion chromatography equipped with AS-HC9 and AGHC9 columns. Ammonia was measured using salicylate method. Ozone concentration was measured using Indigo method that involved a HACH DR/820 colorimeter. DOC and dissolved nitrogen (DN) were measured using a Shimadzu TOC-VCHS analyzer. DON was determined by subtracting NO 3, þ and NH from DN. NO 2 4
3.
Results and discussion
3.1.
HNM formation potentials in drinking waters
Selected characteristics of the test waters are shown in Table 1. The water characterization results showed that natural waters used in the HNM formation potential tests covered a wide range of DOC (0.6–8.7 mg/L), DON (<MRL0.50 mg/L), DOC/DON ratios (5–41 mg/mg), and specific ultraviolet absorbance (SUVA254, 0.9–4.5 L/mg-m) values. The HNM formation in molar yields (nM HNM/mg DOC) categorized by oxidation scenarios are presented in Box-and-Whisker format in Fig. 1. The results are also provided in tabular format in Table 1. The HNM yields were in the order of ozonation– chlorination >> chlorination ozonation–chloramination > chloramination. For the most reactive ozonation–chlorination condition, the HNM yields ranged from 7.5 to 30.9 nM/mg and 12.4 to 39.5 nM/mg, with the mean value of 17.7 and 24.4 nM/ mg for raw and treated waters, respectively. TCNM and BDCNM were the main two HNM species measured during the FP tests at levels above their MRLs. This is because the waters tested in this study had, in general, low bromide levels, and the high dose of chlorine used in the FP tests (high Cl2/Br ratio) suppressed the bromine incorporation. Trace amounts of
108
water research 44 (2010) 105–114
Table 1 – Selected characteristics and HNM formation potential test results of raw and treated drinking waters. Collection date
Sample
Feb. MB R Feb. MB T Jul. MB R Jul. MB T Feb. CH R Feb. CH T Jun. CH R Jun. CH T Jul. CH R Jul. CH T Mar. SP R Mar. SP T May SP R May SP T Jul. SP R Jul. SP T Mar. SJWD R Mar. SJWD T May SJWD R May SJWD T Jul. SJWD R Jul. SJWD T Mar. GV R Mar. GV T May GV T Jul. GV R Jul. GV T b (R) n ¼ 13 samples b (T) n ¼ 14 samples b Overall n ¼ 27 samples
SUVA254a (L/mg-m)
DOCa (mg/L)
4.4 2.0 4.0 2.0 4.0 2.0 4.5 2.9 3.6 2.0 3.4 1.1 1.8 0.9 2.0 0.9 3.5 1.4 1.8 1.4 2.8 2.1 1.9 1.4 1.8 1.6 1.3
8.7 4.0 7.5 3.5 5.5 2.7 3.3 1.4 4.1 2.3 1.8 1.2 2.2 1.2 1.8 1.3 1.7 1.3 2.8 1.9 1.7 1.4 1.0 0.7 0.6 1.2 0.9
DONa (mg/L) 0.21 0.14 0.38 0.22 0.18 0.10 0.18 <MRL 0.19 0.13 <MRL <MRL 0.19 0.13 0.18 <MRL <MRL <MRL 0.45 0.27 <MRL 0.15 <MRL <MRL 0.13 <MRL <MRL
DOC/DONa (mg/mg)
Bra (mg/L)
41 28 20 16 31 27 18 N/C 22 18 N/C N/C 12 9 10 N/C N/C N/C 6 7 N/C 10 N/C N/C 5 N/C N/C
14 31 49 53 54 59 37 39 44 45 13 13 <MRL <MRL 15 <MRL 14 16 14 13 21 20 <MRL <MRL <MRL 13 12
HNM (nM/mg DOC) O3
Cl2
O3-Cl2
NH2Cl
O3-NH2Cl
<MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL
1.5 2.9 3.0 3.0 2.8 4.7 4.2 6.6 3.8 5.1 3.9 5.6 2.5 5.8 2.8 5.5 6.3 3.7 4.8 6.3 5.5 6.0 5.6 9.3 9.5 3.8 6.0 3.9 1.4 5.7 1.9 4.8 1.9
7.6 12.4 9.1 18.0 9.9 19 11.7 17.6 15.1 17.0 20.8 33.8 19.6 22.7 31.1 29.1 14.5 18.6 20.2 35.5 19.6 18.7 26.2 34.9 38.4 26.5 25.3 17.7 7.1 24.4 8.6 21.2 8.5
1.2 <MRL 0.7 2.5 0.7 <MRL 1.2 <MRL 0.9 2.2 2.3 <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL <MRL 0.5 0.7 0.6 1.3 0.6 1.0
1.9 2.5 3.1 4.0 3.3 4.6 3.8 6.9 5.2 5.1 <MRL 2.9 2.6 <MRL 6.4 3.4 2.5 2.7 3.3 10.4 3.1 2.5 3.3 5.4 <MRL 5.3 <MRL 3.4 1.5 3.6 2.9 3.5 2.3
a The values reported in the table are for the waters used in the experiments, accounting for the dilution effects due to spiking of the samples with ozone, chlorine or chloramine solutions during formation potential tests. b The values show the average standard deviation for each parameter. MB: Myrtle Beach, CH: Charleston, SP: Spartanburg, SJWD: Startex, Jackson, Wellford & Duncan, GV: Greenville, R: Raw water, T: Treated water, N/C: Not calculated due to very low level DON. MRL: Minimum reporting level. Therefore, zero was used in the average and standard deviation calculations for samples with MRLs.
other HNM species, mainly DCNM and TBNM, were also detected around the MRL in a few samples. The formation potentials were greater in the low SUVA254 waters (GV, SJWD, SP) than the high SUVA254 waters (CH, MB). The effect of ozone is in agreement with the previous observations that ozonation substantially enhanced HNM formation when combined with chlorination (Merlet et al., 1985; Hoigne and Bader, 1988; Choi and Richardson, 2004; Krasner et al., 2006). Ozonation–chloramination produced significantly less amount of HNMs, sometimes at levels below the MRLs. Chloramination alone resulted in minimal or no formation HNMs, suggesting the oxidation potential of chloramine is not high enough to induce HNM formation, similar to THM formation. Ozonation alone did not form any HNMs; this was expected since there was no chlorination agent and the bromide levels of the waters were usually low or below detection limit. Overall, the results indicate that the use of NH2Cl, alone or after ozonation, significantly reduces the formation of HNMs and regulated THMs and HAAs, as reported in other studies (e.g., Hong et al., 2007, 2008). For a given pair of raw and treated water from the same DWTP, the HNM yield was, in general, greater for treated
waters during ozonation–chlorination and chlorination (Table 1). The greater formation of HNMs in treated waters indicates that the precursors with higher reactivity toward HNM formation were not greatly removed during conventional treatment. It is well-established that HPO and TPH fractions of NOM are preferentially removed during conventional treatment processes (e.g., Kim and Yu, 2005). If such fractions were the major HNM precursors, the treated waters would exhibit lower HNM yields as compared to the raw water samples; however, the opposite was observed in this study. Therefore, these results suggest that the hydrophilic NOM components (i.e., HPI fraction) remaining in water after conventional treatment likely constitute the main precursors of HNMs. Since higher HNM yields were observed, in general, in treated than raw water, this also suggests that HPO and TPH fractions can compete with HPI fraction during ozonation– chlorination and chlorination, probably by forming other DBPs. The reduction in HNM formation potentials are plotted as a function of percent DOC and DON removals by the conventional water treatment processes for five sets of waters that the measured DON values were above the MRL (i.e.,
109
water research 44 (2010) 105–114
40
40
A
15
2C l
40
40
15
O 3N
C
l2
H2 Cl N O 3-
l H2 C N
O 3-
C
Cl 2
0 l2
5
0
H 2C l
10
5
l
10
20
H 2C
15
25
N
20
30
l2
25
D
35
O 3C
30
HNM/DOC (nM/mg)
C
35 HNM/DOC (nM/mg)
H O 3N
O
3N
H2
C l2
l H2 C N
O
C
Cl
0 3C l2
5
0
l
10
5
H2 C
10
20
N
15
25
l2
20
30
O 3C
25
B
35 HNM/DOC (nM/mg)
30
l2
HNM/DOC (nM/mg)
35
Fig. 1 – HNM yields on a DOC basis during different oxidation processes of (A) raw water, (B) treated (i.e., after coagulation, flocculation and sedimentation) water, (C) HPO fractions, and (D) TPH fractions. (Top and bottom of the box are 75th and 25th percentiles, respectively. Top and bottom of the whiskers are 90th and 10th percentiles, respectively. Line in the box shows the median (50th percentile), while the diamond and asterisk are to show the average and outlier, respectively).
DON > 0.1 mg/L) (Fig. 2). In both plots, the data showed a very weak correlation with the y ¼ x line, indicating that DOC or DON removal alone is not an appropriate descriptor to predict the reduction in HNM formation potential. The fact that the majority of the data are located below the line may be viewed as another indication of the importance of hydrophilic NOM components in HNM formation since DOC removal as a result of conventional treatment is mainly due to preferential removal of HPO and TPH fractions. The FP data of individual NOM fractions that were available from a previous project showed similar reactivity patterns with the oxidants tested as compared to the raw or treated waters. However, HNM yields of HPO and TPH fractions were, in general, lower than those of raw and treated water samples (Fig. 1). Lower HNM yields for the fractions as compared to the raw and treated waters suggest that the NOM pool in the HPI fraction (i.e., the missing fraction of NOM in the tested isolates) contains the important precursors of HNMs. In order to verify the hypothesis that HPI fraction contains more reactive precursors of HNMs, additional FP tests were performed with mixed (TPH þ HPI) and HPI fractions of MB and CH raw waters used in this study. Table 2 shows the HNM yields of the raw waters, TPH þ HPI fractions, and HPI fractions obtained from chlorination and ozonation–chlorination FP tests. For chlorination, the results showed that although HNM yields were relatively small, the HNM formation was in the order of HPI > TPH þ HPI > raw water. Substantially higher formation of HNMs was observed for the fractions of both waters when treated with ozonation–chlorination for which
the differences in the yields were more significant. For CH and MB waters, there was approximately 2.5 and 4 times increase in HNM formation for the TPH þ HPI and HPI fractions, respectively, compared to the raw waters. These findings agree with the hypothesis that the likely HNM precursors are some organic moieties with low-molecular weight and hydrophilic characteristics that tend to persist during water treatment processes. The importance of such precursors in HNM formation contrasts to their less significant role in the formation of THMs and HAAs, for which high-molecular hydrophobic moieties are presumed to be more important precursors (Reckhow et al., 1990; Korshin et al., 1997; Kitis et al., 2002).
3.2.
Correlation analyses of HNM formation
Correlations between HNM yields and various parameters (e.g., DOC, DON, DOC/DON ratios, SUVA254) were examined for ozonation–chlorination, the most HNM yielding oxidation conditions, to gain further insight to the potential HNM precursors in natural waters and to assess the prediction capabilities of commonly used water quality parameters in water treatment. HNM concentrations showed an increasing trend with DOC concentrations of raw and treated waters (Fig. S1 in Supplementary Materials). The results of NOM fractions (HPO and TPH) exhibited a wide range of variability in HNM formation at almost the same DOC concentration used in the FP experiments. HNM concentrations showed no clear trend with DON
110
water research 44 (2010) 105–114
100 O3-Cl2
x=y
60
Chlorination
CH raw CH TPH þ HPI CH HPI MB raw MB TPH þ HPI MB HPI Ozonation– CH raw chlorination CH TPH þ HPI CH HPI MB raw MB TPH þ HPI MB HPI
4.14 2.02 1.22 5.60 2.52 1.44 4.14 2.02 1.22 5.60 2.52 1.44
3.6 1.6 1.2 4.0 2.0 1.4 3.6 1.6 1.2 4.0 2.0 1.4
4.2 6.2 10.5 3.3 6.4 9.8 14.8 36.0 59.9 10.1 26.2 36.8
CH: Charleston, MB: Myrtle Beach a The values represent the average of two independent samples.
40 40
A
35
20
0 0
20
40
60
80
100
Percent DOC Removal 100 O3-Cl2
30 y = 27.707e-0.0342x 2 R = 0.6887
25 20 15 10 5
B
Cl2
0
80
0
5
10
15
x=y
20
25
30
35
40
45
DOC/DON (mg/mg)
40
60
B
35
HNM/DOC (nM/mg DOC)
Percent Reduction in HNM Formation
TOC SUVA254 HNM yielda (mg/L) (L/mg-m) (nM/mg DOC)
SP, GV) under chlorine-ozonation condition (Table 1). This suggests that there are certain types of NOM components, probably nitrogenous in character, specifically reacting with ozone-chlorine and exhibiting high yields of HNMs, even at very low levels. This may also explain the relatively large variations in HNM yields observed for ozonation–chlorination cases (Fig. 1).
A
Cl2
80
Table 2 – HNM formation of Charleston and Myrtle Beach water fractions.
HNM/DOC (nM/mg DOC)
Percent Reduction in HNM Formation
concentrations in raw and treated waters, and NOM fractions (Fig. S2 in Supplementary Materials). These observations indicated that both the type and concentrations of organic precursors (e.g., C/N ratio, the structure of (nitrogenous) organic compounds) are important in HNM formation. Since both organic carbon and nitrogen contents in natural waters were expected to be important for HNM formation, the HNM yields were examined with respect to the DOC/DON ratios of the samples (Fig. 3). In the plot, only the data for which DON value is greater than MRL (0.1 mg/L) were included. For raw and treated waters, HNM yields increased with decreasing DOC/DON ratios (i.e., increasing nitrogen content per organic carbon in water). In fact, these were the best correlations among all parameters tested in this study. However, it should be noted that, some of the waters excluded from the plot led to appreciable formation of HNMs despite the low level of DON, especially for low SUVA254 waters (e.g.
40
20
-0.5167
y = 78.554x 2 R = 0.7653
30 25 20 15 10
0 0.0
20.0
40.0
60.0
80.0
100.0
Percent DON Removal Fig. 2 – Percent reduction in HNM formation as a function of percent (A) DOC and (B) DON removals during conventional water treatment.
5 0 0
5
10
15
20
25
30
35
40
45
DOC/DON (mg/mg)
Fig. 3 – HNM yield as a function of DOC/DON ratio during ozonation–chlorination for (A) raw and (B) treated waters.
111
water research 44 (2010) 105–114
H
H C
C
C
H
N
H
H HOCl
O
O
O HOCl
O
C
C OH
OH
H
H
C
H
N H
H
C
C
H
N
O
C
C OH
H2O OH
(1)
C
H O
H2O
(2)
C OH
OH
Cl
Cl
Cl
(3) CO2 HCl
H
C
H
(6)
O
H
C
H
(5)
H C
C
Cl
H
H
HOCl H
(4)
C
C
C
HCl
N
N
H Cl
OH
Cl
OH
OH
CO2
N
H
H
C
C
H
N
Cl
Cl
HOCl
H O
(7) (8) H
OH
H
C
C
C
Cl
(9) H
C
H
Cl
Cl
OH
Cl
C
C
H
N
Cl
N
OH-
N
H
HOCl
Cl
H
HOCl
OH
Cl
OH
OH
(10) (12)
Cl
CH2O
(11)
Cl
Cl
HOCl C
Cl
Cl
C
C
H
Cl
Cl
HCl N
N
N OH
OH
Cl
OH
OH
(13) dehydrogenation
Cl C
Cl
N
O
Cl
+
O
Fig. 4 – A hypothesized pathway of formation of TCNM from chlorination of aspartic acid.
There was no clear pattern between the HNM yields of NOM fractions and DOC/DON ratios. This was in agreement with Lee et al. (2007) who reported no correlation between TCNM yields and DOC/DON ratios for several HPO and TPH fractions during formation chlorination and chloramination potential tests. They also reported that the TCNM yields averaged 2 nM/mg DOC during chlorination and chloramination. In this study, the HNM yields of NOM fractions were 2.8 and 1.2 nM/mg DOC for chlorination and chloramination, respectively. These results further indicate that HPO and TPH fractions make some contribution to HNM formation but they are not the main precursors of HNMs.
For raw waters, HNM formation showed increasing trend with decreasing SUVA254. However, for treated waters and NOM fractions, correlations between HNM yields and SUVA254 did not exhibit a clear correlation (Fig. S3 in Supplementary Materials). It appears that SUVA254 is not a sensitive parameter with respect to HNM formation.
3.3.
Mechanistic pathways of HNM formation
The results of formation potential tests provided insights to the potential precursors of HNMs in natural waters: 1) treated waters exhibited similar or higher reactivity to form HNMs
112
water research 44 (2010) 105–114
than raw waters, 2) low SUVA254 waters yielded higher amounts of HNMs per DOC as compared to high SUVA254 waters, 3) increasing organic nitrogen content of NOM per DOC increased HNM yield, and 4) HPO and THP fractions formed much less HNMs than HPI fractions. Overall, these observations lead to a hypothesis that HPI components of NOM, especially nitrogenous organic compounds, are likely the major precursors of HNMs. Among many types of nitrogenous organic compounds, amino acids constitute an important class of organic nitrogen in drinking water sources. The presence of amino acids in raw and treated waters has been found to exert high chlorine demand (Trehy et al., 1986; Hureiki et al., 1994; Na and Olson, 2007). The chlorine reactivity of amino acids depends on the side chain groups attached to the a-carbon. It has been shown that amino acids containing alkyl groups (e.g. alanine, valine, isoleucine, aspartic acid) exhibit similar chlorine consumption, while those containing thiol group (e.g. methionine, cystein) had high reactivity to chlorine (Na and Olson, 2007). Glycine (H functional group) and proline (alkyl groups with a cyclic structure) showed the least reactivity. The reaction of amino acids and chlorine has been shown to produce DBPs such as haloacetaldehydes, haloacetonitriles and cyanogen chloride. Chlorination of amino acids initially proceeds via chlorination of amino group to form N-chloroamino acids and N,N-dichloroamino acids, which undergo a series of reactions to produce DBPs. To date, the pathways of HNM formation from amino acids have not been reported. However, a similar analogy may be used to deduce the formation pathway of HNM from chlorination of amino acids. Assuming that the initial step of reaction involves chlorination of amino group followed by concomitant elimination of chlorine and carboxylic group, and amino nitrogen of amino acids serves as source of nitrogen of HNM, the formation of HNM requires dissociation of alkyl functional group (except glycine that has a single hydrogen as a functional group), leaving a-carbon behind. In this respect, the likelihood of functional group dissociation would be greater for the amino acids containing short chain methyl group or readily dissociating group present in the side chains. For example, amino acids with short functional group (e.g. alanine with –CH3 functional group) may be more amenable to oxidation and dissociation reaction than those with long-chain side groups (e.g. lycine with CH2CH2CH2CH2NHþ 3 ). Further, acidic amino acids such as aspartic acid and glutamic acid have carboxyl groups on the edge of the side chain of which removal facilitate subsequent oxidation and dissociation of the remaining methyl group from the a-carbon. On the other hand, amino acids containing stable moieties in the functional group (e.g. phenylalanine with phenyl group) are not likely to undergo such reactions because of the high stability of the functional groups. Therefore, in light of likelihood of HNM formation, it is postulated that amino acids with short chain structure and acidic functional groups are likely to serve as precursors of HNM upon chlorination. Fig. 4 shows a hypothesized pathway of TCNM formation of chlorination of aspartic acid. The pathway involves chlorination of amino group (steps 1,2), b-elimination of carboxylate (3), removal of CO2 (4), oxidation of C–N double bond by HOCl (5), b-
elimination of hydrogen and chlorine, forming C–C double bond (6), oxidation of C–C double bond by HOCl (7), b-elimination of hydroxyl group and chlorine, forming C–C double bond (8), oxidation of C–C double bond by HOCl (9), elimination of CH2O group from b-carbon (10), elimination of N-chlorine and hydrogen, forming C–N double bond (11), oxidation of C–N double bond by HOCl (12), and dehydrogenation of N-hydroxyl groups to form TCNM (13). It is proposed in this hypothesized pathway that dissociation of alkyl functional group occurs via alternating oxidation and elimination reactions. It should be noted that the proposed pathway is based on chlorination oxidation and that HNM formation may not necessarily follow the same route under other oxidation scenarios. Furthermore, as indicated by the highest HNM yield from ozonation–chlorination, the presence of strong oxidant may increase the rate and extent of HNM formation. For ozone, its possible role in the proposed scheme may include facilitating reactions involving oxidation of the reaction intermediates, thereby enhancing the overall formation of HNM.
4.
Conclusions
The formation of HNMs in drinking waters with different organic matter characteristics in the presence of typical drinking water oxidants was examined. The results demonstrated that HNM molar yields were the highest for ozonation–chlorination, followed by chlorination, ozonation– chloramination, and chloramination. Ozonation–chlorination significantly enhanced HNM formation, while chloramination, alone or after ozonation, produced the least amount of HNMs. Higher HNM yields were observed in the treated water than the raw water from the same treatment plant, which indicate that the conventional treatment processes do not remove the majority of HNM precursors. In addition, HNM yields of HPO and TPH fractions were lower than those of raw and treated waters, indicating HPO and TPH fractions are not the main precursors of HNMs. Formation potential tests with mixed (TPH þ HPI) and HPI fractions further confirmed that HNM precursors consist of some hydrophilic organic matter with low-molecular weight that tend to persist during conventional water treatment processes. Correlations between HNM yields and various parameters (e.g., DOC, DON, DOC/DON ratios, SUVA254) during ozonation–chlorination showed an increasing trend of HNM formation with DOC, while there was no clear trend with DON. DOC/DON ratios yielded much better correlations in both raw and treated waters, highlighting the importance of nitrogen content in organic matter in the formation of HNMs. However, in some cases, relatively high yields of HNMs were found in very low level DON conditions, suggesting the presence of highly reactive specific NOM components that exhibit high yields of HNMs.
Acknowledgements This work was supported by the Water Research Foundation (Project 4063). Any opinion, findings and conclusions or
water research 44 (2010) 105–114
recommendations expressed are those of the authors and do not necessarily reflect the views of the Water Research Foundation. The authors acknowledge the contributions and assistance of the treatment plant personnel that participated in this study and the constructive inputs of the Water Research Foundation project advisory committee members (Stuart Krasner, Susan Richardson and Benito Marinas).
Appendix A. Supplementary materials Supplementary data associated with this article can be found in the online version at doi:10.1016/j.watres.2009.09.006.
references
Becke, Ch., Maier, D., Sontheimer, H., 1984. Herkunft von Trichlornitormethan in Trinkwasser. Vom Wasser 62, 125–135. Chen, P.H., Richardson, S.D., Krasner, S.W., Majetich, G., Glish, G.L., 2002. Hydrogen abstraction and decomposition of bromopicrin and other trihalogenated disinfection byproducts by GC/MS. Environmental Science & Technology 36 (15), 3362–3371. Choi, J., Richardson, S.D., 2004. Formation studies of halonitromethanes in drinking water. In: Proceedings of the 2004 AWWA Water Quality Technology Conference (WQTC). AWWA, Denver, CO. Coleman, W.E., Lingg, R.D., Melton, R.G., Kopfler, F.C., 1976. The occurrence of volatile organics in five drinking water supplies using GC/MS. In: Keith, L.H. (Ed.), Identification of Analysis of Organic Pollutants in Water. Ann Arbor Science, Ann Arbor, MI, pp. 305–327. Croue´, J.P., Violleau, D., Labouyrie, L., 2000. Disinfection byproduct formation potentials of hydrophobic and hydrophilic natural organic matter fractions: a comparison between a low- and a high-humic water. ACS Symposium Series 761. In: Barrett, S.E., Krasner, S.W., Amy, G.L. (Eds.), Natural Organic Matter and Disinfection By-products. American Chemical Society, Washington, DC, pp. 139–153. Hoigne, J., Bader, H., 1988. The formation of trichloronitromethane (chloropicrin) and chloroform in a combined ozonation chlorination treatment of drinkingwater. Water Research 22 (3), 313–319. Hong, Y., Liu, S., Song, H., Karanfil, T., 2007. HAA formation pathways and kinetics during chloramination. Journal American Water Works Association 99, 57–69. Hong, Y., Liu, S., Song, H., Hu, J., Ates, N., Karanfil, T., 2008. The effects of quenching agents on HAA determination in chloraminated waters. Journal American Water Works Association 100, 89–99. Hu, J., 2009. Formation and Distribution of Halonitromethanes in Drinking Waters. Ph.D. thesis. Clemson University, Clemson, SC. Hureiki, L., Croue´, J.P., Legube, B., 1994. Chlorination studies of free and combined amino acids. Water Research 28 (12), 2521–2531. Karanfil, T., Hong, Y., Song, H., Orr, O., 2007. Exploring HAA Formation Pathways During Chloramination. American Water Works Association Research Foundation Publication, Denver, CO. Karanfil, T., Krasner, S.W., Westerhoff, P., Xie, Y., 2008. Recent advances in disinfection by-products formation, occurrence, control, health effects, and regulation. ACS Symposium Series 995. In: Karanfil, T., Krasner, S.W., Westerhoff, P., Xie, Y. (Eds.), Occurrence, Formation, Health Effects and Control of
113
Disinfection By-products in Drinking Water. American Chemical Society, Washington, DC, pp. 2–19. Kim, H.C., Yu, M.J., 2005. Characterization of natural organic matter in conventional water treatment processes for selection of treatment processes focused on DBPs control. Water Research 39 (19), 4779–4789. Kitis, M., Karanfil, T., Wigton, A., Kilduff, J.E., 2002. Probing reactivity of dissolved organic matter for disinfection by-product formation using XAD-8 resin adsorption and ultrafiltration fractionation. Water Research 36 (15), 3834–3848. Korshin, G.V., Benjamin, M.M., Sletten, R.S., 1997. Adsorption of natural organic matter (NOM) on iron oxide: effect on NOM composition and formation of organo-halide compounds during chlorination. Water Research 31 (7), 1643–1650. Krasner, S.W., McGuire, M.J., Jacangelo, J.G., Patania, N.L., Reagan, K.M., Aieta, E.M., 1989. The occurrence of disinfection by-products in United-States drinking-water. Journal American Water Works Association 81 (8), 41–53. Krasner, S.W., Chinn, R., Hwang, C.J., Barrett, S.E., 1991. Analytical methods for brominated organic disinfection by-products. In: Proceedings of the AWWA Water Quality Technology Conference (WQTC). AWWA, Denver, CO. Krasner, S.W., Weinberg, H.S., Richardson, S.D., Pastor, S.J., Chinn, R., Sclimenti, M.J., Onstad, G.D., Thruston Jr., A.D., 2006. Occurrence of a new generation of disinfection byproducts. Environmental Science & Technology 40 (23), 7175–7185. Krasner, S.W., Westerhoff, P., Chen, B., Amy, G., Nam, S.-N., Chowdhury, Z.K., Sinha, S., Rittman, B.E., 2008. Contribution of Wastewater to DBP Formation. American Water Works Association Research Foundation, Denver, CO. 2008. Krasner, S.W., Westerhoff, P., Chen, B., Rittman, B.E., Nam, S.-N., Amy, G., 2009. Impact of wastewater treatment processes on organic carbon, organic nitrogen, and DBP precursors in effluent organic matter. Environmental Science & Technology 43 (8), 2911–2918. Lee, W., Westerhoff, P., Croue´, J.P., 2007. Dissolved organic nitrogen as a precursor for chloroform, dichloroacetonitrile, N-nitrosodimethylamine, and trichloro-nitromethane. Environmental Science & Technology 41 (15), 5485–5490. Merlet, N., Thibaud, H., Dore, M., 1985. Chloropicrin formation during oxidative treatments in the preparation of drinking-water. Science of the Total Environment 47, 223–228. Na, C., Olson, T.M., 2007. Relative reactivity of amino acids with chlorine in mixtures. Environmental Science & Technology 41 (9), 3220–3225. Plewa, M.J., Wagner, E.D., Jazwierska, P., Richardson, S.D., Chen, P.H., McKague, A.B., 2004. Halonitromethane drinking water disinfection byproducts: chemical characterization and mammalian cell cytotoxicity and genotoxicity. Environmental Science & Technology 38 (1), 62–68. Plewa, M.J., Wagner, E.D., Muellnerm, M.G., Hsu, K., Richardson, S.D., 2008. Comparative mammalian cell toxicity of N-DBPs and C-DBPs. ACS Symposium Series 995. In: Karanfil, T., Krasner, S.W., Westerhoff, P., Xie, Y. (Eds.), Occurrence, Formation, Health Effects and Control of Disinfection By-products in Drinking Water. American Chemical Society, Washington, DC, pp. 36–50. Reckhow, D.A., Singer, P.C., Malcolm, R.L., 1990. Chlorination of humic materials: byproduct formation and chemical interpretations. Environmental Science & Technology 24 (11), 1655–1664. Richardson, S.D., Thruston Jr., A.D., Caughran, T.V., Chen, P.H., Collette, T.W., Floyd, T.L., Schenk, K.M., Lykins Jr., B.W., Sun, G., Majetich, G., 1999. Identification of new ozone
114
water research 44 (2010) 105–114
disinfection by-products in drinking water. Environmental Science & Technology 33 (19), 3368–3377. Richardson, S.D., Simmons, J.E., Rice, G., 2002. Disinfection byproducts: the next generation. Environmental Science & Technology 36 (9), 198A–205A. Richardson, S.D., 2003. Disinfection by-products and other emerging contaminants in drinking water. Trends in Analytical Chemistry 22 (10), 666–684. Song, H., Orr, O.B., Hong, Y., Karanfil, T., 2009. Isolation and fractionation of natural organic matter: evaluation of reverse osmosis performance and impact of fractionation parameters. Environmental Monitoring and Assessment 153 (1), 307–321.
Thibaud, H., Delaat, J., Dore, M., 1988. Effects of bromide concentration on the production of chloropicrin during chlorination of surface waters – formation of brominated trihalonitromethanes. Water Research 22 (3), 381–390. Trehy, M.L., Yost, R.A., Miles, C.J., 1986. Chlorination byproducts of amino acids in natural waters. Environmental Science & Technology 20 (11), 1117–1122. von Gunten, U., 2003. Ozonation of drinking water: part II. disinfection and by-product formation in presence of bromide, iodide or chlorine. Water Research 37 (7), 1469–1487.
water research 44 (2010) 115–122
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
The effects of UV disinfection on drinking water quality in distribution systems Yonkyu Choi, Young-june Choi* Division of R&D for Water, Waterworks Research Institute, Seoul Metropolitan Government, 552-1, Chunho Daero, Kwangjin-Ku, Seoul, Republic of Korea, 143-820
article info
abstract
Article history:
UV treatment is a cost-effective disinfection process for drinking water, but concerned to
Received 27 May 2009
have negative effects on water quality in distribution system by changed DOM structure. In
Received in revised form
the study, the authors evaluated the effects of UV disinfection on the water quality in the
30 August 2009
distribution system by investigating structure of DOM, concentration of AOC, chlorine
Accepted 2 September 2009
demand and DBP formation before and after UV disinfection process. Although UV treat-
Published online 16 September 2009
ment did not affect concentration of AOC and characteristics of DOM (e.g., DOC, UV254,
Keywords:
weight) significantly, the increase of low molecular fraction was observed after UV treat-
UV
ment, in dry season. Chlorine demand and THMFP are also increased with chlorination of
Distribution system
UV treated water. This implies that UV irradiation can cleave DOM, but molecular weights
Molecular weight
of broken DOM are not low enough to be used directly by microorganisms in distribution
AOC
system. Nonetheless, modification of DOM structure can affect water quality of distribution
Chlorine demand
system as it can increase chlorine demands and DBPs formation by post-chlorination.
SUVA254, the ratio of hydrophilic/hydrophobic fractions, and distribution of molecular
ª 2009 Elsevier Ltd. All rights reserved.
DBP
1.
Introduction
Disinfection by ultraviolet light (UV) is considered as a costeffective and easily implementable system for drinking water disinfection. Interest in UV disinfection process has been increased sharply in drinking water industry, since researchers demonstrated that even very low dosage of UV light could inactivate Cryptosporidium effectively in the late 1990s (Bukhari et al., 1999; Clancy et al., 2000). UV spectrum is divided into four regions; vacuum UV (100w200 nm, hereafter VUV), UV-C (200w280 nm), UV-B (280w315 nm), and UV-A (315w400 nm). UV disinfection primarily occurs due to the germicidal action of UV-B and UVC light on microorganisms. Although VUV can disinfect microorganisms, it is not efficient to use VUV for water disinfection because it rapidly dissipates through water in
very short distances (EPA, 2006). VUV is also known to breakdown bonds of organic carbons (Buchanan et al., 2004; Thomson et al., 2004). Two UV systems are generally applied for drinking water disinfection process. Monochromatic low pressure UV (hereafter LPUV) emits single wavelength at 254 nm which is close to the maximum microbial action spectrum. Polychromatic medium pressure UV (hereafter MPUV) emits a wide range of wavelength including UV-A, -B, -C and visible light. Special LPUV emitting two wavelengths at 185 and 254 nm (hereafter LPUV for TOC) is applied to remove TOC for producing ultrapure water. Although these UV systems are inactivate most of microorganisms effectively except for some viruses, they can not guarantee biological safety of tap water because the effect of UV irradiation can not be maintained throughout
* Corresponding author. Tel.: þ82 2 3146 1810; fax: þ82 2 3146 1811. E-mail address:
[email protected] (Y.-j. Choi). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.011
116
water research 44 (2010) 115–122
distribution system. On the contrary, UV disinfection is concerned to have negative effects on water quality by UV photolysis. Many researchers have reported that UV irradiation can modify DOM structure and increase biodegradability (Frimmel, 1998; Thomson et al., 2004; Buchanan et al., 2005; Goslan et al., 2006). Especially, VUV irradiation is known to be more effective than UV-C irradiation in formation of biodegradable compounds and mineralization (Buchanan et al., 2004). UV-A and UV-B can also splits large NOM molecules into organic acids with lower molecular weight (Frimmel, 1998). This change of DOM structure can increase biodegradability, which stimulates microbial regrowth and biofilm formation in distribution system. Increase of biofilm can also cause taste and odor problems and reduction of hydraulic capacity (Shaw et al., 2000). Therefore, sequential disinfection process with additional chemical disinfectant such as chlorine or monochloramine was applied to prevent microbial re-growth in the distribution system. With chlorination as secondary disinfection process, UV treatment is often expected to reduce chlorine demand and DBPs formation. Liu et al. (2006), however, reported that the DBPs formation of four organic waters was increased by chlorination after UV irradiation. The effect of UV irradiation on water quality depends on many factors, such as characteristics of source water quality, UV wavelength and applied dosage. Previous studies have often been carried out under bench-scale conditions, and organic water with relatively high DOC level (5w17.4 mg/L) and high UV dosage of 14w1,000 J/cm2 were used, which were not the conditions used for drinking water disinfection process (Frimmel, 1998; Buchanan et al., 2006; Goslan et al., 2006; Liu et al., 2006). Under drinking water with low DOC level less than 2 mg/L and UV dosage less than 40 mJ/cm2, the impact of UV irradiation on water quality could be different from the results of the previous studies. Moreover, the results under lab scale bench test can hardly reflect the real reactions under full-scale continuous flow system. In this study, the authors used pilot-scale continuous flow UV systems with LPUV, LPUV for TOC, and MPUV, and investigated change of DOM structure, probability of microbial regrowth, chlorine demand and THMs formation before and after UV treatment to evaluate the effects of UV disinfection on water quality in distribution system.
2.1.
UV pilot plant
The UV pilot plant with four UV reactors, LPUV (L85), LPUV for TOC (L90), MPUV (M1300, M350), is installed at the end of sand filters in the WTP. Sand filtered water (SF) was introduced to the reactors, and total capacity of the system was 1080 m3/ day. The experiments were carried out with UV dose of 40 mJ/cm2, which was usually applied for drinking water disinfection process. LPUV for TOC (L90) emitting two wavelengths at 185 and 254 nm is installed to evaluate TOC removal efficiency of vacuum UV. As higher UV dosage is required for TOC mineralization, additional experiments were carried out with UV dose of 150 mJ/cm2. UV dosage of each reactor was calculated from UV intensity by online sensor and contact time at each flow rate. Online sensor of LPUV (L90 and L85) and MPUV (M1300 and M350) can measure at 254 nm and between 200w300 nm, respectively. L90 system emits UV light with 254 nm and 185 nm with the ratio of 3:1. The detailed characteristics of each system were listed in Table 1. The sand filtered water and the five UV treated waters were investigated. The samples were taken from both the inflow and outflow of each reactor.
2.2.
Analytical method
The samples taken from the pilot plant were brought to the laboratory in 2 h and stored in the refrigerator below 4 C. For analyses of THMs already formed by pre-chlorination, ascorbic acid and HCl (1 þ 1) was added instantly to the samples (40 mL) to quench residual chlorine. For THMFP analyses, the samples were chlorinated (TOC : chlorine ¼ 1: 3) and incubated at 25 C for 48 h. After incubation, residual chlorine was quenched with ascorbic acid and HCl (1 þ 1) not to form THMs any more. THMs were analyzed by purge and trap method with GC (Varian, CX3600) equipped with ECD detector according to the EPA 502.2 (EPA, 1995). DOC and UV254 were analyzed with TOC analyzer (Ionics, Sievers 820) and UV/VIS spectrophotometer (Varian Cary 3C), respectively. SUVA254 was calculated from DOC and UV254.
2.3.
Separation of hydrophilic and hydrophobic carbon
DOM was separated into hydrophobic and hydrophilic fractions with resin (Amberitic XAD-7HP, Rohm & Haas Co.,
Table 1 – The characteristics of the UV system in the pilot plant.
2.
Materials and methods
In this study, the characteristics of DOM, biological re-growth potential, chlorine demand, and formation potential of disinfection byproducts before and after UV irradiation were compared to evaluate the effects of UV disinfection on water quality in distribution system. A UV pilot plant was installed at a water treatment plant (WTP) in Seoul, Korea. The samples were taken three times in 2005 and 2006, considering seasonal variation of the raw water quality ; 1) dry season with high algal biomass and BOD from winter to spring, 2) rainy season with high turbidity due to heavy rainfall during summer, and 3) normal times (Fig. 2).
System Lamp type Wavelength of UV Capacity Dosage emission (nm) (m3/h) (mJ/cm2) L90-4 L90-15
90 W Low pressure for TOC
185, 254
180 50
40 150
M1300
1.3 kW Medium pressure 85 W Low pressure 350 W Medium pressure
185w400
650
40
254
120
200w400
260
L85 M350
117
water research 44 (2010) 115–122
Table 2 – DOC, UV254, and SUVA254 of pre- and post-UV treated water. System
UV254 (cm1)
DOC (mg/L) Normal
Dry
Rainy
Normal
Dry
Rainy
Normal
Dry
Rainy
0.96 0.96 0.96 0.97 0.96 0.95
1.26 1.31 1.34 1.25 1.27 1.23
1.27 1.27 1.30 1.24 1.24 1.23
0.014 0.014 0.013 0.014 0.014 0.015
0.015 0.015 0.014 0.015 0.015 0.015
0.022 0.022 0.020 0.022 0.022 0.020
1.45 1.45 1.35 1.44 1.45 1.57
1.19 1.14 1.04 1.20 1.18 1.21
1.73 1.73 1.53 1.77 1.77 1.62
SF L90-4 L90-15 M1300 L85 M350
France). Resin was cleaned with sequential soxhlet extraction method (Ma et al., 2001). XAD-7HP resin was packed in 31 mm (ID) 230 mm (H) glass column and 0.5 N NaOH was introduced into the column to clean the resin. The resin was extracted sequentially with methanol, acetonitrile, and methanol for 12 h. Finally, the column was rinsed with ultrapure water, 0.1 N NaOH, 0.1 N HCl and ultrapure water in order, until the concentration of TOC of the effluent was less than 0.1 mg/L. Each sample was adjusted to pH < 2 by adding (1 þ 1) H3PO4 and passed through clean glass column with flow rate of 15w20 mL/min. The hydrophobic carbon was the fraction that adsorbed to the surface of the resin and the carbon that passed out through the column was determined as hydrophilic fraction. After hydrophilic and hydrophobic fractions were adjusted pH 7 0.2 with 0.1 N H3PO4 and 0.1 N NaOH, DOC was analyzed with TOC Analyzer (Ionics, Sievers 820).
to maintain constant pH and ionic strength for all samples and reduce undesirable interactions. Number-averaged MW (Mn), weight-averaged MW (Mw), and polydispersivity (r) were determined using the following equations. hi and Mi are the height of HPLC-SEC chromatogram and molecular weight. n P
hi Mn ¼ i¼1 n P hi Mi
i¼1
n P
Mw ¼ i¼1
hi Mi
n P
hi
i¼1
r¼
2.4.
SUVA254 (L/mg$m)
Mw Mn
Apparent molecular weight 2.5.
High performance liquid chromatography-size exclusion chromatography (HPLC-SEC) was used to fractionate apparent molecular weight of DOM (Her et al., 2003). Separation by size exclusion was performed using a TSK-50S (Toyopearl HW SOS, 30 mm resin) column prior to sequential on-line detectors consisting of UV/Visble (SPD-20AD, Shimadzu) and DOC (Modified Sievers Total Organic Carbon Analyzer 820 Turbo). Mobile phase solution (pH 6.8 and ionic strength 0.1 M) was made with 4 mM phosphate buffer and 25 mM sodium sulfate. Polyethylene glycols (PEGs, 200 600, 2000, 4000, 8000 dalton) were used for molecular weight (MW) calibration of chromatograms. The pH and ionic strength of each sample were also adjusted with phosphate buffer and sodium sulfate solutions as similar to the mobile phase as possible before analysis
Assimilable organic carbon(AOC)
AOC was analyzed with the method proposed by Kaplan et al. (1993). AOC is defined as the amount of carbon used as energy or converted into biomass by bacteria. Two pure-culture bacterial strains, Pseudomonas fluorescens strain P17 (hereafter, P17) and Spirillum strain NOX (hereafter, NOX) were used. The sample was taken in a glass vial baked at 550 C over 2 h and sodium thiosulfate was added to quench residual chlorine. The sample was pasteurized at 70 C for 30 min in water bath, and spiked with P17 and NOX, and incubated at 15 C for 7 days. The incubated sample was taken out, inoculated in R2A media and incubated at 25 C for 72 h. The colony counts of P17 and NOX in stationary phase were converted into bacterial biomass by multiplying each carbon conversion
Hydrophilic
Hydrophobic
100% 20
22
21
19
20
19
80%
30
29
28
31
30
29
70
71
72
69
70
71
L85
M350
35
34
34
35
33
38
65
66
66
65
67
62
L85
M350
60% 40%
80
78
79
81
80
81
20% 0%
SF
L90-4 L90-15 M1300
Dry season
L85
M350
SF
L90-4 L90-15 M1300
Normal times
SF
L90-4 L90-15 M1300
Rainy season
Fig. 1 – The ratio of hydrophilic and hydrophobic fractions in pre- and post-UV treated water.
118
water research 44 (2010) 115–122
Table 3 – Percentage of each fraction of molecular weight in pre- and post-UV treated water of each system. System
>2 K
1–2 K
0.5–1 K
4 h). 4 C Chlorine demands (mg/L)
SF L90-4 L90-15 M1300 L85 M350
15 C Decay rate (h1)
ID
24 h
48 h
K1
K2
0.10 0.09 0.22 0.12 0.10 0.11
0.50 0.54 0.71 0.56 0.51 0.48
0.60 0.63 0.79 0.68 0.61 0.60
0.52 0.57 0.73 0.67 0.62 0.57
0.005 0.006 0.004 0.005 0.005 0.005
composed of more simple compounds and less complex mixtures. Decrease in molecular weight of DOM was shown with L9015, M1300 and M350 systems in dry season. The observation suggested that short wavelength below 254 nm is more effective to break down the bonds of organic carbons, and various wavelength of light could be related to degradation of DOM. It has been reported that UVA (315–400 nm) and UVB (280–315 nm) splits large DOM molecules to generate lower molecular weight organic acids (Frimmel, 1998).
3.2.
Chlorine demands (mg/L) ID 0.06 0.21 0.22 – 0.16 0.16
24 h
48 h
K1
K2
0.32 0.50 0.55 0.49 0.46 0.49
0.44 0.61 0.66 0.55 0.58 0.61
0.025 0.073 0.079 0.060 0.060 0.074
0.004 0.005 0.006 0.006 0.006 0.006
affect AOC level, since there was not consistent trend of increase in each system. AOC after UV exposure was compared with AOC of the sand filtered water. P17 AOC and NOX AOC of all UV systems were plotted against a line of equal value. More P17 AOC data points fell on or above the line than below, while more NOX AOC data points fell on or below than above the line (Fig. 4). Paired t-tests were carried out in separate group, LPUV (L90-4, L85) and MPUV (M1300, M350). P17 AOC, NOX AOC and AOC of sand filtered water were not different statistically from those of LPUV (p ¼ 0.557, 0.964, 0.545) and MPUV (p ¼ 0.234, 0.053, 0.386) at 95 percent confidence level. Shaw et al. (2000) reported that UV treatment did not appear to affect the AOC concentration, but there were difference in the P17 and NOX data. Only the P17 AOC concentration substantially increased after UV treatment (p value ¼ 0.021) while there was little statistical evidence that UV treatment affected NOX AOC (p value ¼ 0.381).
Effects of UV treatment on AOC
Concentration of AOC, indicator of potential biological regrowth, was investigated before and after UV irradiation. AOC was measured from increased living biomass of P17 and NOX spiked in the samples. AOC of the sand filtered water was 121 mg/L in normal times when DOC was low. There was difference in AOC levels of dry and rainy seasons with similar DOC level. AOC in dry and rainy season were 341 mg/L and 149 mg/L, respectively. The results can be interpreted that DOM in the dry season was much more biodegradable than in the rainy season. Increase of AOC was observed in some cases with L90-15, M1300, and L85 systems after UV treatment (Fig. 3). However, it was not possible to determine if the UV irradiation could
3.3. Effects of UV treatment on chlorine decay and DBPs formation Chlorine demand, chlorine decay rate, THMs and THMFP concentrations were investigated for the samples before and after UV treatment to evaluate the effect of UV disinfection on chlorine demand and DBPs formation in distribution system with post-chlorination process. The chlorine decay rate was
1.6
Residual Chlorine (mg/L)
SF L90-4 L90-15 M1300 L85 M350
SF L90-4 L90-15 M1300 L85 M350
1.4 1.2
Decay rate (h1)
1.0 0.8 0.6 0.4
4 °C
15 °C
0.2 0.0 0
20
40
60
80
100 120 140 160 180 200
0
20
40
60
80
100 120 140 160 180 200
Time (hrs) Fig. 5 – Chlorine decay trends before and after UV treatment.
121
water research 44 (2010) 115–122
winter
spring
Summer
100 90 80 70 60 50 40 30 20 10 0
Fall
winter
spring
Summer
THMFP(ug/L)
Fall
THMs(ug/L)
100 90 80 70 60 50 40 30 20 10 0
SF
L90-4 L90-15 M1300
L85
M350
SF
L90-4 L90-15 M1300
L85
M350
Fig. 6 – THMs and THMFP of pre- and post-UV treated water.
very rapid just after addition of chlorine and became slower with time. The rapid and slow decay rates are likely due to different reactions such as oxidation of inorganic compounds (rapid) and substitution reactions with DOM (relatively slow). In this study, the chlorine consumption in 15 min, 24 h and 48 h were defined as instant demand (ID), 24-h demand, and 48-h demand, respectively. The experiments were conducted at 4 and 15 C considering seasonal variation of temperature. ID, 24-h and 48-h demands were increased in all UV systems at 15 C while there were not significantly different among the systems at 4 C except for L90-15 and M1300 systems (Table 4, Fig. 5). This observation suggested that high energy of UV modify DOM structure and stimulate to react with chlorine at higher water temperature. In this study, rapid chlorine decay rate (K1) and slow chlorine decay rate (K2) were compared before and after UV treatment (Table 4). Rapid chlorine decay rate (K1) was increased after UV irradiation while slow chlorine decay rate (K2) does not change significantly. This suggests that UV disinfection increases the initial rapid chlorine consumption within the clearwell, but it can not affect significantly the slow chlorine decay rate in the distribution system. Chlorine consumption increased after UV irradiation can induce increase of DPBs formation. THMs and THMFP concentrations were investigated seasonally before and after UV treatment. THMs, already formed by pre-chlorination process, were not removed by UV system. On the contrary, THMFP tended to increase after UV exposure up to 16.5 %. Especially, high increases of THMFP were observed in the L90-15 and M1300 systems in summer rainy season (Fig. 6). Paired t-tests were carried out in separate group, LPUV (L90-4, L85), MPUV (M1300, M350) and all UV (L90-4, L90-15, L85, M1300, M350). THMs were not significantly different in all cases (p > 0.072). THMFP of sand filtered water was statistically different from those after UV treatment at 95 percent confidence level (LPUV p ¼ 0.065, MPUV p ¼ 0.039, All UV p ¼ 0.009). This result suggested that UV disinfection process can increase concentration of THMs by post-chlorination to prevent bacterial re-growth in drinking water distribution system, especially in case of UV system with short wavelength. Liu et al. (2006) reported that statistically significant increase in the chloroform, DCAA, TCAA, CNCl formation from chlorination of four organic waters by UV irradiation. The impacts from UV exposure were found to be most
significant in chloroform formation, and MPUV formed slightly more of chloroform than LPUV. The authors attributed the observation to lower molecular weight organic acids generated by the broader band of UV light emitted from MPUV. Buchanan et al. (2006) reported reduction after initial increase of THMFP by UV irradiation. The initial increase of THMFP at relatively low dosage is presumably consequence of halogenation of low molecular weight compounds produced by breakdown of large NOM compounds. But THMFP was reduced at high dosage, which is thought to be primarily due to removal of NOM. VUV irradiation reduced THMFP much faster than UV irradiation, which may be resulted from the faster mineralization and decrease in precursor due to hydroxyl radical produced by VUV. This hydroxyl radicals (OH) formed via water photolysis at 185 nm can mineralize organic matters (Thomson et al., 2004; White, 1999). The destructive capacity of OH radical depends entirely upon the rate of reaction between the OH radicals and the organic substrates. Unfortunately, the reaction rate of OH radical with saturated organic compounds including chloroform is very slow, so THMs can not be removed effectively by OH radical (White, 1999).
4.
Conclusions
The effects of UV disinfection on the quality of drinking water in distribution system were evaluated in three aspects, 1) potential of biological re-growth, 2) chlorine demand and 3) DBPs formation. At 40 mJ/cm2, the dosage applied for drinking water disinfection, UV treatment can not significantly affect DOM characteristics and AOC concentration which is indicator of biological re-growth in distribution system. Although the increase of low molecular portion was observed in dry season in medium pressure and 185 nm emitting low pressure systems, it did not increase AOC concentration significantly. The broken DOM is not likely small enough to be used directly by microorganisms in the distribution system. The chlorine demands and THMFP were increased after UV exposure. This observation differs from general expectation that UV disinfection can reduce post-chlorine demand and DBP formation. Modification of DOM structure by UV irradiation might stimulate reaction with chlorine, and result in increase of DBP formation.
122
water research 44 (2010) 115–122
UV disinfection with low dosage of 40 mJ/cm2 can not mineralize DOM, but might split chemical bonds or change the characteristics of functional groups of DOM. This modification of DOM structure by UV is likely not to stimulate biological regrowth and biofilm formation in distribution system, but can have negative effects on water quality by increase of chlorine demands and DBP formation with following post-chlorination, especially in medium pressure and vacuum UV systems. To guarantee the safety of drinking water from pathogenic microorganisms and harmful DBPs at the same time, the processes to reduce the precursors of DBP are required when considering UV installation.
references
Aiken, G., Cotsaris, E., 1995. Soil and hydrology: their effect on NOM. J. Am. Water Works Assoc. 87 (1), 36–45. Buchanan, W., Roddick, F., Porter, N., 2004. Enhanced biodegradability of UV and VUV pretreated natural organic matter. Water. Sci. Technol. 4 (4), 103–111. Buchanan, W., Roddick, F., Porter, N., Drikas, M., 2005. Fractionation of UV and VUV pretreated natural organic matter from drinking water. Environ. Sci. Technol. 39, 4647–4654. Buchanan, W., Roddick, F., Porter, N., 2006. Formation of hazardous by-products resulting from the irradiation of natural organic matter: comparison between UV and VUV irradiation. Chemosphere 63, 1130–1141. Bukhari, Z., Hargy, T.M., Bolton, J.R., Dussert, B., Clancy, J.L., 1999. Medium-pressure UV for Oocyst inactivation. J. Am. Water. Works. Assoc. 91 (3), 86–94. Clancy, J.L., Bukhari, Z., Hargy, T.M., Bolton, J.R., Dussert, B.W., Marshall, M.M., 2000. Using UV to inactivate Cryptosporidium. J. Am. Water Works Assoc. 92 (9), 97–104. EPA, 1995. Method 502.2 Volatile organic compounds in water by purge and trap capillary column gas chromatography with photoionization and electrolytic conductivity detectors in series. EPA, 2006. Ultraviolet disinfection guidance manual for the final long term 2 enhanced surface water treatment rule. Chapter 2, 1–20. Frimmel, F.H., 1998. Impact of light on the properties of aquatic organic matter. Envrion. Int 24 (5/6), 559–571. Goslan, E.H., Gurses, F., Banks, J., Parsons, S.A., 2006. An investigation into reservoir NOM reduction by UV photolysis and advanced oxidation processes. Chemosphere 65, 1113–1119.
Her, N., Amy, G., McKnight, D., Sohn, J., Yoon, Y., 2003. Characterization of DOM as a function of MW by fluorescence EEM and HPLV-SEC using UVA, DOC and fluorescence detection. Water. Res. 37, 4295–4303. Jadas-He´cart, A., El Morer, A., Stitou, M., Bouillot, P., Legube, B., 1992. The chlorine demand of a treated water. Water Res. 26 (8), 1073–1084. Jeong, Y., Kweon, J., Lee, S., 2007. Characteristics of natural organic matter (NOM) on Han river and criterion of enhanced coagulation. Journal of the Korean Society of Water and Wastewater 21 (6), 653–661. Kaplan, L.A., Bott, T.L., Reasoner, D.J., 1993. Evaluation and simplification of the assimilable organic carbon nutrient bioassay for bacterial growth in drinking water. Appl. Environ. Microbiol. 59 (5), 1532–1539. Kim, S.E., Gu, Y.H., Yu, M.J., Chang, H.S., Lee, S.W., Han, S.H., 2007. Characterization of NOM behavior and DBPs formation in water treatment processes. J. KSWW 21 (4), 395–407. Kitis, M., Karanfil, T., Wigton, A., Kilduff, J.E., 2002. Probing reactivity of dissolved organic matter for disinfection byproduct formation using XAD-8 resin adsorption and ultrafiltration fractionation. Water Res. 36, 3834–3848. Liu, W., Cheung, L.-M., Yang, X., Shang, C., 2006. THM, HAA and CNCl formation from UV irradiation and chlor(am)ination of selected organic waters. Water Res. 40, 2033–2043. Ma, H., Allen, H.E., Yin, Y., 2001. Characterization of isolated fractions of dissolved organic matter from natural waters and a wastewater effluent. Water Res. 35 (4), 985–996. Oh, H.K., Kim, H.C., Ku, Y.H., Yu, M.J., Park, H., Chang, H.S., 2003. Characterization and disinfection by-product formation potential of natural organic matter in drinking water treatment. J. of KSEE 25 (10), 1252–1257. Powell, J.C., Hallam, N.B., West, J.R., Forster, C.F., Simms, J., 2000. Factors which control bulk chlorine decay rates. Water Res. 34 (1), 117–126. Shaw, J.P., Malley Jr., J.P., Willoughby, S.A., 2000. Effects of UV irradiation on organic matter. J. Am. Water Works Assoc. 92 (4), 157–167. Thomson, J., Roddick, F., Drikas, M., 2004. Vacuum ultraviolet irradiation for natural organic matter removal. J. Water SRTAqua 53, 193–206. Wetzel, R.G., 1983. Limnology, second ed. Saunders College, Publishing. 487–518, 667–678. White, M.C., Thompson, J.D., Harrington, G.W., Singer, P.C., 1997. Evaluating criteria for enhanced coagulation compliance. J. Am. Water Works Assoc. 89 (5), 64–77. White, G.C., 1999. Handbook of Chlorination and Alternative Disinfectants, fourth ed. A Wiley-Interscience Publication. 1459–1467.
water research 44 (2010) 123–130
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Pre-treatment mechanisms during thermophilic–mesophilic temperature phased anaerobic digestion of primary sludge Huoqing Ge, Paul D. Jensen, Damien J. Batstone* AWMC, Advanced Water Management Centre, Environmental Biotechnology CRC, The University of Queensland, St Lucia, Brisbane, QLD 4072, Australia
article info
abstract
Article history:
Pre-treatment is used extensively to improve degradability and hydrolysis rate of material
Received 26 June 2009
being fed into digesters. One emerging process is temperature phased anaerobic digestion
Received in revised form
(TPAD), which applies a short (2 day) 50–70 C pre-treatment step prior to 35 C digestion in
2 September 2009
the main stage (10–20 days). In this study, we evaluated a thermophilic–mesophilic TPAD
Accepted 2 September 2009
against a mesophilic–mesophilic TPAD treating primary sludge. Thermophilic–mesophilic
Published online 8 September 2009
TPAD achieved 54% VS destruction compared to 44% in mesophilic–mesophilic TPAD, with a 25% parallel increase in methane production. Measurements of soluble COD and NHþ 4 -N
Keywords:
showed increased hydrolysis extent during thermophilic pre-treatment. Model based
Temperature phased
analysis indicated the improved performance was due to an increased hydrolysis coeffi-
anaerobic digestion
cient rather than an increased inherent degradability, suggesting while TPAD is suitable as
Thermophilic pre-treatment
an intensification process, a larger main digester could achieve similar impact. ª 2009 Elsevier Ltd. All rights reserved.
Mesophilic pre-treatment Primary sludge
1.
Introduction
Waste organic solids are widely produced by domestic and industrial wastewater treatment plants. Anaerobic digestion is a common stabilisation method for treating these solids, which is environmentally beneficial due to production of renewable energy. However, degradability of the feed material needs to be relatively high, to allow good solids destruction, provide gas for heating and mixing, and prevent washout of methanogens. Degradability is particularly poor in longsludge age activated sludge systems (Gossett and Belser, 1982). Many long-sludge age systems are also smaller scale ( 0.80, n ¼ 50), and data from volunteer-collected samples were not more variable (Fig. 2). Our test of the effect of storage in shipping containers showed no significant breakdown of microcystin for the first 2 days (Tukey-Kramer HSD multiple-comparison test, p > 0.05), after which toxin concentrations began to decrease (Fig. 3). Temperatures in the shipping containers remained below 5 C for the first 24 hours and below 20 C for the first w40 hours. Shipping records indicated that none of the samples shipped by volunteers spent more than 2 days in transit and most of the bottles were frozen upon delivery, so it is likely that our method of shipping samples did not result in significant losses of microcystin. Heating samples before analysis resulted in significantly higher microcystin concentrations relative to our standard protocol (Fig. 4, paired t-test, p < 0.05). We were not able to heat samples routinely, so we corrected all our microcystin concentrations as follows to account for underestimation stemming from not heating. A linear regression fitted to the relationship between heated and not-heated samples yielded an intercept that was not significantly different from zero ( p > 0.50), so we fitted a regression with zero intercept to the
Fig. 3 – Effects of storage time in shipping containers on microcystin concentrations (means ± 1 SE). There was no effect of time from day 0 to day 2 (Tukey-Kramer HSD multiple-comparison test, p > 0.05).
data to estimate a factor to account for the underestimation of toxin levels in samples that were not heated ( y ¼ 1.24x, R2 ¼ 0.90, n ¼ 42, p < 0.0001). All microcystin concentrations were multiplied by 1.24 to account for this underestimation. In the methanol-extraction test, we found no significant difference in mean concentration for the two methods (mean SE for standard protocol, 0.44 0.24 mg L1; for methanol extraction, 0.46 mg L1 0.28, paired t-test, p > 0.70).
3.2. Interactive influences of TP and dreissenids on microcystin There was a significant positive influence of TP on euphoticzone chlorophyll a (log chlorophyll ¼ 0.58 þ 1.07 log TP, R2 ¼ 0.43, n ¼ 66, p < 0.0001), which suggests that phosphorus generally limits phytoplankton growth in these lakes, and thus that we should expect TP to influence phytoplankton species composition as well (Watson et al., 1997). In contrast, we found no influence of dreissenid invasion on euphoticzone chlorophyll a nor on the TP-chlorophyll relationship (ANOVA F-tests, p > 0.15). Thus, we could not detect any
0.4 2.0
Microcystin (µg/L)
Microcystin ( g L-1)
0.3
0.2
0.1
0.0
1.5
1.0
0.5
0.0 Volunteer
MSU
Fig. 2 – Mean microcystin concentration (D 1 SE) for samples collected by citizen volunteers versus Michigan State University personnel. Means were not significantly different ( p > 0.80, paired t-test, n [ 50).
Not heated
heated
Fig. 4 – Effects of heat treatment (30 min in a boiling water bath) on microcystin concentrations (means ± 1 SE). Means were significantly different at p < 0.05 (paired t-test, n [ 42).
146
water research 44 (2010) 141–150
Table 1 – Summary of total phosphorus, chlorophyll a and microcystin concentrations (mg LL1) for lakes with and without dreissenid mussels. Without dreissenids
Total phosphorus Microcystin, Euphotic zone Microcystin, Shoreline chlorophyll a, Euphotic zone Chlorophyll a, Shoreline
With dreissenids
n
mean
range
n
mean
range
33 32 33 29 30
18 0.17 0.34 6.5 16.9
6–103 0.03–0.95 0.03–1.88 1.4–21.1 2.1–77.4
44 43 44 37 42
11 0.34 0.83 4.8 9.3
1–36 0.02–8.37 0.02–23.6 0.5–31.0 0.5–142
Shoreline concentrations represent averages of four locations within each lake.
lake with the highest TP (103 mg L1) had only a minor effect on the relationship between microcystin and TP in uninvaded lakes (euphotic-zone microcystin: log-log slope ¼ 0.89, SE ¼ 0.31, R2 ¼ 0.22, p < 0.009, shoreline microcystin: log-log slope ¼ 1.26, SE ¼ 0.44, R2 ¼ 0.21, p < 0.008). Thus, even when
Microcystin (µg L-1)
10
1
0.1
0.01 1
10
100
10
100
10
Microcystin (µg L-1)
influence of dreissenid invasion on total phytoplankton biomass, although mean biomass in invaded lakes was somewhat lower (Table 1). The latter may have been a consequence of lower average TP in invaded lakes (ANOVA F-tests, p < 0.01). Lower average TP in invaded lakes makes the following results for microcystin all the more striking. Microcystin concentrations from depth-integrated euphotic-zone samples ranged up to 8 mg L1 (Table 1), although only 6 of 75 lakes had concentrations above 0.5 mg L1. Shoreline concentrations in individual samples ranged up to 46 mg L1, but this was also an unusual occurrence. Notably, the two lakes with the highest shoreline concentrations (maxima of 46 and 9 mg L1 for individual samples) were both dreissenid-invaded lakes with relatively low TP (9 and 14 mg L1 respectively). Within these two lakes, spatial variation in microcystin along the shore was very high, ranging from 0.3 to 9 mg L1 and 1 to 46 mg L1. We found no statistical differences among microcystin concentrations sampled from the north, south, east and west shores across lakes, so we averaged across the four shoreline sites for all subsequent analyses. There were no significant influences of lake latitude (mean ¼ 43.8 N, median ¼ 43.8 N, range: 41.8–46.5 N) or maximum depth (mean 16 m, ¼ median ¼ 14 m, range: 2–87 m) on microcystin concentrations (ANOVA F-tests, p > 0.15) so these variables were not included in the analyses that follow. As expected, microcystin concentrations in both the euphotic zone and along the shoreline were positively influenced by TP (ANOVA F-test, p < 0.0001), but this influence appeared to differ for lakes with and without dreissenids (Fig. 5). The results of a general linear model suggested that the relationship between log euphotic-zone microcystin and log TP might be different for lakes with and without dreissenids ( p ¼ 0.084 for the TP dreissenid interaction term). Given this suggestive result and our a priori hypothesis based on existing literature, we examined the influence of TP on microcystin separately for each category of lakes. These analyses revealed that the influence of TP on euphotic-zone microcystin was stronger in uninvaded lakes (log-log slope ¼ 0.98, SE ¼ 0.24, R2 ¼ 0.37, p < 0.0002) than in invaded lakes (log-log slope ¼ 0.35, SE ¼ 0.23, R2 ¼ 0.05, p > 0.10). A similar result was obtained for shoreline microcystin versus TP although unexplained error was higher (uninvaded lakes: log-log slope ¼ 1.12, SE ¼ 0.33, R2 ¼ 0.27, p < 0.002, invaded lakes: log-log slope ¼ 0.44, SE ¼ 0.27, R2 ¼ 0.06, p > 0.10). These results were not driven by the fact that maximum TP for uninvaded lakes was almost three times higher than maximum TP for invaded lakes (Table 1, Fig. 5). Excluding the
1
0.1
0.01 1
Total phosphorus (µg L-1) Fig. 5 – Relationships between TP and microcystin in the euphotic zone (top panel) and at the surface (bottom panel). Open circles, solid lines- uninvaded lakes; crosses, dotted lines- invaded lakes. Only regression lines for uninvaded lakes were significant (top panel: log microcystin [ L2.11 D 0.98 log TP, n [ 32, r2 [ 0.37, p < 0.0002; bottom panel: log microcystin [ L2.07 D 1.12 log TP, n [ 33, r2 [ 0.27, p < 0.002).
147
water research 44 (2010) 141–150
3.3. Comparison of microcystin concentrations at the shoreline versus in the euphotic zone As expected, microcystin concentrations were higher at the surface near shore than in depth-integrated samples taken from the euphotic zone (Fig. 9, mean shoreline: 0.62 mg L1, mean euphotic zone: 0.27 mg L1, paired t-test on log-transformed data: p < 0.0001). In a few cases, surface concentrations were an order of magnitude higher than contemporaneous levels in the euphotic zone (Fig. 9). Invasion status had no effect on the relationship between shoreline and euphotic-zone microcystin levels (Fig. 9, ANOVA F-tests, p > 0.50).
4.
Discussion
The effects of dreissenid grazing on the biomass of Microcystis, a major producer of microcystin, appear to be maximally complex in that all possible outcomes have been documented either in before/after invasion studies or in field experiments (Sarnelle et al., 2005). For example, invasion in the Hudson River was followed by a dramatic decrease in Microcystis (Smith et al., 1998), whereas invasion in the Bay of Quinte
Microcystin (µg L-1)
10
1
0.1
0.01 1
10
10
Microcystin (µg L-1)
maximum TP was equalized between the two lake categories, the influence of TP on microcystin was stronger, and only statistically significant, for lakes without dreissenid mussels. In contrast to the positive influence of TP on euphoticzone microcystin, there was no relationship between chlorophyll a and microcystin for depth-integrated samples from the euphotic-zone, regardless of invasion status (Fig. 6). Thus, euphotic-zone chlorophyll a was not a useful predictor of euphotic-zone toxin levels. In contrast, shoreline microcystin was significantly related to shoreline chlorophyll a and the elevation of the relationship was significantly higher in dreissenid-invaded lakes (Fig. 6). A regression tree split the euphotic-zone microcystin data four times, revealing several interesting patterns (Fig. 7). The first split separated 8 lakes (all invaded) with the lowest TP ( 0.65, n ¼ 46). Thus, the regression tree identified an influence of invasion on microcystin concentrations, but only for lakes with TP < 10 mg L1. The lack of uninvaded lakes with TP < 5 mg L1 in the data set made it impossible to be more specific about a lower bound for the positive dreissenid influence.
1
0.1
0.01 1
10
100
Chlorophyll a (µg L-1) Fig. 6 – Relationships between chlorophyll a and microcystin in the euphotic zone (top panel) and at the surface (bottom panel). Open circles, solid lines- uninvaded lakes; crosses, dotted lines- invaded lakes. For surface microcystin, regression slopes were not significantly different between uninvaded and invaded and the influences of chlorophyll a ( p < 0.0001) and invasion status ( p < 0.05) were both significant (ANCOVA, F-tests).
(Lake Ontario) and Gull Lake (southwestern Michigan) was followed by a dramatic increase (Nicholls et al., 2002; Sarnelle et al., 2005). Both positive and negative effects of Dreissena on Microcystis have also been documented in separate field experiments in Gull Lake (Sarnelle et al., 2005), but the exact mechanisms driving these variable effects are not yet understood. Selective mussel grazing (i. e., selective avoidance of Microcystis, Vanderploeg et al., 2001) is not a sufficient explanation, since mussels can sometimes have large negative impacts on Microcystis (Smith et al., 1998; Sarnelle et al., 2005). In further contrast, dreissenid invasion of Oneida Lake was accompanied by no significant change in the relative abundance of cyanobacteria, despite a significant decrease in total phytoplankton biomass during summer (Idrisi et al., 2001). Similarly, previous surveys of Michigan lakes found no difference in the abundance of Microcystis in lakes with
148
water research 44 (2010) 141–150
x = 0.27 n = 75 TP ≥ 5
TP < 5
x = 0.29 n = 67
x = 0.05 n=8
TP ≥ 26
TP < 26 x = 0.28 n = 62
x = 0.44 n=5
TP ≥ 10
Microcystin (µg L-1)
10
1
0.1
TP < 10
x = 0.14 n = 41
0.01
x = 0.56 n = 21
0.01
0.1
1
Microcystin (µg uninvaded
invaded
x = 0.10 n=7
x = 0.80 n = 14
Fig. 7 – Regression tree of log euphotic-zone microcystin versus log TP (mg LL1) and invasion status. x- untransformed mean microcystin, n- number of lakes.
Microcystin (μg L-1)
TP > 25 mg L1, yet a large difference in lakes with TP < 25 mg L1 (Knoll et al., 2008; Raikow et al., 2004). These studies provided the basis for two of the hypotheses we sought to test in this survey. We hypothesized that microcystin would increase faster with increasing TP in lakes without dreissenids than in lakes with dreissenids, given that invasion seems to promote Microcystis in low-nutrient lakes only (Knoll et al., 2008; Raikow et al., 2004). In support of this hypothesis, we found that the slope of log microcystin versus log TP was almost three times higher in uninvaded than invaded lakes (Fig. 5), although the difference in slope between lake types was not statistically significant at p < 0.05. We also found that log TP was a significant predictor of log microcystin in uninvaded lakes ( p < 0.002) but not in invaded lakes ( p > 0.10) despite
1.4
1.4
1.2
1.2
1.0
1.0
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0.0
absent present dreissenids
0.0
absent present dreissenids
Fig. 8 – Influence of dreissenid mussels on mean euphoticzone microcystin concentration (D 1 SE) for lakes with TP between 5 and 10 mg LL1 (left panel) and lakes with TP between 10 and 26 mg LL1 (right panel).
10
L-1)
Fig. 9 – Relationship between microcystin in the euphotic zone (x-axis) and microcystin at the surface (y-axis). Open circles- uninvaded lakes; crosses- invaded lakes. One-toone (dotted) line is depicted for reference. There was no influence of dreissenids on the relationship (log microcystin [ 0.12 D 0.93 log microcystin, n [ 75, r2 [ 0.56, p < 0.0001). a larger sample size for invaded lakes (Table 1). This result was not an artifact of a higher maximum TP in uninvaded lakes, since equalizing maximum TP (by excluding one high-TP uninvaded lake) did not materially affect the relationship for uninvaded lakes. After excluding the uninvaded lake with the highest TP, the range of TP was actually greater in invaded lakes, yet no influence of TP was found. Thus, we suggest that the response of microcystin to eutrophication is weaker in lakes with dreissenids. This conclusion is reminiscent of an earlier survey of phytoplankton species composition which indicated that the response of cyanobacteria to eutrophication was different in lakes with and without dreissenids (Raikow et al., 2004). The latter study found the expected positive influence of TP on cyanobacterial dominance in uninvaded lakes (Kalff, 2002) but not in invaded lakes. We also hypothesized that microcystin would be elevated in invaded lakes with low nutrients based on a previous survey of low-nutrient lakes (Knoll et al., 2008), which could help to explain the lack of positive influence of TP across all invaded lakes. To examine this hypothesis, we ran a regression-tree analysis on euphotic-zone microcystin versus TP and dreissenid presence to objectively categorize lakes with respect to TP. The analysis split 21 lakes with relatively low TP (between 5 and 10 mg L1) into invaded and uninvaded subcategories based on microcystin concentration, with invaded lakes having 8 times higher toxin levels (Fig. 7), a difference that was statistically significant. In stark contrast, at moderate TP levels (between 10 and 26 mg L1), microcystin was almost identical in invaded (0.15 mg L1) and uninvaded (0.19 mg L1) lakes. Notably however, the threshold TP level identified by the regression tree in this survey (10 mg L1) was very different from the level of 25 mg L1 used in previous surveys (Knoll et al., 2008; Raikow et al., 2004). This discrepancy may be in part a function of different methods of measuring TP in the various surveys (for example, TP samples
water research 44 (2010) 141–150
from the entire mixed layer were taken in earlier surveys but only from the surface in this survey). We suggest that further study is required to more clearly identify levels at which dreissenid effects on microcystin shift from positive to neutral. Historically, research aimed at predicting water-quality problems stemming from freshwater cyanobacteria has focused on the role of nutrient loading and in particular phosphorus, because of the powerful influence of phosphorus on the success of these phytoplankton (Kalff, 2002). This perspective has informed recent attempts to predict microcystin concentrations across lakes (Giani et al., 2005; Graham et al., 2004; Kotak et al., 2000). We suggest that this perspective be broadened to account for the effects of dreissenid invasion on Microcystis and microcystin, especially in light of the observation that cyanobacteria in lakes with Dreissena appear to respond differently to nutrient enrichment than in lakes lacking Dreissena (Raikow et al., 2004). In our survey, euphotic-zone microcystin concentrations in invaded lakes with TP between 5 and 10 mg L1 were about double that in uninvaded lakes with TP 26 mg L1. From the perspective of public health, the two lakes with the highest shoreline concentrations of microcystin were dreissenid lakes with low TP (2 log-units. A quantitative description of the processes leading to bacteria removal is
Accepted 6 September 2009
lacking and therefore a model was developed for E. coli removal from secondary clarifier
Available online 12 September 2009
effluent in slow sand filters. Removal was successfully simulated for sands of variable grain size distribution and under a range of hydraulic loading rates compared to data obtained at
Keywords:
pilot-scale filters. The most important process was retention of bacteria at the
Wastewater disinfection
‘‘schmutzdecke’’ and sand surface leading to an enrichment by a factor of up to 600
Tertiary wastewater treatment
compared to the surrounding bulk phase. Bacteria elimination and inactivation both in the
Slow sand filter
bulk phase and the schmutzdecke can be described by a first order kinetic.
Schmutzdecke
ª 2009 Elsevier Ltd. All rights reserved.
Simulation Model
1.
Introduction
Water stress or scarcity will affect more than 2.8 billion people in 48 countries by 2025 (UNEP, 2002) and wastewater is a valuable resource reliably available wherever water is consumed. Since most conventional wastewater treatment releases high number of germs, disinfection of secondary effluent is necessary before wastewater reuse in order to safeguard public health and the environment. Water scarcity affects many developing and emerging countries so that appropriate technologies for wastewater disinfection are needed. Slow sand filtration (SSF) is a simple technology that has been successfully used for over 200 years in drinking water purification. It is credited as a particle and pathogen filter that combines biological, physical and chemical processes (Obst, 1990). Slow sand filters may be adapted for
wastewater disinfection but only a few studies have been conducted on tertiary treatment of wastewater using slow sand filters (Adin, 2003). They showed total coliform bacteria removal of 0.3–3.5 log-units (Ellis, 1987; Farooq and Alyousef, 1993; Adin et al., 1998; Sadiq et al., 2003), fecal coliform removal of 2 log-units (Keraita et al., 2008) and, E. coli reduction of 2.3–3.7 log-units and Enterococci removal of 2.6 logunits (Ma¨lzer, 2006) depending on raw water quality, filter design and hydraulic loading rate. Main advantages of the SSF are its simple and economical construction, operation and maintenance using local materials and skills as well as no requirements for chemicals or energy (Visscher et al., 1987). Up to date filter design and operation mostly rely on experiences gained at lab and full scale over the last centuries. Variable ambient conditions as well as differences in key design and process parameters from one setting to another
* Corresponding author. Tel.: þ49 176 82135124; fax.: þ49 341 235 1471. E-mail addresses:
[email protected] (K. Langenbach),
[email protected] (P. Kuschk),
[email protected] (H. Horn),
[email protected] (M. Ka¨stner). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.019
160
water research 44 (2010) 159–166
impede comparison of the data as well as generalized insights into the filtration process and their effect on performance. In the field of drinking water purification models were developed to overcome these limitations (Ro¨delsperger, 2005; Campos et al., 2006a,b). However, these models were not used to simulate fecal indicator bacteria removal from secondary effluent. Another model is based on total coliform removal from secondary effluent (Sadiq et al., 2004). The use of total coliforms as indicator organisms in slow sand filtration can be problematic, since they have been shown to multiply in filters (Adin et al., 1998; Petry-Hansen, 2005). The goal of this work was to develop a simple model for the removal of fecal indicator bacteria from secondary clarifier effluent in slow sand filters using the software AQUASIM developed at EAWAG (Reichert, 1994). The suitability of the model was evaluated by comparing the simulation results with data of E. coli removal obtained from experimental pilotscale filters with different sand grain sizes operated over a range of hydraulic loading rates. The filters were also subjected to changing ambient conditions such as temperature and composition of the secondary effluent due to seasonal variations. The relevant processes leading to E. coli retention and removal were identified and quantitatively described. The model and its aggregated description of the filtration process should serve as a basis for systematic improvement of filter performance and also as a tool to predict performance under different design, operating and ambient conditions.
2.
Materials and methods
2.1. Experimental setup, analytical procedures and calculations The experimental setup consisted of 2 filter columns named S1 and S4 containing a sand bed of 50 cm height supported by
Variable
gravel (Fig. 1). A minimum supernatant water level of 30 cm was maintained by an outflow weir. The slow sand filters were continuously fed with secondary clarifier effluent of an activated sludge wastewater treatment plant (WWTP) with denitrification and biological P-Elimination located in Langenreichenbach (Saxonia, Germany). Filter material characteristics and operating conditions are listed in Table 1. The sands used varied in effective size (d10) and uniformity coefficient (U ). Hydraulic loading varied from 5 cm/h to 20 cm/h. The specific sand surface area (As) was approximated by As ¼
6000 ð1 pÞ d10 ð1 þ 2logUÞ
with p being porosity (Huisman and Wood, 1974). Phase I of the experiments lasted from September to November 2007 (nI ¼ 16 samples), Phase II from April to July 2008 (nII ¼ 11 samples) and Phase III from July to August 2008 (nIII ¼ 6 samples) following sufficient time for filter ripening to reach the optimum level of bacteria removal. Supernatant water levels were recorded and allowed to rise up to 100 cm above the sand surface before the schmutzdecke on the filter was cleaned by wet harrowing. The schmutzdecke is a layer consisting of inert material, microorganisms and algae that forms on the top of the sand (Huisman and Wood, 1974). However, clogging events of the filter bed did not follow a plausible pattern, probably because of the limited amount of incidents during the runtime of 295 days total. During this time the frequency of clogging for column S5 was five times, for columns S1, S3 and S6 four times and twice for columns S2 and S4. Water samples were taken from the filter influent and effluent as well as from sampling ports placed 2 cm above and 5, 10 and 25 cm below the sand bed surface. Further description of the filters and their performance is available elsewhere (Langenbach et al., 2009). Membrane filtration techniques (GN-6 Metricell, PALL) were used for quantification of E. coli (EN ISO 9308-1) and
Process
Q C_EC_Inf dC _ EC = − k _ EC . C _ EC dt
eps_SD C_EC_SD eps_Sand sand_surf
(1)
dC _ EC = − k _ EC . C _ EC . (1 + Factor_ EC _ SD) dt dC _ EC = − k _ EC . C _ EC . (1 + Factor_ EC . sand _ surf) dt
C_EC_s
Fig. 1 – Compartments, variables and processes in the AQUASIM-model of slow sand filtration (dimensions in cm, not to scale).
161
water research 44 (2010) 159–166
12
Phase d10 (effective sand size) [mm] U (uniformity coefficient) Specific sand surface area [m2/m3]
S1
S4
0.25 1.56 10,388
Hydraulic loading rate HLR [cm/h]
I II III
Temperature [ C]
I II III
ln (CFU/100 ml)
Table 1 – Filter material characteristics and operating conditions of 2 slow sand filters.
0.82 1.51 3,228
5 10 20
8
4
0 0
1 Dec 2007
9.5 5.6 17.6 6.2 20.3 3.2
9.1 5.1 17.9 5.8 20.5 3.4
intestinal Enterococci (EN ISO 7899-2) (DEV, 2007). For indicator bacteria data, all zero counts were replaced by the lowest possible count. Bacteria concentrations were log10transformed and checked for normal distribution using the Kolmogoroff–Smirnov test in SPSS. The software SPSS was then used to calculate arithmetic mean (m), standard deviation (s), standard error and 95%-confidence intervals. Bacteria removal in log-units was calculated from these mean values. 90th percentile values were calculated as described in EU directive 2006/7/EC (EU, 2006): 90th percentile ¼ Antilogðm þ 1:282sÞ
2.2.
(2)
Model
The simulation tool AQUASIM (Reichert, 1994) provides models for several aquatic systems and/or reactor compartments. Compartments, links between compartments, processes and variables were specified in order to model the experimental slow sand filters. The model is based on the following knowledge, simplifications and assumptions: 1. Previous experiments have shown that fecal indicator bacteria removal in SSF of secondary clarifier effluent mainly depends both on the sand surface area and the schmutzdecke (Langenbach et al., 2009). 2. Within the slow sand filter, transport mechanisms and active movement bring bacteria in contact with the schmutzdecke, with the sand grains and with the biofilm on sand grains. Retention of these bacteria in the filter system is due to adsorption and straining (Huisman and Wood, 1974). 3. The concentration of the retained (or immobilised) bacteria is a function of the concentration of bacteria in the water (or bulk) phase surrounding the sand grains and schmutzdecke (mobile bacteria) as well as the sand bed depth and the specific sand surface area. The relationship between mobile and retained bacteria in each filter horizon was calculated from the concentration measured in the bulk phase and in shake-off suspensions of sand samples extracted from the same filter horizon. 4. Elimination of both retained and mobile indicator bacteria follows a first order reaction (see Fig. 2). The reaction rate constant is independent of filter length and encompasses biotic and abiotic processes such as predation, lysis and
y = -1.28x + 10.26 R² = 0.99
2 Time [d]
3
4
Jan 2008
Jul 2008
Aug 2008
y = -1.50x + 10.23 R² = 0.94
y = -1.13x + 7.45 2 R = 0.73
y = -0.44x + 6.32 R² = 0.89
Fig. 2 – Determination of reaction order and rate constants for E. coli-elimination/inactivation in secondary clarifier effluent.
die-off due to a challenging environment. It was estimated from experimental determination within samples taken from the secondary clarifier. The filter was divided into the compartments supernatant water, schmutzdecke (or dirt layer), sand bed and gravel layer. The sand bed was further divided into four compartments, each confined by the location of the sampling ports (e.g. stretching from 5 to 10 cm filter bed depth). Because of the sampling process for determination of bacteria retained in the schmutzdecke, the schmutzdecke was defined in AQUASIM as a compartment of 2 cm height. The compartment supernatant water was depicted with a height of 30 cm. Increase in the supernatant water level was neglected because of the relatively small impact of this compartment on bacteria removal. All compartments were defined as plug-flow reactors (advective diffusive reactor compartment) without dispersion and neighbouring compartments were linked with each other (Fig. 1). In the supernatant water, bacteria elimination was modeled for the water phase. Removal was not considered in the gravel support layer. The compartments, variables and processes are depicted in Fig. 1 and explained below. C_EC: Concentration of E. coli in the bulk phase (dynamic state variable; CFU/100 ml water) C_EC_Inf: Concentration of E. coli in the filter influent (constant variable; CFU/100 ml water) C_EC_Real_Sx: Mean values of measured E. coli concentrations in the filter horizons (real list variable; CFU/100 ml water) C_EC_s ¼ C_EC Factor_EC sand_surf: Concentration of E. coli retained within a sand volume of 100 ml pore volume (formula variable; CFU/100 ml water) C_EC_SD ¼ C_EC Factor_EC_SD: Concentration of E. coli within the schmutzdecke (formula variable; CFU/100 ml water) eps_Sand: Porosity was assumed to be 40 % (constant variable) eps_SD: for the porosity of the schmutzdecke a value of 0.8 0.1 was used (constant variable; active in sensitivity and uncertainty analysis)
162
water research 44 (2010) 159–166
Factor_EC: Retention factor that describes the equilibrium between retained and mobile bacteria related to the specific sand surface area and as a function of filter bed depth (reallist variable; m3/m2; active in sensitivity and uncertainty analysis). All mechanisms of transport, straining and adsorption that lead to retention are summarized herein (Huisman and Wood, 1974; Hendricks, 1991). The values were determined from bacteria concentrations in the bulk phase and those retained on sand and biofilm in the filter horizons of 5 cm, 10 cm and 25 cm depth. Shake-off suspensions had been prepared from approximately 2.5 g of sand sampled from the filter horizons of S1 and S4 at a HLR of 10 cm/h and 20 cm/h. At a HLR of 5 cm/h, sand samples were taken from laboratory columns with similar sands (d10 of 0.25 mm, 0.4 mm and 0.63 mm) that had been operated with secondary clarifier effluent of the same WWTP and were well comparable to the pilot-scale filters. Samples were added to 20 ml of phosphate-buffer-solution in centrifuge tubes of 50 ml and vortexed for 2 minutes. Then the sand was allowed to settle for half a minute and the supernatant water was transferred for membrane filtration of appropriate dilutions. All samples were analyzed in triplicates (see Table 2). Factor_EC_SD: Retention factor that describes the equilibrium between retained and mobile bacteria in the schmutzdecke (real list variable; active in sensitivity and uncertainty analysis). All mechanisms of transport, straining and adsorption that lead to retention in the schmutzdecke are summarized herein (Huisman and Wood, 1974; Hendricks, 1991). The values were determined from bacteria concentrations in the bulk phase directly above the schmutzdecke and within the schmutzdecke. For the latter, samples were taken after draining the supernatant water so that only a layer of 2 cm remained above the sand bed and then mixing this layer. In phases II and III samples from columns S1 and S4 were analyzed in triplicates. For phase I, the sample was taken from a laboratory column with a d10 of 0.8 mm operated with secondary effluent from the WWTP under closely comparable operating conditions (see Table 2).
k_EC: reaction rate coefficient of E. coli-elimination/inactivation (by die-off, predation, lysis) derived from experiments with secondary clarifier effluent from the WWTP in winter and summer (constant variable; active in sensitivity and uncertainty analysis; 1/d). 2-liter-samples were kept at room temperature, protected from light and analyzed for decline of indicator bacteria concentrations. Samples were taken daily during a period of three days and analyzed in triplicates. Logarithmic concentrations were plotted over time and a linear regression analysis was performed in Excel (see Fig. 2 and Table 2). l: filter length measured from the regular headwater level (program variable; m] Q: arithmetic mean of the volume flow measured for each filter column (constant variable; m3/d) sand_surf: specific sand surface area (constant variable; m2/m3) Since parameters like the reaction rate coefficient were subject to variability as expressed by the standard deviation, AQUASIM was also used to perform an uncertainty analysis (Reichert, 1998).
3.
Results and discussion
The goal of this work was to develop a model of bacteria elimination (measured as E. coli) yielding simulation results comparable to the concentrations actually measured over the length of slow sand filters with different sand grain size distribution for three hydraulic loading rates (HLR). In ponds, the elimination or inactivation of E. coli is usually modeled assuming first order kinetics (Von Sperling, 2005). For the secondary clarifier effluent treated by SSF in this study, reaction rate and coefficients of E. coli elimination can be seen in Fig. 2. In three cases elimination closely followed first order kinetics (coefficient of determination > 0.89) with rate coefficients ranging from 0.44 to 1.5. The arithmetic mean of 1.09
Table 2 – Overview of the values of all variables used in the model. Phase I (HLR ¼ 5 cm/h) S1 C_EC_INF [CFU/100 ml]
k_EC [1/d] eps_Sand eps_SD Q [m3/d] sand_surf [m2/m3]
S4
10,082 (10th-Perc.: 3,819) (90th-Perc.: 26,616)
0.03269 10,388
3,228
Phase II (HLR ¼ 10 cm/h) S1
S4
4,380 (10th-Perc.: 1,341) (90th-Perc.: 14,307) 1.09 0.4 0.4 0.8 0.1 0.0594 0.0661 10,388 3,228
Phase III (HLR ¼ 20 cm/h) S1
4,323 (10th-Perc.: 928) (90th-Perc.: 20,133)
0.129 10,388
Factor_EC_SD Schmutzdecke
126 68
465 191 Factor_EC [m3/m2]
Bed depth [cm] 5 10 25
113 64
0.006 0.0034 0.0063 0.0013 0.0038 0.0039
0.0123 0.0033 0.0177 0.0123 0.006 0.0056
S4
0.01 0.0058 0.0041 0.0004 0.01 0.0033
0.127 3,228
water research 44 (2010) 159–166
and standard deviation of 0.4 were used in the model. This is higher than the average bacteria decay rate of 0.16 h1 used by Campos et al. (2006a) and can be ecplained by the fact, that they regarded bacterial biomass and not specifically fecal indicator bacteria that are not adapted to the environment of a slow sand filter. The model and parameters described were applied to simulate E. coli concentrations as a function of filter length in filters S1 and S4 for hydraulic loading rates of 5, 10 and 20 cm/h. The values for all variables used in the model are summarized in Table 2. Fig. 3 shows calculated mean values with 95%-confidenceintervals of measured concentrations of E. coli compared to simulation results with error bounds limiting the range of values of the results plus and minus one standard deviation (uncertainty analysis) as a function of filter length at a HLR of 5 cm/h (a, b), 10 cm/h (c, d), 20 cm/h (e, f) as well as simulated and calculated 90th percentile and 10th percentile concentrations at a HLR of 5 cm/h (g, h). Agreement between simulated concentrations and those calculated from experimental data was found to be satisfactory if confidence intervals overlapped with the corridor of the uncertainty analysis. For S1 this was the case for 12 out of 15 confidence intervals, considering all hydraulic loading rates and excluding the starting points at 0 m filter length. Furthermore, the simulation lay within 8 of 15 confidence intervals. In the case of S4, all confidence intervals overlapped with the corridor generated by the uncertainty analysis and the simulation lay within 7 out of 15 confidence intervals. So the model, its assumptions and simplifications were generally acceptable. It can be concluded that bacteria elimination can be described using a first order reaction depending on bacteria concentration in the mobile bulk phase as well as the concentration of immobilised bacteria retained in the schmutzdecke and within the biofilm attached to the sand surface. Results from fuzzy rulebased modeling also showed that bacteria removal from wastewater by slow sand filters could be adequately expressed in terms of the operational parameters hydraulic loading rate, sand depth and grain size (Sadiq et al., 2004). It is evident from the data in Fig. 3, that filter S1 removed more bacteria from secondary clarifier effluent than S4 and that the model is able to acceptably predict this. The difference is caused by the various surface areas of the filter beds due to the different sands used (Langenbach et al., 2009). A regression model successfully used to predict total coliform removal in SSF showed that a decrease in sand grain size and an increase in bed depth improved the removal of bacteria (Sadiq et al., 2003). These two parameters determine the sand surface area of the filter bed. The simulation results for S1 were not satisfactory at a hydraulic loading rate of 5 cm/h and a sand bed depth of 50 cm (corresponding to the filter effluent at 90 cm filter length) as well as at a HLR of 10 cm/h and the sand bed depths of 25 cm and 50 cm. But for the same hydraulic loading rates, the simulation results for supernatant water, schmutzdecke and upper 10 cm of the sand bed are nearly identical to the calculated confidence intervals. It can be stated, that the model does not exhibit a systematic weakness. Rather, the Factor_EC needs to be measured repeatedly over a longer time. The value had been determined by triplicate measurements of bacteria in shake-off suspensions of samples extracted from
163
three sand bed depths of 3 (phase I) or 4 (phase II) filter columns. The calculated relative standard deviations for Factor_EC were quite high ranging from 21% to 103 % and they did not always decrease with increasing bed depth as expected (see Table 2). Highest standard deviations of 85 % and 103 % were measured at bed depths of 25 cm at hydraulic loading rates of 5 and 10 cm/h. In addition, the model used linear interpolation to determine the retention factors for the whole depth of the filter bed of up to 50 cm relying upon measurements at 5, 10 and 25 cm depth. Linear regression overestimated the retention factors and no samples could be taken from 50 cm depth. These facts may explain the unsatisfactory simulation at deep bed depths for S1. The impreciseness is multiplied by a specific surface area three times higher in the case of S1 compared to S4. Bacteria removal did not substantially decrease with increasing HLR (Fig. 3). This is in accordance with the findings for treatment of surface water (Huisman and Wood, 1974) and can be explained by the much higher concentration of retained bacteria compared to bacteria in the water phase. Modeling of SSF in drinking water purification has already shown the importance of the deposited material in the sand bed that leads to changes of the filtration coefficient and consequently determines the effluent quality (Ojha and Graham, 1996; Campos et al., 2006b). A reduced theoretical hydraulic retention time as a result of an elevated HLR did not affect elimination of immobilised bacteria. Also, reduced retention time between HLR of 5 and 10 cm/h was compensated by an increase in the retention factor (Table 2). In the sand bed, the retention factor (Factor_EC multiplied by specific surface area) was 61 on average ranging between 0 and 187. Comparable enrichment of fecal coliforms by factors of 51–220 between bulk phase and biofilm was found in an artificial stream system (Schultz-Fademrecht et al., 2008). An increase in Factor_EC with increasing HLR seems plausible, because more substrate can be transported deeper into the filter bed. This favours development of biofilm that in turn may lead to improved straining and adsorption of bacteria. However, the factor decreased between HLR of 10 and 20 cm/h. This may be the reason, why the simulated E. coli concentrations in the sand bed were higher than the mean values measured at 20 cm/h. The substantial contribution of the schmutzdecke and upper 5 cm of sand towards bacteria removal can clearly be demonstrated by the data in Fig. 3. Significant contribution of the schmutzdecke to the overall E. coli removal has also been observed in other SSFs treating wastewater (Ma¨lzer, 2005). It is reflected by the high retention factors in the schmutzdecke (Table 2). Compared to the surrounding bulk phase, concentrations of immobilised bacteria were higher by a factor of 113–465 on average. Higher retention factors in the schmutzdecke than in the sand bed were expected, because the accumulated material improves straining and adsorption in the slimy biofilm matrix of this layer. In addition, it can be considered that E. coli adsorbs much better to the schmutzdecke composed of 90 % organic material than to the inorganic sand grain surface. A doubling of HLR was expected to result in a higher retention factor, because doubling the particle load leads to an increase in schmutzdecke thickness. This will be reflected by a higher retention factor due to the experimental
164
water research 44 (2010) 159–166
water research 44 (2010) 159–166
procedure for determining the concentration of immobilised bacteria in the schmutzdecke layer. So the approximate quadruplication in Factor_EC_SD from 126 for a HLR of 5 cm/h to 465 at 20 cm/h is plausible. The value of 113 at 10 cm/h however falls short of the expectations. More samples from schmutzdecke layers are needed to discuss this further. In general, a first order reaction for modeling the elimination of bacteria in the SSF was successfully applied. The reaction rate constant determined for E. coli removal kinetics in the secondary clarifier effluent was used throughout all zones of the filter. Bacteria elimination affects both retained and mobile bacteria. Because the elimination is proportional to the concentration of bacteria and the concentration of retained bacteria was found to be substantially higher than in the surrounding water phase, retention in the SSF is an essential element of the process. The applicability of the same reaction rate coefficient for the secondary clarifier and the whole filter suggests that the concentration of predators in the SSF, especially in the schmutzdecke, must be much higher than in the secondary effluent. In the supernatant water, bacteria removal per filter length was much lower than in the other filter compartments. Therefore, the main role of the supernatant water seems to be the protection of the schmutzdecke from shear forces caused by inflowing water. Faster elimination in the schmutzdecke seems possible since many predators or lytic microorganisms such as Bdellovibrio require minimum concentrations of prey which are only present in the schmutzdecke (Wand et al., 2007). Lower rate coefficients in the sand bed are also conceivable, since bigger predators might not be able to enter the pore channels. In addition, the highly condensed biomass/EPS matrix of the schmutzdecke acts like a ‘‘membrane’’ filter. More detailed investigations of bacteria elimination in samples from the schmutzdecke and other horizons of the filter are needed to determine reaction rate coefficients more closely. The simulation results for the10th and 90th percentile in Fig. 3g–h show that the model is suited for variable concentrations of E. coli in the filter influent. All values lie within the corridor of the uncertainty analysis and are scattered closely above and below the plot of the simulation. This justifies the decision to correlate the concentration of retained bacteria with their concentration in the surrounding bulk phase. The corridors of up to 2.5 log-units bacteria removal determined by uncertainty analysis depict how heavily the variability of some parameters defined in the model affected the simulation result. Sensitivity analysis in AQUASIM showed that the retention factors and reaction rate coefficient most strongly influenced bacteria removal. Slow sand filtration is a process that does rely on biological mechanisms and is thus less determinable than a physical process such as membrane filtration. High variations are also reflected by the results on SSF of surface water in the literature that commonly cites bacteria removals between 2 and 4 log-units (Hendricks, 1991; Huisman, 2004). It has been strongly
165
recommended to conduct pilot-studies prior to establishing a slow sand filter for drinking water purification if in-depth experience in the region is lacking (Ellis, 1985). Some variability may be reduced by determining the retention factors repeatedly. Other parameters like the reaction rate coefficient are likely to vary seasonally depending on composition of the secondary effluent and are expected to decrease with decreasing temperature. To narrow down the corridor of uncertainty and improve the simulation, the dependency could be incorporated into the model after extensive determination of the reaction rate in schmutzdecke and bulk phase at different temperatures. It is recognised that the model described has limited potential as a tool to predict filter performance. Its main contribution is a quantitative description of the most relevant processes leading to bacteria removal in slow sand filters. It should be challenged with and will allow comparison of experimental data from SSF of secondary clarifier effluent obtained under various ambient, design and operating conditions. This will further ensure understanding of the filtration process and could lead to a database of retention factors and reaction rate coefficients to be used in predictive modeling of filter performance. The model could also be applied to bacteria removal from surface water in drinking water purification with SSFs.
4.
Conclusions
The most important process in modeling fecal indicator bacteria removal from secondary clarifier effluent is the retention of bacteria in the schmutzdecke and the filter bed. The concentration of retained bacteria was higher by an average factor of 61 compared to the surrounding water phase. Retention in the filter bed depends on the surface area of the sand that can be chosen by varying the design parameters grain size distribution and bed depth. The schmutzdecke is even more effective in retaining fecal indicator bacteria: The concentration was by a factor of 113– 465 higher than in the surrounding water phase. The creation of a hostile environment for fecal bacteria does not seem to be the main function of the SSF. Bacteria elimination and inactivation in the SSF can successfully be modeled with a first order kinetic using the same reaction rate as in the secondary clarifier. Hydraulic loading rate has no substantial impact on bacteria removal, because retained bacteria are not affected by reduced hydraulic retention times as a result of an elevated HLR. The model allows to better compare fecal indicator bacteria removal from secondary effluent in slow sand filters operated under a variety of process parameters to further enhance understanding of the processes. This will improve the model’s potential as a tool for prediction of filter performance.
Fig. 3 – Calculated mean values with 95%-confidence-intervals of measured concentrations of E. coli compared to simulation results with error bounds limiting the range of values of the results plus and minus one standard deviation (uncertainty analysis) as a function of filter length at a HLR of 5 cm/h (a, b), 10 cm/h (c, d), 20 cm/h (e, f); simulated and calculated 90th percentile and 10th percentile concentrations at a HLR of 5 cm/h (g, h).
166
water research 44 (2010) 159–166
Acknowledgements We would like to acknowledge the great support of the team from the department of Environmental Biotechnology (formerly Bioremediation), Peter Mosig and Katy Bernhard from the Center of Environmental Biotechnology (UBZ) and the staff at Langenreichenbach WWTP. Shyam Janakiraman, Thomas Schindler and Patricia Sommer assisted this work. The Foundation of German Business (SDW) funded this work by a fellowship.
references
Adin, A., 2003. Slow granular filtration for water reuse. Water Science and Technology: Water Supply 3, 123–130. Adin, A., Gerstel, Z., Izakson-Tal, N., 1998. Slow Granular Filtration for Advanced Wastewater Treatment: Design, Performance and Operation. The Hebrew University of Jerusalem, Jerusalem, Israel. Campos, L.C., Smith, S.R., Graham, N.J.D., 2006a. Deterministicbased model of slow sand filtration. 1: model development. Journal of Environmental Engineering-ASCE 132, 872–886. Campos, L.C., Smith, S.R., Graham, N.J.D., 2006b. Deterministicbased model of slow sand filtration. II: model application. Journal of Environmental Engineering-ASCE 132, 887–894. DEV, 2007. Deutsche Einheitsverfahren zur Wasser-, Abwasserund Schlamm-Untersuchung. Wiley-VCH, Weinheim, Germany. Ellis, K.V., 1985. Slow sand filtration. Crc Critical Reviews in Environmental Control 15, 315–354. Ellis, K.V., 1987. Slow sand filtration as a technique for the tertiary-treatment of municipal sewages. Water Research 21, 403–410. EU, 2006. Directive 2006/7/EC of the European Parliament and of the Council Concerning the Management of Bathing Water Quality. http://eur-lex.europa.eu/LexUriServ/site/en/oj/2006/l_ 064/l_06420060304en00370051.pdf 23 January 2009. Farooq, S., Alyousef, A.K., 1993. Slow sand filtration of secondary effluent. Journal of Environmental Engineering-ASCE 119, 615–630. Hendricks, D. (Ed.), 1991. Manual of Design for Slow Sand Filtration. AWWA Research Foundation, Denver, CO, USA. Huisman, L., 2004. Slow Sand Filtration. TU Delft, Delft, The Netherlands. Huisman, L., Wood, W.E., 1974. Slow Sand Filtration. WHO, Geneva, Switzerland. Keraita, B., Drechsel, P., Konradsen, F., Vreugdenhil, R.C., 2008. Potential of simple filters to improve microbial quality of irrigation water used in urban vegetable farming in Ghana. Journal of Environmental Science and Health Part A-Toxic/ Hazardous Substances Environmental Engineering 43, 749–755.
Langenbach, K., Kuschk, P., Horn, H., Kastner, M., 2009. Slow sand filtration of secondary clarifier effluent for wastewater reuse. Environmental Science and Technology 43, 5896–5901. Ma¨lzer, H.-J., 2005. R&D in the Field of Water Supply and Waste Water Treatment Under Regional Conditions, Part I: Drinking Water. In: Recommendations, vol. 2. DVGW Technologiezentrum Wasser, Karlsruhe, Germany. Ma¨lzer, H.-J., 2006. R&D in the Field of Water Supply and Waste Water Treatment Under Regional Conditions, Part I: Drinking Water. In: Recommendations, vol. 2. IWW, Mu¨lheim an der Ruhr, Germany. Obst, 1990. Biotechnologie in der Wasseraufbereitung. Oldenbourg, Mu¨nchen, Germany. Ojha, C.S.P., Graham, N.J.D., 1996. Numerical assessment of microbial interactions in slow sand filtration modelling. In: Graham, N.J.D., Collins, M.R. (Eds.), Advances in Slow Sand and Alternative Biological Filtration. John Wiley and Sons, Chichester, UK. Petry-Hansen, H., 2005. Bakterielle Diversita¨t von Biofilmen in Langsansandfiltern. University Duisburg Essen, Duisburg, Germany. Reichert, P., 1994. Concepts Underlying a Computer Program for the Identification and Simulation of Aquatic Systems (AQUASIM 1.0). Schriftenreihe der. EAWAG, Du¨bendorf. Reichert, P., 1998. Aquasim 2.0-User Manual. EAWAG, Du¨bendorf, Switzerland. Ro¨delsperger, M., 2005. R&D in the Field of Water Supply and Waste Water Treatment Under Regional Conditions, Part I: Drinking Water. In: Recommendations, vol. 2. DVGW Technologiezentrum Wasser, Karlsruhe, Germany. Sadiq, R., Al-Zahrani, M.A., Sheikh, A.K., Husain, T., Farooq, S., 2004. Performance evaluation of slow sand filters using fuzzy rule-based modelling. Environmental Modelling and Software 19, 507–515. Sadiq, R., Husain, T., Al-Zahrani, A.M., Sheikh, A.K., Farooq, S., 2003. Secondary effluent treatment by slowsand filters: performance and risk analysis. Water Air and Soil Pollution 143, 41–63. Schultz-Fademrecht, C., Wichern, M., Horn, H., 2008. The impact of sunlight on inactivation of indicator microorganisms both in river water and benthic biofilms. Water Research 42, 4771–4779. UNEP, 2002. Vital Water Graphics, Freshwater Stress. 23 January 2009. Visscher, J.T., Paramasivam, R., Raman, A., Heijnen, H.A., 1987. Slow Sand Filtration for Community Water Supply. International Reference Centre for Community Water Supply and Sanitation, The Hague, The Netherlands. Von Sperling, M., 2005. Modelling of coliform removal in 186 facultative and maturation ponds around the world. Water Research 39, 5261–5273. Wand, H., Vacca, G., Kuschk, P., Kruger, M., Kastner, M., 2007. Removal of bacteria by filtration in planted and non-planted sand columns. Water Research 41, 159–167.
water research 44 (2010) 167–176
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Impact of microfiltration treatment of secondary wastewater effluent on biofouling of reverse osmosis membranes Moshe Herzberg a,*, David Berry b, Lutgarde Raskin b a
Department of Desalination and Water Treatment, Zuckerberg Institute for Water Research, Ben Gurion University of the Negev, Sede-Boqer Campus, 84990 Israel b Department of Civil and Environmental Engineering, University of Michigan, Ann Arbor, MI 48109-2125, USA
article info
abstract
Article history:
The effects of microfiltration (MF) as pretreatment for reverse osmosis (RO) on biofouling of
Received 12 May 2009
RO membranes were analyzed with secondary wastewater effluents. MF pretreatment
Received in revised form
reduced permeate flux decline two- to three-fold, while increasing salt rejection. Addi-
3 September 2009
tionally, the oxygen uptake rate (OUR) in the biofouling layer of the RO membrane was
Accepted 8 September 2009
higher for an RO system that received pretreated secondary wastewater effluent compared
Available online 11 September 2009
to a control RO system that received untreated secondary effluent, likely due to the removal of inert particulate/colloidal matter during MF. A higher cell viability in the RO
Keywords:
biofilm was observed close to the membrane surface irrespective of pretreatment, which is
Microfiltration
consistent with the biofilm-enhanced concentration polarization effect. Bacterial 16S rRNA
Reverse osmosis
gene clone library analysis revealed dominant biofilm communities of Proteobacteria and
Biofouling
Bacteroidetes under all conditions. The Cramer–von Mises test statistic showed that MF
Biofilm-enhanced osmotic
pretreatment did not significantly change the bacterial community structure of RO
pressure
membrane biofilms, though it affected bacterial community structure of non-membrane-
RO biofilm community
associated biofilms (collected from the feed tank wall). The finding that the biofilm community developed on the RO membrane was not influenced by MF pretreatment may imply that RO membranes select for a conserved biofilm community. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Municipal wastewater is a resource from which high quality water can be produced (Metcalf & Eddy et al., 2003) and reverse osmosis (RO) filtration plays an important role in such wastewater reclamation (Glueckstern et al., 2008; Redondo, 2001). Also, the need for RO filtration is growing due to its high treatment efficiency for removal of endocrine disrupting compounds, pharmaceuticals, personal care products, and other emerging contaminants (Shon et al., 2006). The relatively low ionic strength and related low osmotic pressure of municipal wastewater result in a lower energy cost of RO filtration compared to that of RO desalination applications.
However, the decrease in performance of RO membranes due to fouling, and more specifically biofouling, remains a major challenge in wastewater reclamation and reuse (Belfer et al., 2005; Chen et al., 2004b; Ivnitsky et al., 2005; Ivnitsky et al., 2007; Jarusutthirak and Amy, 2006; Pang et al., 2005; Xu et al., 2006). Fouling requires frequent chemical cleaning and ultimately shortens membrane life, thus imposing a large economic burden on RO membrane plant operation. The major types of fouling in RO membranes are due to inorganic salt precipitation and deposits of organic, colloidal, and microbiological matter. While scaling and organic fouling increase the hydraulic restriction for permeate flux, colloidal fouling and microbial cells decrease the permeate flux due to
* Corresponding author. Tel.: þ972 8 6563520. E-mail address:
[email protected] (M. Herzberg). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.022
168
water research 44 (2010) 167–176
‘‘cake- and biofilm-enhanced’’ osmotic pressure (Herzberg and Elimelech, 2007; Hoek and Elimelech, 2003; Lee and Lee, 2005; Li and Elimelech, 2004). In wastewater treatment effluents, effluent organic matter (EfOM), plays a major role in organic and colloidal fouling of RO membranes (Jarusutthirak et al., 2002). Organic fouling tends to increase the hydraulic resistance of the fouled membrane (Ang and Elimelech, 2007; Ang and Elimelech, 2008) and colloidal fouling tends to increase the transmembrane osmotic pressure and to decrease salt rejection (Hoek and Elimelech, 2003; Lee et al., 2005; Li and Elimelech, 2004; Li and Elimelech, 2006). Organic and colloidal fouling can be reduced by different means. For example, conventional pretreatment for reclamation of wastewater with RO in Fountain Valley, California (Water Factory 21) includes flocculation, lime or alum clarification, re-carbonation, settling, filtration, and granular activated carbon (GAC) adsorption. It is reported that 26% of the total organic carbon (TOC) is removed by lime clarification and that 30–50% of the TOC is removed by adsorption to GAC (Kim et al., 2002). Other pretreatment options include microfiltration (MF), ultrafiltration (UF), hybrid processes of chemical flocculation and powdered activated carbon (PAC) followed by MF, UF, and GAC adsorption (del Pino and Durham, 1999; Gur-Reznik et al., 2008; Reith and Birkenhead, 1998; Teodosiu et al., 1999). While the aforementioned pretreatment processes are appropriate for the removal of organic and colloidal foulants, residual nutrients in the UF permeate or in the product RO feed water from pretreatment processes can still stimulate microbial growth, leading to a high propensity for biofouling of RO membranes used in the reclamation of tertiary wastewater effluents. An improved understanding of the composition of the microbial community responsible for biofouling of RO membranes is an important step in elucidating the mechanisms of biofouling. To investigate the community structure of membrane biofilms, earlier studies generally employed culture-dependent methods. For example, Actinomycetes, Aeromonas, Arthrobacter, Corynebacterium, Acinetobacter, Micrococcus, Flavobacterium, Pseudomonas, Bacillus, Serratia, and Mycobacterium have been recovered from full-scale RO membranes using selective culture media (Ridgway et al., 1984a). Among them, the Mycobacterium isolates were extensively investigated. Their adhesion kinetics were correlated with membrane surface characteristics (Knoell et al., 1999; Ridgway et al., 1984b) and membrane cleaning strategies using surfactants were optimized (Campbell et al., 1999). However, culture-based methods introduce well-known biases and thus have the potential to exclude the detection of important species in the biofouling community. Molecular methods circumvent culturing and provide a more accurate representation of microbial community structure (Amann et al., 1995). A molecular analysis of biofilms on full-scale MF and RO membranes using 16S rRNA gene clone libraries and fluorescence in situ hybridization indicated the dominance of Alphaproteobacteria in these biofilms (Chen et al., 2004a). Only two bacterial genera (Bradyrhizobium, Bosea) were common to both types of membranes out of 17 different bacterial groups observed (Chen et al., 2004a). Another study employing 16S rRNA gene-based denaturing gradient gel electrophoresis followed by sequencing detected the presence of
Flavobacterium in the biofilm community of a lab-scale nanofiltration (NF) membrane system fed synthetic wastewater (Ivnitsky et al., 2005). Characterization of RO membrane biofilms using 16S rRNA gene clone library analysis revealed a high level of bacterial diversity, including representatives from Rhizobiales with Bosea, Rhodopseudomonas, Methylocella, Ochrobactrum, Oligotropha, Shinella, and Xanthobacter (Pang and Liu, 2007). A recent study found that the bacterial community of the biofilm developed on an RO membrane was different from the bacterial community present in biofilms collected from other locations in the same RO plant, suggesting that the conditions on RO membranes select for specific populations (Bereschenko et al., 2008). While the feed water contained Proteobacteria, Cytophaga-Flexibacter-Bacteroides and Firmicutes, the biofilm on the RO membrane contained mainly Proteobacteria and was dominated by Sphingomonas (Bereschenko et al., 2008). The observed differences in the biofilm communities on various membranes suggest that the relationship between the biofilm populations selected for and the environmental conditions on the membrane is poorly understood. In this study, the effects of MF pretreatment for RO filtration on biofouling of RO membranes were analyzed with secondary wastewater effluents. Biofilm-enhanced osmotic pressure had a significant impact on permeate flux decline and decrease in salt rejection. The activity of bacterial cells was greatest for those cells in close proximity to the membrane surface. The structures of the bacterial communities in RO membrane biofilms were similar for experiments with and without MF pretreatment, whereas the structures of the bacterial communities on other surfaces in the RO unit differed depending on pretreatment.
2.
Materials and methods
2.1.
Secondary wastewater effluent
Secondary effluent was collected once a week from the wastewater treatment plant of the town of Wallingford, CT, USA. This plant treats municipal wastewater of approximately 50,000 inhabitants with a rotating biofilm contactor (RBC) facility. The secondary effluent was characterized according to standard methods (APHA, 1999). Secondary effluent characteristics are listed in Table 1.
2.2.
RO membrane
A commercial thin film composite RO membrane, LFC-1 (Hydranautics, Oceanside, CA), was used as a model membrane for the biofouling experiments (Herzberg and Elimelech, 2007; Herzberg and Elimelech, 2008). The hydraulic resistance was determined to be 1.06 (0.018) 1014 m1 at 25 C (Herzberg and Elimelech, 2007; Vrijenhoek et al., 2001). The observed salt passage was 2.11 0.44%, as determined using the synthetic wastewater at an applied pressure of 180 psi (1241 kPa) and a crossflow velocity of 8.5 cm/s (Herzberg and Elimelech, 2007). Note that the RO crossflow cell was not designed to include a feed spacer, therefore spacer effects are not discussed. The membrane was received as a flat sheet and stored in DI water at 4 C. The physical and chemical properties
water research 44 (2010) 167–176
Table 1 – Characteristics of secondary effluent from RBC wastewater treatment plant Component Total suspended solids (TSS) Dissolved organic carbon (DOC) Biochemical oxygen demand (BOD) NHþ 4 NO 3 NO 2 Total Kjeldhal nitrogen (TKN) Total phosphorus Ortho phosphate pH Temperature ( C)
Concentration,a mg/L 7.8 1.5 10.2 1.8 4.5 1.1 1.8 1.3 4.3 1.5 0.2 0.1 1.9 0.4 1.8 0.2 2.1 0.2 7.8 0.6 25 0.5
(as N) (as N) (as N) (as P) (as P)
a Analyses were performed in triplicate and the mean standard deviation are reported.
of the LFC-1 membrane have been described previously (Vrijenhoek et al., 2001). A 7-L laboratory scale test unit, previously described (Ang et al., 2006; Lee et al., 2006), was used for the biofouling experiments. The unit was comprised of a membrane crossflow cell, high-pressure pump (Hydra-Cell, Wanner Engineering Inc., Minneapolis, MN), feed water reservoir, chiller equipped with a temperature control system (Neslab RTE-7, Thermo Electron, Newington, NH), and a data acquisition system (PC interfaced) used to acquire the permeate flow rate (Optiflow 1000 flow-meter, Humonics, CA), conductivity (Accumet AR60, Fisher Scientific, Pittsburgh, PA), and dissolved oxygen concentration (Accumet AR60, Fisher Scientific). Retentate flow rate was monitored with a floating disk rotameter (King Instrument, Fresno, CA). The dimensions of the rectangular, crossflow, channel membrane unit were 7.7 cm 2.6 cm with a channel height of 0.3 cm. Both permeate and retentate were recirculated back to the feed reservoir.
2.3.
Biofouling protocol
A schematic of the completely mixed flow through RO unit is presented in Fig. 1. Using a continuous feed of the treated wastewater to the RO unit coupled with a high recirculation ratio (recirculation flow rate divided by the feed wastewater flow rate of 28) ensured a completely mixed mode of operation and enabled the RO unit to work under relatively constant conditions. Biofouling experiments were performed in duplicate with untreated secondary effluent and secondary effluent treated with MF (0.45 mm). The MF unit consisted of a Millipore disposable filtration capsule, Opticap XL 5 with a hydrophilic PVDF membrane (Millipore Corporation). A thorough cleaning of the unit at the beginning and the end of every experiment was conducted as described previously (Herzberg and Elimelech, 2007). Following the sterilization/cleaning protocol, the membrane was compacted with DI water at a pressure of 300 psi until the permeate flux attained a constant value (usually after 12–18 h). Following compaction of the membrane, a 1 h baseline performance with DI water was conducted at a constant pressure of 180 psi and temperature of 25 C for all experiments. After attaining a stable flux with DI water, the secondary effluent (untreated or after MF
169
treatment) was added to the feed reservoir and was pumped into the RO unit at a constant rate that corresponded to a hydraulic retention time of 8 h. Samples from the permeate and the feed reservoir were collected at all stages. Each continuous experiment was conducted for 8.5 days and major parameters monitored were permeate flux, salt rejection, dissolved oxygen concentration, dissolved organic carbon (DOC), and pH. The dissolved oxygen concentration in the feed reservoir varied from 2.8 to 3.1 mg/L. Microbial growth in the RO unit was monitored continuously by measuring the oxygen uptake rate of the biofouling layer and at the end of the experiment by confocal microscopy and microbial community analysis.
2.4.
Analytical methods
For determination of salt rejection, the conductivity of the feed and permeate were measured during the different stages of the biofouling runs using an Accumet conductivity probe, a four-cell type with cell constant of 1.0 cm1, (Fisher Scientific, Pittsburgh, PA). DOC analysis was conducted with a total organic carbon analyzer (Shimadzu TOC-VCSH, Boulder, CO). Prior to analysis, the 20 mL feed and permeate samples were filtered through a 0.22 mm syringe sterilized PVDF filter (Durapore, Billerica, MA).
2.5.
Laser scanning confocal microscopy (LSCM)
At the end of each biofouling experiment, the membrane coupon was carefully removed and cut into pieces of approximately 5 mm 5 mm for staining with the bacterial dead/live kit (Molecular Probes, Eugene, OR) containing propidium iodide (PI) and SYTO9. Microscopic observation and image acquisition were performed using a spinning disk laser scanning confocal microscope (LSCM; PerkinElmer Life and Analytical Sciences, Boston, MA), equipped with 60/1.4 NA objective (Plan-Apo; Olympus). The LSCM was equipped with detectors and filter sets for monitoring PI/SYTO9 stained cells (excitation wavelengths of 568 and 488 nm, respectively). Three-dimensional reconstruction of the LSCM image stacks was carried out using Imaris software (Imaris Bitplane, Zurich, Switzerland). Cell staining in the biofouling layer was performed with a 100 mL solution of 30 mM PI and 5 mM SYTO9 (prepared in 10 mM phosphate buffer, pH 7.5), which covered the biofilm samples that were incubated in the dark at room temperature for 20 min. Excess PI and SYTO9 solution was carefully drawn off with Kimwipe paper. The excess PI and SYTO9 stains that did not bind to the biofilm samples were then removed by rinsing three times with a 10 mM phosphate buffer at pH 7.5. LSCM images were generated using the BioRad confocal assistant software (version 4.02). Gray scale images were analyzed, and the specific biovolume (mm3/mm2), average thickness (mm), and thickness distribution (number of stained spots in depth location versus thickness) in the biofouling layer was determined by COMSTAT, an imageprocessing software (Heydorn et al., 2000), written as a script in Matlab 5.1 (The MathWorks, Inc., Natick, MA) and equipped with an image-processing toolbox. Thresholding was fixed for all image stacks. At the end of the fouling experiments, between 6
170
water research 44 (2010) 167–176
Completely stirred tank
RETENTATE Flow meter
PERMEATE PC
Recirculating heater/chiller
Needle Valve
Pressure gauge
Floating disk rotameter
Conductivity / DO
Backpressure regulator
Feed High pressure pump
Membrane support for biofilm growth
Fig. 1 – Schematic of the flow through RO biofouling test unit. The main components of the system include: a plate-and frame crossflow membrane cell with a RO membrane coupon as a substrate for biofilm growth, a complete stirred feed tank inoculated with either secondary wastewater effluent or with pre-filtered secondary effluent (0.45 mm) from RBC wastewater treatment plant, a high-pressure feed pump, a heater/chiller equipped with a temperature control system, a permeate flowmeter, and a data acquisition system, which can measure conductivity, DO, and permeate flow rate.
and 9 positions on the membrane were chosen and microscopically observed, acquired, and analyzed.
2.6.
Scanning electron microscopy (SEM)
ESEM (FEI Company, Philips XL30) was used in a conventional high vacuum mode for imaging of the biofouling layers. The biofouling layers were fixed, dehydrated, and coated with a layer of carbon approximately 10–15 nm thick. The fixation method (Fox and Demaree, 1999) involved the following steps: 1) excess electrolyte solution was carefully removed with a filter paper from the specimens (fouled membrane pieces of around 5 mm 5 mm); 2) the fouled membrane specimens were incubated in 0.05 M sodium cacodylate buffer supplemented with 2% gluteraldehyde (Electron Microscopy Sciences, Fisher Scientific) for 1 h; 3) the specimens were incubated for 10 min and rinsed three times with 0.05 M sodium cacodylate buffer; 4) a second fixation step was performed by incubating the specimens in 0.05 M sodium cacodylate buffer supplemented with 1% osmium tetroxide for 1 h (Electron Microscopy Sciences, Fisher Scientific); 5) excess amounts of osmium tetroxide were removed according to the same procedure followed in step 3; 6) specimens were dehydrated during a 20 min incubation period in ethanol/water solutions with increasing ethanol concentrations (25, 50, 75, 95, and 100%); and 7) the specimens were washed once with hexamethyldisilazine (Electron Microscopy Sciences, Fisher Scientific) and dried overnight in a hood at room temperature.
2.7.
Microbial community analysis
Biomass samples were harvested from the feed tank wall and the RO membrane surface of the RO membrane reactor. Total DNA was extracted from biomass using a phenol/chloroform
extraction method and general bacterial primers (27F, 1492R) were used to amplify 16S rRNA genes by PCR (Spear et al., 2005). PCR products from duplicate biofouling experiments were pooled, purified, and cloned into Escherichia coli for sequencing. Sequencing was performed at the Washington University Genome Sequencing Center using an ABI 3700 sequencer employing capillary gel electrophoresis technology (Swerdlow and Gesteland, 1990), and over 90 clones were used for each library. Sequences were aligned using the NAST algorithm for multiple sequence alignments (DeSantis et al., 2006) and classified to the nearest known neighbor from a database of approximately 400,000 16S rRNA gene sequences. The phylogenetic similarity of sequences was calculated using a DNADIST distance matrix algorithm (Felsenstein, 2008) for further analysis in DOTUR and !-LIBSHUFF softwares. Alignment, classification, and distance matrix analysis were executed through the online workbench ‘‘greengenes’’ (http:// greengenes.lbl.gov/cgi-bin/nph-NAST_align.cgi). Species richness was measured using the bias-corrected Chao1 estimator (Chao, 1984) and species diversity was measured using the Shannon-Weaver diversity index (Magurran, 1988). Species richness and diversity measurements were determined using DOTUR software, which uses a distance matrix of sequences to classify sequences into operational taxonomic units based on cutoff distance levels (Schloss and Handelsman, 2005). Comparisons of microbial community structures were determined using an integral form of the Cramer–von Mises test statistic, as described in detail previously (Schloss et al., 2004). Briefly, the statistic calculates the coverage of the membership of one community on another integrated over the range of distance levels present. Significance testing was executed using a Monte Carlo procedure with 10,000 randomizations and a significance threshold of p ¼ 0.05. All calculations were performed in !-LIBSHUFF (Schloss et al., 2004).
water research 44 (2010) 167–176
3.
Results and discussion
3.1.
Permeate flux decline, salt and TOC rejection
171
MF pretreatment of secondary effluent reduced permeate flux decline in the RO unit two- to three-fold, while increasing salt rejection from 94.3–97.0% to 98.2–98.8%, compared to the control, which was not subjected to MF pretreatment (Fig. 2). These interrelated effects are mainly due to the higher particulate and colloidal matter in feed water in the absence of MF pretreatment. Interestingly, both cases showed the strongest flux decline during the first 24 h without a change in salt passage for the MF pretreatment experiment. A full passage of dissolved organic molecules through the MF membrane implies that accumulation of organic matter on the membrane induced a fast flux decline in both cases. Moreover, in the absence of MF pretreatment it can be assumed that additional colloidal material will induce a faster flux decline. It has been shown that organic fouling does not reduce salt rejection, and
1.0
Flux, Non-Filtered Effluents Flux, Filtered Effluents OUR, Non-Filtered Effluents OUR, Filtered Effluents
0.8
1.6 1.2
0.6
0.8
0.4
0.4
0.2
0
50
100 150 Time, Hours
200
0.0
Oxygen Uptake Rate of Biofouling Layer, µg/min
Normalized Permeate Flux
A
B 0.95
1.00
0.90 0.98 0.85 Salt Rej - Non Filtered Effluents Salt Rej - Filtered Effluents
0.96
DOC Rej - Non Filtered Effluents DOC Rej - Filtered Effluents
0
50
100 150 Time, Hours
DOC Rejection
Salt Rejection
Fig. 3 – SEM images of the biofouling layers at the end of the biofouling experiments (A) using non-treated secondary wastewater and (B) using MF pretreated secondary wastewater. Scale bars (white line) are 5 mm in both cases.
0.80 200
Fig. 2 – Effects of MF pretreatment on RO membrane performance during filtration of secondary effluent from RBC wastewater treatment plant: (A) Permeate flux decline, biofouling layer oxygen uptake rate (OUR); (B) Salt and DOC rejection. OUR by the biofilm layer was calculated as the product of the permeate flow rate and the difference between oxygen concentration in the feed and the permeate solution. Initial permeate flux and DOC feed solution concentration were 1.18 3 10L5 m/s (42.5 L/m2 h or 25.0 gal/ft2$day) and 10.2 ± 1.8 mg/L, respectively.
in many cases organic fouling can even improve salt rejection by acting as a secondary barrier that reduces diffusivity of organics and even salts through the membrane (Ang and Elimelech, 2007; Ang et al., 2006). In the present study, a slightly higher rejection for salts was obtained by the membrane fed with MF pretreated secondary wastewater, but was coupled with strong flux decline at the beginning of the run. This fouling behavior at the beginning of the fouling run, in both cases, cannot be attributed to biomass growth, which had a slower effect on flux decline accompanied with an increase in oxygen uptake rate (Fig. 2). Pretreatment with a UF membrane might have reduced the fouling effects observed here in the first 24 h. A fouling layer with an opaque matrix, probably formed by a combination of bacteria, colloids, and organic matter, was observed with SEM using non-pretreated secondary effluents as RO feed water (Fig. 3). In contrast, a typical porous biofilm structure was observed as the fouling layer with MF pretreated feed water (Fig. 3). Colloidal and bacterial deposition on RO membranes is mainly related to membrane surface roughness, with other minor effects related to surface charge and hydrophobicity (Vrijenhoek et al., 2001). LFC-1, the RO membrane used
172
water research 44 (2010) 167–176
Fig. 4 – Three-dimensional reconstructed images from spin laser scanning confocal microscope using Imaris Bitplane software. Top and middle sections, as well as perspective views of biofouling layers from non-filtered (A) and pre-filtered (B) biofouling experiments are presented. Biofouling layers were stained with propidium iodide (red) and with SYTO9 (green) for probing dead and viable cells, respectively. Scale bars are equal to 20 mm and perspective views are 128 3 128 mm (color images are presented online).
in this study, is relatively rough, having an average roughness and RMS (root mean squared) roughness of 52.0 and 67.4 nm, respectively. LFC-1 has a medium hydrophobicity, with contact angle at pH 6.8 of 52.7 in 10 mM NaCl. Therefore, both roughness and hydrophobicity characteristics are likely to enhance adhesion of colloids and particles that were present in the secondary effluent from the RBC wastewater treatment plant. In the absence of prefiltration the feed secondary effluents, the accumulated particles and colloids in the biofouling layer decelerate back diffusion of salts from the membrane surface and are primarily responsible for cake enhanced osmotic
pressure (CEOP), which is believed to be the major mechanism of salt rejection and permeate flux decline (Herzberg and Elimelech, 2007; Hoek and Elimelech, 2003; Ng and Elimelech, 2004). The decrease in salt rejection could not be explained by the decrease in permeate flux and increase of permeate salt concentration. A small effect of the decrease in permeate flux on the salt rejection, namely the ‘‘concentration effect’’, was observed over a wide range of applied pressures and their corresponding permeate fluxes. In a related study, it was observed that a 70% reduction in permeate flux reduced the overall salt rejection by only 1% (Herzberg et al., 2009). Notably, MF
173
water research 44 (2010) 167–176
2.0x10
Spots Measured
4
A
1.5x10
Table 2 – Bacterial community structure as determined by 16S rRNA gene clone library analysis. Groups comprising >1% of each phylogenetic group are shown. Subdivisions of the dominant phyla Proteobacteria and Bacteroidetes are also displayed. Dominant groups (over 25%) are in bold. N.D. [ not detected.
SYTO 9 # 1 PI # 1 SYTO 9 # 2 PI # 2 SYTO 9 # 3 PI # 3
4
4
1.0x10
Bacterial groups
3
5.0x10
Percent composition (%) Membrane biofilm
0.0
Influent treatment
0
20 40 60 80 Biofilm Thickness
2.0x10
Spots Measured
1.5x10
SYTO 9 # 1 PI # 1 SYTO 9 # 2 PI # 2 SYTO 9 # 3 PI # 3
4
4
1.0x10
3
5.0x10
0.0 0
20 40 60 80 Biofilm Thickness
Unfiltered Filtered Unfiltered Filtered
100
4
B
Wall biofilm
100
Fig. 5 – Thickness distribution of propidium iodide (red) and SYTO9 (green) stained spots in three different locations in the biofouling layers taken from non-filtered (A) and pre-filtered (B) biofouling experiments. Average thickness of the biofouling layers were 82.3 ± 31.7 and 57.1 ± 19.3 mm for the non-filtered and pre-filtered biofouling experiments, respectively (color images are presented online).
pretreatment of secondary effluent feed water did not decrease salt rejection, but rather slightly increased it. After the removal of particulate matter with MF, it seems that enhanced osmotic pressure does not take place since salt rejection is slightly improved and CEOP phenomenon is not supported. The small increase in salt rejection only for the biofouling experiment after removal of particles (>0.45 mm) may be due to formation of an additional ‘‘barrier’’ formed by extracellular polymeric substances (EPS) that reduces salt mass transport through the fouled membrane. This effect was also shown during organic fouling in a similar RO laboratory unit by Ang and Elimelech (Ang and Elimelech, 2008). The presence of EPS, as well as the stable salt rejection observed, imply that the main mechanism responsible for permeate flux decline in the case of MF pretreatment is an induced hydraulic resistance by the EPS layer, as reported in our recent study (Herzberg et al., 2009). Note that there was no detectable difference in either conductivity or DOC before and after MF pretreatment of the secondary effluents used as feed solutions. Also, MF pretreatment of the feed solution showed no effect on DOC rejection by the RO membrane (Fig. 2), implying that the concentration of DOC potentially available to microorganisms on the membrane is similar for both cases, irrespective of MF pretreatment.
Phyla Proteobacteria Bacteroidetes Planctomycetes Verrucomicrobia Actinobacteria Firmicutes
40 36 13 7 N.D. 5
31 40 21 3 N.D. 4
46 25 15 8 5 N.D.
51 17 2 3 21 3
Classes within Proteobacteria Alphaproteobacteria 10 Betaproteobacteria 43 Gammaproteobacteria 43 Deltaproteobacteria N.D. Epsilonproteobacteria 3
17 58 25 N.D. N.D.
N.D. 68 25 5 3
9 64 9 18 N.D.
Families within Bacteroidetes Saprospiraceae 23 Flavobacteriaceae 23 Flexibacteraceae 41 Sphingobacteriaceae 14
48 12 32 8
23 23 41 14
48 12 32 8
3.2.
Oxygen uptake rate by the biofouling layer
A substantially lower oxygen uptake rate (OUR) of the biofouling layer was observed when there was no MF pretreatment (Fig. 2). Combined fouling by microorganisms and particulate/colloidal matter most likely resulted in the lower OUR of the fouling layer. Dead/live staining and LSCM analysis showed higher cell viability in biofilms on RO membranes with MF pretreatment, which agrees with the observation that the OUR was higher in the RO membrane biofouling layer with MF pretreatment (Fig. 4). The related observations of cell viability (Fig. 4) and OUR by the biofouling layer (Fig. 2) are affected by two major parameters: surface area for microbial growth and nutrient concentration, which is directly related to permeate flux and to the degree of concentration polarization (Herzberg and Elimelech, 2008). Note that DOC rejection is relatively constant with time and similar (95%) for both cases, with and without MF pretreatment (Fig. 2B). It seems that both biofouling layers, with and without MF pretreatment, are exposed to a similar DOC concentration since DOC bulk concentration and rejection were similar under both conditions. According to the secondary effluent quality (Table 1), we can assume that organic carbon is the limiting nutrient for biofilm growth (a reasonable assumption for secondary effluent of an advanced biological wastewater treatment plant). Therefore, the surface area available for biofilm growth is likely to be the main parameter affecting the viability and oxidative activity of the biofouling layer. In the case of additional deposits on the membrane surface, in the absence of MF pretreatment, a lower OUR was measured (Fig. 2) and less viable biomass was identified on the membranes using non-pretreated feed water, using LSCM
174
water research 44 (2010) 167–176
Table 3 – Comparison of community membership for biofilm location (membrane or wall) and influent treatment (unfiltered or filtered), as determined by the Cramer–von Mises test statistic (Schloss et al., 2004). N.S. indicates a statistically non-significant difference.
Membrane, filtered Membrane, unfiltered Wall, filtered
Membrane, unfiltered
Wall, filtered
Wall, unfiltered
N.S.
p < 0.05
p < 0.05
–
p < 0.05
N.S.
–
–
p < 0.05
microscopy (Fig. 4). Further analysis with COMSTAT software (Heydorn et al., 2000) processing of LSCM images demonstrated that the average specific biovolumes of the spots stained with propidium iodide (red) and with SYTO9 (green) in the biofouling layers using MF pretreated feed water were 40.2 17.8 and 75.3 20.3 mm3/mm2, respectively. In contrast, when non-pretreated feed water were used for the biofouling experiment, the average specific biovolume of dead cells and of viable cells was 65.2 26.8 and 55.3 14.9 mm3/mm2, respectively. These results show that occupying membrane surface with particulate matter can simultaneously enhance concentration polarization and limit the surface area required for biofilm formation. The more evident adverse effects of the combined fouling layer comprised of particulate matter and microbial biofilm than biofilm layer alone highlight the synergistic effects of organic and colloidal fouling, in which membrane performance is extremely reduced (Li and Elimelech, 2006; Li et al., 2007).
3.3. Biofilm viability distribution – enhanced growth in biofilm-membrane interface Notably, due to a relatively low nutrient concentration and concentration polarization effects, higher cell viability was observed in close proximity to the membrane surface using LSCM (Figs. 4 and 5), regardless of MF pretreatment of the feed solution. Since under both conditions concentration polarization dictates higher solute concentration in the base biofilm layers, in both cases cell viability was higher close to the membrane surface and more dead microorganisms were observed close to the bulk solution (Figs. 4 and 5). This result is consistent with our previous study indicating that, when nutrients are limiting biofilm growth, a higher distribution of viable cells in the biofilm is present near the membrane surface due to biofilm-enhanced concentration polarization (Herzberg and Elimelech, 2008; Huertas et al., 2008). Consistent with the supposition that space limitation reduces biological activity of the biofouling layer in the case without prefiltration, a slightly thicker biofouling layer was formed under these conditions (Fig. 5), most likely due to deposition of particulate matter induced by RO filtration.
3.4. Microfiltration pretreatment does not alter biofilm bacterial community structure on the RO membranes 3.4.1.
Dominant community members
The 16S rRNA gene clone library analysis indicated Proteobacteria and Bacteroidetes were dominant in all biofilms sampled,
comprising respectively 31–51% and 17–40% of all libraries (Table 2). Among the Proteobacteria, Betaproteobacteria were the dominant group for all conditions (43–63% of Proteobacteria), which agrees with findings from other studies (Bereschenko et al., 2007; Ivnitsky et al., 2005). Alphaproteobacteria were observed in all samples except feed tank wall biofilms from MF pretreated influent. Some differences in bacterial community composition were observed between feed tank wall biofilms and RO membrane biofilms. For example, feed tank wall biofilms comprised Deltaproteobacteria and membrane biofilms did not, but membrane biofilms contained Epsilonproteobacteria and feed tank wall biofilms did not. Among the Bacteroidetes, Saprospiraceae were observed to be a major group under all conditions, but this group was more dominant in feed tank wall biofilms, while in membrane biofilms Flexibacteriaceae were more abundant. Also, Actinobacteria were dominant in feed tank wall biofilms with MF pretreatment, and were present in feed tank wall biofilms without MF pretreatment, but were absent from RO membrane biofilms irrespective of pretreatment. MF pretreatment caused several shifts in the abundance of bacterial groups. For wall biofilms, MF pretreatment led to a two-fold or greater reduction in Planctomycetes, Verrucomicrobia, several Bacteroidetes groups (Flavobacteriaceae, Flexibacteraceae and Sphingobacteriaceae), and Gammaproteobacteria, and also the disappearance of Epsilonproteobacteria. Wall biofilms with MF pretreatment also had increased abundance (>two-fold) of Actinobacteria and Deltaproteobacteria. MF pretreatment had a lesser impact on membrane biofilm composition, as fewer groups had greater than two-fold changes in abundance, with a reduction in Verrucomicrobia and Gammaproteobacteria, the disappearance of Epsilonproteobacteria, and an increase of Saprospiraceae.
3.4.2.
Species richness and diversity
Species richness remained constant for all tested conditions, as determined by the Chao1 richness estimator (0.03 sequence divergence, estimator ranged 361–1542). Also, biofilm communities from all tested conditions had statistically indistinguishable levels of diversity, as determined by the Shannon-Weaver index of diversity (0.02 sequence divergence, indices ranged from 4.25 to 4.64).
3.4.3.
Community structure
The Cramer–von Mises test statistic was used to determine whether the bacterial communities formed under different conditions were different with statistical confidence (Table 3). As discussed above, differences in the abundance of some microbial groups were observed depending on the influent pretreatment, with a greater number of large (>two-fold) shifts for wall biofilms than for RO membrane biofilms. The statistical analysis revealed that the bacterial communities that fouled the RO membranes were not significantly different irrespective of whether the RO influent underwent MF pretreatment. In contrast, the bacterial communities on the feed tank wall were statistically different for RO treatment of MF pretreated and non-treated secondary effluent (p < 0.05). Finally, this analysis demonstrated that the tank wall and membrane biofilms were not different when RO influents with no pretreatment, but statistically different when using MF pretreated RO influents (p < 0.05). Importantly, these results suggest that MF pretreatment does not significantly change the community structure of
water research 44 (2010) 167–176
membrane biofilms, though it affects community structure of non-membrane-associated biofilms. The identical membrane biofilm assembly implies that there may be a widely conserved stable community structure in RO membrane fouling biofilms. Additional community characterization and longer-term experiments are necessary to confirm these findings.
4.
Conclusions
The mechanic properties of biofilms should be considered to apply biofouling control strategies and cleaning methods and are likely affected by biofilm physiology and by the presence of particulate matter within the biofouling layer. The reduced fouling of RO membranes after pretreatment of the secondary effluents with MF was due to removal of particulate/colloidal matter. Higher oxidative activity and cell viability were observed for the biofouling layer formed during RO treatment of MF pretreated secondary effluents. This result indicates the important effect of pretreatment of wastewater effluent on biofilm activity and viability and suggests that pretreatment impacts biofilm mechanic properties and disinfection/cleaning strategies. Additionally, this research suggests that RO membranes treating secondary effluents select for a unique bacterial community irrespective of pretreatment. Further study of membrane-associated biofilm formation, succession, and ecology is necessary to better understand biofouling and develop biofouling minimization strategies.
Acknowledgements We would like to thank Menachem Elimelech and Aurelio Briones for helpful discussions. This research was made possible by the WaterCAMPWS, a Science and Technology Center of Advanced Materials for the Purification of Water with Systems under the National Science Foundation agreement number CTS-0120978. DB was supported by EPA STAR and Graham Environmental Sustainability Institute (University of Michigan) graduate fellowships.
references
Amann, R.I., Ludwig, W., Schleifer, K.H., 1995. Phylogenetic identification and in situ detection of individual microbial cells without cultivation. Microbiol. Rev. 59 (1), 143–169. Ang, W.S., Elimelech, M., 2007. Protein (BSA) fouling of reverse osmosis membranes: implications for wastewater reclamation. J. Memb. Sci. 296 (1–2), 83–92. Ang, W.S., Elimelech, M., 2008. Fatty acid fouling of reverse osmosis membranes: implications for wastewater reclamation. Water. Res. 42 (16), 4393–4403. Ang, W.S., Lee, S., Elimelech, M., 2006. Chemical and physical aspects of cleaning of organic-fouled reverse osmosis membranes. J. Memb. Sci. 272 (1–2), 198–210. APHA. (1999). Standard Methods for the Examination of Water and Wastewater. 20 ed., Washington, DC.
175
Belfer, S., Gilron, J., Daltrophe, N., Oren, Y., 2005. Comparative study of biofouling of NF modified membrane at SHAFDAN. Desalination 184 (1–3), 13–21. Bereschenko, L.A., Heilig, G.H.J., Nederlof, M.M., van Loosdrecht, M. C.M., Stams, A.J.M., Euverink, G.J.W., 2008. Molecular characterization of the bacterial communities in the different compartments of a full-scale reverse-osmosis water purification plant. Appl. Environ. Microbiol. 74 (17), 5297–5304. Bereschenko, L.A., Stams, A.J.M., Heilig, G.H.J., Euverink, G.J.W., Nederlof, M.M., Van Loosdrecht, M.C.M., 2007. Investigation of microbial communities on reverse osmosis membranes used for process water production. Water. Sci. Technol. 55 (8–9), 181–190. Campbell, P., Srinivasan, R., Knoell, T., Phipps, D., Ishida, K., Safarik, J., Cormack, T., Ridgway, H., 1999. Quantitative structure-activity relationship (QSAR) analysis of surfactants influencing attachment of a Mycobacterium sp. to cellulose acetate and aromatic polyamide reverse osmosis membranes. Biotechnol. Bioeng. 64 (5), 527–544. Chao, A., 1984. Nonparametric estimation of the number of classes in a population. Scand. J. Statist. 11, 265–270. Chen, C.L., Liu, W.T., Chong, M.L., Wong, M.T., Ong, S.L., Seah, H., Ng, W.J., 2004a. Community structure of microbial biofilms associated with membrane-based water purification processes as revealed using a polyphasic approach. Appl. Microbiol. Biotechnol. 63 (4), 466–473. Chen, K.L., Song, L., Ong, S.L., Ng, W.J., 2004b. The development of membrane fouling in full-scale RO processes. J. Memb. Sci. 232 (1–2), 63–72. DeSantis Jr., T.Z., Hugenholtz, P., Keller, K., Brodie, E.L., Larsen, N., Piceno, Y.M., Phan, R., Andersen, G.L., 2006. NAST: a multiple sequence alignment server for comparative analysis of 16S rRNA genes. Nucleic. Acids. Res. 34 (Suppl._2), W394–W399. Felsenstein, J., 2008. DNADIST–Program to Compute Distance Matrix from Nucleotide Sequences. Version 3.68. University of Washington. http://evolution.genetics.washington.edu/ phylip/doc/dnadist.html. Fox, E.N., Demaree, R.S.J., 1999. Quick bacterial microwave fixation technique for scanning electron microscopy. Microsc. Res. Tech. 46, 338–339. Glueckstern, P., Priel, M., Gelman, E., Perlova, N., 2008. Wastewater desalination in Israel. Desalination 222, 151–164. Gur-Reznik, S., Katz, I., Dosoretz, C.G., 2008. Removal of dissolved organic matter by granular-activated carbon adsorption as a pretreatment to reverse osmosis of membrane bioreactor effluents. Water. Res. 42 (6–7), 1595–1605. Herzberg, M., Elimelech, M., 2007. Biofouling of reverse osmosis membranes: role of biofilm-enhanced osmotic pressure. J. Memb. Sci. 295 (1-2), 11–20. Herzberg, M., Elimelech, M., 2008. Physiology and genetic traits of reverse osmosis membrane biofilms: a case study with Pseudomonas aeruginosa. ISME J. 2, 180–194. Herzberg, M., Kang, S., Elimelech, M., 2009. Role of extracellular polymeric substances (EPS) in biofouling of reverse osmosis membranes. Environ. Sci. Technol. 43 (12), 4393–4398. Heydorn, A., Nielsen, A.T., Hentzer, M., Sternberg, C., Givskov, M., Ersboll, B.K., Molin, S., 2000. Quantification of biofilm structures by the novel computer program COMSTAT. Microbiology 146, 2395–2407. Hoek, E.M.V., Elimelech, M., 2003. Cake-enhanced concentration polarization: A new fouling mechanism for salt-rejecting membranes. Environ. Sci. Technol. 37 (24), 5581–5588. Huertas, E., Herzberg, M., Oron, G., Elimelech, M., 2008. Influence of biofouling on boron removal by nanofiltration and reverse osmosis membranes. J. Memb. Sci. 318 (1–2), 264–270. Ivnitsky, H., Katz, I., Minz, D., Shimoni, E., Chen, Y., Tarchitzky, J., Semiat, R., Dosoretz, C.G., 2005. Characterization of
176
water research 44 (2010) 167–176
membrane biofouling in nanofiltration processes of wastewater treatment. Desalination 185 (1–3), 255–268. Ivnitsky, H., Katz, I., Minz, D., Volvovic, G., Shimoni, E., Kesselman, E., Semiat, R., Dosoretz, C.G., 2007. Bacterial community composition and structure of biofilms developing on nanofiltration membranes applied to wastewater treatment. Water. Res. 41 (17), 3924–3935. Jarusutthirak, C., Amy, G., 2006. Role of soluble microbial products (SMP) in membrane fouling and flux decline. Environ. Sci. Technol. 40 (3), 969–974. Jarusutthirak, C., Amy, G., Crou, J.-P., 2002. Fouling characteristics of wastewater effluent organic matter (EfOM) isolates on NF and UF membranes. Desalination 145 (1–3), 247–255. Kim, S.L., Paul Chen, J., Ting, Y.P., 2002. Study on feed pretreatment for membrane filtration of secondary effluent. Separ. Purif. Tech. 29 (2), 171–179. Knoell, T., Safarik, J., Cormack, T., Riley, R., Lin, S.W., Ridgway, H., 1999. Biofouling potentials of microporous polysulfone membranes containing a sulfonated polyether-ethersulfone/ polyethersulfone block copolymer: correlation of membrane surface properties with bacterial attachment. J. Memb. Sci. 157 (1), 117–138. Lee, S., Ang, W.S., Elimelech, M., 2006. Fouling of reverse osmosis membranes by hydrophilic organic matter: implications for water reuse. Desalination 187 (1–3), 313–321. Lee, S., Cho, J., Elimelech, M., 2005. Combined influence of natural organic matter (NOM) and colloidal particles on nanofiltration membrane fouling. J. Memb. Sci. 262 (1–2), 27–41. Lee, S., Lee, C.H., 2005. Scale formation in NF/RO: mechanism and control. Water. Sci. Technol. 51, 267–275. Li, Q., Elimelech, M., 2004. Organic fouling and chemical cleaning of nanofiltration membranes: measurements and mechanisms. Environ. Sci. Technol. 38 (17), 4683–4693. Li, Q., Elimelech, M., 2006. Synergistic effects in combined fouling of a loose nanofiltration membrane by colloidal materials and natural organic matter. J. Memb. Sci. 278 (1–2), 72–82. Li, Q., Xu, Z., Pinnau, I., 2007. Fouling of reverse osmosis membranes by biopolymers in wastewater secondary effluent: role of membrane surface properties and initial permeate flux. J. Memb. Sci. 290 (1–2), 173–181. Magurran, A.E., 1988. Ecological Diversity and Its Measurement. Princeton University Press, Princeton, N.J. Metcalf & Eddy, Tchobanoglous, G., Burton, Franklin L., David Stensel, H., 2003. Wastewater Engineering Treatment and Reuse. McGraw-Hill, United States. Ng, H.Y., Elimelech, M., 2004. Influence of colloidal fouling on rejection of trace organic contaminants by reverse osmosis. J. Memb. Sci. 244 (1–2), 215–226.
Pang, C.M., Hong, P., Guo, H., Liu, W.-T., 2005. Biofilm formation characteristics of bacterial isolates retrieved from a reverse osmosis membrane. Environ. Sci. Technol. 39 (19), 7541–7550. Pang, C.M., Liu, W.T., 2007. Community structure analysis of reverse osmosis membrane biofilms and the significance of rhizobiales bacteria in biofouling. Environ. Sci. Technol. 41 (13), 4728–4734. del Pino, M.P., Durham, B., 1999. Wastewater reuse through dualmembrane processes: opportunities for sustainable water resources. Desalination 124 (1–3), 271–277. Redondo, J.A., 2001. Brackish-, sea- and wastewater desalination. Desalination 138, 29–40. Reith, C., Birkenhead, B., 1998. Membranes enabling the affordable and cost effective reuse of wastewater as an alternative water source. Desalination 117 (1–3), 203–209. Ridgway, H.F., Justice, C.A., Whittaker, C., Argo, D.G., Olson, B.H., 1984a. Biofilm fouling of RO membranes-its nature and effect on treatment of water for reuse. J. Am. Water. Works. Assoc. 76, 94–102. Ridgway, H.F., Rigby, M.G., Argo, D.G., 1984b. Adhesion of a Mycobacterium sp. to cellulose diacetate membranes used in reverse osmosis. Appl. Environ. Microbiol. 47 (1), 61–67. Schloss, P.D., Handelsman, J., 2005. Introducing DOTUR, a computer program for defining operational taxonomic units and estimating species richness. Appl. Environ. Microbiol. 71 (3), 1501–1506. Schloss, P.D., Larget, B.R., Handelsman, J., 2004. Integration of microbial ecology and statistics: a test to compare gene libraries. Appl. Environ. Microbiol. 70 (9), 5485–5492. Shon, H.K., Vigneswaran, S., Snyder, S.A., 2006. Effluent organic matter (EfOM) in wastewater: constituents, effects, and treatment. Crit. Rev. Environ. Sci. Tech. 36 (4), 327–374. Spear, J.R., Walker, J.J., McCollom, T.M., Pace, N.R., 2005. From the cover: hydrogen and bioenergetics in the yellowstone geothermal ecosystem. Proc. Natl. Acad. Sci. U.S.A. 102 (7), 2555–2560. Swerdlow, H., Gesteland, R., 1990. Capillary gel electrophoresis for rapid, high resolution DNA sequencing. Nucleic. Acids. Res. 18 (6), 1415–1419. Teodosiu, C.C., Kennedy, M.D., van Straten, H.A., Schippers, J.C., 1999. Evaluation of secondary refinery effluent treatment using ultrafiltration membranes. Water. Res. 33 (9), 2172–2180. Vrijenhoek, E.M., Hong, S., Elimelech, M., 2001. Influence of membrane surface properties on initial rate of colloidal fouling of reverse osmosis and nanofiltration membranes. J. Memb. Sci. 188 (1), 115–128. Xu, P., Drewes, J.E., Kim, T.-U., Bellona, C., Amy, G., 2006. Effect of membrane fouling on transport of organic contaminants in NF/ RO membrane applications. J. Memb. Sci. 279 (1–2), 165–175.
water research 44 (2010) 177–184
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Radon emanation from radium specific adsorbents Abdulrahman I. Alabdula’aly*, Hamed B. Maghrawy King Abdulaziz City for Science and Technology, P.O. Box 6086, Riyadh 11442, Saudi Arabia
article info
abstract
Article history:
Pilot studies were undertaken to quantify the total activity of radon that is eluted following
Received 27 April 2009
no-flow periods from several Ra-226 adsorbents loaded to near exhaustion. The adsorbents
Received in revised form
studied included two types of barium sulphate impregnated alumina (ABA-8000 and F-1)
10 August 2009
and Dowex MSC-1 resin treated by either barium hydroxide or barium chloride. In parallel,
Accepted 12 September 2009
radium loaded plain activated aluminas and Dowex MSC-1 resin were similarly investi-
Available online 1 October 2009
gated. The results revealed that radon was quantitatively eluted during the first few bed volumes of column operation after no-flow periods. Although similar radon elution profiles
Keywords:
were obtained, the position of the radon peak was found to vary and depended on the
Alumina
adsorbent type. Radon levels up to 24 and 14 kBq dm3 were measured after a rest period of
Dowex MSC-1
72 h from radium exhausted Dowex MSC-1 treated with barium chloride and F-1 impreg-
Radium removal
nated alumina with barium sulphate, respectively. The eluted radon values measured
Radon emanation
experimentally were compared to those calculated theoretically from accumulated radium
Specific adsorbents
quantities for the different media. For plain adsorbents, an agreement better than 10% was obtained. For treated resin-types a consistency within 30% but for impregnated aluminatypes high discrepancy between respective values were obtained. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
In response to its deleterious health impacts on humans, radium removal from drinking water sources has become a vital objective. Many conventional water treatment practices, e.g. chemical and ion exchange softening, and reverse osmosis, bring about the removal of radium with high efficiency (85–95%) (Bennett et al., 1976; Bennett, 1978; Myers and Snoeyink, 1987; Brinck et al., 1978; Subramonian et al., 1990). However, these methods are expensive and suitable only for large capacity public water supplies (>2106 dm3 pd). On the other hand, a need for an efficient, selective, easy, and cost-effective method to remove radium from point-of-entry and small community water supplies is becoming inevitable. The use of radium selective complexer (RSC), since its production in 1984 by Dow Chemical Company (1986) has been appreciated as an ideal example of radium adsorbents
(Clifford et al., 1988; Clifford, 1990). An average capacity of the exhausted RSC at 40 Bq ml1 (133 Bq g1) has been achieved (Clifford et al., 1988). However, disposal problems of the spent RSC has resulted in the discontinuation of its production by DOW in 1987. Other types of specific adsorbents that are characterized by high selectivity towards radium include BaSO4 impregnated alumina are becoming promising as well (Valentine et al., 1992; Fleming, 1986; Garg and Clifford, 1992; Mott et al., 1993). The main disadvantage of these adsorbents is the potential build up of Rn-222 within the media with an ultimate contamination of the product water. The continuously generated radon within these media is either washed out during continuous operation or accumulated during no-flow periods where extremely high radon levels are possible in the first few bed volumes following these periods (Clifford et al., 1988; Clifford, 1990). The problem of radon washout during early periods of operational resumption of
* Corresponding author. Tel.: þ966 1 481 3300; fax: þ966 1 481 3878. E-mail address:
[email protected] (A.I. Alabdula’aly). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.031
178
water research 44 (2010) 177–184
radium specific adsorbent columns has not received much attention. Clifford (1990) has stated that the problem of radon washout upon restarting radium contaminated ion exchange softeners or adsorbers is being quantified at the University of Houston. Abulleif (1989) however, has investigated the elution of radon in case of using simulated ion exchange household softeners and found that with proper aeration the softener contribution to the radon level in household pose minimal health risks. The present paper addresses the problem of Rn-222 emanation from some Ra-226 loaded adsorbents intended for removing radium from water sources. The objectives of the work include determination of the radon elution patterns, quantification of the amount of radon eluted from different types of adsorbents, and comparison of the radon values that already washed out to those calculated from radium decay at different no-flow periods.
2.
Materials and methods
Two types of activated alumina, ABA-8000 (Selecto, Kennesaw, Georgia) and F-1 (Alcoa, New Kensington, Pennsylvania) were used. Dowex MSC-1, a strong acid cation exchange resin in the Naþ form was obtained from Dow Chemical Company, Midland, Michigan, USA. The two activated aluminas were impregnated with BaSO4 as described in the literature (Garg and Clifford, 1992). The procedure of impregnation was as follows: 1. 2.5 d3 of the sieved activated alumina (0.3–0.6 mm) was equilibrated with 25 dm3 of 0.25 N H2SO4 for 30 min. 2. The filtered alumina was treated with 25 dm3 of 0.3 N BaCl2 for 3.5 h. 3. The treated alumina was filtered and washed with a sufficient amount of deionized water. 4. The prepared BaSO4 – impregnated alumina was dried in air at room temperature (22 2 C). The resin was chemically treated to change it into like radium specific complexer as described in the Canadian patent (Hatch, 1984). In the present work the resin was treated by either Ba(OH)2 to give MSC-T1 or BaCl2 to give MSC-T2. The procedure was as follows:
The exhaustion run was divided into two phases. The first one comprised the following media: Plain activated alumina ABA-8000 (ABA-P) BaSO4 – impregnated alumina ABA – 8000 (ABA-T) Plain Dowex MSC-1 resin (MSC-P) Ba(OH)2 treated Dowex MSC-1 resin (MSC-T1). While in the second phase the following media were used: Plain activated alumina F-1 (F1-P) BaSO4 – impregnated alumina F-1 (F1-T) BaCl2 treated Dowex MSC-1 resin (MSC-T2). In addition, the MSC-T1 medium was subjected to a similar investigation again after loading with Ra-226 for extra 3922 h to complete a total of 6916 h. In this stage, it was denoted as MSC-T1A. The used influent water was pretreated (aerated and filtered) groundwater containing Ra-226 at an average concentration of 1.22 Bq dm3. The influent water characteristics are presented in Table 1. The pretreatment process was intended for the removal of dissolved iron and manganese by oxidation (aeration) followed by sand filtration. A schematic diagram of the pilot plant is shown in Fig. 1. The influent flow rate throughout each column was set constant at 400 ml/min (11.7 BV/h). This gives an empty bed contact time (EBCT) of 5 min. Throughout the continuous radium loading run, samples from effluent stream of each column were collected at about every 2 weeks to assess for radon activity during normal operation. After a continuous operation of 2994 h (35030 BV) in case of the first phase and 3080 h (36036 BV) for the second one, the run was terminated. The sampling procedure for radon analysis was started after 12, 18, 24, 48 and 72 h of no-flow periods. After the elapse of each of these periods influent water was passed through the column at a flow rate of 200 ml/ min (5.85 BV/h), through which samples were collected during the first 2 h of operation from the effluent of each column at 1, 4, 7, 10, 15, 20, 30, 40, 50, 60, 80, 100 and 120 min from start of water flow.
Table 1 – Influent water physical and chemical analysis. 1. 2.5 dm3 of the wet resin was soaked in 15 dm3 of 0.15 M Ba(OH)2 (or BaCl2) overnight. 2. The Ba-form resin was washed with deionized water. 3. The filtered Ba-form resin was covered with concentrated H2SO4 solution (5 M) and shaken for 5 h. 4. The filtered resin was washed with deionized water. 5. The treated resin was dried in air at room temperature (22 2 C). The use of BaCl2 as an alternative for Ba(OH)2 was to avoid formation of the white precipitate of BaCO3 during the process of dissolution in water. Pilot-scale adsorption experiments were performed using four 5.1 cm diameter identical plexiglass columns. The Media depth was 101.6 cm supported on 30.5 cm depth graded gravel.
Parameter
Concentration (mg dm3)a
Temp ( C) pH Total alkalinity as CaCO3 Total dissolved solids Total hardness as CaCO3 Calcium hardness as CaCO3 Magnesium hardness as CaCO3 Sodium Iron Chloride Nitrate as N Sulphate Radium-226, pCi/L a Except temperature, pH and Ra-226.
30 6.92 140 670 307 225 82 121 0.16 165 29 153 32.94
water research 44 (2010) 177–184
179
Fig. 1 – Schematics of pilot plant setup.
Meanwhile four representative samples were collected from the influent water during the 2 h operation. In case of resin MSC-T1 column, it was sampled twice; the first after 2994 h (35030 BV) and the second after a total of 6916 h (80917 BV). Radon-222 activity levels in water samples were determined by using the liquid scintillation counting (LSC) technique described by Prichard and Gesell (1977) and Prichard et al. (1992). A water sample of 10 ml was injected under 10 ml of scintillation fluor contained in a 20 ml vial. The used vials are of Teflon coated polyethylene from Zinsser (Germany). The scintillation fluor is of the type NEF-957A from New England Nuclear/Dupont (USA). The vial was recapped, shaken for 30 sec and stored in a refrigerator at 4 C for 8 h before counting. A commercial liquid scintillation counter Wallac 1220 Quantulus (Finland) was used in radon analysis. The counter was frequently calibrated against a standard solution of Ra-226 from the National Institute of Standard and Technology Traceable, USA. The counting time used for samples and background was 120 and 400 min, respectively. The Rn-222 concentration is expressed in Bq dm3 with an associated error of 2-sigma (2s) confidence interval. With respect to Ra-226 analysis, periodical water samples from both influent and effluent streams were collected and analyzed in according to the method recommended by USEPA (method 903.0).
3.
Results and discussion
The results of radon activity levels in the effluent streams during normal operation are shown in Fig. 2. It was observed
that after continuous radium loading for about one month (w8400 BV), the effluent stream showed relatively higher radon concentration than the influent water. This phenomenon was clearly observed in the case of adsorbents that were characterized by high radium capacity, treated resin and impregnated F-1 alumina. Fig. 3 demonstrates the radon elution profiles obtained from the Ra-226 loaded adsorbents after the resumption of operation for different no-flow periods. It is clear that for every adsorbent, similar elution profiles for the different no-flow periods were obtained. Also, radon concentration increases, reaching a peak and then decreases with run time. The height of radon peak, however, is a function of no-flow period. For example, after 72 h of no-flow period, the maximum radon activity in the affluent water from MSC-T2 resin and F1-T alumina was found to be 3704 and 5593 Bq dm3, respectively, in comparison to the respective values of 722.2 and 759.3Bq dm3 for 12 h of no-flow period. It was observed that the behaviour of each adsorbent in flushing radon seems different and could be classified according to the adsorbent’s chemical properties, namely; alumina (plain and treated) and resin (plain and treated). With the aluminas, more than 90% of the total activity of radon on the column bed (experimentally measured) was washed out within the first 2–3 BV throughput. With the resins, the majority of radon activity (>90%) was washed out in the first 5 BV throughput. In addition, the position of the radon peak with respect to the run time is different and belongs to the same adsorbent classification. In case of aluminas, the peak corresponds to about 5 min (0.5 BV) and to 15–20 min (1.5 BV) for the resins. This phenomenon was repeatedly obtained for all the investigated no-flow periods.
180
water research 44 (2010) 177–184
Fig. 2 – Change of radon-222 concentration as a function of effluent bed volume for different media loaded with radium-226 for one month. (A) Resin adsorbents. (B) Alumina adsorbents.
These observations are well represented in a histogram, Fig. 4, showing the radon activity corresponding to each peak for the different adsorbents. It is clear that the radon peak for F1-T is the highest (e.g. 5.6 kBq dm3 for no-flow period of 72 h) in comparison to the adsorbents loaded with radium for almost the same time. The variations observed in the elution patterns of radon contradict the fact that radon is an inert gas and has no chemical affinity to the material of any adsorbent. It seems that in the present study the process of radon flush out depends on the nature of the adsorbent material. In other words, the kinetics of radon elution process may depend on the physical and chemical properties of the adsorbent itself. As an illustration, radon is flushed out at higher rate from alumina than from resin type adsorbents. This phenomenon needs further investigation. The area under each elution curve that corresponds to a certain no-flow period given in Fig. 3 represents well the total amount of radon activity that eluted during the flush out period (Fig. 5). Although the radium loading runs for both groups of adsorbents, as previously mentioned in the experimental part, were different (2994 and 3080 h); no significant differences between the flushed out radon from all adsorbents was observed. This refers to the fact that the loading period of 2994 h was adequate to bring about some adsorbents, namely; ABA-P, F1-P, and MSC-P, to the radium breakthrough. For the other high capacity adsorbents, namely; ABA-T, F1-T, MSC-T1, and MSC-T2, the difference in loading periods between 2994 and 3080 h did not give rise to more than 2% in radium content.
The total activity of radon that eluted from any Ra loaded adsorbent is expected to be a function of the amount of Ra present in the medium and radon ingrowth period. Hence the efficiency of a medium to remove radium plays an important role in quantifying the flushed out radon. From Fig. 5, it is clear that MSC-T2 resin elutes the highest total activity of radon in comparison to the other studied adsorbents loaded for almost the same period with Ra-226. Also, it could be concluded that the chemically treated adsorbents have fairly higher capacity for radium than the corresponding untreated ones. For a certain no-flow period, the studied adsorbents could be arranged according to the decreasing order of flushed out radon activity. Treated adsorbents; MSC-T2 > F1-T > MSC-T1 > ABA-T Plain adsorbents; F1-P > MSC-P > ABA-P It is worthy to compare the total activity of eluted radon from a Ra-loaded medium after a certain no-flow period to the corresponding theoretically calculated value. The general equation that controls the natural ingrowth of Rn-222 from Ra-226 decay is: t1=2 / Rn 222 þ a Ra 226 1620y
t1=2 / 3:82d
Progeny
(1)
i.e. the rate of radon production ¼ rate of its generation rate of its decay.
water research 44 (2010) 177–184
181
Fig. 3 – Radon elution profiles obtained from Ra-226 loaded adsorbent for different no-flow periods. (A) Resin adsorbents (B) Alumina adsorbents.
182
water research 44 (2010) 177–184
Fig. 4 – A histogram showing the radon activity corresponding to each peak for different adsorbents.
In the differential form equation (1) could be expressed as: d½Rn ¼ k1 ½Ra k2 ½Rn dt
(2)
Where k1 and k2 are the rate of decay of both radium and ; t1/2 is the half-life of radon, respectively, which equal to 0:693 t1=2 either isotopes.
equation (3) that the concentration of radon generated is directly proportional to the mass or activity of radium [Ra] and the no-flow period, t. Also, it shows that the generated radon activity approaches that of radium values after no-flow period of approximately 25 days. Equation (3) has been used by Abulleif (1989) to calculate the potential release of radon from radium loaded resins used in home softeners. Also, it has been used by Valentine and Stears (1994) to quantify the radon release from a water distribution system containing radium deposits. From the analytical data concerning Ra-226 activity in both the influent and effluent streams the corresponding activity that retain onto each medium was estimated and used in equation (3). A comparison between the experimental values of radon eluted at different no-flow periods and the corresponding values theoretically calculated using equation (3), is shown in Table 2. This could be accomplished on the assumption that 100% elution of radon produced from the Ra-226 loaded on the column takes place. In general, it was noticed that in all cases, the eluted radon activity was less than the corresponding calculated values. However, the untreated adsorbents showed the least discrepancy among the other chemically pretreated ones. It is convenient to express these inconsistencies between the experimental and theoretical values in terms of percent relative deviation:
Percent relative deviation ¼
The solution of equation (2), after appropriate approximation is: ½Rn ¼ ½Ra 1 e018t
(3)
Where [Rn] is the concentration of Rn-222 given as Bq dm3 of water that is produced after elapsed time t (days) and [Ra] is the radium-226 concentration that is retained onto the medium, given as Bq dm3 of adsorbent. It is clear from
theoretical value experimental value 100 theoretical value
Table 3 demonstrates the percent relative deviation for each adsorbent at the studied no-flow periods. The data show that the untreated adsorbents, namely; MSC-P, ABA-P, and F1-P showed fair agreement, with an average deviation not exceeding 10%. On the other hand, the Ba – treated resin,
Table 2 – A comparison between the flushed out Rnactivities (kBq) experimentally obtained and theoretically calculated for different media. Medium
MSC-P MSC-T1 MSC-T2 MSC-T1A ABA-P ABA-T F1-P
Fig. 5 – The total amount of radon activity that eluted during the flush out periods for different adsorbents.
F1-T
No-flow period (days)
Exptl. Calcd. Exptl. Calcd. Exptl. Calcd. Exptl. Calcd. Exptl. Calcd. Exptl. Calcd. Exptl. Calcd. Exptl. Calcd.
0.5
0.75
1
2
3
1.1 1.0 2.4 5.4 5.0 6.7 7.6 11.3 0.8 0.7 1.0 3.3 1.8 1.9 2.2 6.7
1.4 1.4 3.2 8.0 7.1 9.8 11.2 16.6 0.9 1.0 1.4 4.9 2.7 2.8 4.0 9.8
1.8 1.8 6.9 10.4 9.7 12.8 15.8 21.7 1.2 1.3 1.9 6.4 3.7 3.7 5.9 12.8
3.1 3.4 9.1 19.1 17.2 23.5 27.4 39.9 1.5 2.4 3.4 11.7 6.4 6.7 10.3 23.5
3.7 4.7 10.7 26.3 23.7 32.5 35.1 55.0 2.7 3.2 3.9 16.1 8.2 9.3 13.8 32.5
water research 44 (2010) 177–184
Table 3 – % Relative deviation of radon activity measured experimentally to theoretical values at different no-flow periods. Medium
MSC-P MSC-T2 MSC-T1 MSC-T1A ABA-P ABA-T F1-P F1-T
No-Flow Period (days) 0.5
0.75
1
2
3
Average
11 26 56 33 15 69 7 66
0.3 27 60 33 8 72 5 59
2 24 34 27 5 70 1 54
7 27 53 31 35 71 4 56
20 27 59 36 17 76 12 57
4% 26% 52% 32% 10% 72% 5% 58%
namely; MSC-T1, MSC-T2, and MSC-T1A showed, on the average, an agreement within 52, 26, and 32%, respectively. In case of Ba-treated aluminas, the inconsistency is higher and reached 72% in case of ABA-T. The good agreement found in the untreated adsorbents is due to the fact that their capacities for Ra-226 are lower than that found for the treated ones, which leads to the production of lower radon activities. On the contrary, in case of treated adsorbents, their capacities for radium are much higher, e.g. by a factor of 5.6–7 in case of Dowex MSC and 3.5–5 in case of alumina. More specifically, the Ra-226 loading on the treated adsorbents is in the order of 37 kBq dm3 of adsorbent. The presence of very high amount of Ra-226 on the adsorbent could adversely affect the accuracy of Rn-222 analysis and sampling as well. The phenomenon that resin-type adsorbents gave lower relative deviations than in case of alumina-types is interpreted by the fact that elution of radon from resin-type adsorbents is relatively slower than in case of aluminas, viz. Fig. 3. In other words it is clear in case of resin adsorbents the majority of Rn-activity that eluted is distributed throughout 5 BV, while in case of aluminas the corresponding activity is distributed throughout 1–3 BV. At higher ativities, the concentration gradient between radon in the water and air is much greater, so that it is readily volatilized during sampling process. At lower activities, the driving force is much lower and hence sampling process is less critical (Kinner et al., 1991; Hightower and Watson, 1995). Referring to the published data concerning the release of radon from Ra-226 specific adsorbents, Abulleif (1989) has found considerable disagreement between the flushed out radon found experimentally and the theoretical results on Dow RSC. The RSC was loaded with Ra-226 at 28 Bq dm3. This was ascribed to the high concentration of radium on the resin although the experiments were performed under air tight conditions to ensure no radon leakage to the air.
4.
183
alumina ABA-8000 and F-1 were used as plain and impregnated with barium sulfate. Dowex MSC-1 resin was used in the form of plain and treated with either barium hydroxide or barium chloride. The following conclusions were derived from this study: 1. Radon-222 has been shown to elute in large quantities out of Ra-226 loaded media. Radon elution was found to increase with the increase of both radium concentration and no-flow period. 2. Similar elution profiles were obtained in all of the experimental runs. The position of the radon peak with respect to the run time is different and depends on the adsorbent type. In the case of aluminas, the peak corresponds to about 0.5 BV throughput and in case of resin, it correspondents to 1.5 – 2.0 BV throughput. This phenomenon was repeatedly obtained for all the investigated no-flow periods. 3. For alumina-type adsorbents, more than 90% of the total activity of radon was washed out within the first 2.3 BV throughput. In resin-type, the majority of radon activity (>90%) was eluted during the first five bed volumes throughput. 4. Radon elution kinetic from a Ra-loaded adsorbent depends on its physical and chemical properties. 5. The experimental values of washed out radon were compared to the corresponding theoretically calculated values at different no-flow periods (0.5–3 days). In all cases, the experimental results were less than the corresponding calculated values. The untreated adsorbents showed the least inconsistency (w10%) while impregnated aluminas showed the highest disagreement. This was related to the amount of Ra-226 that burden on the medium and the kinetics of Rn-elution. 6. The total activity of radon flushed out from any adsorbent reflects its capacity in potential removal of Ra-226 from water sources. It was observed that the treated MSC-1 resin with BaCl2 elutes the highest total activity of radon. 7. The relative variation in the position of radon elution peak with respect to BV throughput for both types of adsorbent may be explained by the difference in radium distribution throughout the column bed. Therefore, a further study could be proposed through core sampling and g-activity assessment on radium exhausted bed columns. It is recommended that whenever specific adsorbents are used for radium removal purposes, the first few bed volumes of water after no-flow periods should be discarded and the amount depends on the adsorbent type. This would solve one of the specific adsorbents disadvantages in water treatment applications.
Summary and conclusions Acknowledgement
Two types of radium selective adsorbents were investigated to quantify the total activity of radon that is eluted with respect to different no-flow periods. These included alumina and strong acid cation exchange resin. Two types of activated
This work has been financially supported by King Abdulaziz City for Science and Technology (KACST), Riyadh, Saudi Arabia.
184
water research 44 (2010) 177–184
references
Abulleif, K., 1989. The quantification and removal of radon accumulated on ion exchange resin. M.Sc. thesis, Cullen College of Engineering, Univ. of Houston, USA. Bennett, D.L., 1978. The efficiency of water treatment processes in radium removal. J. AWWA 70 (12), 698–701. Bennett, D.L., Bell, C.R., Markwood, I.M., 1976. Determination of radium removal efficiencies in illinois water supply Treatment processes. technical Note ORP/TAD-76–2. Illinois EPA, Springfield, Illinois, USA. Brinck, W.L., Schliekelman, R.J., Bennett, D.L., Bell, C.R., Karkwood, I.M., 1978. Radium removal efficiencies in water treatment processes. J. AWWA 69 (1), 31–35. Clifford, D.A., 1990. Removal of radium from drinking water. Chapter 16. In: Cothern, C.R., Rebers, P.A. (Eds.), Radon, Radium and Uranium in Drinking Water. Lewis Publishers, Chelsea, Mich, USA, pp. 225–247. Clifford, D., Vijjeswarapu, W., Subramonian, S., 1988. Evaluating various adsorbents and membranes for removing radium from groundwater. J. AWWA 80 (7), 94–104. Dow Chemical Company, 1986. Material safety data sheet: XFS 43230.00 Experimental Radium Complexer. Midland, MI., USA. Fleming, H.L., 1986. Application of aluminas in water treatment. Environ. Progr. 5 (3), 159–166. Garg, D., and Clifford, D., 1992. Removing radium from water by plain and treated activated alumina. USEPA Report, EPA/600/ R-92/048. Hatch, M.J., 1984. Resin particulates capable of removing metal ions from aqueous solution. Canadian Patent 1 (176), 799. Assigned to the Dow Chemical Co.
Hightower, J.H., Watson Jr., J.E., 1995. Rn-222 in water: a study of two sample collection methods, effects of mailing samples, and temporal variation of concentrations in north Caroline groundwater. Health Phys. 69 (2), 219–226. Kinner, N.E., Malley, J.P., Clement, J.A., Quern, P.A., Schell, G.S., Lessard, C.E., 1991. Effect of sampling technique, storage, cocktails, sources of variation and extraction on the liquid scintillation technique for radon in water. Environ. Sci. Technol. 25, 1165–1171. Mott, H.V., Singh, S., Kondapally, V.R., 1993. Factors affecting radium removal using mixed iron-manganese oxides. J. AWWA 85 (10), 114–121. Myers, A.G., Snoeyink, V.L., 1987. Radium removal from drinking water by lime-soda and Ion exchange softening. In: AWWA Seminar Proceedings, ‘‘Radionuclides in Drinking Water’’. Annual Conference, Kansas City, MO., USA, pp. 47–67. Prichard, H.M., Venso, E.A., Dodson, C.L., 1992. Liquid scintillation analysis of Rn-222 in water by alpha-beta discrimination. J. Radioact. Radiochem. 3 (1), 28–36. Prichard, H.M., Gesell, T.F., 1977. Rapid measurements of Rn-222 concentrations in water with a commercial liquid scintillation counter. Health Phys. 33 (12), 577–581. Subramonian, S., Clifford, D., Vijjeswarapu, W., 1990. Radium removal in Lemont, Illinois: results of studies using ion exchange resins. J. AWWA 82 (5), 61–70. Valentine, R.L., Kurt, A., Meyer, J., Walsh, D., Mielke, W., 1992. Radium removal using preformed hydrous manganese oxides. AWWA Research Foundation and American Water Works Assoc., Denver, CO., USA. Valentine, R.L., Stears, S.W., 1994. Radon release from water distribution system deposits. Environ. Sci. Technol. 28, 534–537.
water research 44 (2010) 185–194
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Identifying fouling events in a membrane-based drinking water treatment process using principal component analysis of fluorescence excitation-emission matrices Ramila H. Peiris a, Cynthia Halle´ b, Hector Budman a, Christine Moresoli a, Sigrid Peldszus b, Peter M. Huck b, Raymond L. Legge a,* a
Department of Chemical Engineering, University of Waterloo, 200 University Avenue West, Waterloo, Ontario N2L 3G1, Canada NSERC Chair in Water Treatment, Department of Civil and Environmental Engineering, University of Waterloo, 200 University Avenue West, Waterloo, Ontario N2L 3G1, Canada b
article info
abstract
Article history:
The identification of key foulants and the provision of early warning of high fouling events
Received 25 May 2009
for drinking water treatment membrane processes is crucial for the development of
Received in revised form
effective countermeasures to membrane fouling, such as pretreatment. Principal foulants
10 September 2009
include organic, colloidal and particulate matter present in the membrane feed water. In
Accepted 14 September 2009
this research, principal component analysis (PCA) of fluorescence excitation-emission
Published online 19 September 2009
matrices (EEMs) was identified as a viable tool for monitoring the performance of pretreatment stages (in this case biological filtration), as well as ultrafiltration (UF) and
Keywords:
nanofiltration (NF) membrane systems. In addition, fluorescence EEM-based principal
Principal component analysis
component (PC) score plots, generated using the fluorescence EEMs obtained after just
Fluorescence spectroscopy
1 hour of UF or NF operation, could be related to high fouling events likely caused by
Membrane fouling
elevated levels of particulate/colloid-like material in the biofilter effluents. The fluores-
Drinking water treatment
cence EEM-based PCA approach presented here is sensitive enough to be used at low
Nanofiltration
organic carbon levels and has potential as an early detection method to identify high
Ultrafiltration
fouling events, allowing appropriate operational countermeasures to be taken. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Membrane treatment of surface and ground water by means of ultrafiltration (UF) and nanofiltration (NF) is increasingly being used as an option for the production of drinking water. However, implementation of these membrane-based processes for drinking water treatment is often constrained due to fouling, which may be caused by organic, inorganic, colloidal and particulate matter. In drinking water UF and NF applications, natural organic matter (NOM) is considered to be
the major membrane foulant (Saravia et al., 2006; Jermann et al., 2007). NOM consists of a complex mixture of humic and fulvic acids, proteins, and carbohydrates of various molecular size and functional groups (Her et al., 2003). Characterization of membrane foulant fractions in NOM such as humic substances (HS) and biopolymers (protein and polysaccharides) is indispensable for understanding membrane fouling and for the development of fouling control strategies (Amy, 2008). Application of fluorescence spectroscopy as a tool for characterizing NOM is well documented (Coble et al., 1990;
* Corresponding author. Fax: þ519 746 4979. E-mail address:
[email protected] (R.L. Legge). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.036
186
water research 44 (2010) 185–194
Baker, 2001; Chen et al., 2003; Her et al., 2003; Sierra et al., 2005; Hudson et al., 2007; Henderson et al., 2009). Compared to other available NOM characterization techniques, this technique offers rapid and consistent analyses with high instrumental sensitivity (Peiris et al., 2008). In this study, the fluorescence excitation-emission matrix (EEM) analysis method was used for characterization of NOM and the associated fouling events in UF and NF-based drinking water treatment processes, as this method is able to capture specific fluorescence features that correspond to humic and protein-like materials in a single matrix in terms of fluorescence intensities. The light scattering regions captured in the fluorescence EEMs can also be used to provide information related to the particulate/colloidal matter present in water (Wyatt, 1993; Stramski and Wozniak, 2005). In addition, unlike the fluorescence single scan approach (i.e. scanning only at the fluorescence peak location), the fluorescence EEM method provides a basis for capturing subtle changes in the fluorescence spectra of the water that may occur due to seasonal effects or other changes. The ability of this approach to characterize natural water NOM with a wide range of dissolved organic carbon (DOC) concentrations (i.e. 8.0 DOC-mg/L for raw water to 0.4 DOC-mg/L for NF permeate) without predilution or pre-concentrations steps has also been demonstrated (Peiris et al., 2008). Most reported techniques examine fluorescence EEMs intensity data points at a few excitation-emission coordinate pairs (i.e. main peaks) from fluorescence spectra that may contain thousands of wavelength-dependent fluorescence intensity data points. These techniques lack the ability to capture the heterogeneity of the different NOM fractions in water. The importance of analyzing the full fluorescence EEMs as opposed to individual main peak positions has therefore been highlighted in several studies (Persson and Wedborg, 2001; Chen et al., 2003; Stedmon et al., 2003; Boehme et al., 2004). Due to these reasons, full fluorescence EEMs of the water samples were analyzed in this study. Multivariate data analysis methods such as principal component analysis (PCA) (Persson and Wedborg, 2001; Boehme et al., 2004) and parallel factor analysis (Stedmon et al., 2003) have been used to analyze the full fluorescence EEMs to characterize water samples obtained from different sources/sampling locations. In contrast to these studies, the objective of the present study was to de-convolute the spectral information to identify major foulants present, and thereby to assess the performance of different feed water pre-treatment stages and the subsequent UF/NF stages. Since this objective could be satisfactorily met with PCA, this was the only data-mining technique used in this study. The application of this approach as a potential tool for early detection of high membrane fouling events is also described.
2.
Materials and methods
2.1.
Feed water and pre-treatment
Water from the Grand River (Southwestern Ontario, Canada) was used as the feed water for UF and NF experiments conducted between August 2007 and August 2008. Typical Grand
Table 1 – Grand River water quality parameters from August 2007 – August 2008. Parameters Temperature ( C) pH Turbidity (NTU) DOC (mg/L) Conductivity (ms/cm)
Grand River raw water 1–23 7.30–8.40 1.45–67 5–9 500–1200
River water (GRW) quality parameters recorded during the experimental period are presented in Table 1. Fig. 1 demonstrates a process flow chart of the experimental set-up used in this study. GRW was first filtered through a roughing filter to lower the turbidity level of raw water (RW) prior to biofiltration. The roughing filter was operated in an up-flow mode at 1.1 mh1. More details about the roughing filter can be found in Peiris et al. (2008). The roughing filter effluent (RF) was then processed through one of the two parallel biofilters which consisted of dual media filters (i.e. anthracite and sand) over a support layer of gravel. The biofilters were operated in a downflow mode at 5 mh1. The empty bed contact times (EBCT) of the two biofilters were 5 min (BF1) and 14 min (BF2), respectively. Further details on biofilter design are available (Halle´ et al., 2009). The effluents of BF1 (B1) and BF2 (B2) were then used as the feed for both UF and NF experiments. Table S.1 and S.2 under Supplementary Data summarizes when B1 and B2 were used as the membrane feed for different UF and NF experiments. These tables also identify which UF and NF experiments experienced high fouling events. The biofilters operated continuously, independently of whether the membranes units were in operation.
2.2.
Pilot-scale membrane filtration set-up
2.2.1.
Ultrafiltration membrane
A bench scale UF membrane module made of commercial hollow fibre membranes was used for this study (ZeeWeed – 1
Fig. 1 – Experimental set-up. Circles indicate the sampling points for water samples. The acronyms represent the following: BF1 – biofilter with empty bed contact time (EBCT) of 5 min; BF2 – biofilter with EBCT [ 14 min; RW – raw GRW water; RF – roughing filter effluent; B1- effluent of BF1; B2 – effluent of BF2; UFp – UF permeate; NF_C – concentrate of NF; NF_tank - water in the NF feed tank; NFp – NF permeate.
water research 44 (2010) 185–194
by GE-Zenon, Oakville, Canada). The membrane consisted of PVDF and had a MWCO of 200 kDa. The membrane module had a surface area of 0.047 m2, operated in outside-in mode and was mounted in a cylindrical holder of 1.6 L. The membrane was operated in a dead-end filtration mode at constant flux. The permeate flux was temperature adjusted to correspond to 57.5 LMH at 20 C, and the membrane was operated at a recovery of 94%. The four-step operation cycle was automated and consisted of: (1) permeation for 1 h, (2) back pulsing with air sparging for 20 s, (3) draining 0.4 L from the tank, and (4) filling the tank for 9 min. Each experiment was conducted for a 5 d period during which the UF unit was continuously fed by one of the biofilters. The trans-membrane pressure (TMP) was measured using pressure transducers. Under high fouling conditions (identified later), the TMP required to maintain the preset flux, exceeded the recommended operating range for the UF module and as a consequence, the permeate flux declined. A schematic with a more detailed description of the UF membrane set-up is available elsewhere (Halle´ et al., 2009).
2.2.2.
Nanofiltration membrane
NF experiments were performed using a bench scale module (GE SEPA CFII). The system setup and operational conditions are described elsewhere (Peiris et al., 2008). XN45 and TS80 flat sheet membranes from TriSep Corporation (California, USA) were used. The active layer of the membranes was polyamide and the MWCO provided by the manufacturer was 200 Da for both membranes. XN45 and TS80 are hydrophobic membranes each with a contact angle of 57 1 . Membrane hydrophobicity was characterized in terms of sessile drop contact angle measurement by placing a droplet of ultrapure water (5 mL) onto the membrane surface. The measurement was performed using a VCA2500 XE instrument (AST). Each contact angle was measured three times and an average value was calculated. Prior to the experiment, the membranes were compacted using deionized water until stable permeate flow was achieved.
2.2.
Fluorescence analysis
Fluorescence EEMs of the water samples, obtained from the sampling points indicated in Fig. 1, were acquired using a Varian Cary Eclipse Fluorescence Spectrophotometer (Palo Alto, CA) collecting 301 individual emission intensity values (within the 300 – 600 nm emission range) at sequential 10 nm increments of excitation wavelengths between 250 nm and 380 nm. Disposable UV-grade polymethylmethacrylate (PMMA) cuvettes with four optical windows were used in the analyses. The PMMA cuvettes, used in this study, gradually filter the emission signals captured below the excitation wavelength (Ex): 285 nm and therefore the fluorescence intensities at emission wavelength (Em) range: 300 – 600 nm captured below Ex: 285 nm were seen to be lower than emission intensities captured using quartz cuvettes at the same conditions (results not shown). This approach provides sufficient spectral information necessary to distinguish different fluorescent elements of the NOM and reduces the risk of cuvette contamination as a source of error (Peiris et al., 2008). The following instrument parameters were maintained
187
during the fluorescence signal acquisition: photomultiplier tube (PMT) voltage ¼ 800 V; scan rate ¼ 600 nm/min and excitation/emission slit width of 10 nm each. These parameter settings were identified in a separate study as the optimum instrument settings for obtaining reproducible fluorescence signals, especially for low NOM concentration levels (Peiris et al., 2009). To eliminate water Raman scattering and to reduce other background noise, fluorescence spectra for MilliQ (Millipore) water, obtained under the same conditions, were subtracted from all spectra. The temperature of the samples was maintained at room temperature ( w 25 C) during the analyses. Since the pH of all the water samples did not change significantly (pH w 7.3– 8.4), no pH adjustment was made prior to the fluorescence analysis. A separate study demonstrated that there is no significant difference in the fluorescence EEM intensities (< 2%) of GRW captured in the above pH range (results not shown). This is in agreement with the previously published data (Spencer et al., 2007). Following this procedure, fluorescence EEMs of 128 samples drawn from 15 different UF experiments and 192 samples drawn from 15 different NF experiments were recorded at different filtration time intervals (i.e. 1, 24, 48 and 96 h). During the course of these experiments and before fluorescence analyses, the Raman scattering peak intensity recorded for Milli-Q water at Ex/ Em w 348 nm/396 nm was examined to identify any significant fluctuations in the performance of the spectrophotometer lamp or other hardware. No significant changes in this intensity reading (less than 1%) were observed confirming that there were no significant fluctuations in the performance of the spectrophotometer during this study.
2.3.
Fluorescence data pre-treatment and PC analysis
The fluorescence EEM of each sample contained 4214 excitation and emission coordinate points. The fluorescence intensity values corresponding to all 4214 coordinate points (spectral variables) of each EEM were rearranged to generate data rows of intensity values (Supplementary data Figure S.1). This procedure generated a 128 4214 data matrix from UF experiments (XUF) and 192 4214 data matrix from NF experiments (XNF). Each row of these data matrices corresponded to each sample and the intensity values of the corresponding EEM were arranged over 4214 columns. The XUF and XNF data matrices were then separately subjected to PCA. PCA is a well-known technique for data compression and information extraction from a large number of variables. Essentially, PCA extracts a smaller set of underlying new variables that are uncorrelated, mutually independent (orthogonal) and mathematically represented by linear combinations of original variables in the X matrix (XUF or XNF matrix in this case). These new variables, referred to as principal components (PCs), are calculated to account for much of the variance present in the X matrix (Wold et al., 1987; Eriksson et al., 2001) and therefore are able to describe major trends in the original spectral data sets XUF and XNF. PCA decomposes the data matrix X as the sum of the outer product of vectors ti and pi plus a residual matrix E as presented in Eq. (1).
188
X¼
water research 44 (2010) 185–194
k X
ti ,pi þ E
(1)
i¼1
where k is the number of samples in the X data set. The ti vectors are known as scores (i.e. values) on the PCs (i.e. new variables) extracted by PCA. The pi vectors are known as loadings and contain information on how the variables (spectral variables in this case) relate to each other (Wold et al., 1987; Eriksson et al., 2001). The scores (ti) generated by PCA can be interpreted as projections of the fluorescence spectral variable to a new space spanned by the PCs (i.e. when the fluorescence spectral variables are transformed to PCs, each spectral variable in the X matrix is projected on to the PC space). The coordinates in this PC space are the scores. The set of scores corresponding to a particular PC can be plotted against another set of scores corresponding to another PC and such plots are called score plots. Generally, the score plot is built on the basis of the first two principal components since these explain most of the variability in the data. The PCs in the PC space are related to the spectral variables in the X matrix (i.e. original variable space) by loadings (Persson and Wedborg, 2001). By examining the loading values related to each PC, it is possible to understand which original spectral variables in the X matrix are better explained by each PC. Before performing PCA analysis, both XUF and XNF data sets were auto-scaled, i.e. adjusted to zero mean and unit variance by dividing each column by its standard deviation. To determine the number of principal components that are statistically significant in capturing the underlying features in XUF and XNF data sets, a leave-one-out cross validation method (Eriksson et al., 2001) was implemented. All computations were performed using PLS Toolbox 3.5 (Eigenvector Research, Inc., Manson, WA) within the MATLAB 7.3.0 computational environment (MathWorks, Natick, MA).
3.
Results and discussion
3.1. Typical spectral features in the fluorescence EEM of GRW The fluorescence EEM of GRW water (i.e. RW) shows a peak (a) at Ex/Em ¼ 320 nm/415 nm (Fig. 2), which corresponds to the range reported for fulvic-like HS (Coble et al., 1990; Sierra et al., 2005). The presence of fulvic-like HS in GRW was also independently confirmed by examining the LC-OCD spectra of the same water sample (Peiris et al., 2008). In addition to the primary peak (a), another secondary peak (b) which also corresponds to HS (Sierra et al., 2005; Peiris et al., 2008) appears to be present in the form of a shoulder around Ex/Em ¼ 270 nm/460 nm (Fig. 2). The HS in GRW can be expected to comprise predominantly fulvic acid-type matter compared to humic acid-type matter as reported in other natural waters (Huck, 1999; Sierra et al., 2005). The deviations of the fluorescence EEM contours seen in the region (Ex/Em: 280 nm/330 nm) indicated by d are believed to be due to the presence of protein-like substances in the water. The existence of a fluorescence EEM peak around the same region (d) has been previously observed for protein-like substances (Baker, 2001; Chen et al., 2003; Her et al., 2003). The protein-like
peak in the d region is not clearly visible due to the very low concentration levels of the protein-like substances present in GRW. The light scattering regions (first order Raleigh scattering region and second order Raleigh scattering region) observed in the fluorescence EEM are also important areas that provide information related to the particulate/colloidal matter present in water as will be discussed later.
3.2. PCs that summarize the total variance captured in the fluorescence EEMs PCA analyses were performed separately on XUF and XNF matrices to generate new and fewer numbers of variables or PCs to capture any systematic trends present in the 4214 original spectral variables of both XUF and XNF matrices. The first three PCs alone, generated in this way, were able to capture nearly 90% of the total variance present in the original spectral variables of XUF and XNF matrices separately (Table 2). The remaining variance (w 10%) is due to the combination of unexplained variance by the first three PCs and the instrumental noise in the fluorescence measurements. The instrumental error was however determined to be generally less than 5% for the intensity readings captured by fluorescence EEMs. It is possible to capture this remaining variance by generating more PCs. However, additional PCs were not examined in detail for the reasons explained below (Section 3.3).
3.3.
Physical significance of the PCs generated by PCA
PCA assigns loading values for each original spectral variable in the X matrix. This process therefore establishes a corresponding loading matrix for each PC. The loading values of each PC denote the relative importance of the fluorescence variables (i.e. excitation-emission wavelength combinations) so that the fluorescence variables with higher intensity values (e.g. fluorescence EEM peaks) of the X matrix are associated with large loading values. Hence, by examining loading matrices, one can understand which original spectral variables in the X matrix, i.e. which combinations of excitation and emission wavelengths, would be most dominant within the PCs (Persson and Wedborg, 2001). Fig. 3a, b and c demonstrate the loading values of PC – 1, PC – 2 and PC – 3 that are plotted at their corresponding fluorescence excitation/ emission wavelength coordinates. Similar loading plots were generated in the PCA of XUF and XNF but for brevity only the loading plots generated from XUF are demonstrated here. In the loading plot of PC – 1, a main loading peak (a’) at Ex/Em w 320 nm/415 nm and second loading peak (b’) in the form of a shoulder around Ex/Em ¼ 270 nm/460 nm can be observed (Fig. 3a). The presence of these loading peaks a’ and b’ at the same locations where the fluorescence EEM peaks of a and b (Fig. 2) for HS are situated, therefore indicates that PC – 1 is mostly correlated with the HS content in the water; i.e. samples with high HS content are associated with high PC – 1 scores. The loading plot of PC – 2, on the other hand, demonstrates an array of peaks at the same regions where the light scattering regions (first and second order Raleigh scattering) are situated in the fluorescence EEM of GRW (Fig. 2). The intensity values of these light scattering regions increase
water research 44 (2010) 185–194
189
Fig. 2 – Typical fluorescence features seen in the (a) fluorescence EEM for GRW and (b) 3D view of the same EEM. First order Raleigh scattering (FORS) and Second order Raleigh scattering (SORS) regions are indicated using dashed-lines.
with increasing particulate/colloidal matter present in the water. Hence, samples with high particulate/colloidal content are associated with high PC – 2 scores. The loading plot of PC – 3 demonstrates a distinct valley at the same regions (Ex/Em: 280 nm/330 nm) where the fluorescence EEM peaks related to protein-like substances occur (Chen et al., 2003; Her et al., 2003). Fig. 2 also indicates the presence of protein-like substances in GRW in terms of the deviation in the fluorescence EEM contours around the region highlighted by d. For these reasons, it is reasonable to conclude that PC – 3 is mostly associated with protein-like substances in water. The existence of a valley as opposed to a peak at Ex/Em: 280 nm/ 330 nm of the loading plot of PC – 3, implies that PC – 3 is inversely related to the protein-like content in the water. The loading plots of additional PCs were found to contain largely random variation of the loading values. Also, these loading plots did not contain regions that could be related to the spectral regions in the fluorescence EEM of GRW demonstrated in Fig. 2. For these reasons, the first three PCs were deemed to be the only PCs that contained sufficiently meaningful information and the rest of this discussion therefore focuses on these.
3.4. Performance of the pre-treatment and UF/NF stages as summarized by the score plots of PCs and potential as performance monitoring tool The impact of the pre-treatment on the subsequent UF stages was investigated by classifying the filtration operating conditions as normal, when no rapid permeate flux decline was observed (i.e. minimal fouling), and as high fouling events when a decline in flux was observed. Under normal filtration conditions, a total of 28, 28, 12, 16 and 28 samples of RW, RF, B1, B2 and UFp (Fig. 1), respectively were considered. High fouling events were observed for UF experiments denoted by UF8, UF9, UF12 and UF13. The set of score values for each PC generated by the PCA of the XUF matrix is illustrated in Fig. 4a (PC – 2 vs. PC – 1) and Fig. 4b (PC – 3 vs. PC – 2) according to the sample location and specific UF experiment. Note that each value in this score set is directly related to the fluorescence EEM data of each sample in the XUF matrix. The scores corresponding to samples of RW, RF, B1, B2 and UFp formed groups (or clusters) and these groups are indicated by dashed ellipses based on the 95% joint confidence regions (JCRs) of the
scores in each group. The calculation of these JCRs, based on the PC scores, was done to define regions related to normal operating conditions of the filtration. The PC scores corresponding to the normal operating conditions of the filtration (28, 28, 12, 16 and 28 samples of RW, RF, B1, B2 and UFp, respectively) were considered in the calculation of these JCRs. The horizontal and vertical orientation of these confidence region ellipses in Fig. 4a and b is due to the PCA methodology whereby the resulting PC’s are orthogonal to each other, i.e. there is mathematically zero covariance among them. The points denoted by UF8, UF9, UF12 and UF13 indicate high fouling events captured by fluorescence EEMs after 1 h of UF membrane operation. It should be noted that when only the intensity of the peak maxima of the fluorescence EEMs such as peaks (a), (b), (d) and the Raleigh scattering peaks were used in the PCA as opposed to full fluorescence EEMs, the 95% JCRs of the above mentioned groups were not generally separable (i.e. more overlapping regions) unlike the case presented in Fig. 4a and b (Supplementary data Figure S.2). This is expected and explained by the reasoning provided in previous studies (Chen et al., 2003; Stedmon et al., 2003); a smaller number of fluorescence EEM coordinates lack the ability to capture the heterogeneity of the different NOM fractions in water. Thus, the use of the full spectra results in better sensitivity in separating the data corresponding to normal operating conditions versus the data measured during fouling conditions. The score plot PC – 2 vs. PC – 1 (Fig. 4a) demonstrates the possibility of defining different regions, which can be considered as normal operating regions, for the roughing filter, two biofilters and the UF step. This information can be further investigated in the context of specific NOM fractions and the corresponding pre-treatment and membrane operation. The 95% JCRs of RW, RF, B1, B2 and UFp demonstrate a progressive shift towards lower values (scores) of PC – 1 and PC – 2. The small shift of PC – 1 indicates limited removal of HS corresponding to a slight shift along the PC – 1 axis while the more pronounced shift along the PC – 2 axis indicates a significant removal of particulate/colloidal matter at each pre-treatment stage and by the UF step. It should be recalled that the treatment steps are sequential (Fig. 1), except for B1 and B2 that operate in parallel, with B2 having the longer EBCT. The HS removal in these pre-treatment steps, however,
190
water research 44 (2010) 185–194
Table 2 – Variance captured by the first three principal components. Principal component
1 2 3
UF spectral data
NF spectral data
Variance Cumulative Variance Cumulative captured variance captured variance (%) captured (%) captured (%) (%) 66.6 12.9 8.4
66.6 79.5 87.9
77.5 6.5 5.4
77.5 84.0 89.4
can not be considered as significant due to the overlapping 95% confidence intervals (CIs) on the PC – 1 axis of Fig. 4a. The limited HS removal deduced from the shift along the PC – 1 axis is also supported by LC-OCD analyses and is consistent with the literature (Holzalski et al., 1995; Hallee´ et al., 2009). In particular, the average percentage HS removal by these pretreatment steps, as calculated from LC-OCD measurement during this study, was less than 10% during this study.
The significant removal for particulate/colloidal matter at the pre-treatment stages and by the UF step deduced from the shift along the PC – 2 axis is also supported by the turbidity data presented in Table 3. The 95% CIs of RW, RF, B1, B2 and UFp on the PC – 2 axis of Fig. 4a, are narrow enough to demarcate different operating regions for different treatment stages, with the exception of a small overlapping region between the CIs of B1 and B2, which could be expected. The wider 95% CIs of RW, RF, B1, B2 and UFp, manifested on PC – 1 axis of Fig. 4a, could be due to the large seasonal variation in the humic content in GRW during the study period; in general, higher humic concentration levels were recorded towards the latter part of January 2008 and lower humic concentration levels were recorded in September 2007. The score plot PC – 3 vs. PC – 2 (Fig. 4b) provides information on the reduction of protein-like matter (i.e. higher PC – 3 scores), from RW to UFp, even though the 95% CIs on the PC – 3 axis of Fig. 4b overlap with each other. There is essentially no removal of protein-like matter by the roughing filter, as would be expected. BF2 shows superior performance to BF1 as indicated by the JCRs of B2 and B1, which is consistent with
Fig. 3 – Loading plots of (a) PC – 1 – related to the humic content, (b) PC – 2 – related to the particulate/colloidal content and (c) PC – 3 – related to the protein content in water. PCA assigned loading values for each original spectral variable in the X matrix. These loading values are plotted here at their corresponding fluorescence excitation/emission wavelength coordinates. FORS – First order Raleigh scattering and SORS – Second order Raleigh scattering regions are indicated using dashed lines.
water research 44 (2010) 185–194
a
40
- UF 8
PC - 2 (12.88 %)
30
- U F8
UF 8-RW
- UF8
10
UF12- RW
RW
UF 13 -RW
UF 9-RF UF 8-RF
0
UF 8-B2
protein-like matter removal by different pre-treatment stages located upstream of the NF membrane step. Therefore for brevity, important interpretations of the NF experiments, as explained by PC score plots for NF, are summarized here.
- U F8
UF 9-RW
20
UF 13 -RF
UF 12 -RF
RF
UF 9-B1
UF13- B1 UF 12 -B 2
B1
-10 UF 8-p
UF 9-p
B2
UF12- p UF13- p
-20 UFp
-30 -30
-20
-10
0
10
20
30
PC- 1 (66.6 %) 15
- UF 8 UF p
- UF 8
- U F8
UF 9- p UF 8- p
B1
UF8-RF
UF8-B2 UF12- B2
-10
UF 9-B1
RF RW UF 9- RF UF 8- RW UF 12- RF
UF 9- RW
UF 13 -RW
3.5. Identifying high fouling events by PCA of fluorescence EEMs
UF 12 -RW
-15 -20 -30
-20
-10
1. A significant level of HS and particulate/colloidal matter removal, as would be expected (Her et al., 2007), was seen in the pre-treatment and NF stages. 2. NF was the main contributor to HS removal. Particulate/ colloidal matter, on the other hand, was removed to a large extent by the pre-treatment stages. The turbidity data recorded during NF experiments also support this observation (Table 3). 3. NF permeate quality was consistent in spite of comparatively larger variation in the membrane feed. 4. Higher rejection levels of protein-like matter were demonstrated by the NF step compared to the pre-treatment stages. 5. The rejection of HS and protein-like matter was generally higher with the TS80 membranes compared to XN45 membranes. The same observation was also made in a related study (Peiris et al., 2008).
UF 13- RF
UF 13-B1
0 -5
- U F8
B2
5
UF12- p
PC - 3 (8.41%)
10
UF13- p
b
191
0
10
20
30
40
PC - 2 (12.88%) Fig. 4 – Score plots: (a) PC – 2 vs. PC – 1 and (b) PC – 3 vs. PC – 2. PC scores are grouped and named based on the sampling locations: RW, RF, B1, B2 and UFp (Fig. 1). These groups are indicated by dashed-ellipses showing the 95% joint confidence regions of the scores in each group. UF8, UF9, UF12 and UF13 indicate high fouling events captured within 1 h of UF membrane operation.
removal interpretations and fouling data presented by Halle´ et al. (2009). The removal of protein-like material by the ultrafiltration membrane step is consistent with fouling interpretations provided by Halle´ et al. (2009) and Haberkamp (2008). When PCA was performed on the fluorescence EEMs obtained during the NF experiments (i.e. XNF matrix), PCbased 95% JCRs with similar features to those obtained with UF experiments were generated (Fig. 5a and b). The calculation of these JCRs was based on the PC scores related to normal operating conditions. The PC scores corresponding to a total of 31, 31, 15, 16, 31, 31 and 31 samples of RW, RF, B1, B2, NF_tank, NF_C and NFp (Fig. 1) respectively were considered in the calculation of these JCRs. In contrast to the UF experiments, only one high fouling event (NF8) was recorded during NF experiments. The shapes and the coordinates of these JCRs (Fig. 5a and b) are however dissimilar to those of the JCRs obtained with UF experiments. This dissimilarity is due to differences in the size of PC scales generated by PCA in both cases. As in the PC plots for UF, these JCRs also indicated comparable trends in the HS, particulate/colloidal and
As indicated earlier, the JCRs in Fig. 4a and b were generated from the fluorescence EEMs captured during UF experiments for normal operations, i.e. where incidents of high fouling did not occur. Experiments UF8, UF9, UF12 and UF13, on the other hand, exhibited very high fouling within 30, 60, 15 and 10 h of the start of membrane ultrafiltration, respectively. Experiments UF8 and UF9 were performed when HS content was lower (November 2007) as signified by PC – 1 scores towards the lower end of the 95% CI for RW (Fig. 4a). The HS content in GRW during experiments UF8 and UF9 varied between 3.3 and 3.5 mg C/L, based on LC-OCD determinations. In contrast, experiments UF12 and UF13 were conducted when GRW had a much higher HS content (February 2008) as indicated by the high PC – 1 scores in Fig. 4a. The HS content in GRW varied between 4.6 and 5.4 mg C/L during experiment UF12, and 4.9 and 5.2 mg C/L during experiment UF13, based on LC-OCD determinations. During these experiments, HS, particulate/colloidal and protein-like matter in the RF effluent were similar to those within the normal operating conditions of RF as demonstrated in Fig. 4a and b. This observation is also supported by TOC, DOC and turbidity measurements of RF, recorded during these experiments (Table 3). Nevertheless, the effluents of the biofilters in experiments UF8, UF9, UF12 and UF13 had higher PC – 2 scores than the corresponding JCRs of B2 and B1 in Fig. 4a, indicating reduced particulate/colloidal removal by both biofilters. It is evident that B2 was impacted more than B1. Because of increased levels of particulate/colloidal matter in the UF influent, permeate colloidal levels after the UF stage were also higher as indicated by the much higher PC – 2 scores for UFp, well outside the normal JCR (Fig. 4a). The reduced removal of particulate/colloidal matter in the biofilters and UF stage were however not as clearly demonstrated by the
192
water research 44 (2010) 185–194
RF
B1
B2
Typical UF8 UF9 UF12 UF13 NF8
DOC
Turbidity
(mg-TOC/L)
(mg-DOC/L)
(NTU)
7.4–6.0 5.8 6.6 6.6 5.6 5.8
7.1–5.5 5.8 6.2 6.2 5.4 5.8
60.2–3.2 5.7 2.7 15.6 7.5 5.7
Typical UF-8 UF9 UF-12 UF-13 NF-8
7.1–4.7 5.7 6.3 6.5 5.9 5.7
6.9–4.7 5.4 6.0 6.1 5.5 5.4
7.8–0.7 1.2 1.7 3.8 3.8 1.2
Typical UF8 UF9 UF12 UF13
7.0–5.4
7.0–5.5
2.1–0.1
5.7
5.7
0.9
5.5
5.3
1.0
Typical UF8 UF9 UF12 UF13 NF8
6.7–4.1 5.4
6.5–4.1 5.0
1.7–0.0 0.3
5.7
5.5
2.2
4.9
4.9
1.2
UFp
Typical UF8 UF9 UF12 UF13
7.7–5.2 6.0 5.7 5.3 5.3
7.7–5.5 7.3 5.8 5.4 5.3
0.3–0.0 0.1 0.2 0.4 0.4
NFp
Typical UF8
1.0–0.1 0.3
0.6–0.1 0.3
0.4–0.1 0.4
Typical – denotes the normal filtration conditions; UF8, UF9, UF12, UF13 and NF8 are the high fouling events.
turbidity measurements recorded after one hour of UF operation. Turbidity values recorded for B1, B2 and UFp under these high fouling events fell within the general ranges recorded during normal UF conditions (Table 3) indicating that turbidity is not a suitable parameter to capture the reduced particulate/colloidal matter removal levels. In addition, the effluent concentrations of protein-like matter from both biofilters, were not very different from their RW values (Fig. 4b). Therefore the protein-like matter content of the UF permeate also did not differ from the normal operating range. For these reasons, it is reasonable to conclude that the high fouling incidents experienced during experiments UF8, UF9, UF12 and UF13 were due to the poor removal of particulate/colloidal matter during biofiltration pretreatment. This poor performance was linked to the decrease in biofilter activity at low water temperatures (Halle´ et al., 2009). Similar to the high fouling events during UF, poor removal of particulate/colloidal matter seems to also have contributed to the only high fouling event (NF8) recorded during the NF experiments. The biofilter effluent of this experiment fell outside and above the JCR for B2 in the score plot of PC – 2 vs.
30 NF8-RW
20 15
- TS80
NF8-RF
RF
10 NF8-tank NF8_C NF8-B2
5 0 -5
- NF8
RW
25
PC - 2 (6.48 %)
RW
TOC
a
B1 NF_C
TS80
-10 -15 -20 -100
B2
NF8_p
NF_tank
NFp
-80
-60
-40
-20
0
20
40
60
80
PC -1 (77.51 %)
b
15
TS80
- NF8 NFp
10
PC - 3 (5.37 %)
Table 3 – Comparison of typical TOC, DOC and turbidity values under normal filtration conditions and the values recorded under high fouling events. These values were recorded after one hour of membrane filtration.
NF8_p
5 0
- TS80
B2
B1
NF_C
RW
NF8_B2 NF8_c
-5
RF
NF8_Tank
NF8_RF
NF8_RW
NF_tank
-10 -20
-10
0
10
20
30
PC -2 (6.48 %)
Fig. 5 – Score plot: (a) PC – 2 vs. PC – 1 and (b) PC – 3 vs. PC – 2. PC scores are grouped and named based on the sampling locations: RW, RF, B1, B2, NF_C, NF_tank and NFp. These groups are indicated by dashed-circles/ellipses showing the 95% confidence regions of the scores in each group. NF8 indicates a high fouling event captured within 1 hour of NF membrane operation. Scores of NFp corresponding to the NF experiments run with TS80 are indicated by symbol – ‘‘X’’.
PC – 1 (Fig. 5a) and hence indicated a lower than normal level of particulate/colloidal matter removal by BF2 (since the RW level was in the normal range). Turbidity measurements, recorded after one hour of NF operation, did not provide an indication of this poor removal level. Removal of protein-like material, on the other hand, seems to have been normal (Fig. 5b). The PC scores (Figs. 4 and 5) therefore clearly indicate a relationship between high fouling events for both UF and NF stages and reduced removals by the biofiltration pretreatment. In particular, as mentioned earlier, it is reduced removals of particulate/colloidal matter that contributed to the high fouling events observed in this investigation. Removal of protein-like material by the biofilters was within the normal operating range. The PC scores that demonstrate these deviations in the performance levels of the biofilters and the subsequent membrane stages were generated by PCA of the fluorescence EEMs obtained just after one hour of filtration. The high fouling event for these membranes however became evident only much later in terms of the increase in the TMP. Therefore it is proposed that PCA of fluorescence EEM data could serve as an early detection method to monitor
water research 44 (2010) 185–194
changes in the membrane feed that could lead to high fouling situations.
3.6.
Potential for process monitoring and intervention
In this study, PCA of fluorescence EEM was able to capture the differences in UF and NF membrane feedwater that were responsible for changes in fouling rate. An important reason for this capability is the scope of sensitivity: e.g., as discussed above, a higher sensitivity than turbidity measurements for capturing differences in colloidal/particulate matter in the biofilter effluents. Moreover, with the appropriate instrumental parameter settings, it is possible to obtain reproducible fluorescence EEMs even for the NF permeates (Peiris et al., 2008). This means that this approach could be used to monitor membrane permeate with very low NOM concentration levels. In contrast, most other reported NOM characterization techniques require pre-concentration steps prior to the analysis of water with low NOM concentrations, thereby increasing the chances for higher measurement noise. The fluorescence EEMs obtained during this study were made using off-line measurements, and the signal acquisition time for each EEM was about 5 min. Therefore, as demonstrated above, this approach could be readily used for off-line monitoring of membrane filtration and related pre-treatment processes with relatively inexpensive investments in a spectrophotometer, computer and related software. It is also possible that fibre optics or robotic sampling could be used to develop an on-line approach. Since the time frames involved with membrane fouling in drinking water treatment applications would normally be expected to be on the order of hours or more, the approach discussed here could be used for near real-time or rapid off-line monitoring. A change in membrane feedwater quality leading to accelerated fouling could then be detected in sufficient time that intervention strategies to reduce fouling, such as reducing membrane flux, could be executed.
4.
Conclusions
This investigation employed PCA of fluorescence EEMs to quantify the impact of pre-treatment stages on the removal of foulants for UF/NF membranes. The following conclusions can be drawn: 1. The performance of biofiltration pre-treatment prior to membrane filtration stage could be monitored, in terms of the removal levels of key membrane foulants such as humic substances, protein-like and particulate/colloidallike matter by examining the principal components generated by the PCA. 2. The necessary information could be captured by three principal components. Scores on PC – 1 and PC – 2 were largely related to the humic substances level and the particulate/colloidal-like content respectively. Scores of PC – 3 were inversely related to the protein-like content of the water. 3. The approach was able to provide early warning of high membrane fouling events. The fluorescence EEM-based PC score plots, obtained just after 1 h of UF and NF operation,
193
were able to link the high fouling events seen in this study to reduced removals by the biofilters of high influent levels of particulate/colloidal-like material in certain runs. The turbidity measurements made at the same time did not provide an indication of these high fouling events. 4. The approach is very sensitive, as is evident by its ability to be used in analyzing NF permeates containing low levels of organic carbon. 5. This method has the potential to be used as a monitoring tool for membrane-based water treatment and pre-treatment operations, and as an early detection method to identify high fouling events that may arise. This would allow membrane operational changes to be made proactively. In contrast to chromatographic methods, this off-line monitoring approach allows for nearly real-time monitoring.
Acknowledgments We acknowledge a number of contributors to this work including GE-Zenon for the donation of UF modules, and the financial support of the Canadian Water Network, the Natural Sciences and Engineering Research Council of Canada (NSERC) including an NSERC Postgraduate scholarship to R.H. Peiris and the partners of the NSERC Industrial Research Chair in Water Treatment (P.M. Huck) for funding. The current Chair partners may be found at http://www.civil.uwaterloo.ca/ watertreatment/.
Appendix. Supplementary data The supplementary data associated with this article can be found in the on-line version at doi:10.1016/j.watres.2009.09.036.
references
Amy, G., 2008. Fundamental understanding of organic matter fouling of membranes. Desalination 231, 44–51. Baker, A., 2001. Fluorescence excitation - emission matrix characterization of some sewage-impacted rivers. Environ. Sci. and Technol. 35 (5), 948–953. Boehme, J., Coble, P., Conmy, R., Stovall-Leonard, A., 2004. Examining CDOM fluorescence variability using principal component analysis: seasonal and regional modeling of threedimensional fluorescence in the Gulf of Mexico. Mar. Chem. 89 (1-4), 3–14. Chen, W., Westerhoff, P., Leenheer, J.A., Booksh, K., 2003. Fluorescence excitation-emission matrix regional integration to quantify spectra for dissolved organic matter. Environ. Sci. and Technol. 37 (24), 5701–5710. Coble, P.G., Green, S.A., Blough, N.V., Gagosian, R.B., 1990. Characterization of dissolved organic matter in the Black Sea by fluorescence spectroscopy. Nature 348 (6300), 432–435. Eriksson, L., Johansson, E., Kettaneh-Wold, N., Wold, S., 2001. Multi- and Megavariate Data Analysis, Principles and Applications. Umetrics Academy, Umea, Sweden, ISBN 91973730-1-X, p. 533.
194
water research 44 (2010) 185–194
Haberkamp, J. Organisches Membranfouling bei der Ultrafiltration kommunaler Kla¨ranlagenabla¨ufe-Ursachen, Mechanismen und Maßnahmen zur Verringerung, Ph.D dissertation, Technische Universita¨t Berlin, Berlin, Germany, 2008. http://opus.kobv.de/tuberlin/volltexte/2009/2107/pdf/ haberkamp_jens.pdf Halle´, C., Huck, P.M., Peldszus, S., Haberkamp, J., Jekel, M., 2009. Assessing the performance of biological filtration as pretreatment to low pressure membranes for drinking water. Environ. Sci. and Technol. 43 (10), 3878–3884. Henderson, R.K., Baker, A., Murphy, K.R., Hambly, A., Stuetz, R.M., Khan, S.J., 2009. Fluorescence as a potential monitoring tool for recycled water systems: a review. Water Res. 43 (4), 863–881. Her, N., Amy, G., McKnight, D., Sohn, J., Yoon, Y., 2003. Characterization of DOM as a function of MW by fluorescence EEM and HPLC-SEC using UVA, DOC, and fluorescence detection. Water Res. 37 (17), 4295–4303. Her, N., Amy, G., Plottu-Pecheux, A., Yoon, Y., 2007. Identification of nanofiltration membrane foulants. Water Res. 41, 3936–3947. Holzalski, R.M., Goel, S., Bouwer, E.J., 1995. TOC removal in biological filters. JAWWA 87 (12), 40–54. Huck, P.M., 1999. Development of a framework for quantifying the removal of humic substances by biological filtration. Water Sci. and Technol. 40 (9), 149–156. Hudson, N., Baker, A., Reynolds, D., 2007. Fluorescence analysis of dissolved organic matter in natural, waste and polluted waters – a review. River Res. Appl. 23 (6), 631–649. Jermann, D., Pronk, W., Meylan, S., Boller, M., 2007. Interplay of different NOM fouling mechanisms during ultrafiltration for drinking water production. Water Res. 41 (8), 1713–1722. Peiris, B.R.H., Halle´, C., Haberkamp, J., Legge, R.L., Peldszus, S., Moresoli, C., Budman, H., Amy, G., Jekel, M., Huck, P.M., 2008.
Assessing nanofiltration fouling in drinking water treatment using fluorescence fingerprinting and LC-OCD analyses. Water Sci. and Technol.: Water Supply 8 (4), 459–465. Peiris, B.R.H., Budman, H., Moresoli, C., Legge, R.L., 2009. Acquiring reproducible fluorescence spectra of dissolved organic matter at very low concentrations. Water Sci. and Technol. 60 (6), 1385–1392. Persson, T., Wedborg, M., 2001. Multivariate evaluation of the fluorescence of aquatic organic matter. Anal. Chim. Acta 434, 179–192. Saravia, F., Zwiener, C., Frimmel, F.H., 2006. Interactions between membrane surface, dissolved organic substances and ions in submerged membrane filtration. Desalination 192 (1-3), 280–287. Sierra, M.M.D., Giovanela, M., Parlanti, E., Soriano-Sierra, E.J., 2005. Fluorescence fingerprint of fulvic and humic acids from varied origins as viewed by single-scan and excitation/ emission matrix techniques. Chemosphere 58 (6), 715–733. Spencer, R.G.M., Bolton, L., Baker, A., 2007. Freeze/thaw and pH effects on freshwater dissolved organic matter fluorescence and absorbance properties from a number of UK locations. Water Res. 41 (13), 2941–2950. Stramski, D., Wozniak, S.B., 2005. On the role of colloidal particles in light scattering in the ocean. Limnol. Oceanogr. 50 (5), 1581–1591. Stedmon, C.A., Markager, S., Bro, R., 2003. Tracing dissolved organic matter in aquatic environments using a new approach to fluorescence spectroscopy. Mar. Chem. 82, 239–254. Wold, S., Esbensen, K., Geladi, P., 1987. Principal components analysis. Chemo. and Intell. Lab. Sys. 2, 37–52. Wyatt, P.J., 1993. Light scattering and the absolute characterization of macromolecules. Anal. Chim. Acta 272 (1), 1–40.
water research 44 (2010) 195–204
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Use of fluorescence fingerprints for the estimation of bloom formation and toxin production of Microcystis aeruginosa Markus Ziegmann a,*, Michael Abert b, Margit Mu¨ller c, Fritz H. Frimmel a a
Engler-Bunte-Institut, Chair of Water Chemistry, Universita¨t Karlsruhe (TH), 76131 Karlsruhe, Germany DOC-Labor Dr. Huber, 76229 Karlsruhe, Germany c Wala Heilmittel GmbH, 73087 Bad Boll-Eckwa¨lden, Germany b
article info
abstract
Article history:
The development of methods facilitating the detection of cyanobacterial blooms in
Received 9 April 2009
drinking water reservoirs at an early stage is of great importance. Fluorescence spectros-
Received in revised form
copy could meet these requirements. The study contains the examination of possible
11 September 2009
correlations between the different maxima of a fluorescence excitation-emission matrix
Accepted 15 September 2009
and the amount of produced and excreted toxins of a lab culture of Microcystis aeruginosa at
Available online 19 September 2009
different stages of growth. Various fluorescence signals (protein-like and humic-like substances, pigments) are suited for an estimation of cell density and actual intra- and
Keywords:
extracellular toxin concentration. One signal at 315 nm/396 nm presumably originating
Fluorescence matrix
from protein-like substances might be useful as a tool for the prediction of increasing
Synchronous scan
cyanobacterial toxin concentrations. As the measurement of fluorescence matrices is still
Microcystin production
time consuming, synchronous scans with Dl ¼ 80 nm were tested as a potential alterna-
Toxin release
tive. They accurately depict the course of protein-like and humic-like fluorescence during
Bloom detection
the different stages of growth although especially the latter one is not captured at its maximum. However, due to insufficient separation of chlorophyll a and phycocyanin, the image of the matrix maxima by synchronous scans with Dl ¼ 80 nm can only be used with minor restrictions. Nevertheless, fluorescence spectroscopy seems to be a powerful tool for the evaluation of cyanobacterial blooms. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Surface water reservoirs are one of the most important drinking water resources, especially in arid zones. However, reservoirs can suffer from eutrophic conditions, which might lead to algal and cyanobacterial blooms under certain weather conditions. Some of the bloom forming species, especially cyanobacteria, produce various spurious or harmful substances like taste and odor compounds and toxins. The most prevalent toxins are hepatotoxins called Microcystins (MC). They are cyclic heptapeptides with variable amino acids, of which more than 90 modifications have been detected so
far (Welker and von Dohren, 2006). Oral uptake is the main exposure pathway for humans, e. g. by consumption of contaminated drinking water or of plants irrigated with contaminated water (Crush et al., 2008). For the predominant MC-LR (including leucine (L) and arginine (R)), the WHO has established a threshold value of 1 mg L1 for drinking water (Falconer et al., 1999). In case of a bloom, usually concentrations around 5 mg L1 MC are observed (e. g. Mankiewicz et al., 2005) in the bulk of the lake. Due to accumulation of cells at the surface by their buoyancy and by wind drift, concentrations up to some mg L1 can be reached, which is far beyond the WHO threshold value. Therefore, a monitoring of algal or
* Corresponding author. Tel.: þ49 721 608 7097; fax: þ49 721 608 7051. E-mail address:
[email protected] (M. Ziegmann). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.035
196
water research 44 (2010) 195–204
cyanobacterial species and their metabolites occurring in a reservoir used for drinking water purposes is necessary. However, this is linked with difficulties due to extensive lab work and high expenditure. Finally, it may already be too late to react appropriately after the detection of cyanobacterial cells or metabolites by the current technical methods. Because of this, ways of detecting cyanobacterial blooms at an early stage are an important issue. Microscopic methods developed by Utermo¨hl (1958) are still used for the determination of any planktonic species and their cell concentration. These methods are very specific and give good results but they are also very time consuming and require qualified labor. Alternatively, the chlorophyll a (Chl a) content is often used as a parameter for a rough estimation of the phototrophic plankton biomass. This parameter is easy to measure and can be determined quickly, either by HPLC (Schmid et al., 1998) or by fluorescence (Welschmeyer, 1994), but it does not provide any information about the phytoplankton community composition. To obtain this information, so-called marker pigments can be used (Schmid et al., 1998). These marker pigments, located in the thylakoid membrane of the peripheral antenna of the photosystems I (PS I) and II (PS II), harvest light in regions, where neither Chl a nor Chl b absorb irradiation energy. Afterwards this energy is transferred to the core antenna, where specialized chlorophyll molecules build up biomass. The marker pigments of cyanobacteria are phycobiliproteins, mainly phycocyanin (PC). The proportion between Chl a, PC and other phycobiliproteins varies between the different cyanobacterial species resulting in different fluorescence and reflectance spectra. Reflectance spectra might be used for remote sensing of cyanobacteria (Subramaniam et al., 1999b, Metsamaa et al., 2006). However, fluctuations in the phycobiliprotein concentrations (Seppaila and Balode, 1998; Simis et al., 2005) and the PS I / PS II stoichiometry (Fujita, 1997; Subramaniam et al., 1999a) due to changing light and nutrient regime result in spectral variability and need to be taken into account. Aside from marker pigments cyanobacteria also excrete other fluorescent substances containing aromatic amino acids and quinoid groups e. g. as terminal electron acceptors in PS II (Nitschke and Rutherford, 1991). Henrion et al. (1997) mentioned that during the different stages of growth for each species, only the intensity but not the spectral pattern of its fluorescence spectrum changes. Hence, a principal component analysis of the excitation-emission matrices of surface water samples might be used to characterize the algal or cyanobacterial composition of the sample. Several devices have been developed for the in situ determination and differentiation of phytoplankton communities. Beutler et al. (1998) attempted to evaluate the composition of these communities mathematically by the use of fluorescence intensities at five different wavelengths according to the antenna pigments of five main groups of algal and cyanobacterial species (green, blue, brown, red, and mixed). The method was then enhanced emphasizing shorter wavelengths (370–525 nm) (Beutler et al., 2002), based on the assumption that fluorescence excitation spectra are independent of the physiological status of the cells. Later they tried to account for the influence of different impact factors on the pigment concentration (Beutler et al., 2003). Gregor and
Marsˇa´lek (2004) successfully tested a commercially available fluorescence device (FluoroProbe, bbe-Moldaenke, Kiel, Germany) with several excitation wavelengths based on the work of Beutler et al. (1998). They later focused on the differentiation between algae and cyanobacteria by using only two excitation wavelengths (485 and 570 nm) and evaluating the chlorophyll fluorescence at 670 nm (Gregor and Marsˇa´lek, 2005; Gregor et al., 2007). The principle of the FluoroProbe has also been successfully tested by Izydorczyk et al. (2009) in lakes in Poland. However, Pemberton et al. (2007) found that in samples from Lake Ontario, cyanobacteria were underreported by the above described fluorescence measurements. The aim of the present work was to test whether the influence of age of a cyanobacterial population on the fluorescence spectrum of its extracellular and intracellular substances is indeed negligible or whether it is even of possible use for the early detection of cyanobacterial blooms. Changes in the fluorescence spectrum can best be seen by recording an excitation-emission matrix (EEM) whereas single excitation or single emission spectra are susceptible for wave length changes of the emission or excitation maximum, respectively. However, recording of EEM is time consuming. Therefore, fluorescence synchronous scans (SyncScans) were analyzed as an alternative faster screening method. This technique has been used e. g. by Ferrari and Mingazzini (1995) and Mingazzini (2001) for the characterization of algal organic matter. For further reduction of analysis time, the option of a direct measurement of non-destructed cyanobacterial cells in solution both by EEM and SyncScans was investigated. This application of fluorescence spectroscopy could have a high benefit for online monitoring of algal and cyanobacterial growth in water reservoirs used for drinking water production. By comparing the obtained fluorescence data with measured MC-LR concentrations in the cyanobacterial suspension, a first attempt was made to relate the very selective and sensitive fluorescence spectroscopic measurements of higher concentrated metabolites of a cyanobacterial culture at different growth phases to the toxin concentration.
2.
Material and methods
2.1.
Chemicals
Ultrapure water (Milli Q, 18.2 MU cm), acetonitrile (VWR HiPerSolv Chromanorm), methanol (VWR HiPerSolv Chromanorm) and acetone (Merck LiChrosolv Hypergrade) were used as solvents and eluents for pigment and MC analysis. The MC-LR standard was obtained from ALEXIS Biochemicals. Chl a was purchased from Merck and PC from Aowei Bioengineering.
2.2. Cultivation of cyanobacteria and preparation of cell suspensions The cyanobacterium Microcystis aeruginosa (strain number 14.85) was grown in Erlenmeyer flasks with 20% charging volume at a temperature of 22 C. Zehnder’s medium was used as nutrient solution. The photon radiation density
water research 44 (2010) 195–204
applied was approximately 4 107 Einstein m2 s1 in the visible light region for a period of 12 hours per day. The cultures were shaken twice a day for 5 min with 50 rpm using an orbital shaker to avoid agglomeration. The cell cultures were prepared in triplicate to ensure continuity in the studies of a biological system over the whole period of sampling. Samples were always drawn in the middle of the radiation period and handled identically to minimize external influences, e. g. by light. The cell digestion was accomplished by ultrasonic radiation with a Branson Sonifier Cell Disruptor B15 three times for 30 s each time. Between each radiation step the cell suspension was cooled in ice water to minimize denaturing of cyanobacterial derived proteins. Filtration of the samples was done by 0.45 mm disposable filters (regenerated cellulose, Optiflow), which were prewashed with ultrapure water prior to filtration of the cell culture samples.
2.3.
Fluorescence spectroscopic analysis
Thirteen EEMs and SyncScans of M. aeruginosa were obtained between 2 and 38 days of incubation. The cell suspensions were measured directly after filtration for fluorescent extracellular substances (ES) expressed in solution and after cell digestion by ultrasonic radiation followed by filtration for the sum of extracellular and intracellular substances (IS). For the recording of the EEMs and SyncScans, an Edinburgh Instruments fluorescence spectrometer F900 in steady state mode with L-geometry was used. The system mainly comprised a xenon arc lamp as light source, excitation and emission gratings, a sample chamber for a quartz cuvette (10 10 mm) and a red sensitive photomultiplier tube as detector. EEMs were recorded in the range of excitation wavelengths lex ¼ 235–700 nm with a step width of 5 nm and emission wavelengths lem ¼ 250–750 nm with a step width of 1 nm. The wavelength dependent light intensity of the light source and light sensitivity of the detector were corrected. Fluorescence measurements were not corrected for a possible inner filter effect. SyncScans were recorded at lex ¼ 200–700 nm. For the offset between lex and lem a wavelength gap of Dl ¼ 80 nm was chosen. Slit widths of excitation and emission slits were held constant during all measurements.
2.4.
Biomass determination
Fluorescence measurements were also used for the determination of the cell density of the cultures. The cell suspensions were measured directly and without cell digestion at wavelengths of lex ¼ 400–650 nm and lem ¼ 685 nm. The calibration curve was obtained by relating the integral of the excitation scan to the appropriate cell numbers, which were quantified by cell counting of culture samples in a counting chamber from Neubauer with a depth of 0.1 mm and an area of 0.0025 mm2 at a Zeiss Axio Imager Z1 light microscope. The calculated values for the cell density of the experimental samples were also spot-checked by cell counting of selected samples. For the determination of the biomass, conventional Chl a analysis methods were used. Chl b is reported to lead to overestimation in the determination of Chl a (Welschmeyer,
197
1994). However, this does not affect the present studies since Chl b is not produced by M. aeruginosa. Pheopigments do not disturb the analysis of Chl a. The extraction of the cell pigments was done according to Schmid and Stich (1995). Samples of 5 mL volume were taken under a sterile hood and filtered through a glass fiber filter without binder from Sartorius. The filter was transferred into a 100 mL flask and 10 mL acetone:water (90:10 v/v) were added. The sample was heated in a water bath at 55 C for 5 min and exposed to ultrasound for other 5 min. After cooling, the sample was filtered through a disposable filter (0.45 mm, regenerated cellulose, Optiflow), filled up to 10 mL again with acetone:water (90:10 v/v) and analyzed by HPLC. During the whole procedure light exposure was avoided. The HPLC analysis of Chl a was done according to Schmid et al. (1998). Chl b, a- and b-Carotine, Lutein, Zeaxanthin, and Fucoxanthin can also be determined using this method. An Agilent 1100 HPLC system was used. The injection volume was 50 mL. A column combination of a Nucleosil C18 ODS (250 mm 3 mm) 5 mm (Macherey & Nagel) and a C18 (250 mm 3 mm) 5 mm (MZ-PAH) was applied for separation. The following gradient with eluent A (water:acetonitrile (50:50 v/v) and eluent B (methanol:acetone (60:40 v/v) was used: 0–5 min: 60% A, 17–20 min: 20% A, 40–45 min: 5% A, 15 min post time. Column oven temperature was set to 35 C. Absorption was detected at 444 and 515 nm with a diode array detector and fluorescence signals were recorded with a wavelength programmable fluorescence detector using the following program: 0–23 min (lex: 409 nm/lem: 670 nm), 23– 33 min (465 nm/656 nm), and 33–45 min (409 nm/670 nm).
2.5.
HPLC analysis of microcystins
Toxins were measured by a liquid chromatography system (Agilent HPLC 1100LC) coupled to tandem mass spectrometric detection (API 3000, Applied Biosystems/Sciex) with electrospray ionization (TurboIonSpray, Applied Biosystems/Sciex) A Zorbax RX-C18 column (3 mm 150 mm) 5 mm (Agilent) was used for chromatographic separation at 30 C with an injection volume of 20 mL. The eluents consisted of Milli Q þ 0.05% acetic acid þ 2 mmol ammonium acetate (Eluent A) and acetonitrile þ 0.05% acetic acid þ 2 mmol ammonium acetate (Eluent B). The flow was 0.5 mL min1. The gradient program was as follows: 5 min equilibration70% A, 0–3 min: 70% A, 15 min: 30% A. MC-LR was eluted after 6.48 min and detected in positive mode at a m/z ratio of [M þ H]þ ¼ 995.7 g mol1. The detection limit for MC-LR was 1 mg L1.
3.
Results and discussion
Three cyanobacterial cultures were grown in parallel during the sampling period of 40 days. Between the cultures II and III only minor changes in their growth performance were observed. Culture I temporarily dropped behind. The growth of all three cultures as a function of time is depicted in Fig. 1. Increasing cell numbers – measured as the cell density in solution by fluorescence analysis – fit well with the data for Chl a concentration after extraction of the cells. For reasons of clarity, in the following chapters only the results of culture III
198
water research 44 (2010) 195–204
7
3
cell density of: culture I culture II culture III
4
(Chl a) of:
2.0
2
culture I culture II culture III
7
2.5
1.5
1x10
1.0
1
0
(Chl a) [mg/L]
cell density [cells/mL]
2x10
0.5 0.0
0
5
10
15
20
25
30
35
40
days of cultivation Fig. 1 – Parallel growth of three cell cultures of M. aeruginosa determined by cell density (cells per mL) measured by fluorescence analysis and independently by concentration determination of Chl a in cells measured by extraction and consecutive HPLC determination. Adaptional phase (1), growth phase (2), stationary phase (3) and decay phase (4) are separated by vertical lines.
are presented. This culture grew continuously without greater throwbacks and represents quite well the average of the cultures with regular growth behavior.
3.1.
Excitation-emission-matrix
A full set of analysis (EEM, SyncScan, Chl a, MC-LR) was undertaken for each sampling day. Selected EEMs of the growth phase (14 days after incubation), the stationary phase (28 days) and the decay phase (38 days) of each, extracellular substances (ES) and intracellular substances (IS), are shown in Fig. 2. The EEMs of cell cultures of M. aeruginosa reveal up to six maxima in the wave length range under investigation. All six maxima appear in the EEM of the IS samples. Five of the six maxima can already be assigned to three classes of compounds, as indicated by the dashed ellipses in Fig. 2 d. The signal A at an excitation wave length of 275 nm and an emission wave length of 340 nm belongs to the aromatic amino acid tryptophan as shown by e. g. Chen et al. (2003) as well as by own analysis with the pure compound in water. Other researchers have assigned this region to protein-like fluorescence (Her et al., 2004). This is not a contradiction because due to its fluorescence properties tryptophan dominates the fluorescence of proteins even though other aromatic and fluorescence active amino acids like tyrosine and phenylalanine are present. Freely dissolved tyrosine at 274 nm/303 nm and phenylalanine at 257 nm/282 nm could not be detected in the EEMs at any time. A second region in the EEM is assigned to the so-called humic-like fluorescence. The signals at 250 nm/477 nm and 355 nm/473 nm belong to this region. Sometimes the excitation range is separated into two sub-regions which are attributed to fulvic-like (below an excitation wave length of 250 nm, signal B) and humic-like fluorescence (above 250 nm, signal C). However, fulvic and humic acid represent a complex mixture of different organic compounds of natural origin with irresolvable chemical structure which could be best described
as a continuum rather than an individual fraction. As a consequence, fluorescence analysis of isolates of fulvic as well as of humic acids show signals in both regions with differing intensity ratios (e. g. Alberts and Takacs, 2004). It needs to be mentioned that the terms humic-like and fulviclike fluorescence could be misleading because the fluorescence in these substances originates predominantly from quinone moieties (Klapper et al., 2002), which are also present in extracellular polymeric substances (EPS) derived e. g. from algae (Her et al., 2004; Lee et al., 2006). To distinguish between EPS and humic substances, which represent a step of altered (condensed) organic molecules in the organic carbon cycle, seems to be a more philosophic task. The appearance of the signals A, B and C is also reported for fluorescence analysis of e. g. freshwater planktonic bacteria (Elliot et al., 2006), and for aerobic and anaerobic sludges (Sheng and Yu, 2006). In the border region between protein- and humic-like fluorescence a fourth signal at 315 nm/396 nm arises in the EEM, which is further referred to as signal X. The classification of this signal into one of the above mentioned regions is not clear. It may be caused by tryptophan bound in a protein which leads to a red shift of the emission (towards higher wave length) or other soluble microbial by-products. In fact, a signal was observed in this EEM region for microbially derived fulvic acid in Antarctic water samples. In such a remote area, organic matter derived predominantly from autochthonous microbial processes and, therefore, algal biomass may have been the dominant source of dissolved organic matter (McKnight et al., 2001). Alternatively, this border region may show the start of the humification pathway: Protein material is oxidatively altered during humification resulting in a red shift in the emission wave length into the start of the humic-like fluorescence region. According to Chen et al. (2003) this area in the EEM is related to marine humic acids. Compared to terrestrial humic acids, marine humic acids show a fluorescence signal at shorter wave length, which allows to distinguish marine waters from fresh waters (Coble, 1996). The third marked region above lem ¼ 600 nm is assigned to pigments. The signal at 420 nm/680 nm (signal D) is caused predominantly by Chl a. The broad signal at 605 nm/645– 680 nm (signal E) is composed of phycocyanin (PC) (605 nm/ 645 nm) and Chl a (605 nm/680 nm). Additionally, Chl a shows a minor fluorescence signal at 480 nm/680 nm. Phycoerythrin as the second possible accessory pigment of cyanobacteria is not produced by M. aeruginosa. Therefore, no fluorescence signal around 544 nm/580 nm has been observed. Signals in the lower right corner of the EEM (beneath the diagonal line of second order stray light) were not considered at all. During the growth phase the IS EEM is mainly dominated by the signals A and B. During the stationary phase signal X arises, but the EEM is now dominated by signals D and E (pigments). Chl a initially develops the dominating pigment fluorescence, but later PC appears to become the main component (Fig. 2 f) although the two peaks can not be separated sufficiently. In contrast, EEMs from ES solutions do not show these pigment signals at any period of growth (Fig. 2 a, c, d). This difference exists because a release of pigments into the aqueous solution under normal conditions is not desirable for living phototrophic cells. Furthermore, the pigments and especially Chl a are not stable under light and will decompose
199
water research 44 (2010) 195–204
Extracellular substances (ES)
Intracellular substances (IS)
700
700
600
500
400
300
300
0
b excitation wave length [nm]
excitation wave length [nm]
a
400
500
600
1.250E5
600
2.500E5 3.750E5
500
5.000E5
400
300
300
700
emission wave length [nm] 700
500
600
700
700
c
0
d excitation wave length [nm]
excitation wave length [nm]
400
emission wave length [nm]
600
500
400
300
1.250E5
E
600
2.500E5 3.750E5
500
5.000E5
400
300
D
C X A
B 300
400
500
600
700
300
emission wave length [nm]
500
600
700
emission wave length [nm]
700
0
700
f excitation wave length [nm]
e excitation wave length [nm]
400
600
500
400
300
300
400
500
600
1.250E5
600
2.500E5 3.750E5
500
5.000E5
400
300
700
300
400
500
600
700
emission wave length [nm]
emission wave length [nm]
Fig. 2 – EEM of cell cultures of M. aeruginosa after a growth period of 14 days (Fig. 2 a, b), 28 days (Fig. 2 c, d) and 38 days (Fig. 2 e, f). Fluorescence analyses were undertaken before (ES) and after (IS) cell lysis. The diagonal line in Fig. 2 d shows the course of the SyncScan with Dl of 80 nm, arrows point to the four observed maxima A, X, C and E on the excitation axis of the SyncScan. Dashed lines point to the two observed maxima B and D, which are not covered by the SyncScan. Dashed ellipses indicate different structural classes.
rapidly outside the cell. Signals B and C (humic- and fulvic-like) dominate the ES samples at the beginning of the growth phase. Signal X becomes clearly prevalent between 14 and 38 days, but disappears rapidly afterwards. In the decay phase after 38 days signal A (protein-like) increases disproportionately and finally represents the strongest fluorescent fraction of ES.
3.2.
Synchronous scan
The line in Fig. 2 d marks the SyncScan. The four arrows point to the wave length regions on the excitation axis where the four maxima can be observed. An offset Dl of 80 nm between excitation and emission wave length was chosen to be best
200
water research 44 (2010) 195–204
a fluorescence intensity [counts]
b
Extracellular substances (ES)
5
2x10
X
7d 14 d 28 d 38 d
A
5
A
7d 14 d 28 d 38 d
X
5
C
1x10
Intracellular substances (IS)
5
2x10
1x10
E
C
E 0
0 300
400
500
600
700
300
400
excitation wave length [nm]
500
600
700
excitation wave length [nm]
Fig. 3 – SyncScan of cell cultures of M. aeruginosa at different stages of growth. Fluorescence analyses were undertaken before (open symbols, Fig. 3 a) and after cell lysis (filled symbols, Fig. 3 b). SyncScans were recorded with Dl of 80 nm.
suited. The maxima of A and X are well covered, signals B and D of the EEM are not captured at all. For signal C only the border area is detected, which results in a constant difference between EEM and SyncScan maximum of approximately 20 nm in both, excitation and emission wavelength. The broad signal E is composed of two peaks, which, depending on the intensity, can not be separated. These two signals represent the crucial point for the applicability of EEMs as a method for the detection of cyanobacterial blooms and, furthermore, for the use of SyncScans to replace the EEMs. Therefore, it is of high interest whether the signal heights at the chosen SyncScan wave length difference are linearly correlated to the EEM peak’s maxima. SyncScans of ES samples after 7 days show only two small signals for protein- (signal A) and humic-like fluorescence (signal C, Fig. 3 a). These signals can be considered as leftovers from the inoculation culture. During the growth phase after 14 days the fluorescence intensity of the signals X and C increases
a
ratio maxima EEM/SyncScan
b
Extracellular substances (ES) 5
3
2
2
1
1
0
5
10
15
20
Signal A Signal X Signal C Signal E
4
3
0
Intracellular substances (IS)
5
Signal A Signal X Signal C
4
whereas signal A remains small. This trend is even pronounced during the stationary phase up to 28 days, where, especially for ES, signal X becomes dominant. Surprisingly, the absolute fluorescence intensity of signal X in ES samples is even higher than in the IS samples. This can only be explained by an undesired degeneration during the cell digestion procedure, which is accompanied by ultrasound and a moderate raise of temperature. At the decay phase after 38 days, a disproportionate increase of signal A in the ES samples can be observed, as already observed in the EEM. This signal develops also constantly in the SyncScan of IS samples at all growth phases, whereas signal C remains fairly constant (Fig. 3 b). At the same time signal X drops significantly, which substantiates the impression of a fairly unstable structural group. It is interesting to note that similar to the character of the EEMs the fluorescence intensity of signal E (pigments) for the IS sample increases rapidly at the end of the growth phase. After 14 days of growth two hardly separated maxima
25
days of cultivation
30
35
40
0
0
5
10
15
20
25
30
35
40
days of cultivation
Fig. 4 – Comparability of EEM and SyncScan fluorescence maxima of ES (open symbols, Fig. 4 a) and IS (filled symbols, Fig. 4 b) for the signals A, X, C and E, depicted over days of cultivation of one culture of M. aeruginosa.
201
water research 44 (2010) 195–204
3.3.
Toxins
The increase in fluorescence intensities and MC content of a cyanobacterial culture can be explained in first instance by an increase of cell numbers during growth. To evaluate if there is a specific change in fluorescence intensity or MC content during different growth phases, in the following sections the fluorescence raw data are normalized by a cell number parameter. The Chl a concentration was chosen as an appropriate parameter for this task. In the adaptional phase of the cultures between 0 and 10 days the spectral pattern of both, IS and ES, is likely dominated by leftovers from inoculation and fluorescence counts are quite low. Pigments are almost missing due to photochemical instability and low cell numbers. Little variations in the Chl a concentration emphasize the normalized parameters and, therefore, in the following figures data from the adaptional phase of the cultures are omitted. The major aim of the examination of different types of fluorescence measurements was to find a fast and widely applicable method for the estimation and evaluation of the concentration of hazardous, cyanobacterial derived substances, especially toxins, in surface waters. The mean MC-LR concentration of the cultures II and III related to their Chl a concentration i. e. the cell numbers during growth is depicted in Fig. 5. The relative concentration of intracellular (I)-MC-LR decreases within the first 25 days followed by a drastic increase while for extracellular (E)-MC-LR a more or less continuous increase can be noticed. As expected,
0.20
(I)-MC-LR/Chl a ratio MC-LR/Chl a [µg/µg]
belonging to Chl a and PC are noticed, which combine to one single peak after 38 days. Possible shifts in the peak maxima resulting in a change of the ratio between peak maxima of EEM and SyncScan have been contemplated. For all peaks in the EEM a shift of up to 15 nm could be noticed, which was partly not reflected in the corresponding SyncScans. The ratios of the correlating peak maxima of EEM and SyncScan are presented in Fig. 4. Constant values indicate a good representation of the EEM maxima by the SyncScans. During the adaptional phase between 0 and 10 days, large variations were found for most of the signals. These variations can be explained though by small fluorescence counts of signals, which are not fully covered by the SyncScan. From the beginning of the growth phase constant values were obtained for the signals A, X and C in both, ES and IS samples. Significant deviations were observed for signal E during the growth phase. Therefore, SyncScans seem to be indeed a proper means to gain an estimation of the spectral fluorescence pattern of M. aeruginosa without the need of recording time consuming EEMs. However, they are not precise enough as especially signal E in the IS EEM does not show its maximum at a constant wavelength, which can mainly be explained by the incomplete separation of Chl a and PC in samples at the stationary growth phase. Thus, SyncScans can be used only with minor restriction as a quick and easy to measure alternative to EEMs for the correlation with cyanobacterial toxins. As a consequence, the following comparison of toxin concentration and fluorescence intensities is performed with the EEM data.
(E)-MC-LR/Chl a
0.15
0.10
0.05
0.00
10
15
20
25
30
35
40
days of cultivation
Fig. 5 – Mean normalized ( [ Chl a specific) MC-LR concentration of two cell cultures of M. aeruginosa during growth before ((E)-MC-LR, open symbols) and after ((I)-MCLR, filled symbols) mechanical cell lysis.
at the beginning of the growth phase, the difference between intra- and extracellular toxins was greatest due to a high vitality of the cells and no active excretion. During the stationary phase of the culture, free dissolved and cell bound toxins equalized at a relative MC-LR concentration around 0.05 mg per mg Chl a with a rather constant gap of 0.01 – 0.04 mg mg1 Chl a. At the end of the stationary phase, when cell lysis dominated, surprisingly not only the normalized concentrations for (E)- but also for (I)-MC-LR increased rapidly. This steep increase (factor of 2–3), which is illustrated in Fig. 5, was mainly caused by the increase of MC-LR. Between 24 and 38 days of growth the absolute concentrations changed from 74.4 to 180.1 mg L1 for (E)-MC-LR and from 128.5 to 311.5 mg L1 for (I)MC-LR. But partly the decrease of Chl a in the same period from 2.13 to 1.78 mg L1 due to cell lysis even emphasized the increase of the normalized MC-LR concentrations. Nevertheless, in the decay of their vitality, the cells even seemed to intensify the production of MC. This emphasizes the need of suitable fluorescence marker signals to detect the development of toxin producing cyanobacterial blooms at an early growth stage, which allows to set the time for proper reactions. The present findings for the normalized ( ¼ Chl a specific) MC-LR concentration are in good agreement with other studies. Izydorczyk et al. (2009) monitored cyanobacterial growth, mainly M. aeruginosa, in a drinking water reservoir in Poland and correlated the toxin concentration to cyanobacterial Chl a concentration. They found an average value for intracellular MC of 0.08 mg mg1 (cyanobacterial) Chl a and a maximum value of 0.28 mg mg1 Chl a. The correlation coefficient obtained was 0.70. Looking at the present results (Fig. 5), for the lab culture of M. aeruginosa MC-LR concentrations between 0.06 and 0.18 mg mg1 Chl a were obtained, depending on the growth stage. This emphasizes the range for a possible correlation of cyanobacterial Chl a to MC-LR concentration. Therefore, in a next step possible trends in the ratio between EEM signals and toxin concentration are considered. The MC-LR content of culture III is compared to its maxima of the protein-
202
water research 44 (2010) 195–204
Extracellular substances (ES)
a
2x10
Signal A/Chl a Signal X/Chl a Signal C/Chl a (E)-MC-LR/Chl a
5
Signal A/Chl a Signal X/Chl a Signal C/Chl a Signal E/Chl a (I)-MC-LR/Chl a
0.2
3x10
5
2x10 5
1x10
0.05
0.1 5
1x10
0.00
10
15
20
25
30
35
40
0
0.0
10
15
20
25
30
35
40
ratio intensity/Chl a [L/µg]
ratio MC-LR/Chl a [µg/µg]
Intracellular substances (IS)
b 5
0.10
0
days of cultivation
days of cultivation
Fig. 6 – Normalized ( [ Chl a specific) MC-LR concentration (in asterisks) and normalized fluorescence intensities (EEM) of the signals A, X, C and E for ES (open symbols, Fig. 12 a) and IS (filled symbols, Fig. 12 b) of one culture of M. aeruginosa.
like signal A, signal X and the humic-like signal C in the ES (Fig. 6 a) and IS (Fig. 6 b) samples and additionally to the pigments E in the IS samples. All values are related to the Chl a content. In case of ES, the agreement is very good for all fluorescence signals, the increase of excreted normalized MC-LR at the beginning of the stationary phase as well as the decrease at the end of the stationary phase, i. e. the beginning of the death phase are well depicted. The correlation coefficients amount to r ¼ 0.91 for signal A, r ¼ 0.79 for signal X and r ¼ 0.85 for signal C. The increasing (E)-MC-LR concentrations are therefore not predictable but well traceable. However, (I)-MC-LR is probably the even more crucial toxin parameter from the point of view of surface water treatment. The normalized pigment fluorescence signal E of the IS samples, which is dominated by PC, follows the increased toxin concentration at the beginning of the stationary phase and could therefore be used as tracer for (I)-MC-LR concentrations (r ¼ 0.75). This is also confirmed by Izydorczyk et al. (2005). According to the findings of the present study, the signals A (r ¼ 0.34) and C (r ¼ 0.26) belonging to protein-like and humiclike fluorescence, respectively, do not depict the course of MC-LR production as explicitly as the cyanobacterial pigments do. On the contrary, signal X shows its maximum fluorescence signal at the beginning of the stationary phase after 28 days of growth, coinciding with the beginning of the steep increase in (I)-MC-LR concentration, but 7–10 days earlier than the maximum toxin concentration. This results in a very low correlation coefficient (r ¼ 0.20), but makes it a potential warning signal, i. e. a tool for the prediction of increasing cell number related toxin concentrations. However, absolute toxin concentrations depend on the cell related amount of toxins as well as on the absolute cell number. As the maximum Chla specific intensity of signal X arises at the beginning of the stationary phase, in first instance only the following increasing cell related (I)-MC-LR concentrations can be predicted, which, however, still represent a 3-fold amplification. Yet, one main drawback of the introduced procedure is to set the fluorescence and toxin data in reference to the Chl a concentration, although it is absolutely necessary to perform
a cell number related normalization. But Chl a is produced to a large extent by the majority of phototrophic organisms. This is not a problem as long as it concerns the analysis of bloom samples with one dominating cyanobacterial species, which forms more than 90% of the total biomass. However, for the monitoring of the initial phase of a developing bloom cyanobacterially derived Chl a needs to be measured, which is not possible with conventional methods. Izydorczyk et al. (2009) found a fair correlation (with a correlation coefficient of 0.68) between cyanobacterial Chl a concentration measured by fluorescence according to Beutler et al. (2002) and the cyanobacterial biovolume in lake water samples, supporting the decision to use the Chl a concentration of the cultures to normalize the obtained fluorescence and toxin data.
4.
Conclusions
Fluorescence analysis was successfully applied for monitoring the growth of hepatotoxin producing lab cultures of the cyanobacterium M. aeruginosa. The correlation of EEM fluorescence maxima to Chl a and toxin concentrations revealed the presence of several signals for tracing the growth of the cyanobacterium. Additionally, signal X at 315 nm/396 nm caused by unknown compounds reflects the steep increase in intracellular toxin concentration with a forerun of several days and might be used for early-warning purposes. The introduced analysis procedure is even more attractive, as fluorescence analysis can be accelerated with minor restrictions by SyncScans with Dl of 80 nm – which can be measured online within a few minutes – instead of time consuming recording of EEMs, although minor peak shifts during cyanobacterial growth under lab conditions were found. Additionally, fluorescence spectrometric devices are far more widespread and economically feasible than devices for the quantification of single cyanobacterial metabolites and especially toxins. However, in future studies there is the need to test whether the presented findings are transferable to cyanobacterial species other than M. aeruginosa. Furthermore, the
water research 44 (2010) 195–204
applicability of the proposed method for the detection of bloom formation under field conditions needs to be validated.
Acknowledgements This work was funded by the German Research Foundation (DFG), project no. FR536/31, and by the German Federal Ministry of Education and Research (BMBF), project no. FKZ02WT0480. The authors wish to thank the DFG and the BMBF for the financial support. Further thank goes to Elly Karle for the HPLC analyses.
Appendix. Supplementary data The supplementary data associated with this article can be found in the online version, at doi:10.1016/j.watres.2009.09.035.
references
Alberts, J.J., Takacs, M., 2004. Total luminescence spectra of IHSS standard and reference fulvic acids, humic acids and natural organic matter: comparison of aquatic and terrestrial source terms. Org. Geochem. 35, 243–256. Beutler, M., Wiltshire, K.H., Meyer, B., Moldaenke, C., 1998. Differentiation of Spectral Groups of Algae with Computerbased Analysis of Fluorescence Excitation Spectra. Vom Wasser. 91, 61–74. Beutler, M., Wiltshire, K.H., Meyer, B., Moldaenke, C., Luering, C., Meyerhoefer, M., Hansen, U., Dau, H., 2002. A fluorometric method for the differentiation of algal populations in vivo and in situ. Photosynth. Res. 72, 39–53. Beutler, M., Wiltshire, K.H., Arp, M., Kruse, J., Reineke, C., Moldaenke, C., Hansen, U., 2003. A reduced model of the fluorescence from the cyanobacterial photosynthetic apparatus designed for the in situ detection of cyanobacteria. Biochim. Biophys. Acta, Bioenerg. 1604, 33–46. Chen, W., Westerhoff, P., Leenheer, J.A., Booksh, K., 2003. Fluorescence excitation-emission matrix regional integration to quantify spectra for dissolved organic matter. Environ. Sci. Technol. 37, 5701–5710. Coble, P.G., 1996. Characterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopy. Mar. Chem. 51, 325–346. Crush, J.R., Briggs, L.R., Sprosen, J.M., Nichols, S.N., 2008. Effect of irrigation with lake water containing microcystins on microcystin content and growth of ryegrass, clover, rape, and lettuce. Environ. Toxicol. 23, 246–252. Elliot, S., Lead, J.R., Baker, A., 2006. Characterization of the fluorescence from freshwater plankton bacteria. Water Res. 40, 2075–2083. Falconer, I., Bartram, J., Chorus, I., Kuiper-Goodman, T., Utkilen, H., Burch, M., Codd, G.A., 1999. Safe levels and safe practices. In: Chorus, I., Bartram, J. (Eds.), Toxic Cyanobacteria in Water. A Guide to Their Public Health Consequences, Monitoring and Management. E&FN Spoon, London, England, pp. 155–178.
203
Ferrari, G.M., Mingazzini, M., 1995. Synchronous fluorescence spectra of dissolved organic matter (DOM) of algal origin in marine coastal waters. Mar. Ecol.: Prog. Ser. 125, 305–315. Fujita, Y., 1997. A study on the dynamic features of photosystem stoichiometry: accomplishments and problems for future studies. Photosynth. Res. 53, 83–93. Gregor, J., Marsˇa´lek, B., 2004. Freshwater phytoplankton quantification by chlorophyll a: a comparative study of in vitro, in vivo and in situ methods. Water Res. 38, 517–522. Gregor, J., Marsˇa´lek, B., 2005. A simple in vivo fluorescence method for the selective detection and quantification of freshwater cyanobacteria and eukaryotic algae. Acta Hydrochim. Hydrobiol. 33, 142–148. Gregor, J., Marsalek, B., Sipkova, H., 2007. Detection and estimation of potentially toxic cyanobacteria in raw water at the drinking water treatment plant by in vivo fluorescence method. Water Res. 41, 228–234. Henrion, R., Henrion, G., Bo¨hme, M., Behrendt, H., 1997. Threeway principal components analysis for fluorescence spectroscopic classification of algae species. Fresenius J. Anal. Chem. 357, 522–526. Her, N., Amy, G., Park, H.-R., Song, M., 2004. Characterizing algogenic organic matter (AOM) and evaluating associated NF membrane fouling. Water Res. 38, 1427–1438. Izydorczyk, K., Carpentier, C., Mrowczyski, J., Wagenvoort, A., Jurczak, T., Tarczynska, M., 2009. Establishment of an alert level framework for cyanobacteria in drinking water resources by using the Algae online analyzer for monitoring cyanobacterial chlorophyll a. Water Res. 43, 989–996. Izydorczyk, K., Tarczynska, M., Jurczak, T., Mrowczynski, J., Zalewski, M., 2005. Measurement of phycocyanin fluorescence as an online early warning system for cyanobacteria in reservoir intake water. Environ. Toxicol. 20, 425–443. Klapper, L., McKnight, D.M., Fulton, J.R., Blunt-Harris, E.L., Nevin, K.P., Lovley, D.R., Hatcher, P.G., 2002. Fulvic acid oxidation state detection using fluorescence spectroscopy. Environ. Sci. Technol. 36, 3170–3175. Lee, N., Amy, G., Croue, J., 2006. Low-pressure membrane (MF/UF) fouling associated with allochthonous versus autochthonous natural organic matter. Water Res. 40, 2357–2368. Mankiewicz, J., Komarkova, J., Izydorczyk, K., Jurczak, T., Tarczynska, M., Zalewski, M., 2005. Hepatotoxic cyanobacterial blooms in the lakes of northern Poland. Environ. Toxicol. 20, 499–506. McKnight, D.M., Boyer, E.W., Westerhoff, P.K., Doran, P.T., Kulbe, T., Andersen, D.T., 2001. Spectrofluorometric characterization of dissolved organic matter for indication of precursor organic material and aromaticity. Limnol. Oceanogr. 46, 38–48. Metsamaa, L., Kutser, T., Stroembeck, N., 2006. Recognising cyanobacterial blooms based on their optical signature: a modelling study. Boreal Environ. Res. 11, 493–506. Mingazzini, M., 2001. Synchronous fluorescence spectra as chemical tracers to monitor the organic matter dissolved in north Adriatic waters. In: Faranda, F.M., Guglielmo, L., Spezie, G. (Eds.), Mediterranean Ecosystems: Structures and Processes. Springer, Milan, Italy, pp. 115–122. Nitschke, W., Rutherford, A.W., 1991. Photosynthetic reaction centres: variations on a common structural theme? Trends Biochem. Sci. 16, 241–245. Pemberton, K.L., Smith, R.E.H., Silsbe, G.M., Howell, T., Watson, S.B., 2007. Controls on phytoplankton physiology in Lake Ontario during the late summer: evidence from new fluorescence methods. Can. J. Fish. Aquat. Sci. 64, 58–73. Schmid, H., Stich, H.B., 1995. HPLC-analysis of algal pigments: comparison of columns, column properties and eluents. J. Appl. Phycol. 7, 487–494. Schmid, H., Bauer, F., Stich, H.B., 1998. Determination of algal biomass with HPLC pigment analysis from lakes of different
204
water research 44 (2010) 195–204
trophic state in comparison to microscopically measured biomass. J. Plankton Res. 20, 1651–1661. Seppaila, J., Balode, M., 1998. The use of spectral fluorescence methods to detect changes in the phytoplankton community. Hydrobiologia 363, 207–217. Sheng, G.-P., Yu, H.-Q., 2006. Characterization of extracellular polymeric substances of aerobic and anaerobic sludge using three-dimensional excitation and emission matrix fluorescence spectroscopy. Water Res. 40, 1233–1239. Simis, S.G.H., Peters, S.W.M., Gons, H.J., 2005. Remote sensing of the cyanobacterial pigment phycocyanin in turbid inland water. Limnol. Oceanogr. 50, 237–245. Subramaniam, A., Carpenter, E.J., Karentz, D., Falkowski, P.G., 1999a. Bio-optical properties of the marine diazotrophic
cyanobacteria Trichodesmium spp. I. Absorption and photosynthetic action spectra. Limnol. Oceanogr. 44, 608–617. Subramaniam, A., Carpenter, E.J., Falkowski, P.G., 1999b. Biooptical properties of the marine diazotrophic cyanobacteria Trichodesmium spp. II. A reflectance model for remote sensing. Limnol. Oceanogr. 44, 618–627. Utermo¨hl, H., 1958. Quantitative phytoplankton method (Vervollkommnung der quantitativen PhytoplanktonMethodik). Mitt. Int. Ver. Limnol. 9, 1–32. Welker, M., von Dohren, H., 2006. Cyanobacterial peptide – nature’s own combinatorial biosynthesis. FEMS Microbiol. Rev. 30, 530–563. Welschmeyer, N.A., 1994. Fluorometric analysis of chlorophyll a in the presence of chlorophyll b and pheopigments. Limnol. Oceanogr. 39, 1985–1992.
water research 44 (2010) 205–213
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Modelling the removal of p-TSA (para-toluenesulfonamide) during rapid sand filtration used for drinking water treatment Raffaella Meffe a,*, Claus Kohfahl a, Ekkehard Holzbecher b, Gudrun Massmann a, Doreen Richter c, Uwe Du¨nnbier d, Asaf Pekdeger a a
Institute of Geological Sciences, Freie Universita¨t Berlin, Malteserstr. 74-100, 12249 Berlin, Germany Georg-August-Universita¨t Go¨ttingen, GZG, Goldschmidtstr. 3, 37077 Go¨ttingen, Germany c DVGW-Technologiezentrum Wasser (TZW), Karlsruher Straße 84, 76139 Karlsruhe, Germany d Department of Laboratories, Berliner Wasserbetriebe, 10864 Berlin, Germany b
article info
abstract
Article history:
A finite element model was set-up to determine degradation rate constants for p-TSA
Received 26 March 2009
during rapid sand filtration (RSF). Data used for the model originated from a column
Received in revised form
experiment carried out in the filter hall of a drinking water treatment plant in Berlin
27 August 2009
(Germany). Aerated abstracted groundwater was passed through a 1.6 m long column-
Accepted 31 August 2009
shaped experimental sand filter applying infiltration rates from 2 to 6 m h1. Model results
Available online 6 September 2009
were fitted to measured profiles and breakthrough curves of p-TSA for different infiltration rates using both first-order reaction kinetics and Michaelis–Menten kinetics. Both
Keywords:
approaches showed that degradation rates varied both in space and time. Higher degra-
p-TSA
dation rates were observed in the upper part of the column, probably related to higher
Microbial degradation
microbial activity in this zone. Measured and simulated breakthrough curves revealed an
Reactive transport modelling
adaption phase with lower degradation rates after infiltration rates were changed, followed
Rapid sand filtration
by an adapted phase with more elevated degradation rates. Irrespective of the mathematical approach and the infiltration rate, degradation rates were very high, probably owing to the fact that filter sands have been in operation for decades, receiving high p-TSA concentrations with the raw water. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
The occurrence, fate and potential harms of organic pollutants in the aquatic environment have been an object of growing interest in recent years, as shown by the increasing number of studies in this field (Heberer, 2002a,b; Yu et al., 2006; Kim, 2007; Loss et al., 2007). The presence of organic pollutants in groundwater is of particular concern where the groundwater is used for drinking water production. The principal source of organic micropollutants in the aquatic
environment is municipal wastewater, which is generally treated in wastewater treatment plants (WWTPs) before being discharged into surface water (Daughton and Ternes, 1999). In the past, untreated wastewater was often also irrigated on sewage irrigation farms, causing a significant anthropogenic contamination of the surrounding environment (Grunewald, 1994; Heberer and Stan, 1994; Bechmann and Grunewald, 1995a,b; Abdel-Shafy et al., 2008). Various studies showed that residues of some organic pollutants from human and animal use are not fully eliminated during wastewater treatment and
* Corresponding author. Freie Universita¨t Berlin, Institute of Geological Sciences, Hydrogeology Group, Malteserstr. 74-100, 12249 Berlin, Germany. Tel.: þ49 30 838 70876; fax: þ49 30 838 70742. E-mail address:
[email protected] (R. Meffe). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.08.046
206
water research 44 (2010) 205–213
can be found in the aquatic environment (Heberer, 2002a,b; Derksen et al., 2004). Wastewater-bound, poorly degradable compounds may enter the raw water for drinking water production via bank filtration, or the catchment area of a drinking water treatment plant (DWTP) may receive groundwater affected by former sewage irrigation. The organic pollutant discussed in the following (para-toluenesulfonamide, p-TSA) originates from wastewater. It is applied as a plasticizer, an intermediate for pesticides and drugs, and is the primary degradation product of the common disinfectant Chloramine-T in water. Chloramine-T is used as an antimicrobial agent in the food industry to disinfect surfaces, instruments and machinery. This substance is also used as a therapeutic drug for bacterial gill diseases of fish species and for bacterial diseases of swine and poultry (Beljaars et al., 1994; Meinertz et al., 1999; Haneke, 2002; Gaikowski et al., 2004; Harris et al., 2004; Smail et al., 2004; Richter et al., 2007). According to the German Federal Environmental Agency (UBA), the tolerable concentration limit of p-TSA in drinking water is 0.3 mg L1 (Grummt and Dieter, 2006). P-TSA was found to be ubiquitous in the aquatic environment in Berlin, the largest city of Germany (Richter et al., 2008a). It was detected in Berlin’s untreated and treated wastewater, surface water, groundwater and raw water used for drinking water production (Richter et al., 2008a). The highest concentrations of p-TSA (up to 38 mg L1) were encountered within the catchment area of a DWTP (Friedrichshagen) in the eastern part of the city. This DWTP is located downgradient of a former sewage irrigation farm, where untreated wastewater had been irrigated directly onto the soils until the 1980s, when the farm was closed. Though the concentrations of p-TSA in the raw water of this plant are considerably lower due to dilution with bank filtrate from Lake Mu¨ggelsee, an efficient removal during treatment is still necessary to reach the required limit of 0.3 mg L1 in the final drinking water. Treatment at this DWTP involves aeration and rapid sand filtration (RSF) through open bed filters composed of biologically active sand (Richter et al., 2008b). Using an analytical method described in Richter et al. (2007, 2008b) investigated the behaviour of p-TSA during drinking water treatment with an experimental sand filter (column experiment), which provided the data used for the present modelling approach. Incubation experiments revealed that p-TSA degradation occurs as a result of microbial processes (Richter et al., 2008b). In addition, it appears to be largely limited to oxic conditions, explaining the persistence of p-TSA in the anoxic groundwater downgradient of the former sewage irrigation site (Richter et al., 2009). According to a laboratory experiment carried out by Richter et al. (2008b) sorption and retardation are negligible and can be excluded as a potential removal process. Their conclusion is also supported by data from a recently conducted, unpublished column study, in which p-TSA and a tracer were injected at the same time. P-TSA breakthrough occurred simultaneously with the tracer breakthrough at the outlet. Results of column studies are normally valid only for the specific experimental conditions, making comparison with other experiments and upscaling to field conditions difficult or impossible. Modelling refines and improves the interpretation of experimental studies by providing reaction rate constants which can be applied also to
other sites and experimental condition. This holds true particularly where experimental conditions, such as influent concentrations, are highly variable and transient as in the present case, and it therefore becomes difficult to distinguish different effects. In the literature different approaches were used to simulate microbially mediated reactions in column experiments. The simplest approach neglects kinetics induced by microbial activity and assumes instantaneous chemical equilibrium (e.g. Sabbagh et al., 2007). This method is appropriate if the microbial kinetics are fast compared to the transport timescale. Another group of models accounts for kinetics using zero or first-order reaction (Knudsen et al., 2000; Amondham et al., 2006) or Michaelis–Menten kinetics (Bengtsson and Carlsson, 2001a; Park et al., 2001; Sato et al., 2002). Both approaches are based on the concept of degradation constants, assuming that the microbial population does not change. Monod-type kinetics accounts also for microbial dynamics that describes growth dynamics (Ho¨hener et al., 2006; Kim and Jaffe, 2008; Kinzelbach et al., 1991). The role of microbiology in degradation has been investigated by numerical simulation for several organic compounds, such as antimicrobials (Rooklidge et al., 2004), hydrocarbons (Bengtsson and Carlsson, 2001b; Goedeke et al., 2008), chlorinated organic compounds (Bosma et al., 1988; Corapcioglu et al., 1991) and pesticides (Pang et al., 2005; Magga et al., 2008). However, no quantitative information applicable to other sites and conditions has been obtained yet for the sulfonamide p-TSA. Therefore, this paper provides a modelling framework for the simulation of the column experiment performed by Richter et al. (2008b), and the aim of this study was to (i) to determine the reaction rate constants defining microbial degradation of p-TSA, (ii) to explore and compare two different kinetic approaches to describe the degradation process and (iii) to investigate the dependence of microbial reaction rate constants on infiltration rates.
2.
Methods
2.1.
Column experiment
The column experiment performed by Richter et al. (2008b) simulated rapid sand filtration during drinking water treatment in Berlin. The column was installed in the filter hall of the DWTP Friedrichshagen (Fig. 1) and was operated similarly to the real large rapid sand filters. It was packed with the same silica sand that had been used for decades during RSF and had a length of 1.6 m, corresponding to the real filter length used during drinking water treatment at the DWTP Friedrichshagen. The silica sand had an effective porosity of 0.3, determined with the saturation method, and a uniformity coefficient (ratio d60 over d10) of 1.33 (analysis carried out by the laboratories of Freie Universita¨t Berlin). Sampling ports were installed every 0.25 m along the column (Fig. 1). The aerated raw water used during routine treatment was passed through the column. The infiltration rate was regulated via the effluent flow. After an initial regulation phase of about two weeks using an infiltration rate of 2 m h1, the functional capability of the experimental filter was determined by
water research 44 (2010) 205–213
207
Backwashing of the sand with drinking water was performed every three to four days at an infiltration rate of 40 m h1. During the entire experiment 66 samples were collected both from the inlet and the outlet of the column at intervals between 1 and 3 days. Moreover, to obtain steady state concentration profiles the p-TSA sampled at the intermediate ports was analysed once for each infiltration rate after a minimum of 8 days of constant infiltration rate. The measurements at the outlet and at the sampling ports detected only the removal of p-TSA without accounting for the possibility of p-TSA transformation into an intermediate product. Therefore, the term ‘‘degradation’’ in the present paper refers to a microbial removal of p-TSA and not to a complete mineralization.
2.2.
Model set-up
2.2.1.
Equations
The simulations were performed with Comsol Multiphysics 3.3 (COMSOL Multiphysics, Version 3.3, 2006), a multiphysics software tool for the solution of partial differential equations, which is based on the finite element method. A transient one-dimensional solute transport model was set-up based on the governing equations: qs
vc þ V$½ qs DL Vc þ uc ¼ RL ; vt
DL ¼ al v þ
Fig. 1 – Experimental set-up after Richter et al. (2008b).
measuring the concentration of ammonium, iron and manganese, all efficiently removed in the column. After the initial regulation phase the infiltration rate was raised in steps from 2 m h1 to 6 m h1 (Fig. 2). Results of 5 m h1 are not presented here owing to technical problems during measurements. Considering a given porosity of 0.3, the corresponding residence times in the column range from 9.6 min for the infiltration rate of 3 m h1 and 4.8 min for the infiltration rate of 4 m h1.
(2)
where c (M L3) denotes the solute concentration in the liquid for the studied specie, qS is the porosity, DL stands for hydrodynamic dispersion (L2 T1), al is the dispersivity (L), v is the seepage velocity (L T1), Df is the molecular diffusion coefficient corrected for temperature and pressure (L2 T1), s is the tortuosity, u is the Darcy velocity (L T1), and RL is the reaction term (M L3 T1). The first term of Eq. (1) gives the time rate change in dissolved mass within the porous medium; the expression in brackets is the solute flux.
2.2.2.
Physical parameters
Physical properties used in the model are listed in Table 1 and were assumed to be constant throughout the entire solution domain. The tortuosity was defined as the ratio of the real path length over the shortest path length and was approximated to p/2, assuming a circular shape of the silica grains.
2.2.3.
Fig. 2 – Measured inflow and outflow concentrations during the entire column experiment.
Df s
(1)
Chemical parameters
The initial p-TSA concentration was set to 0.1 mg L1, and the measured concentrations of p-TSA at the column inlet were defined as transient inflow concentrations. Microbial degradation of p-TSA was simulated by (i) first-order kinetics and (ii) Michaelis–Menten kinetics, both defined in the model by the reaction term of Eq. (1). Monod kinetics was not applied due to the lack of input data required for this approach. Firstorder kinetics assumes that the only factor affecting degradation is the concentration of the substrate, without considering a maximum reaction rate. The reaction rate based on first-order kinetics is expressed by the following equation (Appello and Postma, 2007):
208
water research 44 (2010) 205–213
Table 1 – Input parameters of model simulations. Input parameters
Value
Reference
Porosity Dispersivity (m) Tortuosity factor Diffusion coefficient (m2 s1) Infiltration rate (m h1)
0.3 0.01 1.57 1 e-9 3. 4. 6
Experimental data Gelhar et al. (1985) p/2 Frederikse and Lide (1997) Experimental set-up
RL ¼ qs lc
(3) 1
where l(T ) is the degradation rate constant. For the remaining parameters, refer to Eq. (1). The Michaelis–Menten approach describes the dependence of the reaction velocity on the concentration considering a maximum reaction rate (Michaelis and Menten, 1913): RL ¼ qs
Kmax $c ks þ c
(4)
where Kmax is the maximum reaction rate (M T1), and ks is the half-velocity concentration (M L3), also known as Michaelis– Menten constant. For the remaining parameters, refer to Eq. (1). Following the results obtained from the analysis of the filter sludge samples carried out by Richter et al. (2008b), sorption and retardation were not taken into account. The simulations did not take the consumption of oxygen into account because the column experiment was conducted under completely oxic conditions. Hence, parameters are representative for oxic conditions only and are expected to be much lower during anoxic conditions.
2.2.4.
Discretization
The column experiment was modeled by a 1.60 m long, onedimensional solution domain discretized in 120 quadratic elements, corresponding to a degree of freedom of 241. Mesh refinement studies were carried out, and the simulations showed that the results are mesh-independent.
2.2.5.
Calibration procedure
Forward modelling runs with the described model were performed for parameter estimation. Transport parameters were derived from the measurements or from the literature (Table 1), and only the degradation parameters were used as fitting parameters. Inverse modelling was performed manually, minimizing the difference between numerical and experimental values. The mean square error was implemented as a measure of the fit. The difference between simulated and measured results is expressed by: vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi !ffi u m X 1u ðcmeas csim Þ2 R¼ t m n¼1
5
where R is the least square residual, cmeas denotes the measured concentrations, csim is the simulated concentration. The sum in Eq. (5) extends on all values from profiles and breakthrough curves. m is the number of measurements in the profiles and in the breakthrough curves. Note that a weighting factor was not used.
Fig. 3 – Measured steady state depth profiles.
3.
Results
3.1.
Column experiment
The depth profiles measured for infiltration rates of 3, 4 and 6 m h1 showed that p-TSA was almost completely degraded after passage through the column (Fig. 3). Changes of initial concentrations during the experiment correspond to the variability of the inflowing raw water composition. Effluent concentrations show similar values for all infiltration rates. The measured steady state profiles revealed a strong decrease in concentration as far as a depth of 0.5 m, whereas in the lower part of the column only a minor decrease was observed. The breakthrough curves resulting from the experiment showed a total reduction of >89% p-TSA after passage through the filter for all infiltration rates tested (Richter et al., 2008b). A further outcome of breakthrough curve measurements is that variations in the concentration of the inflowing raw water are more attenuated at the outlet of the column after a minimum number of days with constant infiltration rates (Fig. 3). During the first days after a change in the infiltration rate, the attenuation of the breakthrough curves was weaker, indicating lower degradation rate constants during this period. In the following, the term ‘‘adaption phase’’ refers to the first period with lower degradation rates, and the subsequent period is named ‘‘adapted phase’’. The length of the adaption phase ranged from 3 to 20 days (Table 2). The long duration of 20 days after the infiltration rate was reduced from 6 to 4 m h1 may be attributed to strong changes of the inflow concentration in the same period (Fig. 3).
3.2.
Modelling
Measured breakthrough curves and steady state profiles of 3, 4 and 6 m h1 were fitted by first-order kinetics and Michaelis–
209
water research 44 (2010) 205–213
Table 2 – Values of fitted parameters of first-order kinetics and Michaelis–Menten kinetics. Lambda values are given in sL1 and maximum reaction rates (Kmax) in mg LL1 sL1. Velocity (m h1)
First-order kinetics Adaption phase
Adapted phase
Quality of the fit (R)
Upper Lower Upper Lower part (l1) part (l2) part (l3) part (l4) 3 4 6
6.94E-03 2.89E-03 1.62E-02 1.60E-03 8.68E-03 4.05E-03 1.04E-02 3.93E-03 9.49E-03 5.90E-03 1.62E-02 4.98E-03
Michaelis–Menten kinetics Adaption phase
Adapted phase
Quality of the fit (R)
Length adaption phase (day)
0.0074 0.0065 0.0346
9 20 3
Upper part Lower part Upper part Lower part (Kmax1) (Kmax2) (Kmax3) (Kmax4) 0.0256 0.0068 0.0056
8.10E-03 6.37E-03 8.10E-03
Menten kinetics using l, ks and Kmax respectively as fitting parameters. In the initial simulation using first-order kinetics, only one temporary invariant degradation constant for p-TSA for the entire column was assumed, resulting in a single value of 5.55E-03 s1 for l. It was not possible to reproduce the measured data, neither the steady state profiles nor the analysed breakthrough curves as shown for 3 m h1 in Fig. 4. To account for changes of degradation rate constants, the column was divided into an upper part of 0.5 m and a lower part of 1.1 m, allowing the attribution of higher and lower degradation rate constants respectively. With regard to the temporal change of l in the adaption phase and the adapted phase, one degradation rate constant for both time periods (adapted/adaption phase) in both parts of the column (upper/ lower) was considered, resulting in four different values of l. The parameter sets used to fit the different infiltration rates are compiled in Table 2. The lengths of the defined adaption phases are derived from the experimental data and are documented in Figs. 5–7 and in Table 2. No information about the microbiology of the sediment inside the column being available, Kmax and ks values for the two parts of the column were used as fitting parameters. The ks value is specific for a microbial species, and Kmax depends on the individual microbial activity and the number of individuums (Appello and Postma, 2007). Changes of microbial species are not assumed here, and therefore only one ks value was used for the simulation of all filtration rates. To identify the optimal ks value, several simulations were run for the infiltration rate of 3 m h1. The value of ks that provided the
Fig. 4 – Measured and simulated p-TSA breakthrough curve and steady state profile using first-order kinetics for the infiltration rate of 3 m hL1, considering one temporary invariant degradation constant l for the entire column.
1.50E-03 2.78E-03 5.67E-03
1.42E-02 1.16E-02 1.22E-02
4.63E-04 2.08E-03 2.31E-03
best fit for the experimental data (R ¼ 0.0074) was 0.3 mg L1 (Fig. 5). However, ks values between 0.1 mg L1 and 0.5 mg L1 also reproduced the experimental data with sufficient agreement. After fixing the optimum ks value, for each infiltration rate a different Kmax was used to fit simulated to observed results and analogously to first-order kinetics, 4 maximum reaction rates (Kmax) were attributed to the upper/lower column and to the adaption/adapted phase in the Michaelis– Menten approach. For all infiltration rates and both kinetic approaches, simulated and observed data are in good agreement, as illustrated in Figs. 5–7. In the case of the infiltration rate of 6 m h1 (Fig. 7) the simulated peak is related to a strong increase of the inflow concentration at this time, as documented in Fig. 3.
3.3.
Sensitivity study
The sensitivity study was performed to investigate the effects of l and Kmax on degradation, especially because no information on the microbiology within the column is available. The sensitivity study was carried out for two infiltration rates (3 and 6 m h1) and one range of parameter change (50%) using the verified models (Table 3). The sensitivity of l and Kmax was calculated considering the steady state outflow concentration (cout) as model dependent variable, and the adapted phase was analysed. To exclude the effect of different inflow concentrations, a constant inflow concentration of 1.76 mg L1 was used for all
Fig. 5 – Measured and fitted concentrations of p-TSA using first-order kinetics and Michaelis–Menten (MM) kinetics for the infiltration rate of 3 m hL1, considering different values of degradation constants. The dashed circle indicates the adaption phase.
210
water research 44 (2010) 205–213
Fig. 6 – Measured and fitted concentrations of p-TSA using first-order kinetics and Michaelis–Menten (MM) kinetics for the infiltration rate of 4 m hL1, considering different values of degradation constants. The dashed circle indicates the adaption phase.
Fig. 7 – Measured and fitted concentrations of p-TSA using first-order kinetics and Michaelis–Menten (MM) kinetics for the infiltration rate of 6 m hL1, considering different values of degradation constants. The dashed circle indicates the adaption phase.
the simulations. Parameter sensitivity was tested for the upper part of the column because l3 and Kmax3 affect also the lower part and are supposed to have a major impact on steady state outflow concentrations. To enable comparison of sensitivities between different parameters, sensitivities are normalized according to Bennett and Zheng (2002):
4.
Xp ¼
vcout vP=P
(6)
where Xp is the normalized sensitivity, cout is the steady state outflow concentration, and P represents the tested parameter (l3 or Kmax3). The outcome of the sensitivity study is presented in Table 3. The analysis showed a greater sensitivity to Kmax than to l for both infiltration rates, with the sensitivity of Kmax being almost three times higher than the sensitivity of l for the infiltration rate of 3 m h1 and more than two times higher for the infiltration rate of 6 m h1. As expected, the resulting data show a negative correlation between infiltration rate and parameter sensitivity, which is due to the dependence of parameter sensitivity on residence times.
Discussion
Richter et al. (2008b) already presented a rough estimation of degradation rate constants for p-TSA during the column experiment obtained by simple graphical exponential fitting of the steady state depth profiles without considering invariant inflow concentrations, changes in infiltration rate, and breakthrough curves. The approach was only rudimentary, similar to the initial simulation described above, where only one temporary invariant degradation rate constant for p-TSA for the entire column was used. The resulting degradation rate constants of Richter et al. (2008b) were around 6.3E-03 s1, but obtained fits were rather poor. Instead, in the present paper, the degradation rate constants were derived through the application of a model that considered both depth profiles as well as breakthrough curves simultaneously and accounted for the spatial and temporary variability of the parameters, thereby obtaining much better agreement between modeled and measured data and a refined understanding of the processes. Fitted l (and likewise Kmax) values for p-TSA are very high, ranging from 103 (in the lower part of the column) to 102 s1
Table 3 – Sensitivity study. The first-order degradation rate constant values (l) are given in sL1, the maximum reaction rate values (Kmax) in mg LL1 sL1 and the concentration in mg LL1. The constant inflow concentration is 1.76 mg LL1. Run 1D 3 m h1
6 m h1
Reference 1 Reference 2 Reference 3 Reference 4
Parameter
Value in the reference model
Value in the sensitivity
l3
1.62E-02
8.10E-03
Kmax3
1.42E-02
7.12E-03
l3
1.62E-02
8.10E-03
Kmax3
1.22E-02
6.10E-03
% Change in parameter
cout Value
0.00 50.00 0.00 50.00
0.075 0.280 0.076 0.637
0.00 50.00 0.00 50.00
0.166 0.328 0.574 0.953
Normalized sensitivity (Xp) 0.410 1.125
0.324 0.758
water research 44 (2010) 205–213
Fig. 8 – Fitted degradation rate constants in the adapted phase using first-order kinetics and Michaelis–Menten kinetics for all three infiltration rates.
(in the upper part of the column) during the adapted phase (Fig. 8). For comparison, values of degradation constants for other substances such as pesticides range between 106 s1 for the insecticide acephate/orthene and 109 s1 for the herbicide paraquat/Gramoxone (Howard, 2004). Mackay and Shiu (1992) determined half-lives of certain polycyclic aromatic hydrocarbons (PAHs) in the aquatic environments, resulting in degradation rate constants between 107 s1 and 108 s1. We assume that a highly adapted microbial community causes the effective elimination because the sands used in the column have been in operation for decades and p-TSA concentrations in the raw water have been continuously high for a long time. This is supported by
experiments performed by Richter et al. (2008b), who showed that microbial degradation only takes place in filter sands of DWTPs abstracting groundwater polluted with p-TSA. In incubation experiments where drinking water was spiked with p-TSA and backwash water from ‘‘unpolluted’’ sand filters, no degradation occurred. Simulations with both approaches confirmed the assumption based on visual interpretation of the experimental data that the degradation rate constants of p-TSA vary in both time and space (Table 2 and Fig. 8). The degradation rate constants (and likewise Kmax) between the upper and lower parts of the column showed a difference of up to one order of magnitude, which may be related to greater populations of microbes in the upper part of the column owing to the greater availability of nutrients close to the inlet of the column. The reason for the less efficient degradation of p-TSA in the adaption phase could be related to a reduced microbial activity following a change in infiltration rate, owing to the change of ambient conditions. Note that the temporal changes of the degradation rate constants (and likewise Kmax) were limited to the upper part of the column, leading to greater spatial differences during the adapted phase. The infiltration rates were originally varied to determine the optimum operational infiltration rate for p-TSA degradation. Fig. 8 illustrates that no clear correlation exists between degradation rate constants (and likewise Kmax) and infiltration rates. Fig. 9 shows the Michaelis–Menten function (eq. (4)) for parameter settings defined in the upper part of the column during the adaption (Kmax1) and adapted (Kmax2) phases for the infiltration rate of 3 m h1. The figure also includes the ks value as well as the observed concentration range. The fact that the optimized ks value falls within the range of natural concentrations results in high degradation rates already for low concentrations. For concentrations higher than 1 mg L1, the reaction rate RL increases only in minor amounts. Fig. 9 is also representative for the experimental conditions of the other infiltration rates because the values of Kmax are in the same order of magnitude as the values of p-TSA concentration. The two kinetic approaches both support the concept of changing degradation constants in space and time and produce excellent fitting results. In general, the Michaelis–Menten approach shows a higher capacity to reproduce fluctuations of the concentrations during this highly transient experiment. This holds true especially for the adaption phase, where fluctuating inflow concentrations are not attenuated until discharging at the outlet of the column.
5.
Fig. 9 – Applied Michaelis–Menten function for the infiltration rate of 3 m hL1 and range of experimental concentrations.
211
Conclusions
This research presents degradation rate constants for p-TSA removal during RSF, which were previously not available in the literature. The resulting values range between 103 and 102 s1. The model approaches, though simple, illustrate the usefulness of mathematical modelling to determine robust degradation parameters during drinking water treatment processes and can easily be applied to other sites and conditions. Results suggest higher microbial activities in the upper part than in the lower part of the column, where upper part refers to the sector of the column close to the inlet and lower
212
water research 44 (2010) 205–213
part refers to the subsequent column sector. Microbial activity also appears to be temporally variable depending on the change of ambient conditions due to the transition between different infiltration rates. The degradation parameters showed only minor variations for different infiltration rates; therefore an optimal infiltration rate for the removal process could not be determined. Comparison of the two applied kinetic approaches showed that the Michaelis–Menten approach is clearly more appropriate for reproducing highly transient experimental conditions than the more simplistic linear approach. Based on this research, some conclusions may be drawn for the design of treatment plants for p-TSA removal. First, maintenance of oxic conditions appears to be essential to guarantee these high degradation rates. Second, the study has demonstrated that the infiltration rate is not a relevant parameter to optimize future treatment strategies and to obtain more favourable degradation rates the infiltration rate should be maintained constant, avoiding the occurrence of the adaption phase with lower degradation efficiency. Finally, results suggested that the vertical thickness of the filter could be reduced to less than 1 m, because degradation at depths higher than 0.50 m almost vanishes.
references
Abdel-Shafy, H., Guindi, K.A., Tawfik, N.S., 2008. Groundwater contamination as affected by long term sewage irrigation in Egypt. Efficient Management of Wastewater, pp. 53–63. Amondham, W., Parkpian, P., Polprasert, C., Delaune, R.D., Jugsujinda, A., 2006. Paraquat adsorption, degradation, and remobilization in tropical soils of Thailand. J. Environ. Sci. Health Part B Pest. Food Contam. Agric. Waste 41 (5), 485–507. Appello, C.A.J., Postma, D., 2007. In: Balkema, A.A. (Ed.), Geochemistry, Groundwater and Pollution Amsterdam, The Netherlands. Bechmann, W., Grunewald, K., 1995a. Organic pollutants in soils and substrates of the sewage farm area south of Berlin. Zeitschrift fur Pflanzenernahrung und Bodenkunde 158 (8), 543–548. Bechmann, W., Grunewald, K., 1995b. PAH profiles in sewage farm substrates. Chemische Technik. 47 (1), 26–30. Beljaars, P.R., Vandijk, R., Brands, A., 1994. Determination of p-toluenesulfonamide in ice-cream by combination of continuous-flow and liquid-chromatography – summary of collaborative study. J. AOAC. Int. 77 (3), 672–674. Bengtsson, G., Carlsson, C., 2001a. Contribution of suspended and sorbed groundwater bacteria to degradation of dissolved and sorbed aniline. Appl. Microbiol. Biotechnol. 57 (1–2), 234–241. Bengtsson, G., Carlsson, C., 2001b. Degradation of dissolved and sorbed 2,4-dichlorophenol in soil columns by suspended and sorbed bacteria. Biodegradation 12 (6), 419–432. Bennett, G.D., Zheng, C.M., 2002. Applied Contaminant Transport Modeling: Theory and Practice. In: Wiley Interscience, vol. 621. John Wiley and Sons, New York. Bosma, T.N.P., Schnoor, J.L., Schraa, G., Zehnder, A.J.B., 1988. Simulation model for biotransformation of xenobiotics and chemotaxis in soil columns. J. Contam. Hydrol. 2, 225–236. Comsol Multiphysics, August 2006. Earth Science Module User’s Guide. Version 3.3. ª Copyright 1994-2006 by Comsol Ab. Corapcioglu, M.Y., Hossain, M.A., Hossain, M.A., 1991. Methanogenic biotransformation of chlorinated hydrocarbons in ground water. J. Environ. Eng. – ASCE 117 (1), 47–65.
Daughton, C.G., Ternes, T.A., 1999. Pharmaceuticals and personal care products in the environment: agents of subtle change? Environ. Health Perspect. 107, 907–938. Derksen, J.G.M., Rijs, G.B.J., Jongbloed, R.H., 2004. Diffuse pollution of surface water by pharmaceutical products. Water Sci. Technol. 49 (3), 213–221. Frederikse, H.P.R., Lide, D.R., 1997. Handbook of Chemistry and Physics. CRC Press, Boca Raton, FL. Gaikowski, M.P., Larson, W.J., Steuer, J.J., Gingerich, W.H., 2004. Validation of two dilution models to predict chloramine-T concentrations in aquaculture facility effluent. Aquatic. Eng. 30 (3–4), 127–140. Gelhar, L.W., Mantoglou, A., Welty, C., Rehfeldt, K.R., 1985. A Review of Field-Scale Physical Solute Transport Processes in Saturated and Unsaturated Porous Media. Rep. EA-4190. Electro. Power Res. Ins., Palo Alto, CA. Goedeke, S., Vogt, C., Schirmer, M., 2008. Estimation of kinetic monod parameters for anaerobic degradation of benzene in groundwater. Environ. Geol. 55 (2), 423–431. Grummt, T., Dieter, H.H., 2006. Investigation Report on the Compound p-TSA. German Federal Environmental Agency (UBA), Bad-Elster (in German). Grunewald, K., 1994. Investigations of soils in the large sewage farm areas south of Berlin. Zeitschrift fur Pflanzenernahrung und Bodenkunde 157 (2), 125–130. Haneke, K.E., 2002. Toxicological summary for chloramine-T [127-65-1] and p-toluenesulfonamide [70-55-3]. Integrated Laboratory System. http://ntpserver.niehs.nih.gov/ntp/htdocs/ Chem_Background/ExsumPdf/ChloramineT.pdf, pp. 68. Harris, J.O., Powell, M.D., Attard, M., Green, T.J., 2004. Efficacy of chloramine-T as a treatment for amoebic gill disease (AGD) in marine atlantic salmon (Salmo salar L.). Aquac. Res. 35 (15), 1448–1456. Heberer, T., 2002a. Occurrence, fate and removal of pharmaceutical residues in the aquatic environment: a review of recent research data. Toxicol. Lett. 131 (1–2), 5–17. Heberer, T., 2002b. Tracking persistent pharmaceutical residues from municipal sewage to drinking water. J. Hydrol. 266 (3–4), 175–189. Heberer, T., Stan, H.J., 1994. N-(phenylsulfonyl)-sarcosine a new contaminant in sewage farm ground-water. Fresen. Environ. Bull. 3 (10), 639–643. Ho¨hener, P., Dakhel, N., Christophersen, M., Broholm, M., Kjeldsen, P., 2006. Biodegradation of hydrocarbons vapors: comparison of laboratory studies and field investigations in the vadose zone at the emplaced fuel source experiment, Airbase Værløse. J. Contam. Hydrol. 88 (3–4), 337–358. Howard, D., 2004. Pesticide Adsorption and Half-Life. Utah State University Extension. Kim, Y., 2007. Aquatic toxicity of acetaminophen, carbamazepine, cimetidine, diltiazem and six major sulfonamides, and their potential ecological risk in Korea. Environ. Int. 33 (3), 370–375. Kim, H.S., Jaffe, P.R., 2008. Degradation of toluene by a mixed population of archetypal aerobes, microaerophiles and denitrifiers: laboratory sand column experiment and multispecies biofilm model formulation. Biotechnol. Bioeng. 99 (2), 290–301. Kinzelbach, W., Scha¨fer, W., Heryer, J., 1999. Numerical modeling of natural and enhanced denitrification processes in aquifers. Water Resour. Res. 25 (6), 1123–1135. Knudsen, J.B.S., Aagaard, P., Breedveld, G., 2000. A column study on reactive transport of dissolved jet-fuel in a sandy aquifer. 1. Biodegradation and mineralization kinetics of the organic solutes. Groundwater 2000, 175–176. Loss, R., Wollgast, J., Huber, T., Hanke, G., 2007. Polar herbicides, pharmaceutical products, perfluorooctanesulfate (PFOS), perfluorooctanoate (PFOA), and nonylphenol and its
water research 44 (2010) 205–213
carboxylates and ethoxylates in surface and tap waters around Lake Maggiore in Northern Italy. Anal. Bioanal. Chem. 387 (4), 1469–1478. Mackay, D., Shiu, W.Y., 1992. Polynuclear Aromatic Hydrocarbons, Polychlorinated Dioxins and Dibenzofurans. In: Illustrated Handbook of Physical–Chemical Properties and Environmental Fate of Organic Chemicals, vol. II. Lewis Publishers, Boca Raton, FL. Magga, Z., Tzovolou, D.N., Theodoropoulou, M.A., Dalkarani, T., Pikios, K., Tsakiroglou, C.D., 2008. Soil column experiments used as a means to assess transport, sorption and biodegradation of pesticides in groundwater. J Environ. Sci. Health Part-B Pest. Food Contam. Agric. Waste 43 (8), 732–741. Meinertz, J.R., Schmidt, L.J., Stehly, G.R., Gingerich, W.H., 1999. Liquid chromatographic determination of paratoluenesulfonamide in edible fillet tissues from three species of fish. J. AOAC. Int. 82 (5), 1064–1070. Michaelis, L., Menten, M.L., 1913. The kinetics of the invertase activity. Biochem. Z. 49, 333–369 (in German). Pang, L.P., Close, M., Flintoft, M., 2005. Degradation and sorption of atrazine, hexazinone and procymidone in coastal sand aquifer media. Pest. Manag. Sci. 61 (2), 133–143. Park, J., Chen, M.J., Kukor, J.J., Abriola, L.M., 2001. Influence of substrate exposure history on biodegradation in a porous medium. J. Contam. Hydrol. 51 (3–4), 233–256. Richter, D., Duennbier, U., Massmann, G., Pekdeger, A., 2007. Quantitative determination of three sulfonamides in environmental water samples using liquid chromatography coupled to electrospray tandem mass spectrometry. J. Chromatogr. A. 1157 (1–2), 115–121.
213
Richter, D., Massmann, G., Du¨nnbier, U., 2008a. Identification and significance of sulphonamides (p-TSA, o-TSA, BSA) in an urban water cycle (Berlin, Germany). Water Res. 42 (6–7), 1369–1378. Richter, D., Massmann, G., Du¨nnbier, U., 2008b. Behaviour of sulfonamides (p-TSA, o-TSA, BSA) during drinking water treatment. Chemosphere 71 (8), 1574–1581. Richter, D., Massmann, G., Taute, T., Du¨nnbier, U., 2009. Investigation of the fate of sulfonamides downgradient of a decommissioned sewage farm near Berlin, Germany. J. Contam. Hydrol. 106, 183–194. Rooklidge, S.J., Burns, E.R., Bolte, J.P., 2004. Modeling antimicrobial contaminant removal in slow sand filtration. Water Res. 39 (2–3), 331–339. Sabbagh, G.J., Fox, G.A., Ma, L., 2007. Modeling pesticide fate and nonideal transport from seed treated with a slow-release pesticide in a laboratory column. Trans. ASABE 50 (2), 523–532. Sato, T., Kimura, Y., Takamizawa, K., 2002. Modeling transport and biodegradation of PCE in sandy soil. 1st Conference on Brownfields Site: assessment, Rehabilitation and Development. Brownfields Site: Assessment, Rehabilitation and Development, pp. 515–524. Smail, D.A., Grant, R., Simpson, D., Bain, N., Hastings, T.S., 2004. Disinfectants against cultured Infectious Salmon Anaemia (ISA) virus: the virucidal effect of three iodophors, chloramines-T, chlorine dioxide and peracetic acid/hydrogen peroxide/acetic acid mixture. Aquaculture 240 (1–4), 29–38. Yu, J.T., Bouwer, E.J., Coelhan, M., 2006. Occurrence and biodegradability studies of selected pharmaceuticals and personal care products in sewage effluent. Agric. Water Manag. 82 (1–2), 72–80.
water research 44 (2010) 214–222
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Occurrence and fate of synthetic musk compounds in water environment In-Seok Lee, Sung-Hee Lee, Jeong-Eun Oh* Department of Civil and Environmental Engineering, Pusan National University, San 30, Jangjeon-dong, Geumjeong-gu, Busan 609-735, Republic of Korea
article info
abstract
Article history:
Synthetic musk compounds (SMCs) occur widely in water environments. The aims of this
Received 15 April 2009
paper were to investigate the occurrence and fate of SMCs in sewage treatment plants
Received in revised form
(STPs) and surface waters. Total SMC concentrations ranged from 3.69 to 7.33 mg/L
14 July 2009
(influent) and from 0.96 to 2.69 mg/L (effluent) in 10 STPs. The SMC concentrations varied
Accepted 31 August 2009
with the input source and treatment volume of each STP. Biological treatment processes
Available online 6 September 2009
had a greater SMCs removal effect than chemical treatment, filtration and disinfection processes. The SMC concentrations in surface waters ranged from 0.15 to 16.72 mg/L and
Keywords:
exhibited similar SMCs occurrence patterns generally. The fate of SMCs in water envi-
Occurrence
ronments depends on their physical–chemical properties and their concentrations can be
Fate
predicted from other SMC concentrations due to their similar fates. ª 2009 Elsevier Ltd. All rights reserved.
Synthetic musk compound Water environment
1.
Introduction
Nitro musk compounds (NMCs) were first synthesized at the end of the 19th century as fragrance substitutes for natural musk obtained from musk pods of the male musk deer because of the deer’s potential extinction (Heberer, 2002). Another group of fragrance materials, polycyclic musk compounds (PMCs), was developed in the 1950s and 60s (Heberer, 2002). Since then, these synthetic musk compounds (SMCs) have been extensively used as fragrance ingredients in consumer products such as cosmetics, detergents, fabric softeners, shampoos, perfumes and other scented personal care products (Balk and Ford, 1999). NMCs, especially musk xylene (MX; 1-tert-butyl-3,5-dimethyl2,4,6-trinitrobenzene) and musk ketone (MK; 4-acetyl-1-tertbutyl-3,5-dimethyl-2,6-dinitrobenzene), and PMCs, especially HHCB (1,3,4,6,7,8-hexahydro-4,6,6,7,8,8-hexamethylcyclopenta(g)-2-benzopyran; marketed as Galaxolide, Abbalide) and AHTN (7-acetyl-1,1,3,4,4,6-hexamethyltetralin; marketed as
Tonalide, Fixolide), account for approximately 12% and 85%, respectively, of global SMC production (Heberer, 2002). Owing to recent concerns about the effect of NMC toxicities on humans and the environment (Chou and Dietrich, 1999; Tas et al., 1997), their usage has gradually been declining; however, PMC usage has been increasing. Therefore, many studies have identified the entry of these SMCs into the influent of sewage treatment plants (STPs) after household applications and their discharge into the receiving water via the STP effluent due to incomplete removal during the treatment processes (Berset et al., 2004; Bester, 2004; Horii et al., 2007; Reiner et al., 2007; Simonich et al., 2002; Yang and Metcalfe, 2006). STPs have consequently become a potential source of SMCs in water environments (Bester, 2004; Zeng et al., 2007). In sewage treatment processes, sorption and biodegradation play a considerable role in the removal of some SMCs (Bester, 2005; Simonich et al., 2002; Yang and Metcalfe, 2006; Zeng et al., 2007). However, the reported
* Corresponding author. Tel.: þ82 51 510 3513; fax: þ82 51 582 3965. E-mail address:
[email protected] (J.-E. Oh). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.08.049
water research 44 (2010) 214–222
removal efficiencies varied from 50% to 90%, and were even below 50% in some cases, because the SMC concentrations can vary widely, depending on the size of the treatment plant, the size of the population served, the types of waste (domestic, industrial, and/or commercial) and the treatment methods employed (Heberer, 2002; Horii et al., 2007; Simonich et al., 2002). Furthermore, most studies which investigated the removal efficiency of SMCs in STP focused mainly on the conventional activated sludge (CAS) process. Therefore, the SMCs removal efficiency by sewage treatment processes other than CAS needs further research. To answer this and related questions, we attempted to confirm the removal efficiency of the CAS process and investigate the removal effects according to various treatment processes such as modified biological, chemical, filtration and disinfection treatment. In the case of surface water, indirect SMCs originating from point sources such as STP effluent and direct SMCs originating from non-point sources such as households as a result of the lack of sewage gathering systems have led to the widespread occurrence and relatively higher SMC concentrations around urban areas than rural areas (Bester, 2005; Dsikowitzky et al., } ttmann, 2008; Rimkus, 1999; Zhang 2002; Quednow and Pu et al., 2008). To trace their sources in surface water, research on the fate of SMCs is needed. Although several previous studies have attempted to identify the fate of SMCs, the main target compounds were HHCB and ATHN (Heberer, 2002; Dsikowitzky et al., 2002). Therefore, information on other musks such as musk ketone and musk xylene remains inadequate, which necessitates the collection of investigation data from a wide range of various sites to support research capable of providing suitable information and recommending modifications of sewage treatment processes and developments of computational modeling programs. Therefore, the present study objectives focus on (i) the occurrence (concentration) and distribution patterns of SMCs in 10 different STPs, and their removal effects by the various
215
treatment processes of these STPs, (ii) the occurrence of SMCs in the surface waters of urban and rural areas, and (iii) the identification of the fate of SMCs in various water environments. Four species of SMCs (HHCB, AHTN, MK and MX) are analyzed and this is the first survey to investigate the status of SMCs (e.g., usage pattern, occurrence and distribution pattern) in Korea.
2.
Materials and methods
2.1.
Chemicals and reagent
MK, MX, HHCB, AHTN and deuterated (d3) AHTN, with respective purities of 98%, 99%, 51%, 98.5% and 99%, were purchased from Dr. Ehrenstorfer GmbH, Augsburg, Germany. Deuterated (-d10) phenanthrene (100% purity) was obtained from AccuStandard Inc., CT, USA. Dichloromethane (DCM) and n-hexane were analytical grade, and methanol was HPLC grade. Anhydrous sodium sulfate was baked at 450 C prior to use.
2.2.
Study area and sample collection
Ten onsite STPs were selected for this study and their locations and description were presented in Fig. 1 and Table 1, respectively. Ten STPs are located in Busan metropolitan city which is the second largest city with five million inhabitants in Korea. The daily treatment volume and hydraulic retention time (HRT) of each STP varied from 7000 to 330,000 m3/day and from 11.8 to 21.3 h, respectively. Among ten STPs, six STPs received wholly residential sewage as influent while other four STPs treated mixed sewage from industry and households as influent, which are generated from Busan region. All STPs used the physical sorption and settling processes as the primary treatment, with either biological or chemical treatment process as the secondary treatment and additional filtration and/or disinfection treatment processes. Sewage
Fig. 1 – The description of STP locations and surface water sampling points.
216
water research 44 (2010) 214–222
Table 1 – Description of the 10 surveyed STPs. STP
Treatment (m3/day)
HRTa (h)
Source
Treatment processes and sampling pointse (n)
A B C D-1 D-2 E F G H-1 H-2
7000 30,000 80,000 230,000 210,000 330,000 10,000 55,000 200,000 150,000
18.1 17.3 12.0 14.1 15.1 11.8 12.0 12.0 15.8 21.3
100% RSb 100% RS 100% RS 100% RS 100% RS 100% RS 40% ISc þ 60% TWd 30% IS þ 70% TW 50%RS þ 50% TW 50% RS þ 50% TW
Influent (1) / SBR (2) / SF (3) / UV / Effluent / (4) Influent (1) / AS (2) / SF (3) / UV / Effluent / (4) Influent (1) / DE (2) / BF / Effluent / (3) Influent (1) / AS (2) / Cl / Effluent / (3) Influent (1) / AS (2) / UV / Effluent / (3) Influent (1) / AS (2) / Cl / Effluent / (3) Influent (1) / MLE (2) / SF / Effluent / (3) Influent (1) / MLE (2) / UV / Effluent / (3) Influent (1) / AS (2) / Cl / Effluent / (3) Influent (1) / A2O (2) / UV / Effluent / (3)
a HRT ¼ Hydraulic retention time. b RS ¼ Residential sewage. c IS ¼ Industrial sewage. d TW ¼ Treated wastewater from industry. e Acronyms and abbreviations of the treatment processes in this study: Secondary treatment process [Activated sludge, AS; Modified LudzackEttinger, MLE; Anaerobic, anoxic and oxide, A2O; Sequencing batch reactor, SBR; Densadeg, DE], Disinfection [Chlorination, Cl; Ultraviolet (UV) disinfection, UV], Filtration [Biofiltration, BF; Sand filtration, SF].
grab sampling was conducted at the outlet of each treatment process, including influent and effluent in January, 2008. The surface water samples were collected with using grab sampling method twice (October 14, 2007 and April 12, 2008) in the 4 rivers and 10 streams near rural and urban areas, to give a total of 28 samples. The sampling locations are described in Fig. 1. All of the water samples were collected in amber bottles and refrigerated until analyses after appending 0.5% methanol (v/v) to prevent rancidity.
2.3.
The electron impact (EI)-MS was operated in the selective ion monitoring mode and the ionization energy was 70 eV. The following ions were monitored: m/z 243, 258 and 213 for HHCB; 243 and 258 for AHTN; 282 and 297 for MX; 279 and 294 for MK; 246 and 261 for AHTN-d3. The transfer line and ion source temperatures were 300 C and 230 C, respectively. In quantifying the SMCs, the effect of proton exchange of AHTN-d3 was evaluated in procedural blank samples because AHTN-d3 undergoes partial D–H exchange during sample processing and storage (Bester, 2005; Bester, 2009).
Analytical procedures 2.4.
For liquid–liquid extraction, unfiltered water samples (300 mL for sewage influent sample; 500 mL for other sewage samples and surface water samples) were taken in a glass separatory funnel, and extracted with 100 mL of DCM and subsequent n-hexane after spiking the constant amount (200 ng) of internal standard (AHTN-d3) in each sample. After ten minutes of manual shaking and twenty minutes of holding time, the extract was passed through a glass funnel packed with anhydrous sodium sulfate for the removal of moisture and concentrated using a TurboVap II (Zymark, MA, USA) at a temperature of 40 C and a nitrogen-purge concentrator. The final volume of the extract was adjusted to 0.5 mL with DCM after spiking the recovery standard, phenanthrene-d10. Each concentration of the 4 SMCs was determined by gas chromatograph interfaced with a mass spectrometric detector (GC/MSD, Agilent 6890 GC and 5973 series MSD; Agilent Technologies, CA, USA). GC separation was carried out using a 30 m DB-5ms fused silica column (0.32 mm i.d., 0.25 mm film thickness; J&W Scientific, CA, USA). Injection (2 mL) was performed in the splitless mode at 280 C with a constant helium gas flow of 1.0 mL/min. The GC oven temperature was programmed to increase from 50 C (2 min) to 150 C at a rate of 10 C/min, and subsequently to 190 C at a rate of 2 C/min, followed by a third ramp to 300 C at a rate of 25 C/min, and held for 10 min.
Quality assurance and quality control
SMC analysis requires careful laboratory procedure to avoid possible contamination from laboratory personnel due to the widespread occurrence of SMCs in personal care products (hand creams, lotions and perfumes etc) (Kupper et al., 2004; Horii et al., 2007). Procedure blank samples using distilled water were analyzed with samples to check for blank contamination during the sample treatment. None of the target SMCs was detected and proton exchange of AHTN-d3 did not occur in the procedure blank samples. Limit of detection (LOD) and limit of quantification (LOQ) were set on a signal-to-noise ratio of 5 and 10, respectively. The LOD and LOQ were 5 and 10 ng/L (surface water samples) and 10 and 20 ng/L (sewage samples), respectively. Recovery experiments were performed by spiking a mixture of HHCB, AHTN, MX and MK at two concentrations (50, 500 ng/L, n ¼ 3) into tap water and passing through the same analytical procedure as water samples. The average recoveries for HHCB, AHTN, MX and MK were 87 5%, 86 6%, 88 4% and 86 4%, respectively. In water sample extraction, recoveries of AHTN-d3 as surrogate standard, which was spiked into the samples before extraction, were 93 17% in the sewage and surface water samples. To check instrumental stability, a quality control standard was analyzed after every ten samples were injected into the instrument.
water research 44 (2010) 214–222
3.
Results and discussion
3.1. Synthetic musk compounds (SMCs) in sewage treatment plant (STP) 3.1.1. The influent and effluent concentrations and distribution patterns HHCB, AHTN and MK were detected in all influent and effluent samples, while MX was only detected in 4 out of 10 STP influent samples. The total concentrations of the 4 SMCs ranged from 3.69 to 7.33 mg/L in the influents and from 0.96 to 2.69 mg/L in the effluents (Fig. 2). HHCB was the predominant compound in all influent and effluent samples, followed by AHTN, MK and MX. These results coincide with a greater consumption of HHCB (88.0 ton/year), compared with AHTN (below 0.1 ton/year), MK (20.3 ton/year) and MX (0.5 ton/year) in Korea (MoE, 2006). However, AHTN was shown high concentration in influent even though the consumption of AHTN was the lowest. This inconsistency seems to be resulted from the time difference between surveyed year of consumption and this study. Most of previous researches (Balk and Ford, 1999; Heberer, 2002; Horii et al., 2007) have also reported greater production and usage of HHCB than those of
217
other SMCs. The main target compounds of previous research were HHCB and AHTN due to their high usage volume. Therefore, the influent HHCB and AHTN concentrations of this study were compared with those of other countries to assess the present status of SMCs occurrence in Korea (Table 2). The HHCB and AHTN concentrations in this study were similar with those in Germany and Austria, lower than those in the U.S.A., U.K. and Netherlands and higher than those in Switzerland, Spain, Belgium, Canada and China. The occurrence of HHCB and AHTN do not show a specific geographical distribution because this mainly depends on their production and usage volume of each country. Hierarchical cluster analysis was performed with the total SMC concentrations in the influent to investigate their distribution patterns according to input sources and STP treatment volume, so the results could be categorized into two main groups (Fig. 2A): Group A-1 (7 samples) with relatively high SMC concentrations and Group A-2 (3 samples) with low SMC concentrations. Most of the influent samples in Group A1 (except H-1 and H-2) were 100% residential sewage while two of Group A-2 (except A) were industrial sewage, suggesting that household sewage was probably the main source of SMCs. In Table 1, the sewage treatment volume of A (7000 m3/ day) was the lowest among the 10 STPs, so we considered that
Fig. 2 – The SMCs concentrations in sewage (A) influent and (B) effluent samples [Insets: Group A-1 and -2 are categorized by their concentration levels].
218
water research 44 (2010) 214–222
Table 2 – Comparison of influent HHCB and AHTN concentrations (mg/L). Location Korea (10) Germany (3) Austria (3) Switzerland (1) Spain (1) Belgium (1) U.K. (3) and Netherlands (2) Canada (12) U.S.A (12) China (1)
HHCB
AHTN
Source
2.56–4.52 (3.48 0.59) 1.90–5.05 (3.26 1.62) 0.83–4.44 (2.62 1.48) 6.90 2.25 1.34 9.71 5.09 2.03 16.6 10.4 2.30
0.55–1.21 (0.77 0.18) 0.58–1.18 (0.81 0.32) 0.21–1.11 (0.76 0.40) 1.52 0.61 0.20 5.97 3.88 0.80 12.5 7.35 0.72
This study Bester (2005) Clara et al. (2005) Berset et al. (2004) Mitjans and Ventura (2004) Mitjans and Ventura (2004) Simonich et al. (2002) Lishman et al. (2006) Simonich et al. (2002) Zhang et al. (2008)
this low sewage burden caused the relatively low SMC concentrations in A compared to other STPs. H-1 and H-2 received pre-treated wastewater (50% of the influent) from industrial complex and their contribution of residential sewage was also half of the total influent volume (approximately 100,000 and 75,000 m3/day, respectively). Therefore, the high SMC concentrations in H-1 and H-2 influents were attributed to their industrial/residential mix of sewage sources.
3.1.2.
The removal by sewage treatment processes
The removals of the 4 SMCs by various sewage treatment processes were investigated and their removals in each treatment process are shown in Fig. 3. The MX removal could not be obtained in each treatment process because the MX concentration was too low and was detected only in 4 influent samples. Four different biological treatment processes (Activated sludge, AS; Modified Ludzack-Ettinger, MLE; Sequencing batch reactor, SBR; Anaerobic, anoxic and oxide, A2O) were investigated in this study. MLE, SBR and A2O processes are modified versions of the conventional AS process to enhance biological removal efficiency for nutrients (i.e., nitrogen and phosphorus) and decrease solid retention time. Five of the 10 STPs in this study used the AS process as the secondary treatment. The average removals of HHCB, AHTN and MK in the conventional AS process were 53 6%, 56 5% and 53 12%, respectively, which were similar with other studies. The SMCs removal efficiency in the conventional AS process has been reported to range from 50% (in some studies, 70% degradation of the toxin. This exposure time (45 min) is equivalent to a UV dose of 250 mJ/cm2 using the LP UV lamp. Fig. 5 shows the first-order degradation of anatoxin-a in pure water (DI), natural water (EP) and synthetic model water (SY) using the LP UV lamp and the addition of 30 mg/L H2O2. As it can be seen, anatoxin-a was degraded faster in pure water showing that background organic compounds can decrease the efficiency using the UV/H2O2 process. This can be explained by competition between organic matter in the water and anatoxin-a for reaction with OH radicals. The same result was observed using the VUV lamp, indicating that the competition occurs between constituents in the water and OH radicals regardless of the source of the OH radical production.
3.6.
Conclusions
The degradation of anatoxin-a in pure, natural and synthetically produced model waters was investigated using two Advanced Oxidation Processes, VUV photolysis at 172 nm, and combination of UV and H2O2. The second-order rate constant for the reaction between anatoxin-a and the hydroxyl radical was calculated using VUV lamp as a source of OH radical formation and the rate found to be (5.2 0.3) 109 M1 s1 and was independent of pH, temperature, and initial concentration of anatoxin-a. The direct photolysis of anatoxin-a using a medium pressure (MP) UV lamp was also investigated, showing that a UV dose of 1285 mJ/cm2 was required to degrade anatoxin-a by 88% and 50% at concentrations of 0.6 mg/L and 1.8 mg/L of toxin, respectively. Treatment of anatoxin-a with a low pressure (LP) UV lamp in the presence of 30 mg/L of H2O2 was also studied, where it was found that more than 70% of toxin could be degraded at a UV dose of 200 mJ/cm2.
Effect of H2O2 concentration on the AOP treatment
The effect of the H2O2 concentration on the efficiency of degradation of anatoxin-a in ultrapure water was investigated by adding 30, 40 and 60 mg/L of H2O2 to the anatoxin-a solution at the same applied UV dose. Fig. S-4 illustrates the effect of the
Acknowledgments This study was carried out with the financial support provided by the Natural Sciences and Engineering Research Council of
water research 44 (2010) 278–286
Canada (NSERC) and the Alberta Ingenuity Center for Water Research (AICWR). The authors would like to thank Dr. Keisuke Ikehata for his generous help and support during this project. Thanks also go to technicians of the laboratories in Department of Civil and Environmental Engineering, University of Alberta for their technical support. Dr. T. Oppenla¨nder thanks the German Academic Exchange Service (DAAD) for a short term lectureship at the University of Alberta from March 2007 till August 2007. He also is grateful to Drs. J.R. Bolton and M. Gamal El-Din for their financial contribution to the DAAD scholarship.
Appendix. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2009.09.021.
references
Bolton, J.R., Linden, K.G., 2003. Standardization of methods for fluence (UV dose) determination in bench-scale UV experiments. Journal of Environmental Engineering, ASCE 129 (3), 209–215. Bolton, J.R., Stefan, M.I., 2002. Fundamental photochemical approach to the concepts of fluence (UV dose) and electrical energy efficiency in photochemical degradation reactions. Research on Chemical Intermediates 28 (7–9), 857–870. Cabelli, D.E., Bielski, B.H.J., 1985. A pulse-radiolysis study of some dicarboxylic-acids of the citric-acid cycle – the kinetics and spectral properties of the free-radicals formed by reaction with the OH radical. Zeitschrift fu¨r Naturforschung, Section B – A Journal of Chemical Sciences 40 (12), 1731–1737. Carmichael, W.W., 1992. Cyanobacteria secondary metabolites – the cyanotoxins. Journal of Applied Bacteriology 72 (6), 445–459. Falconer, I.R., Humpage, A.R., 2006. Cyanobacterial (blue-green algal) toxins in water supplies: cylindrospermopsins. Environmental Toxicology 21 (4), 299–304. Feng, P.Y., Brynjolf, A., Halliday, J.W., Jarrett, R.D., 1970. Highintensity radiolysis of aqueous ferrous sulfate–cupric sulfate– sulfuric acid solutions. Journal of Physical Chemistry 74 (6), 1221–1227. Hall, T., Hart, J., Croll, B., Gregory, R., 2000. Laboratory-scale investigations of algal toxin removal by water treatment. Journal of the Chartered Institution of Water and Environmental Management 14 (2), 143–149. Hayon, E., Simic, M., 1973. Acid–base properties of organic peroxy radicals, OORH, in aqueous-solution. Journal of the American Chemical Society 95 (20), 6681–6684. Heit, G., Neuner, A., Saugy, P.Y., Braun, A.M., 1998. Vacuum-UV (172 nm) actinometry. The quantum yield of the photolysis of water. Journal of Physical Chemistry 102 (28), 5551–5561. Hitzfeld, B.C., Hoger, S.J., Dietrich, D.R., 2000. Cyanobacterial toxins: removal during drinking water treatment, and human risk assessment. Environmental Health Perspectives 108, 113–122. Huber, M.M., Canonica, S., Park, G.Y., Von Gunten, U., 2003. Oxidation of pharmaceuticals during ozonation and advanced oxidation processes. Environmental Science and Technology 37 (5), 1016–1024. James, K.J., Furey, A., Sherlock, I.R., Stack, M.A., Twohig, M., Caudwell, F.B., Skulberg, O.M., 1998. Sensitive determination of anatoxin-a, homoanatoxin-a and their degradation products by liquid chromatography with
285
fluorimetric detection. Journal of Chromatography A 798 (1–2), 147–157. Kotak, B.G., Kenefick, S.L., Fritz, D.L., Rousseaux, C.G., Prepas, E.E., Hrudey, S.E., 1993. Occurrence and toxicological evaluation of cyanobacterial toxins in Alberta lakes and farm dugouts. Water Research 27 (3), 495–506. Liu, W., Andrews, S.A., Bolton, J.R., Linden, K.G., Sharpless, C., Stefan, M.I., 2002. Comparison of disinfection byproduct (DBP) formation from different UV technologies at bench scale. Water Science and Technology, Water Supply 2 (5/6), 1753–1769. Newcombe, G., Nicholson, B., 2004. Water treatment options for dissolved cyanotoxins. Journal of Water Supply: Research and Technology 53 (4), 227–239. Onstad, G.D., Strauch, S., Meriluoto, J., Codd, G.A., von Gunten, U., 2007. Selective oxidation of key functional groups in cyanotoxins during drinking water ozonation. Environmental Science and Technology 41 (12), 4397–4404. Onstein, P., Stefan, M.I., Bolton, J.R., 1999. Competition kinetics method for the determination of rate constants for the reaction of hydroxyl radicals with organic pollutants using the UV/H2O2 advanced oxidation technology: the rate constants for the tert-butyl formate ester and 2,4-dinitrophenol. Journal of Advanced Oxidation Technology 4, 231–236. Oppenla¨nder, T., Schwarzwalder, R., 2002. Vacuum-UV oxidation (H2O-VUV) with a xenon excimer flow-trough lamp at 172 nm: use of methanol as actinometer for VUV intensity measurement and as reference compound for OH-radical competition kinetics in aqueous systems. Journal of Advanced Oxidation Technology 5 (2), 155–163. Oppenla¨nder, T., Walddorfer, C., Burgbacher, J., Kiermeier, M., Lachner, K., Weinschrott, H., 2005. Improved vacuum-UV (VUV)-initiated photomineralization of organic compounds in water with a xenon excimer flow-through photoreactor (Xe2(*) lamp, 172 nm) containing an axially centered ceramic oxygenator. Chemosphere 60 (3), 302–309. Robertson, P.K.J., Lawton, L.A., Mu¨nch, B., Cornish, B., 1999. The destruction of cyanobacterial toxins by titanium dioxide photocatalysis. Journal of Advanced Oxidation Technology 4 (1), 20–26. Rodriguez, E., Sordo, A., Metcalf, J.S., Acero, J.L., 2007a. Kinetics of the oxidation of cylindrospermopsin and anatoxin-a with chlorine, monochloramine and permanganate. Water Research 41 (9), 2048–2056. Rodriguez, E., Onstad, G.D., Kull, T.P.J., Metcalf, J.S., Acero, J.L., von Gunten, U., 2007b. Oxidative elimination of cyanotoxins: comparison of ozone, chlorine, chlorine dioxide and permanganate. Water Research 41 (15), 3381–3393. Rositano, J., Nicholson, B.C., Pieronne, P., 1998. Destruction of cyanobacterial toxins by ozone. Ozone Science and Engineering 20 (3), 223–238. Rositano, J., Newcombe, G., Nicholson, B., Sztajnbok, P., 2001. Ozonation of NOM and algal toxins in four treated waters. Water Research 35 (1), 23–32. Song, W., Xu, T., Cooper, W.J., Dionysiou, D.D., De La Cruz, A.A., O’Shea, K.E., 2009. Radiolysis studies on the destruction of microcystin-LR in aqueous solution by hydroxyl radicals. Environmental Science and Technology 43, 1487–1492. Stefan, M.I., Bolton, J.R., 2005. Fundamental approach to the fluence-based kinetic and electrical energy efficiency parameters in photochemical degradation reactions: polychromatic light. Journal of Environmental Engineering and Science 4, S13–S18. Svrcek, C., Smith, D.W., 2004. Cyanobacteria toxins and the current state of knowledge on water treatment options: a review. Journal of Environmental Engineering and Science 3 (3), 155–185.
286
water research 44 (2010) 278–286
Wang, D., 2008. Comparison of Disinfection Effects of UV Light at 172, 222, 254 nm on Bacillus subtilis Spores, in Department of Civil & Environmental Engineering. M.Sc. thesis. University of Alberta, Edmonton, AB, Canada, p. 167.
Wood, S.A., Rasmussen, J.P., Holland, P.T., Campbell, R., Crowe, A. L.M., 2007. First report of the cyanotoxin anatoxin-a from Aphanizomenon issatschenkoi (cyanobacteria). Journal of Phycology 43 (2), 356–365.
water research 44 (2010) 287–297
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
When public opposition defeats alternative water projects – The case of Toowoomba Australia Anna Hurlimann a,*, Sara Dolnicar b,1 a b
Faculty of Architecture Building and Planning, The University of Melbourne, Victoria 3010, Australia School of Management & Marketing, University of Wollongong, Northfields Ave, 2522 Wollongong, NSW, Australia
article info
abstract
Article history:
Located approximately 100 km west of Brisbane, Toowoomba is home to approximately
Received 30 January 2009
95,000 people. Surface water from dams is the main source of water for the city. In 2006 the
Received in revised form
residents of Toowoomba were invited to vote in a referendum (plebiscite) concerning
10 August 2009
whether or not an indirect potable wastewater reuse scheme should be constructed to
Accepted 8 September 2009
supply additional water to the area. At that stage dam levels in Toowoomba were at
Published online 10 September 2009
approximately twenty percent of capacity. Toowoomba residents, after intense campaigning on both sides of the referendum debate, voted against the proposal. In July 2008
Keywords:
dam levels dropped to eleven percent. Stage 5 water restrictions have been in place since
Water recycling
September 2006, subsequently mains water must not be used for any outdoor uses. This
Participation
paper describes in detail how public opposition in the case of Toowoomba’s referendum,
Public acceptance
defeated the proposal for a water augmentation solution. Reasons for the failure are
Public opposition
analysed. In so doing, the paper provides valuable insights with respect to public partici-
Toowoomba
pation in indirect potable reuse proposals, and discusses factors including politics, vested
Referendum
interest and information manipulation. This paper is significant because of the lack of
CADS
detailed information published about failed water infrastructure projects. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Australia is in the midst of a water crisis. The water supplies of many of the country’s major urban centres are dwindling. When compared to capital cities, the water situation is often much more critical in regional areas such as Toowoomba. Although many solutions to the water crisis have been proposed, national policy in Australia has predominantly focused on supply side solutions such as water recycling and desalination (Hurlimann, 2006). However, in addition to these sources, a range of other alternative water sources and management options are available including the use of, grey water (domestic wastewater excluding toilet waste), stormwater, and water conservation – a demand side strategy.
In Australia, the use of recycled water for drinking purposes is subject to numerous guidelines including those at a National Level (Natural Resource Management Ministerial Council et al., 2008). However, the viability of alternative water sources also depends on public attitudes. Several recycled water projects in various countries have failed due to lack of community support (Hurlimann and McKay, 2004). These projects include indirect potable reuse schemes in the USA and Australia, and also non-potable reuse projects including one in the Netherlands. Elements contributing to the demise of these projects involved the public’s lack of trust in the institutions charged with delivering the projects (Hurlimann and McKay, 2004). As described by Hurlimann and McKay (2004) anecdotal
* Corresponding author. Tel.: þ61 3 8344 6976; fax: þ61 3 8344 5532. E-mail addresses:
[email protected] (A. Hurlimann),
[email protected] (S. Dolnicar). 1 Tel.: þ61 2 4221 3862. 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.020
288
water research 44 (2010) 287–297
evidence from such projects suggests that factors including timely communication with stakeholders, transparency in the projects’ process and fairness in the way in which it is implemented are critical. In a similar vein Dishman et al. (1989, p. 158) conclude that technical aspects of potable water reuse can be resolved, but ‘‘the issue of public acceptance could kill the proposal’’. Additionally, Postel (1997) highlights a major barrier to reuse of wastewater is psychological not technical. In order to reduce the risk of potential failure of alternative water projects, it is essential to understand the context of such cases. Unfortunately cases where public resistance prevented water augmentation schemes are not well documented. Thus other locations planning the introduction of alternative water sources cannot easily learn from these experiences. Understanding how to facilitate public participation in decision making, and the role that public interest groups have is also important. Public interest groups include those opposed to desalination, such as ‘Sydney community united against desalination, (SCUD), and those opposed to the concept of drinking recycled water such as ‘Citizens against drinking sewage’ (CADS). CADS were present in Toowoomba before the referendum, but this was not the first project the group were opposed to. CADS were first present in an earlier Queensland indirect potable reuse proposal for the area of Maroochy. This plan was driven by community concern for environmental impacts of ocean outfall of sewage (Simpson, 1999). The project was in the final stages of public consultation when CADS campaigned against the project, fearing the effect of the possible presence of ‘gender-bending’ hormones in the water (Stenekes et al., 2001). While the local government (the Council) voted in favour of the proposal, the plans for potable reuse were later abandoned. Stenekes et al. (2001) believe that the Maroochy case was complicated by CADS perceiving a lack of adequate consideration for stakeholders in the consultation process, and feeling that the process was not transparent. CADS believe the Council voted to implement the potable reuse strategy despite evidence that sections of the community would not support potable reuse (Stenekes et al., 2001). The aim of this paper is to fill this gap in understanding of failed potable recycled water projects through three research objectives: (1) to provide a detailed description of one case where public resistance has led to the abandonment of a project aiming to augment water supply through indirect potable reuse (the case of Toowoomba, Australia), (2) to identify factors leading to the Toowoomba community’s opposition to the indirect potable reuse proposal, and (3) assess Toowoomba community attitudes to recycled water two years after the referendum (which was critical to our interpretation of all the data gathered for this research). The paper is structured as follows. In Section 2 we outline our research method. In Section 3, we present Toowoomba’s water history in chronological order. This section contains developments which took place in 2005 and 2006. In Section 4 we present and discuss the situation in Toowoomba two years after the referendum. Finally, in Section 5 we provide overall conclusions which integrate the results from each of the methods employed.
2.
Method
Toowoomba was used as a case study of attempted introduction of indirect potable reuse. As advocated by Eisenhardt (1989) our case study method combined various data collection modes such as archival research, interviews, focus groups, observations and survey. These divergent data collection methods allowed the collection of information about the events that took place in Toowoomba surrounding the referendum. The research consists of three main components: 1) the analysis of a. topical Internet blog sites, and b. information brochures developed by various organisations and which were publicly available, 2) qualitative empirical research, consisting of a focus group and eight in-depth interviews with residents of Toowoomba in July 2008, and 3) quantitative empirical research conducted in January 2009 with 200 Toowoomba residents. The purpose of the qualitative component of the research was to gain an in-depth insight into the current sentiments of the population with regard to alternative water sources and the drought in general. Respondents were recruited by a professional market research company who administered compensation payments. The focus group and interviews were conducted by one of the authors. On average the interviews lasted 45 min. The focus group session was one and a half hours in duration and consisted of ten participants. Responses were entered into a data set and were then coded and categorized by the second author. Krueger and Casey (2000) and Richards (2005) were consulted when analysing the qualitative data. Responses obtained in the qualitative phase informed the question design of the quantitative survey. Data in this latter phase was collected using an Australian permission based Internet panel which recruits respondents through a range of avenues (not only the Internet) to ensure sample representativity. Respondents were paid a small monetary compensation for taking the time to complete the questionnaire. The interviews, focus groups and survey addressed a range of issues and explored various water behaviours including: drinking recycled water and desalinated water, conserving water, talking to others about water issues, purchasing water related products, and joining a water interest group. We used a number of theories to guide our analysis of the topical Internet blog sites and information brochures developed by various organisations, and our synthesis of the three types of data collected. These theories included: information theory (McCornack et al., 1992); the first mover advantage theory (Lieberman and Montgomery, 1988; Robinson and Fornell, 1985; Carpenter and Nakamoto, 1989), and theory regarding referendums and democracy (Heywood, 1999; Smith, 2001). These theories are discussed in detail during our presentation of results.
3.
The recycled water history in Toowoomba
Located approximately 100 km west of Brisbane (the capital city of the state of Queensland), Toowoomba has a population of approximately 95,000 people. Toowoomba is known as ‘Queensland’s Garden City’ (Toowoomba City Council, 2007),
water research 44 (2010) 287–297
hosting an annual ‘Carnival of Flowers’ each spring. In addition to this there are often Camellia and Winter Flower Shows. The city has a famous Park ‘Queens Park’ which is well known for its gardens and flowers (Toowoomba City Council, 2001).
3.1.
Water shortage in Toowoomba
Toowoomba’s water comes from three major storage areas (Lake Cooby, Lake Perserverance and Lake Cressbrook). The supply in these three storage areas has been depleting due to declining rainfall over the catchment areas (Parsons Brinckerhoff Australia Pty Ltd, 2006). Toowoomba’s population is increasing and so is industrial development (Toowoomba City Council, 2005b). In 2005, the average residential water use in Toowoomba was 240 l per person per day, compared to 300 l in South East Queensland (Toowoomba City Council, 2005b). However, since water use restrictions have been in place, per capita water use in Toowoomba and other areas of South East Queensland has decreased. In Toowoomba per capita residential consumption was 151 l/day in January 2009, however it was 123 l/day during the same period in 2008 (Toowoomba Regional Council and Toowoomba Water, 2009). The total water demand in Toowoomba in 2006 was estimated to amount to 17,510 Ml/annum, thus exceeding supply (Parsons Brinckerhoff Australia Pty Ltd, 2006). Because of the critical water situation, Toowoomba residents have been faced with restrictions to water use since 2003. Level 1 restrictions began in 2003, ultimately reaching level 5 restrictions in 2006, which remain today. Restrictions to water use typically involve banning outdoor use of water (for gardens) at certain times of the day, and become increasingly restrictive the higher the level. For example in Toowoomba, Stage 5 water use restrictions prohibit town water use for watering of gardens, topping-up of pools, and washing of vehicles (for further information see: Toowoomba City Council, 2008). Implications of restriction levels vary across water authorities throughout Australia, thus there is not a consistent state or national approach to restrictions. In the financial year 2005/2006 the Toowoomba Council committed AUD850,000 (at 22/06/09 AUD1 ¼ US$0.80 and V0.58) to a Water Demand Management Initiative, as part of this initiative residents were offered rebates for installing rainwater tanks (AUD500), AAAA rated (highly efficient) washing machines (AUD50), and could have their shower heads replaced at no cost. Since 2005 all new developments have to install rainwater tanks (Toowoomba City Council, 2005b).
3.2.
289
plant to provide potable quality recycled water for the town (Toowoomba City Council, 2005b). This was principally a policy document, not a public communication document. However, as part of the proposal, Toowoomba City Council was planning to undertake a three year community engagement program (Thorley, 2007). The Water Futures Initiative was launched by the Federal Member for Groom (including Toowoomba), the Honourable Ian MacFarlane, the then Queensland Premier, the Honourable Peter Beattie, and all three local Members of State Parliament (Toowoomba City Council, 2005a). The Council expected funding to be approved in September or October 2005 (Thorley, 2007).
3.3.
Public opposition to the recycled water proposal
In reaction to the Water Futures Initiative, the CADS Toowoomba group formed on the 21st of July 2005 and held their first public meeting on the 25th of August 2005 (Toowoomba Water Futures Blog, 2006). Half a year later, on the 24th of February 2006, 10,000 people had signed a CADS petition against the potable recycled water initiative (Reynolds, 2006). This public movement against the indirect potable recycled water use politicised the project. Thorley, as mayor of Toowoomba at the time, identified that this moved the focus to be no longer on water but on politics and vested interest, leading to political back-flips and the withdrawal of support of the project by Macfarlane, three Councillors and the local National Party State member (Thorley, 2007). Given that the original Water Futures Initiative proposal was not a specific recycled water communication program, CADS were in fact the first to communicate their view of the recycled water plans, and provide detailed arguments in support of their view to the public. In so doing CADS benefited from a ‘First Mover Advantage’, which is ‘‘the ability of pioneering firms to earn positive economic profits’’ (Lieberman and Montgomery, 1988). In the case of CADS it was not positive economic profits that they earned. Instead, being the first to communicate with the public, they became the benchmark information source for matters relating to the proposed recycling project. This gave CADS significant market power and made it more and more difficult over time, for any positive message about recycled water to be communicated successfully to the residents of Toowoomba. Such consumer information advantages have been achieved through the learning process of consumers are in line with the findings reported by Robinson and Fornell (1985) and Carpenter and Nakamoto (1989).
The recycled water proposal 3.4.
The Toowoomba Council lodged a submission to the National Water Commission for funding towards the project on 30 June 2005. The submission was unanimously supported by all 9 Councillors (elected representatives at local government level), and by all local members of State and Commonwealth Parliaments (Thorley, 2007). On the 1st of July 2005, Toowoomba City Council announced the ‘Water Futures Initiative’. The initiative was launched to address the city’s water challenges. The project includes a range of solutions, most prominently the construction of an advanced water treatment
Announcing the referendum
On the 24th of March 2006, Mr Malcolm Turnbull (Parliamentary Secretary to the Prime Minister) announced that a referendum would be held asking the residents of Toowoomba whether or not they were supportive of the Water Futures Project. In case of a positive vote, the Federal Government was promising to contribute AUD22.9 million towards the project (Mitchell, 2006). Mr Turnbull’s motivation for calling a referendum is unclear, especially given that (1) the National Water Commission had recommended to the Prime Minister that the
290
water research 44 (2010) 287–297
project be approved, and (2) Commonwealth funding for a simular project in Goulburn NSW was approved without a referendum subject to a six month consultation with the public, and (3) the Queensland government had to make a special regulation to allow the vote to proceed (Thorley, 2007). Thorley (2007) views the Commonwealth Government’s decision to approve funding for the project subject to a referendum as a dangerous precedent, stating that ‘‘The decision was an abrogation of political leadership and usurped the democratically elected Council’s mandate for making decisions relating to its community’’ (p. 50). It is possible that Mr Turnbull’s decision was motivated by the increasing public opposition developing in Toowoomba. Toowoomba City Council was not pleased with the referendum. In fact, they had actively campaigned to Mr Turnbull against the referendum, pointing to poor records of referendums without bipartisan political support, and cognisant of the fear campaigns that tend to dominate political debate (Thorley, 2007). These arguments are partially supported by theory on democracy and referendums. According to Heywood (1999) models of democracy range from the classical idea of direct democracy in which people literally govern themselves, through to more modern forms of representative democracy where professional politicians govern on behalf of people. Referendums are a form of direct democracy, which are used widely in some countries such as Switzerland (Heywood, 1999). The way in which referendums are implemented, and the influence they have on decision making, varies from jurisdiction to jurisdiction (Ashworth, 2001). As discussed by Smith (2001) there are positive and negative aspects of all methods of deliberation. Those in favour of referendums believe that they have the capacity to widen the political agenda; are more likely to overturn established pro-business policy than normal parliamentary proceedings; and are a mechanism by which groups within civil society can challenge the government to defend status-quo (Smith, 2001). Common arguments against referendums include the belief that ordinary people lack the time, maturity and specialist knowledge to rule wisely on their behalf (Heywood, 1999). However, on the contrary most studies suggest that voters exercise shrewd judgement despite the complexity of measures and the deceptions of some campaigns (Heywood, 1999). Additionally in opposition to referendums, it has been highlighted that consulting the general public on each and every issue could paralyse decision making and make a country ungovernable (Heywood, 1999). Importantly, as highlighted by Heywood (1999), referendums suffer the effects of material and social inequalities. These such issues include but are not limited to 1) uneven participation in referendums by minority groups, 2) a growing influence of money, paid petition circulators, direct mail deception and deceptive advertising campaigns, and 3) media manipulation – particularly when business interests are threatened. Many of these problems identified by Heywood were present in the Toowoomba referendum.
3.5. Council’s attempt to rescue the Water Futures Initiative When the referendum became unavoidable, Toowoomba City Council started 10-week information campaign. On the 20th of
March 2006, they distributed a Water Futures booklet which contained explanations about the water cycle, the current level of water supply as well as possible water alternatives (Donaghey, 2006). This put Toowoomba City Council in the situation of (1) having to condense a proposed three year community engagement program – consisting of public fora, flyers, taste testings of recycled water and on-request public presentations (Toowoomba City Council, 2006a, c) – into a three month local political campaign (Thorley, 2007), and (2) face the substantial first mover advantage of CADS. By the time Council started informing the public, CADS had been communicating with Toowoomba residents for more than half a year. The main proponents of the Water Futures Project were Toowoomba Council, the Mayor of Toowoomba at the time (Ms Dianne Thorley), Mr Malcolm Turnbull, as well as State and Federal Governments. Examples of the ‘yes’ campaign material are referenced in Table 1. These were predominantly produced by the Council and were factual. Personal testimonies by upstanding members of the community were used to promote the scheme. It should be noted that, as opposed to CADS, Council were bound by Codes of Conduct, and thus had to ensure that campaign content was at all time ‘above board’ (Thorley, 2007). In response to the CADS campaign arguments, the Council presented the following messages: 1) Communities around the world use recycled water for drinking. Examples were given including Orange County and Virginia in the USA since the 1970s, Singapore since 2003 and Namibia since 1968 (multiple campaign brochures including the prominent: Toowoomba City Council, 2006b). 2) The reputation of the Toowoomba food industry will not be at risk: Water used in food processing is required to meet Australian Drinking Water Guidelines. The six star recycled water treatment far exceeds these guidelines (multiple campaign brochures including the prominent: Toowoomba City Council, 2006b). 3) Recycled water is safe and will produce water as safe as current drinking water because of the ‘Advanced Water Treatment Plant Purification Process’. Academics and General Practitioners (doctors) were quoted about safety in multiple campaign brochures (including items listed in Table 1). Diagrams of the ‘seven barriers of water futures – Toowoomba’ were provided in multiple Council brochures. It should be noted that when the Australian national recycled water guidelines were first drafted (Natural Resource Management Ministerial Council et al., 2006) they did not include indirect potable reuse as a possible option, this has since been addressed in phase two of the guidelines (Natural Resource Management Ministerial Council et al., 2008).
3.6.
More public opposition
While Council commenced its campaign, CADS continued to use public meetings, petitions and Internet blogs to activate residents to vote ‘‘no’’ at the referendum (O’Malley, 2006). The key opponents of the Water Futures Project who were rallying for a ‘‘no’’ vote were CADS (led by Rosemary Morely, a past
291
water research 44 (2010) 287–297
Table 1 – Select pictorial messages from both sides of the Toowoomba potable recycled water referenduma. No.
Date
No campaign 1 21/12/2005 2
28/11/2005
3
30/4/2006
4
19/4/2006
5
7/5/2006
6
27/5/2006
7
28/5/2006
8 9
16/7/2006 31/7/2006
10
28/1/2007
11
02/2007
Yes campaign 1 03/2006
Title
Organisation
The Downstream Boys
Water Futures Blog
I don’t know what is going through Council Vote no
Water Futures Blog
Will the guinea pigs drink? Straight from sewage plant for you to drink Save your children now Trick or turd
BlogToowoomba
Clive says ‘NO’ Truth told in pictures to the people I don’t want to die mummy Think before you agree to drink
Web address http://waterfutures.blogspot.com/2005/12/ downstream-boys.html http://waterfutures.blogspot.com/2005/11/ humour-this-cartoon-has-appeared-in.html#links
BlogToowoomba
http://www.blogtoowoomba.com/entry.php? w¼toowoombawatervote&e_id¼97 http://www.blogtoowoomba.com/entry.php? w¼toowoombawatervote&e_id¼69 http://www.blogtoowoomba.com/entry.php? w¼toowoombawatervote&;e_id¼103
BlogToowoomba
BlogToowoomba
http://www.blogtoowoomba.com/entry.php? w¼toowoombawatervote&e_id¼110 http://www.blogtoowoomba.com/entry.php? w¼toowoombawatervote&;e_id¼113 http://www.valscan.com.au/webpaper.pdf http://www.blogtoowoomba.com/entry.php? w¼toowoombawatervote&e_id¼235
BlogToowoomba Water Futures Blog BlogToowoomba
BlogToowoomba Privately produced document
The Water Booklet
Water Futures Toowoomba Water Futures Toowoomba
2
04/2006
3
04/2006
4
04/2006
5
04/2006
6
04/2006
7
04/2006
What does recycled water mean for me? What are our water options? How safe is recycled water? How will our water be recycled? No new dam – how Toowoomba’s water recycling will work It’s a dam good thing!
8
04/2006
Water futures facts
Water Futures Toowoomba
9
04/2006
Councillors statements
Water Futures Toowoomba
10
25/07/2006
Water Futures Toowoomba
11
03/2006
The Chronicle (newspaper) advertisements Other materials
Water Futures Toowoomba Water Futures Toowoomba Water Futures Toowoomba Water Futures Toowoomba
Water Futures Toowoomba
Water Futures Toowoomba
http://www.blogtoowoomba.com/entry.php? w¼toowoombawatervote&e_id¼565 http://www.valscan.com.au/tbyatdBris.pdf
http://www.toowoombawater.com.au/dmdocuments/ TCC-WaterFuturesLORES.pdf http://www.toowoombawater.com.au/index.php? option¼com_docman&;task¼cat_view&gid¼58&Itemid¼23 http://www.toowoombawater.com.au/index.php? option¼com_docman&;tas¼cat_view&gid¼58&Itemid¼23 http://www.toowoombawater.com.au/index.php? option¼com_docman&;task¼cat_view&gid¼58&Itemid¼23 http://www.toowoombawater.com.au/index.php? option¼com_docman&;task¼cat_view&gid¼58&Itemid¼23 http://www.toowoombawater.com.au/index.php? option¼com_docman&;task¼doc_view&gid¼174&Itemid¼23
http://www.toowoombawater.com.au/dmdocuments/ StreetsAheadInsP1Page1.pdf http://www.toowoombawater.com.au/index.php? option¼com_content&;task¼view&id¼210&Itemid¼20 http://www.toowoombawater.com.au/index.php? option¼com_docman&;task¼cat_view&gid¼61&Itemid¼23 http://www.toowoombawater.com.au/index.php? option¼com_docman&;task¼cat_view&gid¼0&Itemid¼23 http://www.toowoombawater.com.au/index.php? option¼com_docman&task¼cat_view&gid¼57&Itemid¼23
a All websites were viewed and verified 20 January 2009.
president of the Chamber of Commerce), Clive Berghofer (a millionaire property developer and former local mayor) as well as members of the public who posted their concerns in Internet blogs (of which there were more than three). One blog (waterfutures.blogspot.com) claimed to be impartial, yet the majority of contributions were arguing against the recycled water scheme. Some water experts from industry and University contributed to the blogs.
Examples of initiatives from the ‘no’ campaign include a newspaper printed by Clive Berghofer called ‘‘Water Poll’’ which was dedicated solely to arguing against the recycled water scheme (Berghofer, 2006). Table 1 provides more extensive references to pictorial material produced by the ‘no’ campaign. As can be seen from this material, much of it was driven by emotions, and at discrediting sources of factual information. In addition to pictorial material, there was
292
water research 44 (2010) 287–297
reading material and videos produced by each side of the campaign. The main reasons against the recycled water scheme stated by the opponents were as follows: (1) People were concerned about the image of Toowoomba. They were worried that their image as Garden City would change to an image of being the ‘‘Shit City’’ or ‘‘Poowoomba’’ (Balderson, 2006). (2) As a consequence of such an image residents were concerned that Toowoomba would become less attractive to businesses, industry, families, retirees and travellers both as a tourism destination and as a place to live (Concerned Ratepayer, 2006; Frew, 2005). One illustrative case was that of an ice cream factory in Toowoomba which claimed that it could never use Toowoomba’s town water for production because the market would not tolerate any question mark over the water quality (SBS Network, 2005). The same was claimed to be true for all businesses in the food industry (Clark, 2006). (3) Residents had health concerns. They were not sure if they could trust science; they were irritated that the Toowoomba Council refused to state that the water was 100% safe and stated that they felt like ‘‘lab rats’’ (Berghofer, 2006). Furthermore they were concerned that there were no official guidelines for the quality of recycled drinking water and that a twenty-five percent component of recycled water in tap water is very high by international standards (Concerned Ratepayer, 2006). Laurie Jones, an Australian plumber interviewed on television (SBS Network, 2005), summarized these fears: ‘‘Well, the problem with the purifying, and my biggest concern, is that the impact of drinking treated sewage wastewater will have on my family and all other families. And I’m concerned because there is no guarantee, there is absolutely no evidence that the treated sewage wastewater is free of all contaminants. And along those lines, in Australia, there’s no health department that approves it presently.’’
3.7.
Toowoomba votes
On the 29th of July 2006 the referendum was held in Toowoomba. The majority, 62% of residents, voted against the proposed recycled water scheme. As a consequence the Water Futures Project was abandoned (Australian Associated Press, 2006). The Internet blog sites have continued, in light of a new indirect potable recycled water proposal for Brisbane with implications for Toowoomba – The Western Corridor Recycled Water Scheme (described in Section 3.3). CADS have reproduced campaign material for Brisbane households (Water Futures Blog, 2007). As reported by the Science Media Centre (2006), one water engineer from Toowoomba City Council said he was frustrated, angry and disappointed. He was especially frustrated that the debate was ‘‘.not based on science. It was not a debate about water, but about politics and vested interests’’. Another water expert was quoted as saying: ‘‘The No in Toowoomba is ultimately a failure in communication, first on the safety
and reliability and second on the urgency of Australian water crisis’’ (Science Media Centre, 2006). The conclusion Thorley (2007), as the Mayor of Toowoomba at the time of the referendum, draws from the events, is that the way forward for indirect potable reuse is for governments to forget referendums, plebiscites and polls which will always be at the mercy of negative campaigns and are thus likely to fail. Instead, politicians need to have vision and leadership and decide to implement such schemes, or else, alternative ways of measuring community acceptance need to be developed. Interestingly more recent research by Miller and Buys (2008) through which 410 household questionnaires conducted in South East Queensland found that the majority of respondents believed that the general community did not have adequate knowledge to vote on indirect potable reuse. The majority of respondents were found to be supportive of the government’s decision to implement the recycled water decision without a referendum. It is clear that political/decision making processes have been a significant influence in the indirect potable reuse plan outcomes in Toowoomba.
4. Toowoomba two years after the referendum 4.1.
Political developments
On the 28th of January 2007, Peter Beattie, the then Premier of Queensland, publicly announced his decision not to let the public vote on whether or not to proceed with a large scale recycled water project for the State’s capital city Brisbane. This was contrary to his prior commitment to a referendum. The Premier argued that even if the public were opposed, there is no other option than to put in place ways to augment water such as recycling (Australian Associated Press, 2007). The project soon began construction and was completed at the end of 2008. It involves six wastewater treatment plants (WWTPs) (Luggage Point, Gibson Island, Bundamba, Oxley Creek, Goodna and Wacol), connected to Wivenhoe dam (Brisbane’s main dam). Three separate Advanced Water Plants have been constructed: one at Luggage Point (receiving water from the Luggage Point WWTP), Gibson Island (receiving water from the Gibson Island and Luggage Point WWTPs) and Bunamba (receiving water from the other four WWTPs). For further information see Western Corridor Recycled Water Project (2008). In response, CADS members distributed a booklet titled ‘think before you agree to drink’ to 500,000 Brisbane households in early 2007 (Roberts, 2008). In July 2008 the Member for Toowoomba South, Mike Horan announced that a pipeline would be constructed from Wivenhoe Dam (Brisbane’s main dam to which the above recycled water would be delivered) to Lake Cressbrook in order to address Toowoomba’s water demand (Australian Associated Press, 2007). Consequently Toowoomba will be supplied with recycled water (Western Corridor Recycled Water Project, 2008) despite the negative referendum vote. However, more recently, the current Queensland Premier Anna Bligh announced that treated wastewater will only go into the dams when they fall below 40% of capacity (ABC News, 2008).
water research 44 (2010) 287–297
Brisbane’s dams were at 74% of capacity at 29th May 2009 after significant rainfall over the past 12 months, thus the recycled water will not be put into the dam at present. Based on the referendum history, it could be expected that Toowoomba residents hold negative attitudes towards the Western Corridor Recycled Water Project. It would logically be expected that Toowoomba residents would be concerned that the State Government has ended up building a recycling plant which will feed into their water supply system despite the negative referendum. Interestingly these feelings were not expressed by the respondents who participated in the interviews and focus group during July 2008 and the survey in January 2009, the results of which are described in detail below.
4.2.
Residents’ attitudes
Details of the empirical results from both the qualitative and quantitative study are now presented. Firstly respondents’ opinions about Toowoomba’s water situation are detailed, followed by their attitudes to the use of recycled water, and the referendum which was held. Lastly information needs of respondents are identified. With respect to residents’ opinions about Toowoomba’s water situation, they generally agreed that Toowoomba will run out of water unless action of some kind is taken. Having a rainwater tank is a common solution to the problem, many participants use tank water for multiple purposes, including drinking. Respondents were attuned to the fact that with below average rainfalls, rainwater tanks may no longer be a solution to the water problem. They were also aware that the tank size they currently have would not cover all their needs if the water situation got worse. Respondents felt strongly about people who break water restrictions and/or steal other people’s water tanks or tank water (which is reportedly common). There was a perception that the Council was not actually enforcing whether or not people comply with the water restrictions, and thus respondents held a belief that offenders are not prosecuted. Respondents proposed that there should be more control and fines for offenders. Some respondents thought that making above average water use very expensive would be an appropriate and indirect way of punishing people for excessive use. Water conservation was an expressly important concern for all participants. It appeared that all respondents were actively conserving water. Stated water conservation measures included, but were not limited to: taking short showers, reusing washing machine water on the garden, using water saving shower heads, and fixing leaks. As stated by one respondent: ‘‘I am absolutely disgusted by people who do not save water, I want to drown them in their own water.’’ This demonstrates the strong emotions surrounding water and its status as a public resource. Table 2 contains results from questions about water conservation asked in the survey of 200 Toowoomba residents. Respondents were presented with a series of statements about water conservation and asked to state whether they agree or disagree with the statements. As can be seen from Table 2, the attitude of Toowoomba residents towards water conservation is overwhelmingly positive with 99% of respondents stating that it is important,
293
95% stating they conserve water wherever they can and only 10% or less feeling no pressure to conserve water or feeling that it is not their responsibility. With respect to residents’ attitudes to water recycling, five interview respondents stated they have no reservations about recycled water at all. One respondent stated they dislike the chlorine (but admits that this is a problem not only related to recycled water but also the current tap water – they prefer to drink ‘‘the shit and leaves in the tank water’’). Another respondent had no concerns, as long as the recycled water had been approved by scientists. Only one respondent categorically refused to use recycled water for drinking, stating: ‘‘I won’t drink it – just me personally, I don’t think I would let my children drink it either. Because you can buy bottled water, but now they are saying it might not be that good either. Well it’s like any machine, how it works and everything . if it doesn’t work properly or it leaks a little bit, it only needs to let a little bit in, doesn’t it?’’ One interview respondent directly mentioned the referendum. When asked how they feel about recycled water the respondent replied: ‘‘It doesn’t bother me – they are going to stick other germs in it to get it the same. How do they know that with the normal water you drink, someone hasn’t gone and crapped in it. It is not going to impact it. Fish and turtles swim in it. Some people just don’t think about it. That was when the vote was in. It was stupid, it just should have gone ahead. I don’t think things would change now – people are still afraid of getting turds in their water, I think it is stupid.’’ This shows that the respondent acknowledged that water from dams also has impurities at source, but is managed in the treatment process. When asked whether they would drink recycled water if the drought got worse, the majority of respondents said that they would be quite happy to use and drink it now. Arguments made by respondents in support of their view included that recycling water would simply increase water supply and thus allow water uses which under current restrictions are not permitted. For example one respondent made the following comment: ‘‘My husband and I thought it was the best thing coming. When I had my first daughter the restrictions weren’t so bad. You could fill up her little pool and have a little splash but with my second one there is none of that you can’t go out and have fun like that – like we did when we were kids’’ Other respondents commented that recycled water may in fact represent an improvement over current solutions. For example: ‘‘They have just scientifically proven that recycled water is better than tank water. I’m drinking pesticides’’ Respondents mentioned that while there might be a little risk of some contamination of the recycled water, it is rather unlikely:
294
water research 44 (2010) 287–297
Table 2 – Attitudes towards water conservations expressed by Toowoomba survey respondents. Attitudinal statement
Average agreement (%)
Water conservation is important Water conservation is necessary because of water scarcity More attention to water conservation is needed I conserve water wherever I can I advocate water conservation among my friends and family I could make more effort to conserve water I only conserve water if water conservation does not cause additional expenses for me Water conservation ALONE can solve Australia’s water problem I only conserve water if water conservation does not take more time I only conserve water if water conservation does not inconvenience me I feel no pressure to conserve water at the moment Water shortage issues don’t affect me Water conservation isn’t my responsibility I am not concerned at all with water conservation
‘‘We are going to have to do it eventually, and it really doesn’t worry me. The scientists have said it is ok. There is only one thing: sometimes scientists say something, then 10 years later they say, oh we were wrong. Can they guarantee 100% that the water is safe, not one little micro organism. It might come and bite them, but the possibility of that is very, very rare’’. Most respondents who first expressed a negative reaction to recycled water use, subsequently changed their attitude when asked to consider necessity. Only one of the respondents, who expressed a negative attitude towards recycled water originally, indicated that they would not change their attitude even if the drought got worse: ‘‘I won’t drink it, I would bath in it and everything else. You’ve got Gatorade and other things for drinking. If they put recycled water in the supply I would buy other water for drinking’’ Table 3 presents the results from the survey which relate to attitudes to recycled water use. Respondents were presented
99 97 95 95 80 75 23 21 12 11 10 4 3 3
with a series of statements about water recycling and were asked to state whether they agree or disagree with the statements. As can be seen from Table 3, most of the statements that have achieved high agreement levels relate to safety issues relating to recycled water. Strict controls of recycled water are demanded by 96% of respondents and two thirds state that they would like to have more information about how recycled water is treated and how safe it is. Despite the stated safety concerns almost half of the Toowoomba residents agree that recycled water is safe to drink. About one third of respondents had very negative feelings about recycled water, agreeing that it is disgusting and that it tastes/smells bad. Another interesting finding, a likely consequence of the referendum in Toowoomba, is that 28% of the respondents agreed with the statement ‘‘They should supply recycled water without asking the public’’. When asked about the referendum, it was clear that the information campaigns from both sides of the referendum had an impact on the emotions of participants. One participant (P1) in the focus group was against the use of recycled
Table 3 – Attitudes towards recycled water expressed by Toowoomba residents (n [ 200). Recycled water attitudinal statement Recycled water would have to be strictly controlled Those who don’t like recycled water can install a rainwater tank to use I am cautious of what is actually in recycled water It’s OK as long as it’s clean I need more information on how recycled water is treated/how safe it is It’s OK if it’s absolutely necessary Those who don’t like recycled water can buy bottled water I think it’s OK if scientists approve of it for human consumption It’s OK for other uses but not as drinking water I am sceptical of how clean/safe recycled water is I have no problem with recycled water I think recycled water is safe for everyone to drink I don’t like the idea of recycled water There are too many health risks Recycled water is disgusting It is wrong to supply recycled 2water to people’s homes They should supply recycled water without asking the public The taste/smell of recycled water is bad
Average agreement (%) 96 76 70 67 66 66 66 65 63 62 50 49 46 45 37 32 28 27
295
water research 44 (2010) 287–297
water for drinking purposes based on concerns about radioactive material (from hospitals). The interaction between participants at this point is found below: P1: ‘‘If they worked out the radioactive business I wouldn’t have a problem’’ P2: ‘‘As I understood it, you know the little booklet that came out in opposition to CADS, well all the filters, those molecular filters will not let molecules through, those molecules carrying radioactive charge . they will be stopped there. I think the radioactive argument stops there because those filters – and there are seven of them – each one is designed to filter out something specific. Even atoms can’t get through’’. P1: ‘‘How big is an AIDS virus?’’ P3: ‘‘We have a friend who is a pharmacist who says you can’t get all of it out, the hormones etc.’’ P1: ‘‘It has to be stopped at source’’ P2: ‘‘I disagree with that because a virus is much bigger than a molecule’’ P4: ‘‘If there was no water, I’d drink anything’’ P2: ‘‘Two atoms of hydrogen and one of water is not very big’’ Respondents clearly felt that the Council information was a reaction to CADS. It also confirms the first mover advantage CADS appears to have had with having their message in public before the Council. The discussion above shows how important ‘expert friends’ (pharmacists), are in shaping attitudes to recycled water. When asked about barriers to drinking recycled water, the main barrier identified by participants was the need for accurate information which was ‘untarnished’, ‘unbiased’, ‘scientific’, and ‘the truth’. When asked about incentives to drinking recycled water, respondents again identified information. P5: ‘‘Good information on what filters remove. Are men going to become women? Scientific information from someone from a University who is not funded by a company building the plant.’’ ALL: ‘‘Agree’’ P5: ‘‘I would really like Australian information at least in relation to our temperature and humidity’’ [regarding the treatment process] P6: ‘‘It would be interesting to have one brochure on all drinking alternatives: desalination, recycled, tank, bore, and have the information on all of them so you could decide which to drink.’’ P1: ‘‘The information should not be a sales pitch from one party or the other. Because the information we got here was a sales pitch from one side or the other’’ P4: ‘‘It was very biased’’ P8: ‘‘It was a scare campaign’’ P7: ‘‘Scare mongering. This is what happens a lot. People with vested interests’’ P8: ‘‘We won’t mention any names, but certain land developers’’ P4: ‘‘Didn’t want to scare anyone from buying in Toowoomba’’ This excerpt from the focus group demonstrates the need to provide unbiased and impartial information. It is clear that respondents were not satisfied with the information campaign surrounding the Toowoomba referendum, and did not seem to trust ‘either side’. This relates to Heywood’s (1999) identified limitations to referendums as discussed in
Section 3.2. A number of respondents indicated the need for information about the cleaning process that takes place with recycling (specifically scientific information from someone who has no conflict of interest) and comparative information about all kinds of water from alternative sources. The results from the survey confirm the sentiments of the focus group. As shown in Table 3, sixty six percent of respondents stated that they need more information on how recycled water is treated and how safe it is. Sixty five percent stated that it would be acceptable to them is if scientists approved of it for human consumption (see Table 3). Respondents were asked who would influence their opinion about recycled water use. About half of the interview respondents stated that nobody would influence them. The following sources of influence were mentioned by other respondents: scientists, their General Practitioner, information on the Internet and information obtained from locals who are seen as having no particular agenda with respect to recycled water. One respondent provided an illustrative example: ‘‘Well, we were about to vote. We were thinking of no, but a scout leader we knew in the area said by voting no we were not going to get the federal government money, so vote yes. He did clarify a lot. We had a good talk about it. With the medication he said we wouldn’t even know. He told us that Dolby [a near by town] has had it for years and you wouldn’t even know.’’ The responses indicated that those participants who were open to consideration (who had not already formed a firm opinion about recycled water), were interested in obtaining more information. They sought information from a wide range of sources including from experts, in general, on the Internet, or even interested respected non-experts from within the community.
Table 4 – Factors/people influential to respondent attitudes to water. Factor/person Research findings News/facts/other publicised information Consideration of future generations An individual or organisation qualified in water management A scientist Family An ecologist The water authority Friends My partner An environmentalist/an environmental group Conservation advertisements The media Neighbours The government A recognisable personality No one A politician
Average agreement (%) 89 86 84 78 78 72 71 69 62 60 55 49 39 33 32 21 17 9
296
water research 44 (2010) 287–297
When asked about what others would think about them drinking recycled water, there were a number of responses from participants of the focus group, with one saying that it would ‘‘depend which side of the fence they are on’’. One respondent clearly stated they don’t mind what others think: ‘‘I don’t think I would let someone else’s opinion worry me actually. If I was thirsty, it is simple as that.’’ Another participant questioned: ‘‘Who cares?’’ Results from the survey regarding people of influence to respondent attitudes towards water related matters, are presented in Table 4. Respondents were asked ‘‘Who or what could influence your attitude towards water related matters (e.g. the use of water efficient appliances, the use of recycled water etc.)?’’ A list of people/factors were presented and respondents were asked to indicate whether each was an influence (yes/no), these were drawn from results of the in-depth interviews. The results in Table 4 indicate that objective sources of information are perceived as more influential be Toowoomba residents. Politicians received the lowest rating with only nine percent of the Toowoomba population agreeing that they would influence their attitudes. In sum, the insights gained through the focus group, the interviews and the survey indicate that overall, respondents were open-minded about recycled water and in many instances regretted that indirect potable reuse was voted against. People were well aware of their dependence on water (especially having a very strong garden city culture) and acknowledged that insufficient water supply may well force them to relocate.
5.
Conclusions
The referendum on indirect potable reuse in Toowoomba was perceived by the Council to be forced upon them, a condition of Commonwealth Government funding. The Council’s preferred approach was a three year consultation program. As such, the Council’s resultant public consultation was rushed and the government information campaign commenced many months after public interest groups started mobilising the residents of Toowoomba to vote against the recycling scheme. The impact of this was evidenced in the focus group discussion and could be one explanation for the negative vote. Another explanation could be information in general and the difficulty participants had in trusting information sources. Participants raised concerns about information and sources of bias on both sides of the referendum Interestingly, the public resistance clearly expressed at the referendum was not mirrored in people’s attitudes towards recycled water as evidenced in this study conducted 2–2.5 years post referendum. Participants were very aware of water issues and were found to actively contribute to local solutions (such as water conservation and the use of rainwater tanks). Given that the Queensland government is building a large scale recycling plant, the Toowoomba residents may end up with indirect potable reuse. Perhaps knowledge of this was a contributing factor to the more positive attitudes towards recycled water found in this study. Many media statements made by CADS in the lead up to the referendum mentioned that Toowoomba did not want to be the first, or the only location in Australia to drink recycled water. Thus knowing
Brisbane (the State’s capital city) would also be drinking recycled water may have alleyed some concerns. The research conducted and presented in this paper indicates that the failure of the Toowoomba indirect potable reuse plans, cannot just be attributed to public opposition to the plans. Politics, timing, vested interests and information manipulation also played a part. The case of Toowoomba raises fundamental questions regarding public participation in government decisions and the way in which democracy is exercised. As a consequence of the Toowoomba referendum, the Queensland state government chose not to put critically needed alternative water projects to a public vote. Currently a large scale recycled water scheme is being implemented, which will in fact lead to recycled water being fed into the dams that are the source of Toowoomba’s water supply. It may well be that such an approach is more effective in achieving the ‘public interest’. A question this raises is how should the public be involved in decisions which have unavoidable consequences for them? It would be beneficial to conduct research in the future to better understand the impact politics, vested interests, information manipulation, and timing each had on the Toowoomba referendum, and the potential impact such factors may have in future projects.
Acknowledgements This study was funded through Australian Research Council (ARC) Discovery Grant (DP0878338). We thank Sarah Oberklaid, Ben Posetti, Katrina Matus and Sharon Lum for research assistance provided. The helpful comments of blind reviewers of the article are appreciated.
references
ABC News, 1 December 2008. Bligh rejects recycled water pipeline a waste of money. Australian Broadcasting Commission. Brisbane (accessed 17.09.08). Ashworth, G.J., 2001. Planning by referendum: empowerment or anarchy in Groningen, the Netherlands. Local Environment 6 (3), 367–372. Australian Associated Press, 30 June 2006. Toowoomba says no to recycled water. Sydney Morning Herald. Sydney (accessed 17.09.08). Australian Associated Press, 28 January 2007. Qld Govt expects recycled water by 2008. Sydney Morning Herald. Sydney (accessed 16.09.08). Balderson, S., 25 May 2006. Anti-recycling campaigners voice their defiance. Toowoomba Chronicle. Toowoomba (accessed 19.01.09). Berghofer, C., 2006. Clive says NO. Toowoomba Decides – Water Poll. (accessed 19.09.08). Carpenter, G.S., Nakamoto, K., 1989. Consumer preference formation and pioneering advantage. Journal of Marketing Research 26 (3), 285–298.
water research 44 (2010) 287–297
Clark, D., 5 May 2006. Recycling debate – transcript. Stateline. Australia. (accessed 14.09.08). Concerned Ratepayer, 28 May 2006. It is ok to say no. Toowoomba (accessed 16.09.08). Dishman, M., Sherrard, J.H., Rebhum, M., 1989. Gaining support for direct potable water reuse. Journal of Professional Issues in Engineering 115 (2), 154–161. Donaghey, K., 21 March 2006. Booklet is latest weapon in water war of words. Toowoomba Chronicle. Toowoomba (accessed 19.01.09). Eisenhardt, K.M., 1989. Building theories from case study research. Academy of Management Review 14 (4), 532–550. Frew, W., 26 October 2005. The yuk factor. Sydney Morning Herald, 5. Sydney, News and Features. Heywood, A., 1999. ‘‘Democracy, Representation and the Public Interest’’. Political Theory: an Introduction. Macmillan, London (Chapter 8). Hurlimann, A., 2006. Water, water, everywhere – which drop should be drunk? Urban Policy and Research 24 (3), 303–305. Hurlimann, A., McKay, J., 2004. Attitudes to reclaimed water for domestic use: Part 2. Trust. Water, Journal of the Australian Water Association 31 (5), 40–45. Lieberman, M.B., Montgomery, D.B., 1988. First-mover advantages. Strategic Management Journal 9, 41. Krueger, R.A., Casey, M.A., 2000. Focus Groups: a Practical Guide for Applied Research. Sage, California. McCornack, S.A., Levine, T.R., Solowczuk, K.A., Torres, H.I., Campbell, D.M., 1992. When the alternation of information is viewed as deception: an empirical test of information manipulation theory. Communication Monographs 63, 83–92. Miller, E., Buys, L., 2008. Water-recycling in South East Queensland, Australia: what do men and women think? Rural Society 18 (3), 220–229. Mitchell, S., 25 March 2006. Town to vote on drinking sewage. The Weekend Australian, 10. Canberra. Natural Resource Management Ministerial Council, Environment Protection and Heritage Council, Australian Health Ministers’ Conference, 2006. National Guidelines for Water Recycling: Managing Health and Environmental Risks (Phase One). Natural Resource Management Ministerial Council and Environment Protection and Heritage Council, Canberra. Natural Resource Management Ministerial Council, Environment Protection and Heritage Council, National Health and Medical Research Council, 2008. Australian Guidelines for Water Recycling Managing Health and Environmental Risks (Phase Two): Augmentation of Drinking Water Supplies. Environment Protection and Heritage Council, the National Health and Medical Research Council and the Natural Resource Management Ministerial Council, Canberra. O’Malley, B., 10 June 2006. Waste not, water not: Toowoomba is the first battleground in the recycled water debate. The Courier Mail, 60. Brisbane, Focus. Parsons Brinckerhoff Australia Pty Ltd, 2006. Future Water Supply Options for Toowoomba City and Customer Shires Prefeasibility Study. Department of Natural Resources, Mines and Water, South East Queensland Regional Water Supply Strategy, Brisbane. (accessed 19.09.08). Postel, S., 1997. Last Oasis. W.W. Norton and Company, New York. Reynolds, M., 24 February 2006. Recycling receives a spray. The Australian, Adelaide. Finance, 28. Richards, L., 2005. Handling Qualitative Data. Sage, London. Roberts, G., 30 October 2008. Flush then drink in the sunshine state. The Australian, Canberra. (accessed 27.01.09).
297
Robinson, W.T., Fornell, C., 1985. Sources of market pioneer advantages in consumer goods industries. Journal of Marketing Research 22 (3), 305–317. SBS Network, 1 November 2005. Running dry – transcript. Insight. SBS Australia. (accessed 19.09.08). Science Media Centre, 2006. Toowoomba votes ‘no’ to waste water recycling. Clean Air & Environmental Quality 40 (4), 21–22. Simpson, J.M., 1999. Changing community attitudes to potable reuse in South-East Queensland. Water Science and Technology 40 (4–5), 59–66. Smith, G., 2001. Taking deliberation seriously: institutional design and green politics. Environment Politics 10 (3), 72–93. Stenekes, N., Schaefer, A.I., Ashbolt, N., 2001. Community involvement in water recycling – issues and needs. In: Schaefer, A.I., Waite, T.D., Sherman, P. (Eds.), Proceedings from the Recent Advances in Water Recycling Technologies Workshop. Centre for Water and Waste Technology, Civil and Environmental Engineering The University of New South Wales, and the Queensland Government Environmental Protection Authority, Brisbane, pp. 113–124. Thorley, D., 2007. Toowoomba recycled water poll. Reform 89 (Summer), 49–51. Toowoomba City Council, 2001. Queens Park Gardens: Where History and Contemporary Garden Design Connect (Brochure). Toowoomba City Council, Toowoomba. Toowoomba City Council, 2005a. Toowoomba Water Futures Project, Media Release. Toowoomba (accessed 20.09.08). Toowoomba City Council, 2005b. Water futuresdToowoomba: taking control of Toowoomba’s future briefing paper. (accessed 12.09.08). Toowoomba City Council, 2006a. Media release: yes campaign up and running. (accessed 27.07.09). Toowoomba City Council, 2006b. "Peter Beattie Insert to the Toowoomba Chronicle. (accessed 20.01.09). Toowoomba City Council, 2006c. Water Futures Toowoomba. Toowoomba (accessed 19.09.08). Toowoomba City Council, 2007. Toowoomba and Surrounds Visitors Guide. Toowoomba City Council, Toowoomba. Toowoomba City Council, 2008. Water restrictions. (accessed 12.12.08). Toowoomba Regional Council, and Toowoomba Water, 2009. Weekly water supply statistics. (accessed 19.01.09). Toowoomba Water Futures Blog, 2006. How we got this far. Toowoomba: Posted 25th March (accessed 15.01.09). Water Futures Blog, 2007. Over a million Brisbane people will have their say on drinking sewage water. Toowoomba: Posted 20th March (accessed 20.01.09). Western Corridor Recycled Water Project, 2008. Western Corridor Recycled Water Project Home Page. Queensland Government, Brisbane. (accessed 22.06.09).
water research 44 (2010) 298–308
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Enhanced transformation of triclosan by laccase in the presence of redox mediators Kumarasamy Murugesan a, Yoon-Young Chang b, Young-Mo Kim a,1, Jong-Rok Jeon a, Eun-Ju Kim a, Yoon-Seok Chang a,* a
School of Environmental Science and Engineering, Pohang University of Science and Technology (POSTECH), San 31, Hyoja-dong, Nam-gu, Pohang 790-784, Republic of Korea b Department of Environmental Engineering, Kwangwoon University, Seoul 139-701, Republic of Korea
article info
abstract
Article history:
Triclosan (TCS), an antimicrobial agent, is an emerging and persistent environmental
Received 2 June 2009
pollutant that is often found as a contaminant in surface waters and sediments; hence,
Received in revised form
knowledge of its degradability is important. In this study we investigated laccase-mediated
5 August 2009
TCS transformation and detoxification, using laccase (from the fungus Ganoderma lucidum)
Accepted 10 September 2009
in the presence and absence of redox mediators. Transformation products were identified
Available online 1 October 2009
using HPLC, ESI-MS and GC–MS, and transformation mechanisms were proposed. In the absence of redox mediator, 56.5% TCS removal was observed within 24 h, concomitant
Keywords:
with formation of new products with molecular weights greater than that of TCS. These
Triclosan
products were dimers and trimers of TCS, as confirmed by ESI-MS analysis. Among the
Antimicrobial compound
various mediators tested, 1-hydroxybenzotriazole (HBT) and syringaldehyde (SYD) signifi-
Ganoderma lucidum
cantly enhanced TCS transformation (w90%). The presence of these mediators resulted in
Laccase
products with lower molecular weights than TCS, including 2,4-dichlorophenol (2,4-DCP;
Syringaldehyde
confirmed by GC–MS) and dechlorinated forms of 2,4-DCP. When SYD was used as the
Natural redox mediator
mediator, dechlorination resulted in 2-chlorohydroquinone (2-CHQ). Bacterial growth inhibition studies revealed that laccase-mediated transformation of TCS effectively decreased its toxicity, with ultimate conversion to less toxic or nontoxic products. Our results confirmed the involvement of two mechanisms of laccase-catalyzed TCS removal: (i) oligomerization in the absence of redox mediators, and (ii) ether bond cleavage followed by dechlorination in the presence of redox mediators. These results suggest that laccase in combination with natural redox mediator systems may be a useful strategy for the detoxification and elimination of TCS from aqueous systems. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Triclosan (2,4,40 -trichloro-20 -hydroxydiphenyl ether; TCS) is a synthetic antimicrobial compound that is present in a wide range of health care products, and in consumer products
including textiles and plastics (Bhargava and Leonard, 1996; Schweizer, 2001). The wide use of products containing TCS has resulted in the entry of this compound into the environment; it has been detected in various environmental matrices including wastewaters, freshwater, seawater and sediments,
* Corresponding author. Tel.: þ82 54 279 2281; fax: þ82 54 279 8299. E-mail address:
[email protected] (Y.-S. Chang). 1 Present Address: Biological Sciences Division Pacific Northwest National Laboratory Richland, WA 99352, USA. 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.058
water research 44 (2010) 298–308
and biotic samples such as fish and human breast milk (Okumura and Nishikawa, 1996; Adolfsson-Erici et al., 2002; Singer et al., 2002; Halden and Paull, 2005; Miller et al., 2008). TCS has also been detected in agricultural soils following land application of biosolids from a wastewater treatment plant (Cha and Cupples, 2009). TCS kills a wide range of microorganisms, resulting in poor biodegradation and its accumulation and long-term persistence in the environment. For example, TCS was found to persist for more than 40 years in estuary sediment (Miller et al., 2008). The presence in aquatic environments of TCS over certain concentrations has been reported to lead to harmful effects to aquatic organisms, including changes in the thyroid hormone in tadpoles, and estrogenic activity and death in fish (Ishibashi et al., 2004; Veldhoen et al., 2006). In chemical structure, TCS resembles other halogenated diphenyl ethers and dioxins that are known to be highly persistent in the environment. In addition, on exposure to UV or heat, TCS is a potential precursor for toxic chlorinated dioxins (Latch et al., 2005; Rule et al., 2005). The presence of TCS may also inhibit the nitrification process in activated sludge systems (Stasinakis et al., 2008). Consequently, the removal of contaminating TCS is received increasing attention. A photocatalytic process has been suggested as a potential method to eliminate TCS, but in practice produced 2,7/2,8-dichlorinated dioxins (Mezcua et al., 2004). Certain bacteria known to transform halogenated diphenyl ether compounds (Schmidt et al., 1992; Kim et al., 2007) failed to metabolize TCS (Schmidt et al., 1993). Due to its inhibitory activity against a wide range of bacteria, bacterial degradation of TCS is very limited. The toxicity of TCS is attributed to inhibition of the bacterial fatty acid biosynthetic enzyme, enoyl (acyl-carrier protein) reductase, which occurs in Gramnegative and Gram-positive bacteria, as well as in mycobacteria (McMurry et al., 1998; Levy et al., 1999; Rule et al., 2005). Nevertheless, some bacterial strains are able to survive in the presence of TCS due to target mutations, increased target expression, active efflux from the cell, and/or enzymatic inactivation/degradation (Schweizer, 2001). Meade et al. (2001) reported that very few bacteria are able to inactivate TCS. In contrast to bacteria, transformation of TCS by the white rot fungi Trametes versicolor and Pycnoporus cinnabarinus has been reported (Hundt et al., 2000). Although their growth was markedly inhibited by TCS, even at low concentrations, they were able to decrease the cytotoxic and microbicidal effects of TCS by converting it to methylated and glycosyl conjugated forms. In addition, transformation of TCS by T. versicolor produced only small amounts of toxic chlorophenols, such as 2,4-dichlorophenol (2,4-DCP), and glycosyl conjugation of 2,4-DCP further reduced the toxicity (Hundt et al., 2000). As an alternative to fungal cultures, the use of isolated enzymes may be effective in degrading toxic halogenated compounds. Among fungal enzymes, laccases (EC 1.10.3.2) from the white rot fungi can detoxify a wide range of organic pollutants (Dec and Bollag, 1995; Bollag et al., 2003; Murugesan et al., 2009b), and laccase-catalyzed processes have the advantage of involving molecular oxygen as an electron acceptor. Laccase-mediated oxidation and detoxification of TCS have recently been demonstrated using laccases isolated from T. versicolor and Coriolopsis polyzona (Cabana et al., 2007, 2009; Kim and Nicell, 2006). Laccase-mediated degradation of
299
chlorophenols and chlorohydroxy diphenyl ether compounds is mainly due to oligomerization of oxidized substrate via radical–radical coupling (Dec and Bollag, 1995; Schultz et al., 2001). Cabana et al. (2007) reported the first evidence for laccase-mediated TCS degradation by oligomerization, which is a well known mechanism. Although TCS is transformed by oligomerization, the TCS structure remains in the oligomer, which could make destruction of this compound more difficult. However, breakdown of the TCS into a mono-ring compound and elimination of chlorine atoms could yield less toxic or nontoxic products. The aim of the present study was to elucidate the mechanisms of laccase-mediated TCS transformation in the presence and absence of redox mediators. We used a laccase isolated from the white rot fungus Ganoderma lucidum (Murugesan et al., 2007). In addition to the known synthetic mediator, we compared various natural phenolic compounds as redox mediators for TCS transformation (Camarero et al., 2005). We demonstrate that in the presence of laccase and redox mediators, TCS degradation proceeds by ether bond cleavage and partial dechlorination.
2.
Experimental procedures
2.1.
Chemicals
Triclosan (2,4,40 -trichloro-20 -hydroxydiphenyl ether) (TCS), 2,4-dichlorophenol (2,4-DCP), 3-chlorophenol (3-CP), 2-chlorohydroquinone (2-CHQ), and redox mediators, 1-hydroxybenzotriaozole (HBT), acetovanillone (ACV), syringaldehyde (SYD), vanillin (VAN), p-coumaric acid (PCA), 2,4-dimethoxyphenol (DMP), and guaiacol (GUI) were purchased from Sigma– Aldrich (St. Louis, MO; Milwaukee, MI). Ferulic acid (FA) and 2,20 azino-bis-(3-ethylbenzothiazoline-6-sulfonate) (ABTS) were supplied by Fluka. All other chemicals are the highest purity of analytical grade.
2.2.
Laccase (1.10.3.2)
Laccase production from the white rot fungus G. lucidum KMK2 was performed as described previously (Murugesan et al., 2007). The crude laccase was purified through ammonium sulfate precipitation, ion-exchange and gel filtration chromatography using FPLC system (BIO-RAD BIOLOGIC). The molecular mass of purified laccase was 43 kDa as confirmed by SDSPAGE. The purified laccase was filtered and stored at 4 C. for further use. Laccase activity was measured at 30 C using 1 mM ABTS as the substrate (Wolfenden and Wilson, 1982) as described in a previous report (Murugesan et al., 2007).
2.3.
TCS transformation
TCS stock (100 mM) was prepared in acetonitrile and appropriate dilution of this stock was used for transformation experiments. TCS transformation was carried out in glass vials (2 mL) using citrate–phosphate buffer (50 mM; pH 4.0) and purified laccase. The final reaction volume 1 mL contained laccase (5 U) and 0.2 mM of TCS from acetonitrile stock. The final concentration of acetonitrile in reaction mixture was not exceeded 2.0% at which no laccase inhibition was
300
water research 44 (2010) 298–308
observed. The reaction vials were incubated at 30 C under dark. At pre-determined intervals, the reaction vials were removed and the reaction was stopped immediately by acidifying the reaction mixture using acetic acid and then 250 mL acetonitrile was added, and analyzed through HPLC. Control sample was maintained without laccase.
2.4.
Effect of redox mediators on TCS transformation
To study the effect of redox mediators, we screened different redox mediators namely HBT, ABTS, and natural compounds such as syringaldehyde, acetovanillone, vanillin, p-coumaric acid, 2,4-dimethoxyphenol and guaiacol each at 1 mM concentration. The reaction mixture containing 1 mM redox mediator, 0.2 mM TCS and 5 U laccase was incubated at 30 C for 12 h and TCS removal was measured. After, screening, HBT and SYD were chosen for time course studies and product identifications.
2.5.
Dechlorination of TCS and 2,4-DCP
Since dechlorinated product 2-CHQ was detected from TCS transformation, we conducted experiments to monitor the dechlorinating activity of G. lucidum laccase using 0.2 mM TCS and 2,4-DCP as the initial reaction substrate with 5 U laccase in 1 mL 10 mM sodium citrate buffer (pH 4.0). After 24 h incubation the reaction was stopped by adding 250 mL acetonitrile and the samples were analyzed by ion-chromatography for estimation of chloride.
2.6.
Analytical methods
2.6.1.
HPLC analysis
The residual of TCS and its transformation products were quantified using HPLC (1100 Series, Agilent, Germany) fitted with ZORBAX SB-C18 column. The reaction mixture was mixed with 250 mL of acetonitrile, vortexed vigorously and filtered through 0.45 mm filter (Pall Cooperation, MI), and then 10 mL sample was injected by auto-injector port. The elution was performed using 70% acetonitrile in 0.1 phosphoric acid as solvent at constant (1 mL/min) flow rate. TCS and metabolites were monitored at 277 nm by DAD-UV detector.
2.6.2.
GC–MS and ESI-MS analysis
To identify the products of TCS transformation, the acidified reaction mixture was extracted thrice with equal volume of ethyl acetate. Then the extract was dehydrated with anhydrous sodium sulfate, dried under N2 gas, and dissolved in acetone. This extract was analyzed through GC–MS using 60 m DB-5 column or 30 m DB-5 column (Trace GC system coupled with Polaris Q Iontrap MS – Thermoquest, Jan Jose, CA). The GC conditions were set as initial 60 C for 3 min followed by 10 C increment up to 230 C and remain constant for 10 min. To identify the oligomeric products, TCS (200 mM) transformation was conducted in 10 mL reaction for 48 h. Then the products prepared as described by Cabana et al. (2007) and analyzed by ESI-MS (A triple quadrupole instrument; API 2000 liquid chromatography/MS/MS system; Applied Biosystems, Foster City, CA). The ESI source was selected in the negative voltage mode at 4500 to detect metabolites.
2.6.3.
Ion-chromatography
Chloride concentration was analyzed using a Dionex ion chromatograph (IC, DX-120) that was equipped with a conductivity detector, Dionex Ionpac AS-14 (4 mm 250 mm; G-14 Guard, 4 mm 50 mm) for anion analysis. The eluent composition was 3.5 mM Na2CO3/1 mM NaHCO3 for the anion analysis. Chloride ions were quantified using NaCl calibration curve.
2.7.
Bacterial growth inhibition studies
To access whether laccase-catalyzed TCS transformation can reduce the toxicity of TCS, we performed the growth inhibition studies using Escherichia coli and Sphingobium sp. PH-07 in the presence of TCS and its transformation products. Reactions containing 50 and 200 mM TCS with 5 U laccase in the presence and absence of 1 mM HBT and SYD were incubated for 24 h. Then, 100 mL of 0.1 OD cells of over night grown culture of the above bacteria were mixed with each reaction vials and incubated for 1 h and then inoculated into Erlenmeyer flasks containing 9 mL nutrient broth. Flasks were incubated in an incubator at 30 C with shaking at 160 rpm. The growth was monitored at OD600 nm and results were presented in percentage inhibition of growth relative to control culture.
3.
Results and discussion
3.1.
TCS transformation by G. lucidum laccase
Widespread use of the broad spectrum antimicrobial agent TCS has led to it becoming a common environmental contaminant. In the United States, TCS has been detected at concentrations up to 21.9 mg/L in influent and effluent of wastewater treatment plants (Schweizer, 2001), indicating that it is not easily removed by conventional wastewater treatment processes. As TCS can kill a wide range of bacteria, bacterial degradation of this compound is very poor, and even bacteria capable of degrading chlorodiphenyl ether have failed to metabolize TCS (Schmidt et al., 1993). Thus, high concentrations of TCS have recently been found in sediments (Singer et al., 2002), and have been reported to have persisted for over 40 years in estuary sediments (Miller et al., 2008). Enzymatic detoxification of TCS by laccase, an enzyme able to detoxify toxic halogenated phenols, has recently been considered as an alternative to microbial degradation (Kim and Nicell, 2006; Cabana et al., 2007). However, information on the byproducts and mechanism of laccase-mediated TCS transformation is very limited. Hence, we studied the enzymatic transformation of TCS in an aqueous system using laccase from G. lucidum, to investigate the transformation mechanism under different reaction conditions, including the presence and absence of redox mediators. Following incubation of TCS (0.2 mM) with 5 U mL1 laccase for 24 h, we found there had been 56.5% removal of TCS, with concomitant formation of a new product, as determined by RP-HPLC (Fig. 1). Under the RP-HPLC conditions used, TCS was eluted with a retention time (Rt) of 3.1 min (Fig. S1), whereas the new product had an Rt of 5.8 min, indicating that it is more
water research 44 (2010) 298–308
Fig. 1 – Transformation of TCS by G. lucidum laccase. Residual TCS concentration (-); TCS product (6). Reaction was carried out in citrate–phosphate buffer (pH 4.0) with 0.2 mM TCS and laccase 5 U mLL1. Data are the mean ± SD of triplicate experiments.
hydrophobic than TCS. We assumed that this product could be a dimer of TCS formed by oxidative coupling of the phenoxy radicals of TCS. Previous studies have reported that laccase effectively transforms hydroxylated diphenyl ethers into an oligomer through the C–C or C–O bonds formed between the phenoxy radicals of diphenyl ether (Jonas et al., 2000). Cabana et al. (2007) also observed oligomerization of TCS by laccase from C. polyzona. Our result also indicates oligomerization of TCS by G. lucidum laccase. Although TCS was transformed by oligomerization, the basic structure of the monomer would remain identical in the oligomer, as reported by Jonas et al. (2000). As oligomerization makes TCS more complex, this could hinder its biodegradation. Thus, destruction of the chemical structure of TCS, such as through ether bond cleavage and elimination the chlorine atoms, is important and could yield products that are accessible to other microorganisms. To assess this possibility we used various redox mediators to attempt to break down the TCS, as high redox potential mediators are known to destroy recalcitrant compounds (Li et al., 1998).
301
synthetic mediators HBT and ABTS, and included several lignin-related monomers because they have been reported to act as potential redox mediators in the transformation of recalcitrant pollutants (Camarero et al., 2005, 2007; Can˜as et al., 2007; Jeon et al., 2008; Murugesan et al., 2009a,b). Screening of the various mediators, each at 1 mM concentrations, revealed that the natural compound syringaldehyde (SYD) was the best redox mediator, resulting in 84.2% removal of TCS (Fig. 2). Of the two synthetic mediators tested, HBT was more effective than ABTS. This result stands in contrast to those of Kim and Nicell (2006) and Cabana et al. (2007), who found that HBT did not enhance TCS removal. This discrepancy may be due to differences in the ratio between the mediator and substrate concentrations. The mediator:substrate molar ratio is an important factor in the effective removal of pollutants in laccase-catalyzed reactions. We used a molar ratio of 5:1, whereas in the studies noted above the ratio was 1:1 or less. Can˜as et al. (2007) observed the effective removal of a polycyclic aromatic hydrocarbon when the mediator:substrate ratio was increased 10-fold. Balakshin et al. (2000) reported that a true nonenzymatic reaction between an oxidized mediator and veratryl alcohol only occurred at a molar ratio of 2:1 or higher. Among the several mediators tested in the present study, significant removal of TCS relative to the treatment without mediator was observed with HBT ( p < 0.01) and SYD ( p < 0.001), indicating that the latter, a natural phenolic compound, is as effective as the efficient synthetic mediator HBT. Recent studies have also demonstrated the potential role of SYD in enhancing laccasecatalyzed decolorization of recalcitrant synthetic dyes (Camarero et al., 2005; Murugesan et al., 2009a,b). For time course studies of TCS transformation and identification of degradation products, we conducted experiments with HBT and SYD. HPLC analysis revealed that, in the presence of these mediators, the extent of TCS transformation was
3.2. TCS transformation by laccase in the presence of redox mediators Inclusion of low molecular weight redox mediators in the reaction mixture is the general approach taken to enhance laccase-catalyzed transformation of recalcitrant pollutants. Kim and Nicell (2006) and Cabana et al. (2007) used the synthetic mediators ABTS, HBT and TEMPO (2,20 ,6,60 tetramethoxy piperidine-1-oxyl), and the natural phenolic compound syringic acid for TCS removal. They found that with the exception of ABTS, these compounds did not significantly enhance the degradation of TCS. However, the oxidized ABTS radical was highly toxic in the Microtox toxicity test (Kim and Nicell, 2006). In the present study we used the
Fig. 2 – Effect of various redox mediators on TCS transformation by G. lucidum laccase. Reaction was carried out in citrate–phosphate buffer (pH 4.0) with 0.2 mM TCS and laccase 5 U mLL1. Data are the mean ± SD of triplicate samples.
302
water research 44 (2010) 298–308
enhanced by 88% (HBT) and 92% (SYD) after 24 h incubation (Fig. 3). In the presence of either mediator, TCS was transformed to a new product with a retention time (Rt ¼ 2.01) (Fig. S1) lower than that for TCS (Rt ¼ 3.1 min), suggesting that the product is less hydrophobic than TCS. We hypothesized that the product may be 2,4-dichlorophenol (2,4-DCP), formed by ether bond cleavage, and using authentic compound as a standard we confirmed 2,4-DCP was present in samples amended with HBT or SYD. In addition to 2,4-DCP, trace amounts of monochlorophenol and chlorohydroquinone were detected. Although 88% and 92% of the initial 0.2 mM TCS was transformed, the concentration of 2,4-DCP detected was about 10–15 times lower than the amount predicted on the basis of stoichiometry. This may have been due to rapid removal of 2,4-DCP by laccase through polymerization or degradation, as laccase is very effective in removal and detoxification of chlorophenols from aqueous systems (Roy-Arcand and Archibald, 1991; Dec and Bollag, 1994, 1995; Zhang et al., 2008). We observed that with the laccase used, chlorophenol removal was faster than that of TCS. Stability of residual enzyme is important for the potential transformation of phenolic compounds. Buchanan and Nicell (1998) reported that the inactivation of peroxidase during oxidation of phenolic compounds is due to the interaction of enzyme with the reaction products. The use of additives like polyethylene glycol has been used to minimize the enzyme inactivation (Kim and Nicell, 2006). In our study, we measured the residual laccase activity during TCS transformation and observed the decrease in residual activity in the range between 40 and 67% in TCS added samples after 24 h. The significant decrease of residual enzyme may have influenced the differences in TCS conversion between different set of experiments. The highest lose in residual activity (67%) was observed in presence of HBT, which is known to inhibit the laccase activity (Garcı´a et al., 2003).
Fig. 3 – TCS transformation by G. lucidum laccase in presence of 1 mM HBT and SYD. Residual TCS concentration in the presence of HBT (-) and SYD (,). 2,4DCP concentration in the presence HBT (B) and SYD (C). Reaction was carried out in citrate–phosphate buffer (pH 4.0) with 0.2 mM TCS and laccase 5 U mLL1. Data are the mean ± SD of triplicate experiments.
Important features of a potential redox mediator are the reversibility and stability of radicals during catalytic reactions. The mediators HBT and SYD belong to N–OH and C–OH groups, respectively, and hydrogen atom transfer is the mechanism of radical formation in both cases. The half-life of the N–O radical of HBT is very short due to its high reactivity, and it rapidly decays to benzotriazole (Li et al., 1998). Thus, a relatively high intensity benzotriazole peak was detected in the GC profile of the HBT-treated samples. In contrast, the phenoxy radicals (C–O) of phenolic compounds have long half-lives and reversible reactions (Ferna´ndez-Sa´nchez et al., 2002). The stability and reversibility of SYD radicals is attributed to the presence of two methoxy groups in ortho positions to the phenolic group (Camarero et al., 2007), explaining the enhanced transformation of TCS that was observed with SYD in this study. However, Kim and Nicell (2006) observed no degradation enhancement when syringic acid, an analog of SYD, was used as a redox mediator. This may be due to the low mediator:substrate ratio (1:1) used. Our study clearly suggests that SYD is a redox mediator with potential application in the removal of TCS from wastewater and soil, as SYD is naturally available.
3.3. Mass spectrometry analysis of TCS transformation products With the exception of oligomer formation, observed by Cabana et al. (2007), no information is available regarding the products formed in laccase-mediated TCS transformation. Our RP-HPLC analysis of TCS transformation by G. lucidum laccase clearly showed that the transformation product profiles varied in the absence or presence of redox mediators. To identify the products and transformation mechanism, we used LC–MS and GC–MS. To identify the oligomer, we extracted 10 mL reaction samples in chloroform after they had been incubated for 48 h. HPLC profiles of the chloroform extracts showed two new peaks in addition to the TCS peak, with retention times of 7.1 min and 11.8 min (Fig. 4A). These peaks indicated the presence of compounds more hydrophobic than TCS. ESI-MS analysis also revealed the presence of two new mass peaks, in addition to the TCS mass peak (m/z 287), with molecular masses of 575 and 863.0; these masses correspond to those of the dimer and trimer of TCS, respectively (Fig. 4B). This result is similar to that obtained by Cabana et al. (2007) using C. polyzona laccase. Transformations of diphenyl ether and halogenated diphenyl ethers have been demonstrated in T. versicolor and P. cinnabarinus (Hundt et al., 1999, 2000; Jonas et al., 2000). Stationary phase cultures of P. cinnabarinus metabolized the diphenyl ether substrate into the hydroxylated diphenyl ether, and cell-free laccase derived from this fungus transformed 2-hydroxy diphenyl ether into its dimer through C–C bonds in the position para to the hydroxy group of the monomers (Jonas et al., 2000). T. versicolor hydroxylated the halogenated diphenyl ether compounds at the nonhalogenated ring, and then cleaved the hydroxylated ring (Hundt et al., 1999). This fungus also metabolized TCS, mainly producing glycosyl conjugated metabolites including 2-o-(2,4,40 -trichlorodiphenyl ether)-b-D-xylopyranoside, 2-o(2,4,40 -trichlorodiphenyl ether)-b-D-glucopyranoside, and
water research 44 (2010) 298–308
303
Fig. 4 – Identification of products from TCS transformation by laccase. A. HPLC profile of chloroform extract showing TCS (peak 1) and TCS products (peaks 2, 3); B. ESI-MS chloroform extract showing mass peak of TCS (m/z 287), dimer (m/z 575) and trimer (m/z 863) of TCS.
a small amount of 2,4-DCP, thereby reducing the toxicity of TCS (Hundt et al., 2000). Under the same conditions P. cinnabarinus produced a glucoside conjugate and methylated TCS, and no ether bond cleavage product (2,4-DCP) was observed (Hundt et al., 2000). T. versicolor mainly produces laccase that transforms hydroxylated diphenyl ethers and TCS to oligomer. However, in the above study little 2,4-DCP was observed. We assumed that a redox mediated reaction may have occurred or other enzymes were involved. The use of redox mediators in laccase-catalyzed TCS transformation would enable assessment of these possibilities. The product formed in the presence of redox mediators was less hydrophobic than TCS. We extracted the product with ethyl acetate and analyzed it using GC–MS. In a control sample a single peak (Rt ¼ 22.84 min) with a mass spectrum identical to TCS (confirmed from the standard mass spectral library) was detected (data not shown). In the case of HBT-treated samples, new major GC peaks were apparent at Rt ¼ 14.74 min and Rt ¼ 17.47 min (Fig. 5a), in addition to the TCS peak. The mass spectrum of the compound giving rise to the peak at Rt ¼ 14.74 min was identical to that of 2,4-DCP (Fig. 5b), and the Rt ¼ 17.47 min peak matched benzotriazole. Therefore, the
GC–MS results confirmed that 2,4-DCP was the main TCS product when HBT was used as a redox mediator, and proved that ether bond cleavage had occurred. The experiments were repeated and the products were again identified in the N,O-bis(trimethylsilyl)-trifluoro acetamide (BSTFA) derivatized form, as shown in Table S1 (see Supplementary Material). Peaks eluted at Rt ¼ 31.24 and Rt ¼ 18.78 had masses corresponding to TCS and 2,4-DCP, respectively. Similar results were obtained when SYD was added as the redox mediator (Table S1). However, in the presence of SYD an additional product identified as 2-CHQ (Rt ¼ 10.27 min) was formed along with 2,4-DCP (Rt ¼ 6.22 min) (Fig. 6a), based on comparison of mass fragmentation (Fig. 6b) with data from the mass spectrometry library and an authentic standard. This was confirmed by BSTFA derivatization, which showed mass fragmentation corresponding to 2-CHQ (Table S1). These findings are consistent with the dechlorination of 2,4-DCP due to laccase-catalyzed oxidation of 2,4-DCP. From the GC–MS analysis results, it is evident that oxidized HBT and SYD mediators cleave the ether bond linkage and produce 2,4-DCP. The mediated enzymatic transformation of TCS resulted in its structural dehalogenation of TCS, thereby its toxicity is being reduced.
304
water research 44 (2010) 298–308
17.47
100
Benzotriazole
a
80 70 60 50 40
Triclosan
2,4-dichlorophenol
Relative abundance (%)
90
14.74
30
22.86
20 10 0
11.17
12.34 13.80
.21
10
12
16
14
251.3 24.04 25.54 26.88 27.60
17.76 19.31 20.43 21.41
16.07
18
20
22
24
26
28
Time (min)
b
162.04
100
OH
Relative abundance (%)
90
Cl
80 70 Cl
60
164.01
50 40 30
62.98 98.02 126.05
20
166.11
71.11
10 0 50
100
150
200
m/z Fig. 5 – Identification of products from TCS transformation in the presence of 1 mM HBT by GC–MS. (a) GC–MS Chromatogram; (b) Mass spectrum of 2,4-DCP (peak 14.74 min of chromatogram). GC was performed using 60 m DB-5 column.
3.4. Dechlorination of TCS and 2,4-DCP by laccase from G. lucidum Previous studies have demonstrated laccase-mediated dehalogenation of chlorophenols, chloroanilines and chlorobiphenyl ethers as a result of free radical coupling or nucleophilic attack (Dec and Bollag, 1994, 1995; Schultz et al., 2001). To confirm the occurrence of dechlorination, we conducted experiments with TCS and 2,4-DCP as initial substrates in the absence and presence of HBT and SYD. No chloride release was detected with TCS as the initial substrate in the absence of a redox mediator (Table 1), indicating that TCS removal occurred without loss of chlorine atoms. ESI-MS results supported this finding, as the dimer and trimer peaks of TCS showed masses with no loss of chlorine atoms (Fig. 4b). In the presence of the mediators HBT and SYD, chloride ion concentrations of 344 mM and 97.3 mM were detected in the reaction, respectively. The extent of TCS removal was also greater with redox mediators present than in their absence.
In contrast, dechlorination was observed in both the absence and presence of redox mediators when 2,4-DCP was used as the initial substrate. Laccase alone showed 96.0% 2,4DCP removal with the release of 92.7 mM chloride ions, whereas 100% of the 2,4-DCP was removed in the presence of HBT and SYD, with the release of 323.0 mM and 147.0 mM chloride ions, respectively. The dechlorination results suggest that in our experiments chloride ion release only occurred from 2,4-DCP. Dec and Bollag (1995) reported that release of chloride ions depends on the mode of 2,4-DCP radical coupling, and proposed that C–C coupling would release 2 chloride ions, and C–O–C coupling would release a single chloride ion. Assuming the molar concentration of chlorine in 2,4-DCP, our results indicate that C–C coupling occurred in the HBT-treated sample, as the amount of chloride detected was two-fold more in the presence of HBT than in its absence. This result was corroborated by the absence of 2-CHQ formation with HBT. In contrast, 2-CHQ was observed in the presence of SYD, indicating the removal of a single chloride ion from the
305
water research 44 (2010) 298–308
1 5 .4 2
100
SYD 2-Chlorohydroquinone
90
70 60 50
1 0 .2 7
2,4-dichlorophenol
Relative abundance (%)
80
40 30 20 10 0
4 .3 7
4
5 .9 1
6 .2 2
6
2 0 .8 3
Triclosan
a
2 2 .0 0
6 .5 7
7 .1 8
1 0 .7 7 1 1 .6 9 1 3 .9 8 1 4 .9 6
1 0 .1 5
8
10
12
14
1 5 .6 7 1 7 .8 6 1 8 .4 6 1 8 .8 9
16
18
20
2 2 .3 6
22
2 4 .7 6
24
2 6 .7 4 2 7 .3 0 2 8 .7 0 2 9 .56
26
28
30
Time (min)
b
144.11
100
OH
Relative abundance (%)
90
Cl
80 70 OH
60 50 40
5 1 .9 6
30
1 146.16 8 0 .0 4 1 0 8 .1 1
20 10 0 50
100
150 m/z
Fig. 6 – Identification of products from TCS transformation in the presence of 1 mM SYD by GC–MS. (a) GC–MS Chromatogram; (b) Mass spectrum of 2-CHQ (peak 10.27 min of chromatogram). GC was performed using 30 m DB-5 column.
para position of 2,4-DCP. This suggests that instead of a coupling reaction, oxidative dechlorination occurred. Oxidative dechlorination of halogenated diphenyl ether has been reported in Coriolus versicolor cultures (Hiratsuka et al., 2001). Direct dechlorination of chlorophenols by laccase has also been shown for some chlorophenols and chloroguaiacols (Roy-Arcand and Archibald, 1991).
3.5.
Bacterial growth inhibition by TCS and its products
TCS and its products observed in this study, 2,4-DCP and 2-CHQ, are highly toxic to microbial cells. TCS inhibits bacterial growth even at concentrations as low as 0.24 mg L1 (0.83 mM), because it blocks lipid synthesis (McMurry et al.,
1998) thereby preventing the formation of new cell membrane (Levy et al., 1999). We assessed the toxicity of a laccasecatalyzed TCS reaction mixture in a bacterial growth inhibition test. Our results revealed that bacterial growth was inhibited when the nutrient broth medium was amended with reaction mixture containing 200 mM TCS (Table 2). The inhibition was attributed to the presence of residual TCS, because in our experiment removal of TCS was incomplete at an initial concentration of 200 mM, and the residual TCS level was much higher than the reported inhibitory concentration. Therefore, we tested lower concentrations and found that TCS at 50 mM was completely removed within 24 h incubation, with no growth inhibition in either E. coli or Sphingomonas sp. PH-07 cultures (Table 2). Detoxification of
306
water research 44 (2010) 298–308
O
Table 1 – Chloride ions detected from TCS and 2,4-DCP transformation by G. lucidum laccase. Substrate
Treatment
Substrate removal (%)
Chloride release (mM)
Laccase Laccase þ HBT Laccase þ SYD
59.6 83.6 85.3
0 344.0 97.3
Laccase Laccase þ HBT Laccase þ SYD
96.0 100 100
92.7 323.0 147.0
O
i c La
OH
TCS 200 mM
2,4-DCP 200 mM
se ca
Cl
TCS
Cl
TCS
Cl
Cl
Cl
HB
ii
La
To
TCS
TCS
Dimer, trimer ….oligomer
O
Cl
TCS
cca s rS e YD Cl
Data presented are average of duplicate experiments.
BT e +H cas Lac l OH - 2C
Oligomer
Cl
-C Lac cas l e+ SY D
OH Cl
- Cl-
Oligomer
OH
a variety of chlorophenols by laccase has been shown previously. Our results confirm the detoxification of TCS with laccase from G. lucidum. Different pathways for the degradation of diphenyl ether compounds by white rot fungi have been reported. For example, the commercial herbicide chloronitrofen (2,4,6trichlcoro -40 -nitrodiphenyl ether) was metabolized by C. versicolor via four different pathways including hydroxylation, oxidative dechlorination, reductive dechlorination and nitro-reduction. These reactions eventually led to the formation of mono-ring compounds, but there was no direct evidence of ether bond cleavage (Hiratsuka et al., 2001). In a cell-free enzyme system, hydroxylated diphenyl ether compounds are oligomerized (Jonas et al., 2000; Schultz et al., 2001; Cabana et al., 2007). Although bacterial degradation of diphenyl ether is difficult, several bacteria are able to mineralize this compound (Schmidt et al., 1992; Kim et al., 2007) and transform the halogenated diphenyl ethers by hydroxylation and ring fission, leading to cleavage of the ether bond. Based on HPLC, GC–MS and ESI-MS analyses, the pathway for TCS transformation by G. lucidum laccase is shown in Fig. 7. It is evident that two transformation routes were involved in our study: (i) oligomerization, which mainly occurred in the absence of redox mediators; and (ii) ether bond cleavage and dechlorination, which occurred in the presence of redox mediators. Ether bond cleavage has also been demonstrated in a photocatalytic process (Latch et al., 2005), but the generation of toxic chlorinated dioxins is a disadvantage of this process. Considering the difficulties involved in bacterial and
Table 2 – Inhibition of bacterial growth in the presence of TCS and laccase-catalyzed TCS products. Reactions/substrates incubated
Control TCS TCS þ Laccase TCS þ HBT þ Laccase TCS þ SYD þ Laccase
Fig 7 – Proposed pathways of TCS transformation by G. lucidum laccase. Route (i) Transformation in the absence of redox mediator. Route (ii) Transformation of in the presence of redox mediator HBT or SYD.
photocatalytic degradation of TCS, laccase-mediated degradation could be an effective method for TCS removal from contaminant soils and wastewaters, as humic substances enhance TCS degradation.
4.
Conclusions
Degradation and detoxification of TCS is important because of its environmental persistence and toxicity. In this study, enzymatic transformation and detoxification of TCS was demonstrated using laccase isolated from G. lucidum, and the addition of redox mediators. Two modes of TCS transformation were observed, depending on the reaction conditions: i) oligomerization in the absence of redox mediators, and ii) ether bond cleavage followed by dechlorination in the presence of redox mediators. Using bacterial growth inhibition studies, it was confirmed that TCS was detoxified by enzymatic transformation. To our knowledge this is the first report providing direct evidence of ether bond cleavage of diphenyl ether by laccase. Although this study indicates the potential for TCS detoxification in aqueous systems, the efficiency of laccase and natural mediator systems needs to be investigated in wastewater and soil samples in situ, due to the complexity of natural environments.
Acknowledgements
Relative growth (%) E. coli
PH-07
200 mM TCS
50 mM TCS
200 mM TCS
50 mM TCS
100.0 0.0 0.6 0.6 2.4
100.0 0.0 100.0 100.0 100.0
100.0 0.0 0.5 0.5 2.2
100.0 0.0 100.0 100.0 100.0
Data presented are average of duplicate experiments.
This project was funded by Ministry of Environment of Korea as ‘‘The GAIA Project’’ and ‘‘BK-21 Project’’.
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2009.09.058.
water research 44 (2010) 298–308
references
Adolfsson-Erici, M., Pettersson, M., Parkkonen, J., Sturve, J., 2002. Triclosan, a commonly used bactericide found in human milk and in the aquatic environment in Sweden. Chemosphere 46, 1485–1489. Balakshin, M.Y., Chen, C.L., Gratzl, J.S., Kirkman, A.G., Jakob, H., 2000. Kinetic studies on oxidation of veratryl alcohol by laccase-mediator system. Part 1. Effects of mediator concentration. Holzforschung 54, 165–170. Bhargava, H.N., Leonard, P.A., 1996. Triclosan: applications and safety. American Journal of Infection Control 24, 209–218. Bollag, J.M., Chu, H.L., Rao, M.A., Gianfreda, L., 2003. Enzymatic oxidative transformation of chlorophenol mixtures. Journal of Environmental Quality 32, 63–69. Buchanan, I.D., Nicell, J.A., 1998. Kinetics of peroxidase interactions in the presence of protective additive. Journal of Chemical Technology and Biotechnology 72, 23–32. Cabana, H., Jiwan, J.L.H., Rozenberg, R., Elisashvili, V., Penninckx, M., Agathos, S.N., Jones, J.P., 2007. Elimination of endocrine disrupting chemicals nonylphenol and bisphenol A and personal care product ingredient triclosan using enzyme preparation from the white rot fungus Coriolopsis polyzona. Chemosphere 67, 770–778. Cabana, H., Alexandre, C., Agathos, S.N., Jones, J.P., 2009. Immobilization of laccase from the white rot fungus Coriolopsis polyzona and use of the immobilized biocatalyst for the continuous elimination of endocrine disrupting chemicals. Bioresource Technology 100, 3447–3458. Camarero, S., Ibarra, D., Martic´nez, M.J., Martic´nez, Ac´.T, 2005. Lignin-derived compounds as efficient laccase mediators for decolorization of different types of recalcitrant dyes. Applied and Environmental Microbiology 71, 1775–1784. Camarero, S., Ibarra, D., Martı´nez, A.T., Romero, J., Gutie´rrez, A., del Rı´o, J.C., 2007. Paper pulp delignification using laccase and natural mediators. Enzyme and Microbial Technology 40, 1264–1271. ´ .T., Can˜as, A.I., Alcalde, M., Plou, F., Martı´nez, M.J., Martı´nez, A Camarero, S., 2007. Transformation of polycyclic aromatic hydrocarbons by laccase is strongly enhanced by phenolic compounds present in soil. Environmental Science & Technology 41, 2964–2971. Cha, J., Cupples, A.M., 2009. Detection of the antimicrobials triclocarban and triclosan in agricultural soils following land application of municipal biosolids. Water Research 43, 2522–2530. Dec, J., Bollag, J.M., 1994. Dehalogenation of chlorinated phenols during oxidative coupling. Environmental Science & Technology 28, 484–490. Dec, J., Bollag, J.M., 1995. Effect of various factors on dehalogenation of chlorinated phenols and anilines during oxidative coupling. Environmental Science & Technology 29, 657–663. Ferna´ndez-Sa´nchez, C., Tzanov, T., Gu¨bitz, G.M., Cavaco-Paulo, A., 2002. Voltammetric monitoring of laccase-catalysed mediated reactions. Bioelectrochemistry 58, 149–156. ´ .T., Martı´nez, M.J., Garcı´a, O., Camarero, S., Colom, J.F., Martı´nez, A Monje, R., Vidal, T., 2003. Optimization of a laccase-mediator stage for TCF bleaching of flax pulp. Holzforschung 57, 513–519. Halden, R.U., Paull, D.H., 2005. Co-occurrence of triclocarban and triclosan in US water resources. Environmental Science & Technology 39, 1420–1426. Hiratsuka, N., Wariishi, H., Tanaka, H., 2001. Degradation of diphenyl ether herbicides by the lignin-degrading basidiomycete Coriolus versicolor. Applied Microbiology and Biotechnology 57, 563–571. Hundt, K., Jonas, U., Hammer, E., Schauer, F., 1999. Transformation of diphenyl ethers by Trametes versicolor and characterization of ring cleavage products. Biodegradation 10, 279–286.
307
Hundt, K., Martin, D., Hammer, E., Jonas, U., Kindermann, M.K., Schauer, F., 2000. Transformation of triclosan by Trametes versicolor and Pycnoporus cinnabarinus. Applied and Environmental Microbiology 66, 4157–4160. Ishibashi, H., Matsumura, N., Hirano, M., Matsuoka, M., Shiratsuchi, H., Ishibashi, Y., Takao, Y., Arizono, K., 2004. Effects of triclosan on the early life stages and reproduction of medaka Oryzias latipes and induction of hepatic vitellogenin. Aquatic Toxicology 67, 167–179. Jeon, J.R., Murugesan, K., Kim, Y.M., Kim, E.J., Chang, Y.S., 2008. Synergistic effect of laccase mediators on pentachlorophenol removal by Ganoderma lucidum laccase. Applied Microbiology and Biotechnology 81, 783–790. Jonas, U., Hammer, E., Haupt, E.T.K., Schauer, F., 2000. Characterisation of coupling products formed by biotransformation of biphenyl and diphenyl ether by the white rot fungus Pycnoporus cinnabarinus. Archives of Microbiology 174, 393–398. Kim, Y.J., Nicell, J.A., 2006. Laccase-catalysed oxidation of aqueous triclosan. Journal of Chemical Technology and Biotechnology 81, 1344–1352. Kim, Y.M., Nam, I.H., Murugesan, K., Schmidt, S., Crowley, D.E., Chang, Y.S., 2007. Biodegradation of diphenyl ether and transformation of selected brominated congeners by Sphingomonas sp. PH-07. Applied Microbiology and Biotechnology 77, 187–194. Latch, D.E., Packer, J.L., Stender, B.L., VanOverbeke, J., Arnold, W.A., McNeill, K., 2005. Aqueous photochemistry of triclosan: formation of 2,4-dichlorophenol, 2,8-dichlorodibenzop-dioxin, and oligomerization products. Environmental Toxicology and Chemistry 24, 517–525. Levy, C.W., Roujeinikova, A., Sedelnikova, S., Baker, P.J., Stuitje, A.R., Slabas, A.R., Rice, D.W., Rafferty, J.B., 1999. Molecular basis of triclosan activity. Nature 398, 383–384. Li, K., Helm, R.F., Eriksson, K.E.L., 1998. Mechanistic studies of the oxidation of a non-phenolic lignin model compound by the laccase/I-hydroxybenzotriazole redox system. Biotechnology and Applied Biochemistry 27, 239–243. McMurry, L.M., Oethinger, M., Levy, S.B., 1998. Triclosan targets lipid synthesis. Nature 394, 531–532. Meade, M.J., Waddell, R.L., Callahan, T.M., 2001. Soil bacteria Pseudomonas putida and Alcaligenes xylosoxidans subsp denitrificans inactivate triclosan in liquid and solid substrates. FEMS Microbiology Letters 204, 45–48. Mezcua, M., Gomez, M.J., Ferrer, I., Aguera, A., Hernando, M.D., Fernandez-Alba, A.R., 2004. Evidence of 2,7/2,8dibenzodichloro-p-dioxin as a photodegradation product of triclosan in water and wastewater samples. Analytica Chimica Acta 524, 241–247. Miller, T.R., Heidler, J., Chillrud, S.N., Delaquil, A., Ritchie, J.C., Mihalic, J.N., Bopp, R., Halden, R.U., 2008. Fate of triclosan and evidence for reductive dechlorination of triclocarban in estuarine sediments. Environmental Science & Technology 42, 4570–4576. Murugesan, K., Nam, I.H., Kim, Y.M., Chang, Y.S., 2007. Decolorization of reactive dyes by a thermostable laccase produced by Ganoderma lucidum in solid state culture. Enzyme and Microbial Technology 40, 1662–1672. Murugesan, K., Yang, I.H., Kim, Y.M., Jeon, J.R., Chang, Y.S., 2009a. Enhanced transformation of malachite green by laccase of Ganoderma lucidum in the presence of natural phenolic compounds. Applied Microbiology and Biotechnology 82, 341–350. Murugesan, K., Kim, Y.M., Jeon, J.R., Chang, Y.S., 2009b. Effect of metal ions on reactive dye decolorization by laccase from Ganoderma lucidum. Journal of Hazardous Materials. doi:10. 1016/j.jhazmat.2009.02.075. Okumura, T., Nishikawa, Y., 1996. Gas chromatography mass spectrometry determination of triclosans in water, sediment
308
water research 44 (2010) 298–308
and fish samples via methylation with diazomethane. Analytica Chimica Acta 325, 175–184. Roy-Arcand, L., Archibald, F.S., 1991. Direct dechlorination of chlorophenolic compounds by laccases from Trametes (Coriolus) versicolor. Enzyme and Microbial Technology 13, 194–203. Rule, K.L., Ebbett, V.R., Vikesland, P.J., 2005. Formation of chloroform and chlorinated organics by free-chlorinemediated oxidation of triclosan. Environmental Science & Technology 39, 3176–3185. Schmidt, S., Wittich, R.M., Erdmann, D., Wilkes, H., Francke, W., Fortnagel, P., 1992. Biodegradation of diphenyl ether and its monohalogenated derivatives by Sphingomonas sp. strain SS3. Applied and Environmental Microbiology 58, 2744–2750. Schmidt, S., Fortnagel, P., Wittich, R.M., 1993. Biodegradation and transformation of 4,40 - and 2,4-dihalodiphenyl ethers by Sphingomonas sp. strain SS33. Applied and Environmental Microbiology 59, 3931–3933. Schultz, A., Jonas, U., Hammer, E., Schauer, F., 2001. Dehalogenation of chlorinated hydroxybiphenyls by fungal laccase. Applied and Environmental Microbiology 67, 4377–4381. Schweizer, H.P., 2001. Triclosan: a widely used biocide and its link to antibiotics. FEMS Microbiology Letters 202, 1–7.
Singer, H., Muller, S., Tixier, C., Pillonel, L., 2002. Triclosan: occurrence and fate of a widely used biocide in the aquatic environment: field measurements in wastewater treatment plants, surface waters, and lake sediments. Environmental Science & Technology 36, 4998–5004. Stasinakis, A.S., Mainais, D., Thomaidis, N.S., Danika, E., Gatidou, G., Lekkas, T.D., 2008. Inhibitory effect of triclosan and nonylphenol on respiration rates and ammonia removal in activated sludge systems. Ecotoxicology and Environmental Safety 70, 199–206. Veldhoen, N., Skirrow, R.C., Osachoff, H., Wigmore, H., Clapson, D.J., Gunderson, M.P., Van Aggelen, G., Helbing, C.C., 2006. The bactericidal agent triclosan modulates thyroid hormoneassociated gene expression and disrupts postembryonic anuran development. Aquatic Toxicology 80, 217–227. Wolfenden, B.S., Wilson, R.L., 1982. Radical cations as reference chromogens in studies of one-electron transfer reactions: pulse radio analysis studies of 2,20 azinobis-(3ethlbenzthiazoline-b-ulfonate). Journal of the Chemical Society Perkin Transaction II, 805–812. Zhang, J.B., Liu, X.P., Xu, Z.Q., Chen, H., Yang, Y.X., 2008. Degradation of chlorophenols catalyzed by laccase. International Biodeterioration & Biodegradation 61, 351–356.
water research 44 (2010) 309–319
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effects of lanthanum and lanthanum-modified clay on growth, survival and reproduction of Daphnia magna Miquel Lu¨rling a,*, Yora Tolman a,b a
Aquatic Ecology & Water Quality Management Group, Department of Environmental Sciences, Wageningen University, P.O. Box 47, 6700 AA Wageningen, the Netherlands b Waterboard Delfland, P.O. Box 3061, 2061 DB Delft, the Netherlands
article info
abstract
Article history:
The novel lanthanum-modified clay water treatment technology (Phoslock) seems very
Received 10 February 2009
promising in remediation of eutrophied waters. Phoslock is highly efficient in stripping
Received in revised form
dissolved phosphorous from the water column and in intercepting phosphorous released
21 July 2009
from the sediments. The active phosphorous-sorbent in Phoslock is the Rare Earth Element
Accepted 10 September 2009
lanthanum. A leachate experiment revealed that lanthanum could be released from the clay,
Available online 17 September 2009
but only in minute quantities of 0.13–2.13 mg l1 for a worst-case Phoslock dosage of
Keywords:
that lanthanum, up to the 1000 mg l1 tested, had no toxic effect on the animals, but only in
Eutrophication control
medium without phosphorous. In the presence of phosphorous, rhabdophane (LaPO4 $ nH2O)
Lake management
formation resulted in significant precipitation of the food algae and consequently affected
Lake restoration
life-history traits. With increasing amounts of lanthanum, in the presence of phosphate,
Lanthanum
animals remained smaller, matured later, and reproduced less, resulting in lower population
Life-history
growth rates. Growth rates were not affected at 33 mg La l1, but were 6% and 7% lower at 100
250 mg l1. A life-history experiment with the zooplankton grazer Daphnia magna revealed
Modified clay
and 330 mg l1, respectively, and 20% lower at 1000 mg l1. A juvenile growth assay with
Phoslock
Phoslock tested in the range 0–5000 mg l1, yielded EC50 (NOEC) values of 871 (100) and 1557 (500) mg Phoslock l1 for weight and length based growth rates, respectively. The results of this study show that no major detrimental effects on Daphnia are to be expected from Phoslock or its active ingredient lanthanum when applied in eutrophication control. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Cyanobacterial proliferation and accumulation of biomass in nuisance scums are an obvious symptom of anthropogenic nutrient over-enrichment of surface waters (Fogg, 1969; Reynolds, 1987; Reynolds and Walsby, 1975; Paerl, 1988, 2008). Such cyanobacterial blooms may cause high turbidity, anoxia, fish kills, bad smells and pose potentially serious environmental and human health problems, because several cyanobacteria can produce a variety of very potent toxins (Codd
et al., 2005; Dittmann and Wiegand, 2006; Paerl, 2008; Paerl and Huisman, 2008). Climate change is expected even to aggravate hazardous blooms (Paerl and Huisman, 2008), while safe and aesthetically acceptable water is a growing need in a modern society (Steffensen, 2008). Hence, water management is faced world-wide with a call for reducing this vulnerability to the threats of harmful cyanobacterial blooms. This means that eutrophication control remains one of the key challenges to global environmental sustainability for the 21st Century (Sharpley and Tunney, 2000; Schindler, 2006).
* Corresponding author. Tel.: þ31 317 483898; fax: þ31 317 484411. E-mail address:
[email protected] (M. Lu¨rling). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.034
310
water research 44 (2010) 309–319
Inasmuch as the most important cause of lake eutrophication is phosphorous pollution (Schindler, 1974, 1975, 1977; Correll, 1998), phosphorous (P) control is critical to mitigating eutrophication (Carpenter, 2008; Schindler et al., 2008). This requires both input control from point and nonpoint sources as well as the P-removal from the water column and P-retention in the bottom sediments (Welch and Cooke, 1995; Carpenter et al., 1998; Søndergaard et al., 2003; Mehner et al., 2008). In the Netherlands, from the early 1980s a variety of restoration techniques have been employed. However, more long-term failures than successes have been recorded that are largely related to inadequate treatment of or neglect of in-lake P control (Gulati and Van Donk, 2002). As the European Union Water Framework Directive (2000/60/ EC) aims to restore all waters to a good ecological status or potential by 2015, it is obvious that additional remedial measures are needed to reduce in-lake P concentrations to low levels and to overcome P-release from the P-rich bottom sediments (Gulati and Van Donk, 2002). Here, the novel lanthanum-modified clay water treatment technology (Phoslock) developed by CSIRO (Australia) seems very promising in remediation of degraded water. Phoslock is highly efficient in stripping dissolved P from the water column and in intercepting P released from the sediments (Douglas et al., 1999; Robb et al., 2003; Akhurst et al., 2004; Ross et al., 2008). The active P-sorbent in Phoslock is the Rare Earth Element lanthanum which is absorbed to or complexed with the clay. This element may be released from the bentonite clay–La complex when added to water. La3þ-ions could be toxic to some aquatic organisms, particularly cladocerans such as Daphnia (Barry and Meehan, 2000; NICNAS, 2001). Hence, the potential liberation of La3þ-ions from the bentonite could mean a significant environmental risk (Akhurst et al., 2004), but Phoslock has been classified as not hazardous (Martin and Hickey, 2004). It should be noted, however, that there is no consistency in the results of the few studies on the effects of lanthanum on cladocerans (Barry and Meehan, 2000; Sneller et al., 2000; Stauber, 2000; NICNAS, 2001; Martin and Hickey, 2004). In addition, the effects of Phoslock have not been tested as such, rather an indirect so-called Toxic Characteristic Leachate Procedure has been employed (Stauber, 2000; NICNAS, 2001; Martin and Hickey, 2004). The purpose of this study was: 1) to establish a dose response relationship between Phoslock and the growth of Daphnia magna, 2) to determine the amount of lanthanum released from Phoslock, and 3) to test the effects of lanthanum on life-history characteristics of D. magna in artificial P-free and P-containing medium. Based on the very strong binding of lanthanum to oxyanions and especially phosphates (e.g. Haghseresht, 2005a,b; Biswas et al., 2007; Ross et al., 2008), we hypothesize that in the presence of phosphate the formation of the insoluble mineral rhabdophane will dramatically mitigate toxicity of lanthanum.
2.
Materials and methods
2.1.
Test organisms
Experiments were carried out with the cladoceran D. magna Straus that has been isolated from Lake Zwemlust
(The Netherlands) and has been maintained for more than 10 years in our laboratory. Here the Daphnia are kept at 20 C in 1 l jars containing 800 ml artificial RT-medium with a pH of 7.6, a conductivity of 270 mS cm1 and a total hardness of 88 mg CaCO3 l1 (Tollrian, 1993). The animals are fed three times a week with the green alga Scenedesmus obliquus (Turpin) Ku¨tzing (w4 mg C l1). S. obliquus SAG 276/3a originated from the culture collection of the University of Go¨ttingen (Germany). S. obliquus was maintained in 1.0 l chemostat systems in continuous light of 120 mmol quanta m2 s1 at 20 C on a slightly modified WC medium (Lu¨rling and Beekman, 1999) and with a dilution rate of 1.1 d1.
2.2.
Phoslock leachate experiment
Two batches (25 kg each) of Phoslock were obtained from Phoslock Water Solutions Ltd. (Australia). About 0.5 g Phoslock was added to Erlenmeyer flasks that contained 100 ml nanopure water. Each batch was tested in triplicate (0.5033 0.004 g of batch 1 and 0.5022 0.002 g of batch 2). Three additional Erlenmeyers contained only 100 ml nanopure water. The Erlenmeyers were closed with Parafilm and placed for 48 h in an incubator in darkness, at 22 C with continuous orbital shaking (200 rpm). After this the material was centrifuged for 5 min at 3000 rpm, followed by filtration through a 0.45 mm membrane filter. Filtrates were analyzed for metals (Al, Cd, Cu, Hg, La, Pb, Zn) using AAS (Hg) and ICP-MS (Al, Cd, Cu, La, Pb, Zn) in the Chemical–Biological Soil Laboratory of the Department of Soil Sciences (Wageningen University).
2.3. Effect of lanthanum on life-history traits of D. magna Juvenile Daphnia born on the same day were collected from the stock cultures and placed individually in separate 125 ml test tubes containing 100 ml of Scenedesmus food suspension with a concentration of 5 mm3 l1 (equivalent to w2.5 mg C l1). These Daphnia were transferred daily to new tubes with fresh food and newborns from the third broods were used as experimental animals. The newborns were placed in a 500-ml beaker with RT-medium. For each treatment ten neonates were randomly selected and transferred individually into 125 ml test tubes containing 100 ml of a food suspension (in RT-medium) with different concentrations of lanthanum. Stock solutions of lanthanum were made from La(NO3)3 $ 6H2O at 3.3 mg La l1, 10 mg La l1, 33 mg La l1 and 100 mg La l1 in nanopure water. Concentrations of La in the water were measured by inductively coupled plasma mass spectrometry (ICP-MS) in the Chemical–Biological Soil Laboratory of the Department of Soil Sciences (Wageningen University). Lanthanum was tested at the following nominal concentrations: 0, 33, 100, 330 and 1000 mg l1 in the absence and presence of phosphate (330 mg l1), yielding 5 La concentrations 10 replicates 2 phosphate levels ¼ 100 experimental units. Each test tube contained only one experimental animal to avoid density effects (Martı´nez-Jero´nimo et al., 2000). The test tubes were incubated in a temperature-controlled room at 20 C. The animals were transferred daily to clean tubes with
water research 44 (2010) 309–319
fresh food suspensions and lanthanum. Before Daphnia were transferred into these new tubes their body length was measured using a stereo-binocular microscope. The number of survivors, time to reproduction, and number of newborns were recorded. Growth and reproduction were recorded until animals had reached the fourth adult instar and consequently released their third brood, because the first three broods largely determine population growth rate (Vanni and Lampert, 1992). The instantaneous rates of population increase (r) were estimated from abbreviated life-tables (three broods) using the equation: ln ry
PN
x¼0 lx mx
T
;
r ¼ rate of population increase (d1), x ¼ age class (0, ., N), lx ¼ probability of surviving to age x, mx ¼ fecundity at age x, and T ¼ the generation time. A Jack-knifing method was used to calculate standard errors of r (Meyer et al., 1986). The increase in body-size over time for the different treatments was statistically analyzed running repeated measures ANOVAs in the toolpack SPSS version 16.0.1. When the ANOVA indicated significant differences a Tukey post-hoc comparison test was run to distinguish means that were significantly different (P < 0.05). Age and size at first reproduction and brood sizes were compared running one-way ANOVA.
2.4.
this would reflect an application dosage to water with maximally 1000 mg FRP l1. Each treatment consisted of three replicates with three animals per beaker. All beakers received Scenedesmus (at a concentration of 10 mm3 l1, which is equivalent to w5 mg C l1) as food to the animals. The beakers were incubated in a temperature-controlled room at 20 C in darkness. At the start of the experiment the body lengths of 15 newborns were measured using a stereo-binocular microscope. Body length is defined as the distance from the most posterior point on the eye to the base of the junction of the tail spine with the carapace. Five groups of three specimens were transferred in small pre-weighed aluminium boats, dried at 105 C for 24 h, and weighed on an electronic balance (Mettler UMT 2; 0.1 mg). After 5 days of incubation, experimental animals were collected from the beakers, rinsed in RTmedium after which their body lengths and dry-weights were determined. The juvenile somatic growth rates ( g) were determined as the increase in dry mass (W ) and body length (BL) from the beginning of the experiment (X0) to day 5 (Xt) using the equation: gW ¼ ðlnXt lnX0 Þ=t For both endpoints growth rates were statistically compared running one-way ANOVA in the statistical toolpack SPSS Release 16.0.1. Differences between means were distinguished by Tukey’s post-hoc comparison (P < 0.05). EC50 values (i.e. Phoslock concentration causing a 50% inhibition of growth) were determined using non-linear regression by fitting a 3 parameter sigmoidal function in the toolpack SigmaPlot 2000, version 6.00 (Cleuvers, 2003).
Effect of lanthanum on S. obliquus
Because precipitation was observed in medium containing P and lanthanum, the effect of lanthanum on the food availability was examined. The duration of the experiment, i.e. max 25 h, was kept similar to the daily medium renewal regime in the life-history experiment. Separate 125 ml test tubes were filled with 100 ml suspensions of Scenedesmus (with a concentration of 5 mm3 l1) in P-free or P-containing (330 mg l1) RT-medium. Lanthanum (1000 mg l1) was added to three tubes with P-free medium and to three with P-containing medium, while for each medium three replicate flasks served as controls (no lanthanum added). Initially and after 2, 18 and 25 h, biovolume was determined using an electronic cell counter (CASY, Scha¨rfe System Gmbh., Reutlingen, Germany). Biovolumes in the different treatments, measured after 2, 18 and 25 h, were statistically compared running oneway ANOVAs and were followed by Tukey post-hoc comparison test (P < 0.05).
2.5.
311
Effect of Phoslock on growth of Daphnia
A 5 d juvenile growth experiment was conducted with thirdclutch juveniles ( ZH for the humic-like fractions. SUVA of the WWTP isolates was lower than those obtained for IHSS standards (SRFA, 4.5 L mg C1 m1; SRHA, 7.8 L mg C1 m1) indicating the weaker aromaticity of humic-like fractions of EfCOM. WWTP colloidal isolates are characterized by C/N ratios in the range of 8 to 18 (Table 2), values more than five times lower than those generally found for pedogenic HS with C/N ratios of 50 to 100. For WWTP effluent isolates, the C/N ratio of biopolymers is lower than that for the VD lake water isolate. These differences indicate that the biopolymers could be of a proteinaceous nature in WWTPs, due to bacterial production in the activated sludge. The above observations are consistent with a lower aromatic moiety of WWTP humiclike substances, in comparison with pedogenic HS (Westerhoff et al., 2001). The tryptophan-like fluorescence is a characteristic of high microbial productivity (Westerhoff et al., 2001; Baker et al., 2003; Chen et al., 2003; Saadi et al., 2006; Leenheer et al., 2007), and is consistent with the high nitrogen content of WWTP HS and microbial products originating from the activated sludge treatment.
3.2. Molar mass distribution characterization of colloidal fractions Two major size (molar mass) fractions of the EfCOM were determined by the AFlFFF, defined operationally as low molar mass (LMM) and high molar mass fractions (HMM). The molar mass distribution characteristics of the different components in the LMM fraction (Mn, Mw, PDI) are given in Table 3. The number- and weight-average molar masses ranged from 800 to 1800 Da (Mn) and from 1600 to 3400 Da (Mw) and decreased in the order DB > ZH > HW. Vidy Bay COM exhibited slightly lower Mw (1600 Da) than the EfCOM. The obtained LMM fraction characteristics for DB and ZH were in the same range as those obtained for standard SRHA and SRFA, while HW and VD LMM fractions exhibit lower masses (Table 3). The obtained values were in the range generally measured by FlFFF for freshwater (Thang et al., 2001) and standard humic substances (Beckett et al., 1987; Reszat and Hendry, 2005), although slightly lower than previously published values for Mn (690–2900 Da) and Mw (1600–14,350 Da) for WWTP HS fractionated by HPSEC (Perminova et al., 2003). Although similar molar mass distributions were obtained with both AFlFFF setups, the distribution characteristics of the LMM fractions obtained with the Postnova system (Fig. 4a) were slightly lower than the Wyatt setup (Fig. 4b). This could be attributed to the different type of membrane used as the accumulation wall in the AFlFFF (Thang et al., 2001). The molar mass distribution of the HMM fraction of the EfCOM was more complex. Three major populations of high molar mass components with Mp1 ¼ 45 kDa, Mp2 ¼ 265 kDa and Mp3 ¼ 456 kDa were detected. Their sphere equivalent diameters were d1 ¼ 7 nm, d2 ¼ 20 nm and d3 ¼ 30 nm (Fig. 4c). The molar masses obtained for the HMM fraction components were similar to those determined by HPSEC for isolated extracellular polymeric substances (polysaccharides and proteins) from activated sludge of WWTPs, which can have molar masses above 200 kDa (Garnier et al., 2005; Comte et al., 2007).
3.3. water
Metal distributions in WWTP effluents and lake
The distributions of Al, Cr, Cu, Fe, Mn, Ni, Pb and Zn among particulate, colloidal and truly dissolved fractions for the WWTP effluent samples obtained by TFF are given in Fig. 5. With the exception of Al and Pb, the total dissolved metal
Table 3 – Molar mass corresponding to the peak maximum (Mp); number-average (Mn), weight-average (Mw) molar masses and polydispersity index (PDI) of LMM fraction obtained by AFlFFF for Duebendorf (DB), Hinwil (HW) and Zurich (ZH) WWTP effluents, Vidy Bay lake water (VD) and standard Suwannee River fulvic and humic acids. Samples
Postnova AFlFFF 3
Wyatt AFlFFF
Mp (10 Da)
Mn (10 Da)
Mw (10 Da)
PDI
Mp (10 Da)
Mn (103 Da)
Mw (103 Da)
PDI
2.5 1.4 2.5 1.2 1.7 2.7
1.8 0.8 1.6 1.1 1.5 2.4
3.4 1.8 2.9 1.6 2.6 4.4
1.9 2.3 1.8 1.5 1.8 1.9
2.9 2.1 2.5 1.9 nd nd
2.2 1.6 1.9 1.4 nd nd
3.6 2.6 3.6 2.9 nd nd
1.7 1.7 1.7 2.1 nd nd
DB HW ZH VD SRFA SRHA nd, not determined.
3
3
3
346
water research 44 (2010) 340–350
in the colloidal fraction and low concentrations in the particulate form. The DB effluent had a total dissolved Mn concentration as low as the lake water. Al and Pb were mainly in particulate form in the HW effluent, whereas Al was found predominantly in the truly dissolved fraction in the ZH effluent. Cu, Cr, Ni and Zn were found mainly within the truly dissolved fraction. Significant proportions of Cu, Cr and Zn were also found in both colloidal and particulate fractions, however their proportions varied between the 2007 and 2008 samples. The metal contents in the colloids were higher in the effluents sampled in 2007 than in those sampled in 2008, in agreement with the higher COC in the first sampling. Moreover, the colloidal metal-to-COC concentration ratios (Fig. 5) showed little difference between the two sampling dates, considering the variability of the TFF. Overall, the obtained results demonstrated the important role of the EfCOM for trace metal distribution in the total dissolved fraction.
3.4.
Fig. 4 – Molar mass distribution of EfCOM by AFlFFF. (a) LMM fractions of DB, HW, ZH, VD colloidal isolates and Suwannee River fulvic and humic acids (SRFA and SRHA) using the Postnova AFlFFF-UV (Vxf [ 3 mL minL1; 10 mM NaNO3, pH [ 5.4; logMw [ 1.8913 logtrw D 0.178 for PSS calibration). (b) LMM fractions of DB, HW, ZH, and VD using the Wyatt AFlFFF-UV (Vxf [ 3 mL minL1; 10 mM NaNO3, pH [ 5.5; logMw [ 2.0 logtrw L 0.6 for PSS calibration). (c) HMM fractions of ZH colloidal isolate, using Postnova AFlFFF MALS and a cross flow value of Vxf [ 0.25 mL minL1. The 908 angle LS signal was deconvoluated in four distinct components (1,2,3,4) using Origin software. Sphere equivalent diameter (indicated on the upper x-axis) was obtained from latex standard calibration. concentrations in the WWTP effluents were higher than those measured in the lake water (VD). Iron was present predominantly in particulate form. Manganese was mostly found in the truly dissolved fraction in all samples, with nearly no Mn
Distribution of Metals within the COM fractions
Using AFlFFF coupled to ICP-MS, most of the metals associated with the colloidal pool in the WWTP effluent isolates were detected within the LMM fraction (as exemplified by the DB sample in Fig. 6a). However, the distribution pattern was metal dependent. Ag, Cr and Zn ICP-MS signals corresponded to the maximum of the LMM distributions. Al, Mn and Cu signals were slightly shifted to the lower end of the molar mass distribution. They contrast with the previous finding that Al signals followed the UV signal corresponding to the higher end of molar mass distribution (Wu et al., 2004; Bolea et al., 2006). Furthermore, the maximum of Fe and Pb signals was closer to the higher end of the distribution. Similarly, Pb was shown to bind to the higher end molar mass components in compost leachate organic matter (Bolea et al., 2006). Addition of a mixture containing 50 nM Cu, Cd and Pb (gray lines in Fig. 6a) increased the Cd signal, which corresponded to the higher end of LMM distribution, although it was in the lower range for stream water (Wu et al., 2004), suggesting some specificity of WWTP LMM fraction binding sites to cadmium. No increase in the ICP-MS signal and no changes in the LMM distribution of the colloidal fraction were found for Cu and Pb. Such behavior can be due to the complexation of these metals by the small organic ligands not totally separated from COM by TFF (Wilding et al., 2004), as detected by LC-OCD (data not shown), that are lost through the FFF channel during the fractionation. Among the measured trace metals, only Al, Fe and Pb were found to be associated with the HMM fractions (Fig. 6b). ICPMS signal of lead followed both 90 angle LS and UV signals, but Al and Fe elute within the lower end of molar mass of the HMM fraction. Pb and Al preferential binding to high molar mass colloids in natural water has previously been shown (Lyven et al., 2003; Dubascoux et al., 2008). Estimations of the distribution between the LMM and HMM fractions of the EfCOM, based on the AFlFFF-ICP-MS signal, show that iron was preferentially found within the LMM fraction (DB, 94%; HW, 68%; ZH, 90%), whereas an intermediate situation occurred for aluminum with 64%, 54% and 53%, for DB, HW and ZH, respectively, associated with the LMM fraction. To the contrary, lead was predominantly associated with the HMM fraction, representing 65%, 70% and 70% for DB, HW and
water research 44 (2010) 340–350
347
Fig. 5 – Metals distribution between particulate (black bar), colloidal (gray bar) and truly dissolved (white bar) phases and colloidal metal to colloidal organic carbon concentration ratios obtained after mass balance of results obtained from TFF procedure for DB, HW and ZH WWTP effluents sampled in 2007 and 2008 and in VD lake water.
Fig. 6 – Metals associated with LMM (a,c) and HMM (b,d) fractions of the EfCOM. UV absorbance (UV), light scattered at 908 (LS 908) fractograms and metals’ relative signals obtained with AFlFFF-UV-MALS-ICP-MS for Duebendorf colloidal isolate LMM (a) and HMM (b) fractions (Huber and Frimmel, 1994). Analyses were done before (black lines) or after (gray lines) a simultaneous spike of 50 nM Cd, Cu and Pb. Similar distributions were obtained for HW, ZH and VD colloidal isolates. Relationships between total ICP-MS Al (,), Fe () and Pb (;) signals and (c) humic-like fraction absorbance intensities for LMM components or (d) biopolymers concentrations for HMM components of the three WWTP colloidal isolates. Both UV absorbance intensities and biopolymers concentrations were obtained from LC-OCD analysis.
348
water research 44 (2010) 340–350
ZH colloidal isolates. Furthermore, as is shown for Al, Fe and Pb (Fig. 6c), ICP-MS signal intensities of the different metals P obtained for the LMM fraction ( (hi)LMM) were well correlated with UV absorbance of humic-like substances obtained from LC-OCD, but not directly with their concentrations measured by this technique. The quantities of iron associated with HMM P components ( (hi)HMM Fe) are found to be relatively constant among the three WWTP isolates and independent of the concentration of biopolymers found by LC-OCD (Fig. 6d). For Al or Pb, an inverse correlation between biopolymer concentraP tions and total signal ( (hi)HMM) intensity was obtained. Considering that LMM components are of humic-like nature, while the HMM held more biopolymers, difference in affinity for trace metals must be observed (Lamelas et al., 2005). The results obtained herein suggest that LMM and HMM components compete for the binding of Pb and Al, with the LMM fraction having a higher affinity for metals than the HMM fraction.
4.
Implications for the receiving ecosystems
The obtained results of the characterization of colloidal organic matter and its role in metal partitioning have important implications in improving our understanding of the fate of EfCOM and associated trace metals in the receiving ecosystems. The HMM fraction of effluent colloidal isolates is expected to degrade within the receiving compartment, due to their polysaccharide or proteinaceous nature. By contrast, metal associated with the LMM fraction of the EfCOM could be transported for larger distances in the receiving water. Differences in the chemical composition and hydrophobicity of the WWTP humic-like substances and pedogenic HS will also result in different adsorption behavior to the surfaces. Effluent colloidal organic matter is more hydrophilic than naturally occurring HS of pedogenic origin. This will decrease the adsorption of EfCOM and associated pollutants to the surfaces and consequently will increase their mobility within and between different environmental compartments (e.g. river water, sediment, soil) despite their similar sizes/ molar mass distributions.
5.
Conclusions
Better understanding of the role of colloidal organic matter from wastewater treatment plants in metal distribution was provided by using tangential flow ultrafiltration, liquid chromatography coupled with organic carbon and UV detectors, and an asymmetrical flow field-flow fractionation (AFlFFF) multidetection platform. Liquid chromatography, coupled with organic carbon and UV detectors, and fluorescence demonstrated that the humic-like fraction of low aromaticity was the main colloidal component, whereas the biopolymers were present in much lower proportions. Asymmetrical flow field-flow fractionation multidetection platform showed that the low molar mass fractions (1600 Da < Mw < 2600 Da) dominated the colloidal isolates. However, three major populations were also detected in the wide range of molar mass from 20 kDa to more than 500 kDa.
AFlFFF coupled on-line with ICP-MS revealed that the low molar mass fraction of the colloidal pool controls the distribution of metals in the colloidal phase, although the high molar mass fractions are important for Al and Pb binding, and probably their dispersion, in receiving ecosystems.
Acknowledgements The authors gratefully acknowledge the financial support provided by Swiss National Science Foundation project PP002102640, COST action 636 Xenobiotics in urban water cycle and PAI ‘‘Germaine de Stael’’. Warm thanks are extended to Prof. H. A. Lashuel (EPFL) for providing access to the fluorescence spectrometer, and David Kistler (Eawag) for his valuable help in sample preparation and ICP-MS measurement.
references
Baalousha, M., Kammer, F.V.D., Motelica-Heino, M., Baborowski, M., Hofmeister, C., Le Coustumer, P., 2006. Sizebased speciation of natural colloidal particles by flow field flow fractionation, inductively coupled plasma-mass spectroscopy, and transmission electron microscopy/X-ray energy dispersive spectroscopy: colloids–trace element interaction. Environmental Science & Technology 40, 2156–2162. Baker, A., 2001. Fluorescence excitation-emission matrix characterization of some sewage-impacted rivers. Environmental Science & Technology 35, 948–953. Baker, A., Inverarity, R., Charlton, M., Richmond, S., 2003. Detecting river pollution using fluorescence spectrophotometry: case studies from the Ouseburn, NE England. Environmental Pollution 124, 57–70. Beckett, R., Jue, Z., Giddings, J.C., 1987. Determination of molecular-weight distributions of fulvic and humic acids using flow field-flow fractionation. Environmental Science & Technology 21, 289–295. Bolea, E., Gorriz, M.P., Bouby, M., Laborda, F., Castillo, J.R., Geckeis, H., 2006. Multielement characterization of metalhumic substances complexation by size exclusion chromatography, asymmetrical flow field-flow fractionation, ultrafiltration and inductively coupled plasma-mass spectrometry detection: a comparative approach. Journal of Chromatography A 1129, 236–246. Buffle, J., Wilkinson, K.J., Stoll, S., Filella, M., Zhang, J.W., 1998. A generalized description of aquatic colloidal interactions: the three-colloidal component approach. Environmental Science & Technology 32, 2887–2899. Cabaniss, S.E., Zhou, Q.H., Maurice, P.A., Chin, Y.P., Aiken, G.R., 2000. A log-normal distribution model for the molecular weight of aquatic fulvic acids. Environmental Science & Technology 34, 1103–1109. Chen, W., Westerhoff, P., Leenheer, J.A., Booksh, K., 2003. Fluorescence excitation-emission matrix regional integration to quantify spectra for dissolved organic matter. Environmental Science & Technology 37, 5701–5710. Comte, S., Guibaud, G., Baudu, M., 2007. Effect of extraction method on EPS from activated sludge: an HPSEC investigation. Journal of Hazardous Materials 140, 129–137. Dignac, M.F., Ginestet, P., Rybacki, D., Bruchet, A., Urbain, V., Scribe, P., 2000. Fate of wastewater organic pollution during activated sludge treatment: nature of residual organic matter. Water Research 34, 4185–4194.
water research 44 (2010) 340–350
Dubascoux, S., Le Hecho, I., Gautier, M.P., Lespes, G., 2008. On-line and off-line quantification of trace elements associated to colloids by AFlFFF and ICP-MS. Talanta 77, 60–65. Garnier, C., Gorner, T., Lartiges Bruno, S., Abdelouhab, S., de Donato, P., 2005. Characterization of activated sludge exopolymers from various origins: a combined size-exclusion chromatography and infrared microscopy study. Water Research 39, 3044–3054. Gustafsson, O., Gschwend, P.M., 1997. Aquatic colloids: concepts, definitions, and current challenges. Limnology and Oceanography 42, 519–528. Hassellov, M., Lyven, B., Haraldsson, C., Sirinawin, W., 1999. Determination of continuous size and trace element distribution of colloidal material in natural water by on-line coupling of flow field-flow fractionation with ICPMS. Analytical Chemistry 71, 3497–3502. Huber, S.A., Frimmel, F.H., 1992. A new method for the characterization of organic carbon in aquatic systems. International Journal of Environmental Analytical Chemistry 49, 49–57. Huber, S.A., Frimmel, F.H., 1994. Direct gel chromatographic characterization and quantification of marine dissolved organic-carbon using high-sensitivity DOC detection. Environmental Science & Technology 28, 1194–1197. Imai, A., Fukushima, T., Matsushige, K., Kim, Y.H., Choi, K., 2002. Characterization of dissolved organic matter in effluents from wastewater treatment plants. Water Research 36, 859–870. Lamelas, C., Wilkinson, K.J., Slaveykova, V.I., 2005. Influence of the composition of natural organic matter on Pb bioavailability to microalgae. Environmental Science & Technology 39, 6109–6116. Lead, J.R., Wilkinson, K.J., 2006. Aquatic colloids and nanoparticles: current knowledge and future trends. Environmental Chemistry 3, 159–171. Leenheer, J.A., Dotson, A., Westerhoff, P., 2007. Dissolved organic nitrogen fractionation. Annals of Environmental Science 1, 45–56. Liu, R., Lead, J.R., Baker, A., 2007. Fluorescence characterization of cross flow ultrafiltration derived freshwater colloidal and dissolved organic matter. Chemosphere 68, 1304–1311. Lyven, B., Hassellov, M., Turner, D.R., Haraldsson, C., Andersson, K., 2003. Competition between iron- and carbon-based colloidal carriers for trace metals in a freshwater assessed using flow field-flow fractionation coupled to ICPMS. Geochimica et Cosmochimica Acta 67, 3791–3802. McKnight, D.M., Boyer, E.W., Westerhoff, P.K., Doran, P.T., Kulbe, T., Andersen, D.T., 2001. Spectrofluorometric characterization of dissolved organic matter for indication of precursor organic material and aromaticity. Limnology and Oceanography 46, 38–48. Pedrot, M., Dia, A., Davranche, M., Bouhnik-Le Coz, M., Henin, O., Gruau, G., 2008. Insights into colloid-mediated trace element release at the soil/water interface. Journal of Colloid and Interface Science 325, 187–197. Perminova, I.V., Frimmel, F.H., Kudryavtsev, A.V., Kulikova, N.A., Abbt-Braun, G., Hesse, S., Petrosyan, V.S., 2003. Molecular weight characteristics of humic substances from different environments as determined by size exclusion chromatography and their statistical evaluation. Environmental Science & Technology 37, 2477–2485. Pernet-Coudrier, B., Clouzot, L., Varrault, G., TusseauVuillemin, M.H., Verger, A., Mouchel, J.M., 2008. Dissolved organic matter from treated effluent of a major wastewater
349
treatment plant: characterization and influence on copper toxicity. Chemosphere 73, 593–599. Pote, J., Haller, L., Loizeau, J.L., Bravo, A.G., Sastre, V., Wildi, W., 2008. Effects of a sewage treatment plant outlet pipe extension on the distribution of contaminants in the sediments of the Bay of Vidy, Lake Geneva, Switzerland. Bioresource Technology 99, 7122–7131. Prestel, H., Schott, L., Niessner, R., Panne, U., 2005. Characterization of sewage plant hydrocolloids using asymmetrical flow field-flow fractionation and ICP-mass spectrometry. Water Research 39, 3541–3552. Ren, J., Packman, A.I., 2004. Modelling of simultaneous exchange of colloids and sorbing contaminants between streams and streambeds. Environmental Science & Technology 38, 2901–2911. Ren, J., Packman, A.I., 2005. Coupled stream-subsurface exchange of colloidal hematite and dissolved zinc, copper, and phosphate. Environmental Science & Technology 39, 6387–6394. Reszat, T.N., Hendry, M.J., 2005. Characterizing dissolved organic carbon using asymmetrical flow field-flow fractionation with on-line UV and DOC detection. Analytical Chemistry 77, 4194–4200. Saadi, I., Borisover, M., Armon, R., Laor, Y., 2006. Monitoring of effluent DOM biodegradation using fluorescence, UV and DOC measurements. Chemosphere 63, 530–539. Schafer, T., Artinger, R., Dardenne, K., Bauer, A., Schuessler, W., Kim, J.I., 2003. Colloid-borne americium migration in Gorleben groundwater: significance of iron secondary phase transformation. Environmental Science & Technology 37, 1528–1534. Schmitt, D., Muller, M.B., Frimmel, F.H., 2001. Metal distribution in different size fractions of natural organic matter. Acta Hydrochimica et Hydrobiologica 28, 400–410. Shon, H.K., Vigneswaran, S., Kim, I.S., Cho, J., Ngo, H.H., 2004. The effect of pretreatment to ultrafiltration of biologically treated sewage effluent: a detailed effluent organic matter (EfOM) characterization. Water Research 38, 1933–1939. Shon, H.K., Kim, S.H., Erdei, L., Vigneswaran, S., 2006a. Analytical methods of size distribution for organic matter in water and wastewater. Korean Journal of Chemical Engineering 23, 581–591. Shon, H.K., Vigneswaran, S., Snyder, S.A., 2006b. Effluent organic matter (EfOM) in wastewater: constituents, effects, and treatment. Critical Reviews in Environmental Science and Technology 36, 327–374. Stolpe, B., Hassellov, M., 2007. Changes in size distribution of fresh water nanoscale colloidal matter and associated elements on mixing with seawater. Geochimica et Cosmochimica Acta 71, 3292–3301. Stolpe, B., Hassellov, M., Andersson, K., Turner, D.R., 2005. High resolution ICPMS as an on-line detector for flow field-flow fractionation; multi-element determination of colloidal size distributions in a natural water sample. Analytica Chimica Acta 535, 109–121. Thang, N.M., Geckeis, H., Kim, J.I., Beck, H.P., 2001. Application of the flow field flow fractionation (FFFF) to the characterization of aquatic humic colloids: evaluation and optimization of the method. Colloids and Surfaces A 181, 289–301. Town, R.M., Filella, M., 2002. Size fractionation of trace metal species in freshwaters: implications for understanding their behaviour and fate. Metal size fractionation and speciation. Reviews in Environmental Science and Biotechology 1, 277–297. Warren, L.A., Haack, E.A., 2001. Biogeochemical controls on metal behaviour in freshwater environments. Earth-Science Reviews 54, 261–320.
350
water research 44 (2010) 340–350
Weng, L., Fest, E.P., Fillius, J., Temminghoff, E.J., van Riemsdijk, W.H., 2002. Transport of humic and fulvic acids in relation to metal mobility in a copper-contaminated acid sandy soil. Environmental Science & Technology 36, 1699–1704. Westerhoff, P., Chen, W., Esparza, M., 2001. Fluorescence analysis of a standard fulvic acid and tertiary treated wastewater. Journal of Environmental Quality 30, 2037–2046.
Wilding, A., Liu, R., Zhou, J.L., 2004. Validation of cross-flow ultrafiltration for sampling of colloidal particles from aquatic systems. Journal of Colloid and Interface Science 280, 102–112. Wu, F.C., Evans, D., Dillon, P., Schiff, S., 2004. Molecular size distribution characteristics of the metal-DOM complexes in stream waters by high-performance size-exclusion chromatography (HPSEC) and high-resolution inductively coupled plasma mass spectrometry (ICP-MS). Journal of Analytical Atomic Spectrometry 19, 979–983.