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
<10 gCl m3 10–20 gCl1 m3 >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] <10 10–30 30–150 >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 (<20 C) and process combination with chemical methods (Oehmen et al., 2007; Lopez-Vazquez et al., 2008).
3.2.3.2. Nitrogen removal. There is more literature on biological nitrogen removal than phosphorus removal. Of the four case studies presented in Table 2 that involved nutrients removal, three included biological nitrogen removal in the studies and only one investigated both biological nitrogen and phosphorus removal. This may be explained as follows. Biological systems have essentially replaced the physical/ chemical systems and become the process for nitrogen removal in used water treatment; whereas chemical systems
remain attractive for phosphorus removal due to its simplicity of operation and ease of implementation (Task Force, 1998). The nitrification process comprises two steps involving the oxidisation of reduced nitrogen compounds (e.g. ammonia) to nitrite and then to nitrate. Consequently, the nitrifiers can be divided into two groups of microorganisms as the ammonia oxidisers (also known as the nitroso-bacteria and exemplified by the nitrosomonas genus) and as the nitrite oxidisers (also known as the nitro-bacteria and exemplified by the nitrobacter genus). Nitrification is a sensitive process. The nitrifiers have slow growth rates, and are vulnerable to environmental conditions such as temperature, DO concentration, pH and a number of inhibitory compounds including ammonia and salt (Seviour and Blackall, 1999; Moussa et al., 2006; Campos et al., 2007). Nitrification is the more critical process in the entire scheme of biological nitrogen removal, and it has been pointed out that the cause of failure for nitrogen removal is often due to poor nitrification (Campos et al., 2007). There are two opposing views in the literature as to which group of nitrifiers is more susceptible to increasing salt level. The first view holds that the nitrite oxidisers are more susceptible than the ammonia oxidisers to salt effects, based on the accumulation of nitrite in some studies (Campos et al., 2002). The second view is that the nitrite oxidisers are less affected by salt stress than the ammonia oxidisers, at least at the lower salt levels (Moussa et al., 2006; Sharrer et al., 2007). It was explained that the accumulation of nitrite might be attributed to DO limitation, phosphorus limitation and/or the presence of toxicants rather than the salt stress itself. Elevated salt levels could however exacerbate the limited availability of oxygen (Section 3.1.1). There could also be some basis to the second view. Firstly, the ammonia oxidisers are known to be obligate chemoautotrophic aerobes, whereas some strains of the nitrite oxidisers can grow as both autotrophs and heterotrophs (Seviour and Blackall, 1999). The wider range of metabolic pathways may imply that the nitrite oxidisers would be more resilient to environmental stresses than the ammonia oxidisers. Secondly, it is known that the low DO inhibition effect has been greater for Nitrobacter – a nitrite oxidiser, than for Nitrosomonas – an ammonia oxidiser (Tchobanoglous et al., 2004). As nitrification is part of the marine nitrogen cycle, it is accepted that nitrification can take place in an elevated salt environment as high as that of the seawater (Yu et al., 2002; Yoshikawa et al., 2005). In the field of used water treatment, however, different levels of salt tolerance for the nitrifiers have been reported (Lefebvre and Moletta, 2006). A study on the bioenergetic aspects of halophilism suggested that there could be a salinity limit for nitrification, as the nitrifiers were yet to be found at salinities above 100 to 150 g L1. This may be explained by bioenergetic constraints, whereby the energy obtainable from the nitrification biochemical reaction is not able to make up for the high energetic cost of adapting the intracellular environment to the outside salinity (Oren, 1999). In another study, it was pointed out that although there are nitrifiers that are halotolerant or even halophilic, their growth optimum with respect to salt concentration could be at around the concentration of the marine environment up to
water research 44 (2010) 21–40
about 700 mM or 40 g L1 NaCl (Koops and PommereningRoser, 2001). In our context of HRMBR systems, this could imply a practical upper limit on CF of 40–100 for the occurrence of nitrification, depending on the salt content of the influent used water. In line with the above findings, a number of studies involving the acclimation of conventional activated sludge biomass to an increasing saline environment suggested that the nitrifiers could adapt to salt environment up to around 30 g L1 for used water application (Oren, 2002b; Uygur and Kargi, 2004). However, it seemed that severe inhibition would occur, when the salt level increased to around 40 g L1 and beyond (Rothschild and Mancinelli, 2001). Addition of salttolerant culture (halobacter halobium) to the biomass was able to improve nitrification with a higher ammonia removal rate than the original biomass, but the trend of reducing removal rate with increasing salt concentration up to 60 g L1 could not be averted (Uygur, 2006). Another observation to be made is that in the course of acclimation, the effect of an increasing saline environment on the nitrifiers is more pronounced at the lower salt concentrations up to around 10 g L1. In this concentration range, it was reported that the magnitude of loss of nitrification activity was larger (Panswad and Anan, 1999). This effect is of significance to real operations, as it means that a significantly longer acclimation period and start-up will be required before steady-state nitrification and hence nitrogen removal may be achieved again (Sharrer et al., 2007). This finding is corroborated by other studies, which attributed it to a transition in the microbial population (Intrasungkha et al., 1999; Rothschild and Mancinelli, 2001). Possible strategies to enhance the nitrification process include the incorporation of suspended support media in the aeration tank for the attached growth of the nitrifiers as biofilm (GarzonZuniga and GonzalezMartinez, 1996; Artiga et al., 2008), and operation of the bioreactor at longer SRT (Rene et al., 2008). Both strategies work at overcoming the slower growth rate of the nitrifiers. However, longer SRT increases CF (Eq. (1)) and raises the salt concentration, so a ‘trade off’ is needed. Denitrification is normally not the limiting process in the entire scheme of nitrogen removal in an elevated salt environment. The prevailing observation that denitrifiers are not as inhibited as the nitrifiers by increasing salinity was presented in a number of studies (Glass and Silverstein, 1999; Sharrer et al., 2007). There is no known bioenergetic constraint for denitrifiers in saline environment, and it was estimated that denitrification could occur at or close to NaCl saturation up to the concentration of 300 g L1 (Oren, 1999). This may be explained by the phenomenon that a wide taxonomic range of bacteria can denitrify (Seviour and Blackall, 1999). The denitrifiers also include a number of halophilic and halo-tolerant species (Glass and Silverstein, 1999). A recent review paper by McAdam and Judd (2008) included discussion on nitrate removal from saline ion-exchange brine using biological and MBR processes.
3.2.4. Biomass characteristics and biological operating conditions The biomass characteristics and biological operating conditions in an MBR system are important. They not only
31
determine the efficiency of the biological treatment process, but also influence the performance of the membrane separation (Judd, 2006) (see Section 3.3.2). It is therefore important to consider the effect of salt on the biomass characteristics and their relation to the operating conditions such as the organic loading rate, the F/M ratio, the SRT and the HRT. Measuring and analysing biomass characteristics under elevated salt condition is not without difficulty. This may be attributed to the lack of reliable and simple methods of measurement (Henze, 1991), and the variability in parameter estimates often encountered in biological experiments (Grady et al., 1996). The difficulty is further compounded by the presence of salt, which could potentially create interferences to some measurement methods such as the chloride ion interference in the measurement of the COD (Sawyer et al., 2003), and the possible crystallisation of dissolved solids in the measurement of the suspended solids (SS) (APHA et al., 1998). Caution needs to be exercised when analysing samples with high salt content. Other analytical methods used for elevated salt studies include total organic carbon (TOC) measurement (Walker and Clifford, 2000), gas chromatography/ mass spectrometry (GC/MS) techniques and respirometry methods (Hamoda and Alattar, 1993; Tellez et al., 1995; Dan et al., 2003). Scanning electron microscopy (SEM), polymerase chain reaction followed by denaturing gradient gel electrophoresis (PCR– DGGE) and fluorescence in situ hybridisation (FISH) are also techniques that are used in biological studies under elevated salt conditions (Chen et al., 2003; Choi et al., 2005, 2007; Figueroa et al., 2008). Studies have shown that increasing salinity of the used water affects the microbial community in it. Microscopic studies have shown that the microbial morphology and the dominant species of the population change with increasing salt concentration (Ng et al., 2005). In addition, higher microorganisms such as the protozoa and rotifiers, which are normally present in freshwater used water treatment systems and graze on the dispersed microorganisms and the microbial flocs, tend to be absent when the salt concentration is increased beyond 10 g L1. FISH studies on the nitrifying microorganisms have also confirmed that there can be changes to the microbial community when subjected to increasing salt concentration (Chen et al., 2003; Moussa et al., 2006). This means that, within the context of this study, the application of HRMBR systems, the salt accumulation would cause changes to the microbial community and that, as a result of salinity selection, only microorganisms with adequate salt tolerance could adapt and thrive in an elevated salt environment. At this point, it may be interesting to note that a study using PCR–DGGE technique did not observe significant changes to the microbial diversity between the conventional MF-MBR and the dense membrane NF-MBR (Choi et al., 2007). However, the study was conducted at relatively low salinities: 1.68 0.33 mS/cm (equivalent to about 0.9 g L1 NaCl) for the NF-MBR and 1.11 0.35 mS/cm (equivalent to about 0.6 g L1 NaCl) for the MF-MBR, and this would preclude any straightforward conclusion to be drawn. The foregoing discussion indicates the importance of salt tolerance to the biological treatment of used water under elevated salt condition. For the salt range of interest in this study, it appears possible to build up salt tolerance and
32
water research 44 (2010) 21–40
acclimatise activated sludge cultures from a fresh water environment to a saline environment (Section 3.2.2). However, the process of acclimation takes place gradually over a relatively long time (several weeks to months) (Sharrer et al., 2007; Rene et al., 2008). An acclimated microbial system also tends to be sensitive to changes in the ionic strength of the used water. Shock loads of salt or rapid fluctuations in salt concentration would reduce system performance . It is therefore important to operate an acclimated microbial system at relatively constant salt concentration. In practical terms, this may indicate the need for an equalisation step for influent with fluctuating salt concentrations. Microbial kinetics including stoichiometry and mass balances are important for the mechanistic understanding of the biological treatment. This understanding could derive practical benefits, when it is translated into effective modelling that can be applied in the design and operation of biological systems (Henze et al., 2000), and lead to further insights into the scientific fundamentals of the process. At present, modelling of conventional biological used water treatment systems has progressed to a mature stage and typical input data for non-saline used water are available (Gujer, 2005). However, biological fundamental and kinetics data for saline used water at higher salt concentrations are comparatively limited. As in the case of industrial used waters (Orhon et al., 1999), many studies conducted under elevated salt condition have tended to adopt a ‘‘black box’’ approach that focuses on general performance appraisal based upon a few parameters and input-output functions. Although there are reports on biological kinetic data under elevated salt conditions, these are usually case specific, and depend on the type of substrate and microorganisms, and different salt and experimental conditions (Tellez et al., 1995; Peyton et al., 2002). In this regard, there is a need for more effort in the modelling of biological used water treatment systems under elevated salt condition with the objective of gaining a better understanding of it. In spite of the knowledge gaps in the literature on the biological treatment of used water under elevated salt conditions, there are some general observations. For example, increasing salt concentration tends to result in slower microbial kinetics, lower growth yield coefficient (Y ) and higher endogenous decay or death rate coefficient (kd) (Woolard and Irvine, 1995; Kargi et al., 2000). Consequently, a biological treatment process in an elevated salt environment will need to be operated at lower food to microorganisms (F/M ) ratios. As the F/M ratio is directly dependent on the organic loading rate (OLR) and inversely proportional to the concentration of the microorganisms (X ), this implies that increasing salt concentration requires smaller OLR or larger X as shown in Eq. (8). This conclusion was substantiated experimentally on a rotating biodisc contactor (RBC) system, where it was found that the system’s performance improved with increasing disc surface area or number of discs, but reduced with increasing COD loading rate (Kargi and Dincer, 1999; Dinc¸er and Kargi, 2001). F=M ¼
QS0 OLR ¼ VX X
(8)
The solids retention time (SRT) is another important operating parameter for biological processes (Rittmann and McCarty, 2001).
The SRT is a measure of the time that the microorganisms reside in the bioreactor system, and fundamentally related to the specific growth rate (m) of the microorganisms and other process parameters as shown in Eq. (9). In an elevated salt environment with smaller Y and higher kd, the SRT would need to be adequately long in order to achieve lower F/M and higher removal efficiency (E ) of the organic matter. m¼
1 E ¼ YðF=MÞ kd SRT 100
(9)
The above discussion seems to suggest that it could be appropriate and possibly advantageous to use MBRs to treat saline used water. With membrane separation, the biomass can be effectively retained in a MBR, thereby achieving larger X. A MBR system would also be able to operate at longer SRT and lower F/M ratios (Stephenson et al., 2000), and thus would be suitable for treating saline used water. It is to be pointed out that conventional activated sludge process typically applies SRT between 4 and 14 d, and F/M ratio between 0.25 and 0.5 (Rittmann and McCarty, 2001), whereas conventional MBR process may operate with SRT from 10 to 30 d or longer, and F/M ratio <0.12 (Judd, 2006). It is to be further noted that for adequate nitrification to take place, SRT would need to be reasonably long, possibly around 10 d or longer (Huang et al., 2001). For snapshot references, the examples listed in Table 2 give SRT ranging from 15 to 73 d, and F/M ratios between 0.021 and 0.48. In general, a longer SRT of the MBR could bring about benefits to the biological process. These benefits include lower sludge production for disposal and enabling slow-growing and specialised microorganisms such as the nitrifiers to establish in the bioreactor. In addition, longer SRT could serve in the removal of slow-to-degrade organics and persistent organic pollutants (Chen et al., 2003; Dan et al., 2003). However, it is to be pointed out that excessively long SRT could also lead to undesirable effects in the MBR, such as excessively high MLSS and increased production of extracellular polymeric substances (EPS) due to higher level of endogenous metabolism and cell lysis. These effects would result in inefficient oxygen transfer, increased viscosity and aggravated fouling on the membrane (Wu, 2007). As an enhancement to the biological treatment, a possible strategy would be to incorporate attached-growth (biofilm) biomass, either as fixed-bed or moving-bed, into the suspended-growth based MBR process (Ng, 2006; Leiknes and Ødegaard, 2007). The dual system biological treatment would be able to optimise performance by exploiting the competitive advantages of different groups of microorganisms with individual specifically conducive environments. This strategy has shown to work and achieved high removal efficiency on used water from a fish canning factory in the example given in Table 2 (Artiga et al., 2008). Another point to note is that the SRT would not be an independent variable in an HRMBR system, as salt accumulation takes place. Rather, the SRT is coupled with the HRT in the CF as given in Eq. (1). When the HRT is constrained by the system dimension, increasing SRT will invariably result in higher CF. This means that inert materials, as well as organic and inorganic substances will be accumulated in the bioreactor. The accumulation of organic substances would exert
water research 44 (2010) 21–40
negative effects on both the biological process and the permeability of the membrane (Huang et al., 2000) as detailed in Section 3.3. The accumulation of the inorganic substances further increases the salt concentration in the bioreactor. This would not only further exacerbate the mentioned negative impacts, but also add on other physicochemical and biological impacts as previously discussed. The SRT, HRT and their ratio (CF) are therefore important parameters to be optimised in an HRMBR system under elevated salt conditions (Section 3.4).
3.3.
Membrane aspects
3.3.1. Driving force, concentration polarisation, flux and product quality Salinity will also impact the driving force, concentration polarisation, flux and product quality of an HRMBR system. As in all membrane separation processes, where parts of the constituents selectively permeate, and the rest are retained, concentration polarisation arises (Baker, 2004). When the retained constituents become adhered to the membrane and form a cake layer due to adsorption, precipitation or other interactions, and are not back-transported to the bulk solution, this effect is known as fouling (Section 3.3.2). Whereas the effect of concentration polarisation (CP) depends on filtration and dissipates when the water transport stops, the effect of fouling does not. Fouling and concentration polarisation tend to be mutually exacerbating. The larger the CP, the more intense would be the fouling, because the constituents at higher concentration are able to interact and deposit more strongly on to the membrane. The key equations for membrane processes are shown in Table 5. Eq. (10) is the well-known film model and shows how concentration polarisation (CP) depends on water flux (Jw) and the boundary layer mass transfer (k ¼ D/d). For inorganic constituents, CP is typically <2, but for larger and less diffusive molecules, CP could be >5 (Fane, 2005). It should further be noted that CP can be greatly aggravated by the fouling cake due to the hindered back diffusion of the retained solutes – k can be greatly reduced as given in Eq. (11), which is known as
33
cake enhanced concentration polarisation (CECP) or cake enhanced osmotic pressure (CEOP) (Hoek and Elimelech, 2003; Chong et al., 2008). From the above relationships, the effects of salt on the membrane process become apparent. First, a higher salt concentration of the mixed liquor increases the osmotic pressure (Dp [) as predicted by the Van’t Hoff equation (2). This means that a larger driving force (Dp [) is necessary in order to operate the process. Next, the higher salt concentration will also increase the viscosity (h [) (Section 3.1.2) and the total resistance across the membrane due to aggravated fouling (Rf [) (Section 3.3.2). In turn, the fouling cake layer intensifies CP due to the CECP effect (CP [). For illustration: for Jw ¼ 20 L/m2/h, k ¼ 8 105 m/s, d ¼ 20 mm, dcake ¼ 100 mm, e ¼ 0.3 s ¼ 3, CPCECP/CP w 30 as predicted by Eq. (12). This will cause the membrane flux (Jw) to reduce or a higher pressure to be applied, as derived from Eq. (13). Concomitantly, the salt flux (Js [) will increase with the higher wall concentration and impair the product quality as predicted by Eq. (14). In short, a higher salt concentration in the mixed liquor is expected to reduce the effective driving force, increase the hydraulic resistances, reduce the water flux, and impair the product water quality. Consequently, a saline environment may imply that lower sustainable fluxes are attainable compared to nonsaline environment. These effects have practical implications, and more research will be necessary to confirm and elucidate the various issues involved. The foregoing discussion uses the pressure-driven NF-MBR system (Fig. 3 - adapted from Chong (2007)) as a reference, and establishes the detrimental effect of elevated salt concentration on membrane processes. Although the degree of the salt effect may vary from one HRMBR process to the other, higher salt concentration also affects the other high rejection membrane processes, and is briefly outlined as follows. For the OMBR, an increase in the salt concentration of the feed will reduce the effective driving force and the water flux in a similar manner. Theoretically, this could be overcome by increasing the driving force with a more concentrated draw
Table 5 – Governing equations for a pressure-driven NFMBR system
CP ¼
k ¼
cW cp Jw ¼ exp cb cp k
D3 s dcake
CPCECP Jw k ¼ exp 1 CP k k Jw ¼
Dp CP Dp h Rm þ Rf
JS ¼ B ðcw cP Þ
(10)
(11)
(12)
(13)
(14)
Fig. 3 – Transport phenomena in a pressure-driven NF-MBR process.
34
water research 44 (2010) 21–40
solution. However, a more concentrated draw solution would intensify the internal concentration polarisation (ICP) due to the logarithmic relationship between the draw concentration and the water flux (Gray et al., 2006), and may cause ‘‘osmotic deswelling’’ of the membrane (Mehta and Loeb, 1978), thereby reducing the process performance. Furthermore, if the draw solution is an inorganic salt solution such as the NaCl solution, the salt concentration gradient will be in the opposite direction to the driving force in an OMBR. In this instance, a more concentrated draw solution would result in a larger salt flux permeating from the draw side to the feed side, thereby enhancing the external concentration polarisation (ECP) and further diminishing the water flux. For MDBR, which is a thermally based membrane process where the separation is driven by the vapour pressure difference, the effect of the salt concentration on the feed side has a lesser impact than pressure-driven processes. However, the impact is still evident. Increasing salt concentration of the feed will reduce both the vapour pressure of the solvent (water) and the activity coefficient of the components in the feed. Consequently, the driving force would decrease, though rather modestly, and reduce the water flux (Schofield et al., 1987; Lei, 2005). At very high salt concentrations and under the influence of concentration polarisation, crystal formation as scaling may also occur on the membrane surface, resulting in further flux decline and membrane wetting (Tun et al., 2005).
3.3.2.
Membrane fouling
Fouling refers to the general phenomenon of loss of performance of a membrane due to the deposition of suspended or dissolved substances on its external surfaces, at its pore openings, or within its pores as defined by (Koros et al., 1996). Therefore, membrane fouling encompasses inorganic scaling, organic fouling, colloidal fouling as well as biofouling. Fouling has always been a central issue in all membrane processes used for water and used water applications due to its significance to the operation and economics of the process. The extent of fouling determines the sustainable flux that is achievable under real operation (Section 3.3.1). Fouling is all the more significant for a MBR system, because of the direct interactions between the membrane and the activated sludge components. Fouling in conventional MBR systems is primarily attributed to fouling that is biopolymeric in nature, and also fouling that is colloidal and particulate in nature. Fouling in conventional MBR systems is a complex subject that is still receiving a large number of ongoing research efforts (Le-Clech et al., 2006; Meng et al., 2009). In an HRMBR system, fouling could be more complex, due to the added impact of elevated salinity. In general, as is observed in practice, increased salinity aggravates fouling and requires more cleaning for a membrane process (Tam et al., 2006). For HRMBR systems, things are more complicated, because the salt level is directly determined by the CF, and this in turn is linked to the other process parameters (SRT and HRT). As a result of the complexity, it is difficult to quantify the likely additional impact of fouling due to salinity. However, a few observations can be made. Higher salt concentration would result in greater scaling propensity and aggravated colloidal fouling due to the higher supersaturation index and double layer compression
respectively (Section 3.1.3). Furthermore, higher salt levels would cause both soluble microbial products (SMP) and extracellular polymeric substances (EPS) in the biological system to increase (Reid et al., 2006). Both SMP and EPS are known to exert great impact on membrane fouling (Le-Clech et al., 2006), though it is to be pointed out that soluble EPS and SMP can be considered as identical (Laspidou and Rittmann, 2002). Membrane fouling decreases the achievable permeability by increasing the fouling resistance. In practical terms, this means that either the flux would reduce over time in the operation mode of constant driving force, or the driving force would need to be increased steadily over time to maintain constant flux. When the operating limit is reached, the fouled membrane would need to be chemically cleaned or even replaced, if the fouling is severe and the membrane life is spent. As in the case of conventional MBR systems, fouling would need to be mitigated and limited. For HRMBR systems, where higher fouling propensity under elevated salt conditions is expected, a number of options may be considered: The first option is to adopt a process that would have inherently low membrane fouling tendency. In this instance, it has been hypothesised that the OMBR may be one such process, as it does not entail the use of hydraulic pressure (Cath et al., 2006; Oo et al., 2008). However, the role of pressure, per se, rather than flux is a moot point at this stage. The second option is to mitigate fouling either by removing the foulants on the membrane or by optimising system parameters to alleviate fouling. While elaborate pretreatment methods are practised for the dense membrane processes (Tanninen et al., 2005), these would not be applicable to the MBR application. More over elementary sedimentation and / or chemical dosage of coagulants, flocculants or adsorbents to be discussed later, the choice of pretreatment for MBR systems is plainly limited. This means that no more than conventional fouling mitigation methods that are used in conventional MBR systems could be applied to HRMBR systems. Physical and chemical cleaning are used to remove foulants from the membrane. Current physical techniques include air scouring, back washing, relaxation, but these would need to be optimised for salt accumulating MBR systems. Furthermore, cleaning chemicals and protocols would need to be developed. It is also possible to optimise system parameters by improving the anti-fouling properties of membrane and optimising the operating parameters of the MBR system. The former would involve membrane modification by increasing its hydrophilicity or by precoating it with substances that would limit the contact between the foulants and the membrane. The latter would include optimising aeration, hydrodynamics, SRT and to operate at a sustainable flux to avoid excessive fouling. It should be further noted that the choice of flux is important; note the exponential influence on concentration polarisation. The effect on fouling of operating at a flux closer to 10 L m2 h1 rather than 20–30 L m2 h1 (more typical of conventional MBR systems, Section 3.3.1) would be very significant; for example, CP ¼ 2 at 10 L m2 h1 becomes CP ¼ 8 at 30 L m2 h1. To date, the reported fluxes of the innovative HRMBRs have been around or below 10 L m2 h1 (Choi et al., 2007; Cornelissen et al., 2008; Phattaranawik et al., 2008),
water research 44 (2010) 21–40
potentially providing fouling relief. The fact that these HRMBRs tend to operate at lower fluxes is compensated for by the improved permeate quality. The third option is to use strategic chemical methods that exploit the greater tendency of precipitation and coagulation in an elevated salt environment for process enhancement. The strategy is to chemically modify the characteristics of the mixed liquor to alleviate fouling (Judd, 2006). There has been a number of studies devoted to this area, and methods that have been successfully demonstrated include the addition of coagulants or flocculants (Itonaga et al., 2004; Wickramasinghe et al., 2004) such as ferric chloride (FeCl3) or aluminium sulphate (alum), and/or adsorbent agents such as powdered activated carbon (PAC) (Ng et al., 2006; Zhang et al., 2008). An added benefit of using PAC is that it could enable better biodegradation of toxic compounds in the biological system in addition to ameliorating membrane fouling (Lesage et al., 2005). It is also noted that coagulation could result in the concomitant removal of phosphorus (Song et al., 2008), and in a study whereby sodium aluminate (SAL) was used as the coagulant, reduced calcium carbonate scaling was observed (Katz and Dosoretz, 2008). Another approach for process enhancement uses what is known as ‘accelerated seeded precipitation’ (Sanciolo et al., 2008) or ‘accelerated precipitation softening’ (Kedem and Zalmon, 1997; Masarwa et al., 1997). This method has been applied successfully to increase the overall recovery of reverse osmosis (RO) from 83% to 95% (Gabelich et al., 2007). It works by increasing the pH of the feed to induce precipitation of the scaling compounds prior to the membrane separation. Hence, the scaling propensity of the membrane would reduce and the recovery would increase. It may be expected that the method would also work favourably for an HRMBR system at very high CF and TDS values (high SI) due to co-precipitation mechanisms and ionic effects. The described chemical methods could potentially enhance an HRMBR system with elevated salt environment. However, it is to be pointed out that this chemical method is not infallible. In the same study as above (Gabelich et al., 2007), the observed membrane fouling and scaling of silicates and CaSO4 after 2 months of operation indicated the importance of using both the appropriate antiscalant and maintaining consistent efficiency of the demineralisation process. In addition, there are other drawbacks in the form of higher chemical cost and the additional chemical waste sludge that needs disposal.
3.4.
Concentration factor
From the foregoing discussions, it is apparent that the concentration factor CF is an important operating parameter to be determined and optimised in salt accumulating HRMBR systems. CF is intrinsically linked to the other important process parameters, and has direct influence on all three aspects of physicochemistry, microbiology and membrane performance. Therefore, CF will play an integral role in the optimisation of HRMBR systems. Higher CF values are desirable to achieve higher water recovery (Eq. (1)). In the context of the HRMBR systems, the CF is directly related to the SRT:HRT ratio, such that at a given CF, the SRT is fixed by the HRT, and vice versa. When the CF is low
35
and at a given HRT, the resultant shorter SRT may not be sufficient to allow for adequate biological activities. Conversely, when the SRT is set to be longer, a commensurate longer HRT is necessary at the same CF, implying larger system volume and hence greater capital costs. The alternative would be to increase the CF to allow for longer SRT and shorter HRT. However, there could be an upper limit to CF, because increasing the CF increases the salt (TDS) concentration in the mixed liquor correspondingly, due to the retention effect of the high rejection membrane. The elevated salt concentration would have detrimental effects on the process as previously discussed. An optimum CF, where both reasonable water recovery and acceptable elevated salt condition needs to be identified, for a given application. As the HRMBR systems are in the early stage of development, there is limited literature discussing the role of CF and the resulting effect of salt elevation for such systems. As such, there is no reference value for CF, though the importance of CF was discussed by Phattaranawik et al. (2008) for the case of the MDBR. More study is required to establish the fundamental basis for optimising CF and elucidating the various effects associated with it. However, it is possible to apply rudimentary methods based on practical considerations to estimate the possible range of values for optimum CF. The first consideration is with regard to the application objective of the HRMBR systems. Considering that the HRMBR systems could be applied for the purpose of water reclamation to produce high quality water suitable for augmenting water supply, their viability would need to be compared against other water supply options. In most applications, there is scarcity of the natural water resource, so membrane-based seawater desalination processes would serve as a reasonable benchmark for comparison. It is known that the costs of water reclamation is generally about half of that of seawater desalination for membrane based processes (Coˆte´ et al., 2005). When this is related to the TDS concentration (higher TDS leads to higher energy and capital costs), then the appropriate TDS concentrations for the HRMBR systems to be practically viable would be limited to around 15 gL1 or lower. The second consideration is with regard to the relation of CF to the SRT:HRT ratio. For adequate biological activities, real MBR applications would require SRT ¼ 10–30 d (Section 3.2.4). Considering that HRT needs to be kept small to reduce capital outlay and is typically <1 d, then CF ¼ SRT/HRT would range between 10 and 30, which would correspond to the range of TDS concentration between 5 and 15 g L1 for the reference used water of about 500 mg L1 TDS (Table 1). At these salt concentrations, most biological activities (carbon and nitrogen removal) would still be functional by acclimation. However, biological phosphorus removal may be impaired (Section 3.2.3). The above preliminary assessment therefore suggests that the CF optimum could be between 10 and 30, corresponding to the range of TDS concentrations between 5 and 15 g L1. At these salt concentrations, the physicochemical impacts also appear moderate and may be manageable (Section 3.1). However, it is to be pointed out that CF considers the ratio of SRT to HRT, but not their actual magnitudes. For example, CF ¼ 20 might be considered reasonable, but when HRT is 2 h, then SRT < 2 d is clearly not sufficient. In this instance, for
36
water research 44 (2010) 21–40
CF ¼ 20: HRT ¼ 12 h and SRT ¼ 10 d may be acceptable. Therefore, appropriate SRT and HRT are to be taken into consideration in optimising CF. Finally, the presented preliminary assessment is based on practical, rather than scientific considerations. The assessment considers some general technical issues, and at the suggested range of values (CF ¼ 10–30, TDS ¼ 5–15 g L1, appropriate SRT and HRT), the HRMBR systems would probably provide adequate performance. However, the conditions depicting optimal system performance (maximum benefit and lowest cost) remains elusive, and needs further research for the range of HRMBR systems.
4.
Conclusions
This study has reviewed and elucidated the various salinity issues associated with the use of innovative HRMBR systems for treatment and reclamation of used water of domestic origin. Examples of such systems are the NFMBR, OMBR and MDBR. Compared to conventional MBR systems, the HRMBR systems would be able to produce higher quality product water and potentially possess comparative economical advantage in removing pollutants from the used water more effectively. However, as these high rejection systems retain dissolved solids, they will need to be operated under elevated salt condition. With increasing salt concentration, there will be challenges associated with the use of HRMBR systems. These challenges cover physicochemical, microbiological and membrane aspects, and are complex in nature due to their dynamic interaction. In this paper, the TDS level up to 50 g L1 is considered. With regard to the physicochemical aspects, the impact of elevated salt conditions is generally detrimental. Increasing salt concentration tends to yield unfavourable physicochemical parameters for oxygen transfer, density, turbidity, viscosity, salt precipitation, solute interaction and colloid chemistry. Considerable physicochemical challenges would be expected at the higher salt concentrations, but may be tolerable at lower salt concentrations. With regard to the biological aspects, increasing salt concentration changes the microbial community due to salinity selection. Acclimation and introduction of halotolerant or halophilic microorganisms are feasible strategies for an increasing saline environment. In general, biological carbon removal is not an issue for the salt range considered, but different strategies may be more appropriate for different salt concentrations. 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. However, biological nutrients removal could be more affected by the elevated salt conditions. While the denitrifiers would not be affected in the salt range considered, nitrifiers and the phosphorus accumulating organisms are sensitive to the increasing salt concentration. For used water applications, nitrification may be expected to function adequately up to around 30 g L1. However, acclimation for nitrification could be difficult and could take significantly longer periods even at the lower salt concentrations around 10 g L1. Biological phosphorus removal may be impaired by elevated salt concentration from around 5 g L1,
indicating that chemical methods are likely to be needed to remove phosphorus for higher salt concentrations. For biological operations, an elevated salt level could lead to slower microbial kinetics, lower growth yield and greater endogenous decay. The use of MBR systems to treat saline used water could therefore be appropriate, by operating at longer SRT and lower F/M ratio. Biological treatment may be improved by providing a dual system environment (attached growth and suspended growth) to exploit the competitive advantages of different groups of microorganisms. Prolonging SRT could be another method to improve biological treatment, but this would increase MLSS and the production of EPS, which result in inefficient oxygen transfer, increased viscosity and aggravated fouling on the membrane. The SRT is coupled to the HRT via CF. As the acclimated microbial system is sensitive to fluctuating salt concentration, it is important to operate at constant salt concentration. More research is needed to obtain biological fundamental and kinetics data, and for modelling of biological used water treatment system under elevated salt conditions with the objective of gaining a better understanding of it. With regard to the membrane aspects, increasing salt concentration would reduce the effective driving force, aggravate fouling, reduce the water flux, and impair the product water quality. As elaborate pretreatment methods are not available to MBR systems, the salinity promoted fouling could be a central operational issue for the membrane, and would diminish production by reducing the achievable sustainable flux. Some fouling alleviation and process enhancement strategies are possible, but these are not without cost. When chemicals are added, drawbacks in the form of the higher chemical cost and the additional chemical waste sludge that requires disposal need consideration. The concentration factor, CF, is an important operating parameter to be determined and optimised in the salt accumulating HRMBR systems. CF is intrinsically linked to the other important process parameters, and has direct influence on all three aspects of physicochemistry, microbiology and membrane. Based on practical considerations, a preliminary assessment for the treatment of used water of domestic origin indicates that the range for optimum CF could be between 10 and 30, corresponding to the TDS range between 5 and 15 g L1, and subject to appropriate SRT and HRT. Further study is required to establish the fundamental basis for optimising CF and elucidating the various effects associated with it.
Acknowledgement The authors acknowledge and thank the Environment and Water Industry Development Council (EWI) and the PUB of Singapore for support to the Singapore Membrane Technology Centre (SMTC) and for the PhD scholarship awarded to Winson C.L. Lay.
references
APHA, AWWA, WEF, 1998. Standard Methods for the Examination of Water and Wastewater. APHA.
water research 44 (2010) 21–40
Artiga, P., Garcı´a-Toriello, G., Me´ndez, R., Garrido, J.M., 2008. Use of a hybrid membrane bioreactor for the treatment of saline wastewater from a fish canning factory. Desalination 221 (1–3), 518–525. Baker, R.W., 2004. Membrane Technology and Applications. John Wiley & Sons, Ltd. Belfort, G., Davis, R.H., Zydney, A.L., 1994. The behavior of suspensions and macromolecular solutions in crossflow microfiltration. Journal of Membrane Science 96 (1–2), 1–58. Bird, R.B., Stewart, W.E., Lightfoot, E.N., 2007. Transport Phenomena. John Wiley & Sons, Inc. Campos, J.L., Garrido, J.M., Mosquera-Corral, A., Mendez, R., 2007. Stability of a nitrifying activated sludge reactor. Biochemical Engineering Journal 35 (1), 87–92. Campos, J.L., Mosquera-Corral, A., Sa´nchez, M., Me´ndez, R., Lema, J.M., 2002. Nitrification in saline wastewater with high ammonia concentration in an activated sludge unit. Water Research 36 (10), 2555–2560. Cath, T.Y., Childress, A.E., Elimelech, M., 2006. Forward osmosis: principles, applications, and recent developments. Journal of Membrane Science 281 (1–2), 70–87. Chen, G.H., Wong, M.T., Okabe, S., Watanabe, Y., 2003. Dynamic response of nitrifying activated sludge batch culture to increased chloride concentration. Water Research 37 (13), 3125–3135. Choi, J.H., Dockko, S., Fukushi, K., Yamamoto, K., 2002. A novel application of a submerged nanofiltration membrane bioreactor (NF MBR) for wastewater treatment. Desalination 146 (1–3), 413–420. Choi, J.H., Fukushi, K., Yamamoto, K., 2005. Comparison of treatment efficiency of submerged nanofiltration membrane bioreactors using cellulose triacetate and polyamide membrane. Water Science and Technology 51 (6, 7), 305–312. Choi, J.H., Lee, S.H., Fukushi, K., Yamamoto, K., 2007. Comparison of sludge characteristics and PCR–DGGE based microbial diversity of nanofiltration and microfiltration membrane bioreactors. Chemosphere 67 (8), 1543–1550. Chong, T.H., 2007. Reverse osmosis desalination and reclamation – control of colloidal and biofouling (PhD thesis), Singapore, Nanyang Technological University. Chong, T.H., Wong, F.S., Fane, A.G., 2008. Implications of critical flux and cake enhanced osmotic pressure (CEOP) on colloidal fouling in reverse osmosis: Experimental observations. Journal of Membrane Science 314 (1, 2), 101–111. Colt, J., 1984. Computation of dissolved gas concentrations in water as functions of temperature, salinity, and pressure. Special Publication by the American Fisheries Society. Cornelissen, E.R., Harmsen, D., de Korte, K.F., Ruiken, C.J., Qin, J.J., Oo, H., Wessels, L.P., 2008. Membrane fouling and process performance of forward osmosis membranes on activated sludge. Journal of Membrane Science 319 (1, 2), 158–168. Coˆte´, P., Siverns, S., Monti, S., 2005. Comparison of membranebased solutions for water reclamation and desalination. Desalination 182 (1-3), 251–257. Dan, N.P., Visvanathan, C., Basu, B., 2003. Comparative evaluation of yeast and bacterial treatment of high salinity wastewater based on biokinetic coefficients. Bioresource Technology 87 (1), 51–56. Dan, N.P., Visvanathan, C., Polprasert, C., Aim, R.B., 2002. High salinity wastewater treatment using yeast and bacterial membrane bioreactors. Water Science and Technology 46 (9), 201–209. Dinc¸er, A.R., Kargi, F., 2001. Performance of rotating biological disc system treating saline wastewater. Process Biochemistry 36 (8, 9), 901–906. Faibish, R.S., Elimelech, M., Cohen, Y., 1998. Effect of interparticle electrostatic double layer interactions on permeate flux decline in crossflow membrane filtration of colloidal
37
suspensions: an experimental investigation. Journal of Colloid and Interface Science 204 (1), 77–86. Fane, A.G., 2005. Module design and operation. In: Schaefer, A.I., Fane, A.G., Waite, T.D. (Eds.), Nanofiltration: Principles and Applications. Elsevier, pp. 67–88. Fane, A.G., Phattaranawik, J., Wong, F.S., 2005. Contaminated inflow treatment with Membrane Distillation Bioreactor (Patent WO/2006/137808), Singapore. Figueroa, M., Mosquera-Corral, A., Campos, J.L., Mendez, R., 2008. Treatment of saline wastewater in SBR aerobic granular reactors. Water Science and Technology 58 (2), 479–485. Gabelich, C.J., Williams, M.D., Rahardianto, A., Franklin, J.C., Cohen, Y., 2007. High-recovery reverse osmosis desalination using intermediate chemical demineralization. Journal of Membrane Science 301 (1, 2), 131–141. GarzonZuniga, M.A., GonzalezMartinez, S., 1996. Biological phosphate and nitrogen removal in a biofilm sequencing batch reactor. Water Science and Technology 34 (1, 2), 293–301. Gerday, C., Glansdorff, N., 2007. Physiology and Biochemistry of Extremophiles. ASM Press. Germain, E., Nelles, F., Drews, A., Pearce, P., Kraume, M., Reid, E., Judd, S.J., Stephenson, T., 2007. Biomass effects on oxygen transfer in membrane bioreactors. Water Research 41 (5), 1038–1044. Gillot, S., Capela, S., Heduit, A., 2000. Effect of horizontal flow on oxygen transfer in clean water and in clean water with surfactants. Water Research 34 (2), 678–683. Glass, C., Silverstein, J., 1999. Denitrification of high-nitrate, highsalinity wastewater. Water Research 33 (1), 223–229. Gloede, M., Melin, T., 2008. Physical aspects of membrane scaling. Desalination 224 (1–3), 71–75. Grady, C.P.L., Smets, B.F., Barbeau, D.S., 1996. Variability in kinetic parameter estimates: a review of possible causes and a proposed terminology. Water Research 30 (3), 742–748. Gray, G.T., McCutcheon, J.R., Elimelech, M., 2006. Internal concentration polarization in forward osmosis: role of membrane orientation. Desalination 197 (1–3), 1–8. Green, D.W., Perry, R.H., 2007. Perry’s Chemical Engineers’ Handbook. McGraw-Hill. Gregory, J., 2006. Particles in Water: Properties and Processes. CRC Press. Gujer, W., 2005. Activated sludge modelling: past, present and future. In: 3rd IWA Leading-Edge Conference Water and Wastewater Treatment Technologies, Sapporo, Japan. Hamoda, M.F., Alattar, I.M.S., 1993. Effects of high sodium chloride concentrations on activated sludge treatment. In: Asian Waterqual 93 – the 4th IAWQ Asian Regional Conference, on Water Conservation and Pollution Control, Jakarta, Indonesia. Henze, M., 1991. Characterization of wastewater for modelling of activated sludge processes. Specialized Seminar on Interactions of Wastewater, Biomass and Reactor Configurations in Biological Treatment Plants. PergamonElsevier Science Ltd, Copenhagen, Denmark. Henze, M., Gujer, W., Mino, T., Van Loosdrecht, M., 2000. In: IWA Task Group on mathematical modelling for design and operation of biological wastewater treatment (Ed.), Activated Sludge Models ASM1, ASM2, ASM2d and ASM3. IWA Publishing. Hoek, E.M.V., Elimelech, M., 2003. Cake-enhanced concentration polarization: A new fouling mechanism for salt-rejecting membranes. Environmental Science and Technology 37 (24), 5581–5588. Horikoshi, K., 1999. Alkaliphiles: some applications of their products for biotechnology. Microbiology and Molecular Biology Reviews 63 (4), 735. Huang, X., Gui, P., Qian, Y., 2001. Effect of sludge retention time on microbial behaviour in a submerged membrane bioreactor. Process Biochemistry 36 (10), 1001–1006.
38
water research 44 (2010) 21–40
Huang, X., Liu, R., Qian, Y., 2000. Behaviour of soluble microbial products in a membrane bioreactor. Process Biochemistry 36 (5), 401–406. Hwang, H.J., 1979. The effect of surface active agents on oxygen transfer. M.S. thesis, University of California, Los Angeles. Intrasungkha, H., Keller, J., Blackall, L.L., 1999. Biological nutrient removal efficiency in treatment of saline wastewater. Water Science and Technology 39 (6), 183–190. Itonaga, T., Kimura, K., Watanabe, Y., 2004. Influence of suspension viscosity and colloidal particles on permeability of membrane used in membrane bioreactor (MBR). Water Science and Technology 50 (12), 301–309. Judd, S., 2006. The MBR Book: Principles and Applications of Membrane Bioreactors in Water and Wastewater Treatment. Elsevier. Kargi, F., Dincer, A.R., 1996a. Effect of salt concentration on biological treatment of saline wastewater by fed-batch operation. Enzyme and Microbial Technology 19 (7), 529–537. Kargi, F., Dincer, A.R., 1996b. Enhancement of biological treatment performance of saline wastewater by halophilic bacteria. Bioprocess Engineering 15 (1), 51–58. Kargi, F., Dincer, A.R., 1999. Salt inhibition effects in biological treatment of saline wastewater in RBC. Journal of Environmental Engineering – ASCE 125 (10), 966–971. Kargi, F., Dincer, A.R., Pala, A., 2000. Characterization and biological treatment of pickling industry wastewater. Bioprocess Engineering 23 (4), 371–374. Katz, I., Dosoretz, C.G., 2008. Desalination of domestic wastewater effluents: phosphate removal as pretreatment. Desalination 222 (1–3), 230–242. Kedem, O., Zalmon, G., 1997. Compact accelerated precipitation softening (CAPS) as a pretreatment for membrane desalination I. Softening by NaOH. Desalination 113 (1), 65–71. Koops, H.P., Pommerening-Roser, A., 2001. Distribution and ecophysiology of the nitrifying bacteria emphasizing cultured species. FEMS Microbiology Ecology 37 (1), 1–9. Koros, W.J., Ma, Y.H., Shimizu, T., 1996. Terminology for membranes and membrane processes (IUPAC Recommendation 1996). Journal of Membrane Science 120 (2), 149–159. Krampe, J., Krauth, K., 2003. Oxygen transfer into activated sludge with high MLSS concentrations. Water Science and Technology 47 (11). Lange, R., Staaland, H., Mostad, A., 1972. The effect of salinity and temperature on solubility of oxygen and respiratory rate in oxygen-dependent marine invertebrates. Journal of Experimental Marine Biology and Ecology 9 (3), 217–229. Laspidou, C.S., Rittmann, B.E., 2002. A unified theory for extracellular polymeric substances, soluble microbial products, and active and inert biomass. Water Research 36 (11), 2711–2720. Le-Clech, P., Chen, V., Fane, A.G., 2006. Fouling in membrane bioreactors used in wastewater treatment. Journal of Membrane Science 284 (1, 2), 17–53. Le Gouellec, Y.A., Elimelech, M., 2002. Calcium sulfate (gypsum) scaling in nanofiltration of agricultural drainage water. Journal of Membrane Science 205 (1, 2), 279–291. Lefebvre, O., Moletta, R., 2006. Treatment of organic pollution in industrial saline wastewater: A literature review. Water Research 40 (20), 3671–3682. Lei, Z., 2005. Special Distillation Processes. Elsevier. Leiknes, T., Ødegaard, H., 2007. The development of a biofilm membrane bioreactor. Desalination 202 (1–3), 135–143. Lesage, N., Sperandio, M., Cabassud, C., 2005. Performances of a hybrid adsorption/submerged membrane biological process for toxic waste removal. Water Science and Technology 51 (6, 7), 173–180.
Lide, D.R. (Ed.), 2008. CRC Handbook of Chemistry and Physics, 89th ed. CRC Press/Taylor and Francis (Internet Version). Lopez-Vazquez, C.M., Hooijmans, C.M., Brdjanovic, D., Gijzen, H.J., van Loosdrecht, M.C.M., 2008. Factors affecting the microbial populations at full-scale enhanced biological phosphorus removal (EBPR) wastewater treatment plants in the Netherlands. Water Research 42 (10, 11), 2349–2360. Madigan, M.T., Oren, A., 1999. Thermophilic and halophilic extremophiles. Current Opinion in Microbiology 2 (3), 265–269. Masarwa, A., Meyerstein, D., Daltrophe, N., Kedem, O., 1997. Compact accelerated precipitation softening (CAPS) as pretreatment for membrane desalination II. Lime softening with concomitant removal of silica and heavy metals. Desalination 113 (1), 73–84. McAdam, E.J., Judd, S.J., 2008. Biological treatment of ionexchange brine regenerant for re-use: a review. Separation and Purification Technology 62 (2), 264–272. Mehta, G.D., Loeb, S., 1978. Performance of permasep B-9 and B-10 membranes in various osmotic regions and at high osmotic pressures. Journal of Membrane Science 4, 335–349. Melin, T., Rautenbach, R., 2007. Membranverfahren. SpringerVerlag. Meng, F., Chae, S.-R., Drews, A., Kraume, M., Shin, H.-S., Yang, F., 2009. Recent advances in membrane bioreactors (MBRs): membrane fouling and membrane material. Water Research 43 (6), 1489–1512. Moussa, M.S., Sumanasekera, D.U., Ibrahim, S.H., Lubberding, H.J., Hooijmans, C.M., Gijzen, H.J., van Loosdrecht, M.C.M., 2006. Long term effects of salt on activity, population structure and floc characteristics in enriched bacterial cultures of nitrifiers. Water Research 40 (7), 1377–1388. Ng, C.A., Sun, D., Fane, A.G., 2006. Operation of membrane bioreactor with powdered activated carbon addition. Separation Science and Technology 41 (7), 1447–1466. Ng, H.Y., Ong, S.L., Ng, W.J., 2005. Effects of sodium chloride on the performance of a sequencing batch reactor. Journal of Environmental Engineering (ASCE) 131 (11). Ng, W.J., 2006. Industrial Wastewater Treatment. Imperial College Press. Oehmen, A., Lemos, P.C., Carvalho, G., Yuan, Z.G., Keller, J., Blackall, L.L., Reis, M.A.M., 2007. Advances in enhanced biological phosphorus removal: From micro to macro scale. Water Research 41 (11), 2271–2300. Oo, H., K.A. Kekre, J.J. Qin, G. Tao, C.L. Lay, C.H. Lew, E.R. Cornelissen, C.J. Ruiken, K.F. de Korte, L.P. Wessels, 2008. Osmotic membrane bioreactor: preliminary pilot study on effects of osmotic pressure on membrane flux and air scouring on fouling. In: IWA Regional Conference ‘‘Membrane Technologies in Water and Waste Water Treatment’’, Moscow. Oren, A., 1999. Bioenergetic aspects of halophilism. Microbiology and Molecular Biology Reviews 63 (2), 334. Oren, A., 2002a. Diversity of halophilic microorganisms: Environments, phylogeny, physiology, and applications. Journal of Industrial Microbiology and Biotechnology 28 (1), 56–63. Oren, A., 2002b. Molecular ecology of extremely halophilic Archaea and Bacteria. FEMS Microbiology Ecology 39 (1), 1–7. Oren, A., 2007. Biodiversity in Highly Saline Environments. In: Gerday, C., Glansdorff, N. (Eds.), Physiology and Biochemistry of Extremophiles. ASM Press. Oren, A., Gurevich, P., Azachi, M., Henis, Y., 1992. Microbial degradation of pollutants at high salt concentrations. Biodegradation 3 (2, 3). Orhon, D., Tasli, R., Sozen, S., 1999. Experimental basis of activated sludge treatment for industrial wastewaters – the state of the art. Water Science and Technology.
water research 44 (2010) 21–40
¨ zbek, B., Gayik, S., 2001. The studies on the oxygen mass O transfer coefficient in a bioreactor. Process Biochemistry 36 (8, 9), 729–741. Panswad, T., Anan, C., 1999. Impact of high chloride wastewater on an anerobic/anoxic/aerobic process with and without inoculation of chloride acclimated seeds. Water Research 33 (5), 1165–1172. Peyton, B.M., Wilson, T., Yonge, D.R., 2002. Kinetics of phenol biodegradation in high salt solutions. Water Research 36 (19), 4811–4820. Phattaranawik, J., Fane, A.G., Pasquier, A.C.S., Bing, W., 2008. A novel membrane bioreactor based on membrane distillation. Desalination 223 (1–3), 386–395. PUB, 2002. Singapore Water Reclamation Study: Expert Panel Review and Findings. C.N. Ong, Singapore. Rahardianto, A., Gao, J., Gabelich, C.J., Williams, M.D., Cohen, Y., 2007. High recovery membrane desalting of low-salinity brackish water: integration of accelerated precipitation softening with membrane RO. Journal of Membrane Science 289 (1, 2), 123–137. Rautenbach, R., 1993. Mechanische Verfahrenstechnik (Partikelund Separationstechnologie), Vorlesungsumdruck. IVT, RWTH Aachen. Rautenbach, R., Mellis, R., 1994. Waste water treatment by a combination of bioreactor and nanofiltration. Desalination 95 (2), 171–188. Reid, E., Liu, X., Judd, S.J., 2006. Effect of high salinity on activated sludge characteristics and membrane permeability in an immersed membrane bioreactor. Journal of Membrane Science 283 (1, 2), 164–171. Rene, E.R., Kim, S.J., Park, H.S., 2008. Effect of COD/N ratio and salinity on the performance of sequencing batch reactors. Bioresource Technology 99 (4), 839–846. Rittmann, B.E., McCarty, P.L., 2001. Environmental Biotechnology: Principles and Applications. McGraw-Hill. Rodriguez-Valera, F., Ruiz-Berraquero, F., Ramos-Cormenzana, A., 1981. Characteristics of the heterotrophic bacterial populations in hypersaline environments of different salt concentrations. Microbial Ecology 7 (3). Rothschild, L.J., Mancinelli, R.L., 2001. Life in extreme environments. Nature 409 (6823), 1092–1101. Sanciolo, P., Zou, L., Gray, S., Leslie, G., Stevens, D., 2008. Accelerated seeded precipitation pre-treatment of municipal wastewater to reduce scaling. Chemosphere 72 (2), 243–249. Sawyer, C.N., McCarty, P.L., Parkin, G.F., 2003. Chemistry for Environmental Engineering and Science. McGraw-Hill. Schaefer, A.I., Andritsos, N., Karabelas, A.J., Hoek, E.M.V., Schneider, R., Nystrom, M., 2005a. Fouling in nanofiltration. In: Schaefer, A.I., Fane, A.G., Waite, T.D. (Eds.), Nanofiltration: Principles and Applications. Elsevier, pp. 169–240. Schaefer, A.I., Fane, A.G., Waite, T.D. (Eds.), 2005b. Nanofiltration: Principles and Applications. Elsevier Ltd. Schofield, R.W., Fane, A.G., Fell, C.J.D., 1987. Heat and mass transfer in membrane distillation. Journal of Membrane Science 33 (3), 299–313. Seviour, R.J., Blackall, L.L. (Eds.), 1999. The Microbiology of Activated Sludge. Kluwer Academic Publishers. Sharrer, M.J., Tal, Y., Ferrier, D., Hankins, J.A., Summerfelt, S.T., 2007. Membrane biological reactor treatment of a saline backwash flow from a recirculating aquaculture system. Aquacultural Engineering 36 (2), 159–176. Song, K.G., Kim, Y., Ahn, K.H., 2008. Effect of coagulant addition on membrane fouling and nutrient removal in a submerged membrane bioreactor. Desalination 221 (1–3), 467–474. Stephenson, T., Judd, S., Jefferson, B., Brindle, K., 2000. Membrane Bioreactors for Wastewater Treatment. IWA Publishing. Tam, L.S., Tang, T.W., Leung, W.Y., Chen, G.H., Sharma, K.R., 2006. A pilot study on performance of a membrane bio-reactor
39
in treating fresh water sewage and saline sewage in Hong Kong. Separation Science and Technology 41 (7), 1253–1264. Tanninen, J., Kamppinen, L., Nystrom, M., 2005. Pretreatment and hybrid processes. In: Schaefer, A.I., Fane, A.G., Waite, T.D. (Eds.), Nanofiltration: Principles and Applications. Elsevier, pp. 241–262. Tarabara, V., 2007. Coupled effects of scaling and colloidal deposition on the performance of reverse osmosis membrane filters, NWRI Final Project Report. Task Force, W.E.F., 1998. Biological and Chemical Systems for Nutrient Removal, Special Publication. Water Environment Federation. Tchobanoglous, G., Burton, F.L., Stensel, H.D., 2004. In: Metcalf, Eddy (Eds.), Wastewater Engineering, Treatment and Reuse, fourth ed. McGraw-Hill. Tellez, G.T., Nirmalakhandan, N., Gardeatorresdey, J.L., 1995. Evaluation of biokinetic coefficients in degradation of oil-field produced water under varying salt concentrations. Water Research 29 (7), 1711–1718. Thomas, D.G., 1965. Transport characteristics of suspension: VIII. A note on the viscosity of Newtonian suspensions of uniform spherical particles. Journal of Colloid Science 20 (3), 267–277. Thompson, M., Powell, D., 2003. Case Study – Kranji High Grade Water Reclamation Plant. Singapore, IMSTEC ’03, Sydney. Tun, C.M., Fane, A.G., Matheickal, J.T., Sheikholeslami, R., 2005. Membrane distillation crystallization of concentrated salts – flux and crystal formation. Journal of Membrane Science 257 (1, 2), 144–155. Uygur, A., 2006. Specific nutrient removal rates in saline wastewater treatment using sequencing batch reactor. Process Biochemistry 41 (1), 61–66. Uygur, A., Kargi, F., 2004. Salt inhibition on biological nutrient removal from saline wastewater in a sequencing batch reactor. Enzyme and Microbial Technology 34 (3, 4), 313–318. Ventosa, A., Nieto, J.J., 1995. Biotechnological applications and potentialities of halophilic microorganisms. World Journal of Microbiology and Biotechnology 11 (1). Ventosa, A., Nieto, J.J., Oren, A., 1998. Biology of moderately halophilic aerobic bacteria. Microbiology and Molecular Biology Reviews 62 (2), 504. Vrouwenvelder, J.S., Manolarakis, S.A., Veenendaal, H.R., van der Kooij, D., 2000. Biofouling potential of chemicals used for scale control in RO and NF membranes. Desalination 132 (1–3), 1–10. Walker, K.R., Clifford, R.H., 2000. High accuracy measurement of TOC in 28% brine samples. Pollution Equipment News. Wang, F., Tarabara, V.V., 2007. Coupled effects of colloidal deposition and salt concentration polarization on reverse osmosis membrane performance. Journal of Membrane Science 293 (1, 2), 111–123. Wicaksana, F., Fane, A.G., Chen, V., 2006. Fibre movement induced by bubbling using submerged hollow fibre membranes. Journal of Membrane Science 271 (1, 2), 186–195. Wickramasinghe, S.R., Han, B., Akeprathumchai, S., Chen, V., Neal, P., Qian, X., 2004. Improved permeate flux by flocculation of biological feeds: comparison between theory and experiment. Journal of Membrane Science 242 (1, 2), 57–71. Woolard, C.R., Irvine, R.L., 1995. Treatment of hypersaline wastewater in the sequencing batch reactor. Water Research 29 (4), 1159–1168. Wu, B., 2007. Enhanced Membrane Bioreactor (MBR) Operation by Optimisation of Bacterial Population, Ph.D. thesis, Singapore, Nanyang Technological University. Yoshikawa, C., Yamanaka, Y., Nakatsuka, T., 2005. An ecosystem model including nitrogen isotopes: Perspectives on a study of the marine nitrogen cycle. Journal of Oceanography 61 (5), 921–942. Yu, S.M., Leung, W.Y., Ho, K.M., Greenfield, P.F., Eckenfelder, W.W., 2002. The impact of sea water flushing
40
water research 44 (2010) 21–40
on biological nitrification-denitrification activated sludge sewage treatment process. Water Science and Technology 46 (11, 12), 209–216. Zhang, L.L., Yang, D., Zhong, Z.J., Gu, P., 2008. Application of hybrid coagulation-microfiltration process for treatment of
membrane backwash water from waterworks. Separation and Purification Technology 62 (2), 415–422. Zlokarnik, M., 1979. Sorption characteristics of slot injectors and their dependency on the coalescence behaviour of the system. Chemical Engineering Science 34 (10), 1265–1271.
water research 44 (2010) 41–57
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Review
Low-pressure membrane integrity tests for drinking water treatment: A review H. Guo a,1, Y. Wyart a,2, J. Perot b,3, F. Nauleau b,4, P. Moulin a,* a Universite´ Paul Ce´zanne Aix Marseille, Laboratoire de Me´canique, Mode´lisation et Proce´de´s Propres (M2P2 – UMR-CNRS 6181), Europoˆle de l’Arbois, BP. 80, Baˆtiment Laennec, Hall C, 13545 Aix en Provence Cedex 04, France b SAUR, 1 avenue Euge`ne Freyssinet, 78064 Saint Quentin En Yvelines Cedex, France
article info
abstract
Article history:
Low-pressure membrane systems, including microfiltration (MF) and ultrafiltration (UF)
Received 15 April 2009
membranes, are being increasingly used in drinking water treatments due to their high
Received in revised form 10
level of pathogen removal. However, the pathogen will pass through the membrane and
September 2009
contaminate the product if the membrane integrity is compromised. Therefore, an effec-
Accepted 12 September 2009
tive on-line integrity monitoring method for MF and UF membrane systems is essential to
Available online 24 September 2009
guarantee the regulatory requirements for pathogen removal. A lot of works on lowpressure membrane integrity tests have been conducted by many researchers. This paper
Keywords:
provides a literature review about different low-pressure membrane integrity monitoring
Low-pressure membrane
methods for the drinking water treatment, including direct methods (pressure-based tests,
Microfiltration
acoustic sensor test, liquid porosimetry, etc.) and indirect methods (particle counting,
Ultrafiltration
particle monitoring, turbidity monitoring, surrogate challenge tests). Additionally, some
Membrane integrity
information about the operation of membrane integrity tests is presented here. It can be
Drinking water
realized from this review that it remains urgent to develop an alternative on-line detection technique for a quick, accurate, simple, continuous and relatively inexpensive evaluation of low-pressure membrane integrity. To better satisfy regulatory requirements for drinking water treatments, the characteristic of this ideal membrane integrity test is proposed at the end of this paper. ª 2009 Elsevier Ltd. All rights reserved.
Contents 1. 2. 3.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Sensitivity evaluation for membrane integrity tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Low-pressure membrane integrity tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
* Corresponding author. Tel.: þ33(0) 4 4290 8501; fax: þ33(0) 4 4290 8515. E-mail addresses:
[email protected] (H. Guo),
[email protected] (Y. Wyart),
[email protected] (J. Perot), fnau@saur. fr (F. Nauleau),
[email protected] (P. Moulin). 1 Tel.: þ33(0) 4 4290 8504; fax: þ33(0) 4 4290 8515. 2 Tel.: þ33(0) 4 4290 8508; fax: þ33(0) 4 4290 8515. 3 Tel.: þ33(0) 1 3060 1655. 4 Tel.: þ33(0) 1 3060 1655. 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.032
42
water research 44 (2010) 41–57
Pressure-driven tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 3.1.1. PDT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 3.1.2. DAF test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 3.1.3. The Memsure process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 3.1.4. Bubble point test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 3.1.5. Vacuum decay test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 3.2. Binary gas integrity test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 3.3. Acoustic sensor test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 3.4. Liquid porosimetry technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 3.5. Novel membrane-based sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 3.6. Particle counting and particle monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 3.7. Turbidity monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 3.8. Surrogate challenge tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 3.8.1. Spiked integrity monitoring system – SIM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 3.8.2. Microbial challenge tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 3.8.3. Nanoscale probe challenge tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 3.8.4. Magnetic particle challenge tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Operation of membrane integrity tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 3.1.
4. 5.
1.
Introduction
Low-pressure membrane systems, including microfiltration (MF) and ultrafiltration (UF) membranes, are being increasingly used for the drinking water treatment. The global installed volume of low-pressure membranes has grown at an impressive rate during the last 10 years, and 60% of applications are for drinking water (David et al., 2008). This rapid growth of water treatment low-pressure membranes, both in terms of capacity and number of installations, is mainly due to their high level of pathogen removal, such as viruses, bacteria and protozoa cysts (Giardia and Cryptosporidium). Complete removal of coliform bacteria, Giardia spp. and Cryptosporidium spp. by MF and UF membranes has been demonstrated by many researchers (Adham and Jacangelo, 1994; Edwards et al., 2001; Freeman et al., 1996; Hirata and Hashimoto, 1998; Jacangelo et al., 1991). Compared to MF, UF technology is able to remove viruses due to its low cut-off and thus it can take the place of the disinfection step. In addition to concerns over microbial contaminants, low-pressure membranes are becoming more attractive for the drinking water industry due to some other reasons, such as stricter regulatory requirements, operation easiness, minimum staffing requirements, competitive cost and independence of water source quality. However, pathogens may pass through membranes and enter the public water supply system if the membrane system integrity is compromised (e.g., broken fiber, fiber degradation, O-ring failure, etc.). Failure of membrane fibers or sheets results from four main reasons: (1) chemical corrosion such as oxidation; (2) faulty installation and maintenance; (3) membrane stress and strain from operating conditions, such as backwashing or excessive movement due to vigorous bubbling; and (4) damage by sharp objects not removed by pretreatment. Zondervan et al. (2007) concluded that the fouling status of a membrane, the number of applied back pulses and the combination of these two factors are significant aging factors and responsible for the membrane failure.
Accurate and efficient integrity tests of the membrane system can guarantee the quality of filtered products. Therefore, it is important for a membrane plant to perform an effective membrane integrity monitoring as well as reachable flux, membrane retention and membrane permeability recovery. An increasing number of regulatory agencies in North America and Europe (e.g., U.S. EPA in United States, DWI in United Kingdom, etc.) require membrane water treatment utilities to conduct membrane integrity monitoring on a regular basis. Membrane integrity tests are specific for the membrane type and depend on the membrane manufacturer and membrane system supplier. Generally, membrane integrity monitoring techniques are divided into two main groups: direct methods and indirect methods. Direct methods refer to tests directly applied to the membrane or the membrane module, i.e., pressure decay test (PDT) (Adham et al., 1995; Johnson, 1997, 1998), diffusive air flow (DAF) test (Adham et al., 1995; Johnson, 1997, 1998), bubble point test (Randles, 1997), acoustic sensor test (Adham et al., 1995; Laıˆne et al., 1998), liquid–liquid porosimetry test (DiLeo and Phillips, 1994, 1995; Phillips and Dileo, 1996; Gekas and Zhang, 1989) and binary gas integrity test (Giglia and Krishnan, 2008). Indirect methods refer to tests applied to water quality parameters in the permeate solution, i.e., particle counting (Adham et al., 1995; Landsness, 2001; Panglisch et al., 1998), particle monitoring (Adham et al., 1995), turbidity monitoring (Adham et al., 1995; Banerjee et al., 1999, 2001), and different surrogate challenge tests such as spiked integrity monitoring system – SIM, microbial challenge tests and some new surrogate challenge tests (Van Hoof et al., 2001; Kruithof et al., 2001; Brehant et al., 2008; Gitis et al., 2006a,b; Trimboli et al., 2001; Moulin, 2008; Rajagopalan et al., 2006; Deluhery and Rajagopalan, 2008). Among these methods, the PDT and DAF test are the most frequently used in drinking water treatments due to their advantages of simplicity, low maintenance, reliability and high sensitivity to detect membrane breaches. In addition, particle counting, turbidity monitoring and routine microbial
43
water research 44 (2010) 41–57
analysis are the most frequently used indirect methods (Crozes et al., 2002). All these methods have their own advantages and disadvantages. For example, the PDT and DAF test are sensitive enough to meet the regulatory requirements of drinking water treatment, but they are conducted off-line and provide no measurement of filtrate water quality. It should be noted that on-line monitoring is very important to guarantee the filtered quality of the membrane system. Plants that only rely on pressure integrity tests (e.g., PDT and DAF test) are likely to continue their operations beyond the point where their membrane integrity is compromised. In contrast, indirect methods, such as particle counting and turbidity monitoring, are very convenient for routine qualitative monitoring. Changes in filtrate quality are monitored by comparing measured data with a previously established baseline level. These indirect tests are able to detect membrane integrity in continuous and on-line mode. In addition, the same indirect methods and testing instruments can be applied to any membrane system, independent of membrane manufacturers, system configurations, and other parameters intrinsic to the system. They are also likely to remain applicable to any new system that is developed. But the main problem of these two techniques is their low detection sensitivity. It has been reported that these two tests failed to detect the individual fiber pin-holes in membrane system occurring as a result of challenging operating conditions (Walsh et al., 2005). Furthermore, current membrane integrity tests in practical drinking water treatments are not sufficiently sensitive to detect nanometric breaches for capital and operational reasons. Then they are not effective to meet removal requirements for viruses (w20 nm). The absence of a reliable, sensitive, and on-line detection method for monitoring membrane integrity is currently hampering more significant virus removal credits of UF membranes. Therefore, it is necessary to develop an alternative on-line detection technique for an accurate, quick, simple, and affordable evaluation of the UF membrane integrity for drinking water plants. In this paper, different methods for monitoring low-pressure membrane integrity are reported. In addition to commonly used tests (pressure-based tests, acoustic sensor test, particle counting, particle monitoring, turbidity monitoring) which have been summarized in some previous reports (U.S. EPA, 2001, 2003), other integrity monitoring technologies which are mostly in lab-scale study so far are also presented here (liquid–liquid porosimetry, binary gas integrity test, new membrane-based sensor and different surrogate challenge tests). Pointing out advantages and disadvantages of these methods, it can be realized that the development of a new membrane integrity test up to the level of industrial applications is still important and urgent.
Cryptosporidium (w4–6 mm) are required for a complete removal. Considering that it is not practical for a 100% guaranteed removal, generally the drinking water industry measures the pathogen removal efficiency in terms of the log removal value (LRV), which is defined as Eq. (1) (Bennett, 2008): LRV ¼ log10
Cf Cp
(1)
where Cf is the concentration of the retained species in feed solution and Cp is the concentration of the species in permeate solution. Of course, a stated LRV is related to a particular particle size or particle size distribution. When there are membrane failures, the impact of a liquid leak on the retention can be calculated as Eq. (2) (Giglia and Krishnan, 2008): DLRV ¼ LRV1 log10
Cf V T Cp Vp þ Cf Vd
(2)
where LRV1 is the LRV of the unimpaired portion of membranes, Vp is the feed volume passing through the unimpaired portion of the membrane, Vd is the feed volume passing through the defect and VT is the total feed volume passing through the membrane. Here, it is assumed that the defect does not partially retain the considered species. The United States Environmental Protection Agency has specified pathogen removal (or inactivation) rates (Cryptosporidium, Giardia and viruses) through the enhanced surface water treatment rule (ESWTR), as shown in Table 1 (Faber and Pearce, 2004). In general, the purposes of membrane integrity tests include verification of high filtered water quality, demonstration of regulatory compliance and detection of equipment/filtration problems. Considering that membrane integrity tests are function of the different membrane suppliers, the American Society for Testing and Materials (ASTM) sub-committee designed the Standard Practice for Integrity Testing of Water Filtration Membrane Systems (ASTM – D6908-03), which describes four integrity test methods (PDT, vacuum decay test, soluble dye test (SDT) and total organic carbon monitoring test (TOCMT)) that can be applied to all membrane systems, regardless of application (Moch and Paulson, 2003). The sensitivity of a membrane integrity test is very important to guarantee the pathogen removal credit. According to the membrane filtration guidance manual (MFGM) (U.S. EPA, 2003), the sensitivity of membrane integrity tests refers to the maximum log removal value that can be reliably verified by integrity tests associated with a given membrane filtration system. It is expressed in terms of a LRV,
Table 1 – Inactivation/removal requirements for pathogen in portable water(Faber and Pearce, 2004).
2. Sensitivity evaluation for membrane integrity tests According to the DWI requirements, the continuous removal or retention of particles greater than 1 mm should be obtained during the membrane operation for drinking water treatments (Jackson, 2001). This can be interpreted as an absolute removal, which means that Giardia (w6–20 mm) and
Pathogen Cryptosporidium parvum Giardia lamblia Viruses
Log removal value Percentage removal (LRV) (%) 2
99.00
3 4
99.90 99.99
Based on ESWTR – adapted from Faber and Pearce (2004).
44
water research 44 (2010) 41–57
which must be equal to or greater than the required pathogen removal credit. The sensitivity of some other tests can be calculated in different forms. For pressure-based tests, the sensitivity can be expressed as Eq. (3): Qp LRV ¼ log VCF ,Qbreach
(3)
where Qp is the designed filtrate flow of the membrane unit, Qbreach is the flow from the breaches associated with the smallest integrity test response that can be reliably measured and VCF is the volume concentration factor (i.e., the ratio of concentration in the retentate to concentration in the influent). For surrogate challenge tests, the sensitivity can be expressed as Eq. (1). Based on the practical results, Johnson (1998) compared the sensitivity of different integrity tests (PDT, DAF test, particle counting and turbidity monitoring) and ranked them as shown in Fig. 1.
3.
Low-pressure membrane integrity tests
3.1.
Pressure-driven tests
Pressure-driven tests are non-destructive for membrane systems. All pressure-driven membrane integrity monitoring tests, including PDT, DAF test, vacuum decay test and bubble point test, are theoretically similar since they are based on the bubble point pressure concept but they differ in operating protocols and measured parameters. Here, bubble point is typically defined as the minimum pressure required to overcome the capillary forces and surface tension of a liquid in a fully wetted membrane filter and force air flow through the filter pores. Briefly, for a wetted membrane, liquid can be forced out of the filter pores by applying gas pressure. The removal of liquid from the largest pores creates a passage way through which bulk air flow takes place. The minimum pressure at which this bulk flow through the membrane is detected is referred to as the bubble point. The bubble point is related to the diameter of the largest pore or defect, which can be estimated by the capillary equation as Eq. (4) (Farahbakhsh and Smith, 2004):
(4)
where k is the correction factor for the largest pore shape, d is the diameter of the largest pore, q is the contact angle between the liquid and the membrane and s is the surface tension of liquid. It can be seen that the air pressure required to force liquid from the pores is inversely proportional to the pore diameter. The larger the pore is, the lower the pressure is required. If the membrane has no defect, any air pressure below the bubble point cannot force water from the membrane pores. In contrast, defects such as holes or broken fibers have a comparatively low bubble point. Assuming that the pore shape factor k is 1 and the liquid/membrane contact angle q is close to 0 it is possible to calculate the maximum bubble point pressure (Bennett, 2005). For UF membranes (pore size <0.1 mm), the bubble point is generally in the range of 3000–30,000 KPa. For a hole in the membrane comparable to a cyst at 4 mm mean diameter, the bubble point of the membrane is 73 KPa. This means that any hole with a diameter larger than 4 mm can be identified if the test pressure is above 80 KPa. For current PDT and DAF test, a typical test pressure of 100 KPa, which can identify a defect of 2.8 mm, guarantee a complete barrier against pathogens such as Cryptosporidium and Giardia but not against viruses. Diffusive air flow through a wetted membrane can be expressed as Eq. (5) modified from the Fick’s law of diffusion (Giglia and Krishnan, 2008): Q¼
A3DS Pf Pp sL
(5)
< 0.3 log Turbidity
3.1.1.
6.5 to7 log Diffusive
6
Log Removal value
4k cos q s d
where Q is the diffusive air flow rate, A is the membrane area, 3 is the membrane porosity, D is the gas diffusivity in the liquid, S is the gas solubility coefficient, Pf and Pp are the feed and permeate side pressures respectively, s is the pore tortuosity and L is the liquid thickness in the membrane. A measured gas flow rate in excess of that predicted by Eq. (5) or higher than a flow rate empirically established for an unimpaired membrane indicates the presence of defects in the membrane. As shown in Eq. (5), the amount of diffused air through a wetted membrane at a given applied pressure is a function of the membrane porous surface area (3A). During a pressure decay test, an oversized defect or hole contributes to a more diffusive air flow because of an increase in membrane porous surface area. It has been demonstrated that the dilution effect observed during pressure decay tests on membranes with large surface can be mainly attributed to air diffusion through intact pores (Giglia and Krishnan, 2008; Farahbakhsh and Smith, 2004). This may produce false– negative results. In other words, membrane breaches will not be detected until the defect is significant enough to produce a noticeable pressure decay rate above that resulting from diffusion. Therefore, accounting for and estimating the air diffusion contribution to pressure decay during a pressure decay test would produce much more reliable and sensitive results.
7
Air Flow 5 log
5
Pressure 4
3.5 log
Decay
On-line
3
Particle 2 1
P¼
Counter
0
Fig. 1 – Relative sensitivity of various monitoring methods (Johnson, 1998).
PDT
The main principle of the PDT is based on the measurements of pressure drop in the feed side after draining and pressurizing. PDT can be performed with PDT-filled or PDT-drained
water research 44 (2010) 41–57
mode. An outline of the PDT-filled involves: (1) draining the water from one side of the membrane; (2) applying pressurized air to the drained side of the fully wetted membrane at a predetermined level below the bubble point; and (3) holding the pressure for a specific time duration (2–10 min) and monitoring the pressure decay. The predetermined pressure, ranging from 20 to 200 KPa (typically 100 KPa during practical operation), directly relates to the defect size under investigation. It also determines the smallest hole that can be detected by the PDT. During the test, a small decrease (e.g., 0.5–1.5 KPa per minute) is considered acceptable due to air diffusion across the intact membrane structure. A faster pressure decrease in excess of a critical threshold indicates a faulty membrane. PDT-drained is similar to ‘‘PDT-filled’’, but in this case the module is drained on both sides of the element so that only the membrane is ‘‘wetted’’. After the detection of membrane defects, the bubble test or sonic test can be used to locate defects. Bubble tests are performed with a specific test pressure (e.g., 30–50 KPa) in water filled basin or using soap on the outer ends of a membrane element. Bubbles will appear from any defective fiber. Then, the fiber can be isolated and repaired. Sonic tests for defect location will be presented in the Memsure process. Generally, the PDT is a standard part of most UF and MF membrane systems and it is highly automated during the operation. PDT is a reliable method for membrane integrity monitoring but it cannot be operated on-line and continuously. PDT appears to be very sensitive to detect leaks and integrity breaches. Adham et al. (1995) reported that a considerable loss of pressure was observed due to a needle puncture of 0.6 mm internal diameter in the lumen wall of one out of over 22,000 fibers in a membrane module. Johnson (1998) determined that this method is able to detect a single broken fiber in a membrane array containing over one million fibers. PDT is able to detect changes in the membrane integrity at a level up to 4.5–5 of Giardia or Cryptosporidium log removal, independent of feed water quality and without relying on filtered water quality monitoring such as turbidity and particle counters. However, the sensitivity of PDT is limited by the minimum detectable excess air flow. It has been shown that the PDT effectiveness for the detection of membrane integrity changes is affected by the membrane surface (U.S. EPA, 2001). This phenomenon becomes more important as the membrane surface increases. It has been reported that, for a rack of 90 modules of the same type of membrane tested by Adham et al. (1995), PDT results have exhibited significant variation from the case of intact membrane system until six fibers were cut (Landsness, 2001). Hence, the greater the number of fibers tested, the more sensitive the pressure transducers need to be so as to differentiate between background noise spread out over a larger membrane area and losses due to integrity breaches. Additionally, PDT may sometimes cause membrane breaks during the test and may yield false–positive results due to a non-fully wetted membrane (U.S. EPA, 2003).
3.1.2.
DAF test
DAF test is fundamentally similar to the PDT. However, rather than measuring pressure decay rate, DAF test measures the diffused gas filtrate flow or displaced water flow through the fully wetted membrane pores when applying a constant feed
45
side gas pressure below the bubble point of the selected whole size. The most commonly cited DAF tests refer to air diffusion measurement (Cheryan, 1998). But one associated difficulty is the sensitivity of air diffusion rates to temperature, which will directly affect the air viscosity. Seasonal variations in temperature along a year may cause fluctuations in the results of this test. The other DAF test, which measures the displaced water flow, has been described in a 1997 report by the American Water Works Association Research Foundation (AWWARF) (Jacangelo et al., 1997). Due to its advantages of easiness and accuracy of measurements, the DAF test measuring displaced water flow was widely used in practical membrane water treatment plants. Overall, the DAF test shows more sensitivity to detect changes in membrane integrity than the PDT (Trimboli et al., 2001). It is able to detect integrity changes at levels >6 LRV. Despite of its increased sensitivity, the DAF test requires some additional pipe work and fittings in order to measure displaced liquid flow rate. The DAF test is subject to the same disadvantages as the PDT. In addition, it is not included as part of the standard equipment in most MF and UF systems and thus it has not been automated.
3.1.3.
The Memsure process
As a typical example of application of PDT and DAF test, the Memsure process, which was developed by Simens–Memcor for industrial application, is an integrity monitoring technique for continuous microfiltration (CMF) system (Johnson, 1997 and 1998). This technique involves three key steps: (1) monitoring the integrity using the Memsure PDT or the Memsure DAF test; (2) identification of leaks using sonic analysis (The Memcor Sonic Analyzer); and (3) isolation of faulty modules using module isolating valves for later repair. Generally, the lumen is pressurized at 100 KPa during the Memsure PDT and the Memsure DAF test, the test duration being around 5 min. One significant advantage of this PDT test is that it requires no additional equipment and it makes use of the CMF air backwash system. Thus the test can be automatically conducted by the control system, including the logging of results and generating an alarm when results are beyond limits. The Memcor Sonic Analyzer is a sensitive listening device tuned to the sound made by bubbles escaping through fiber defects or leakages. Air leaking through defects creates a distinctive sound that can be picked up by the device and displayed as a sound level on the front of the unit. In this way the identified modules can be isolated using the in-built module isolation valves at the top and bottom of each Memcor CMF module to be removed and repaired on site later on.
3.1.4.
Bubble point test
A bubble point test is a test designed to determine the bubble point pressure of the membrane. Theoretically, the bubble point pressure decreases with the presence of breaches in the membrane. When a bubble point test is conducted, the module to be tested is first removed from the rack. The internal shell of the module is then drained and pressurized. The membrane must be wetted uniformly. A dilute surfactant solution is applied to the open ends of the membrane fibers at the end of the module. When reaching the bubble point pressure, the liquid is expelled from one or more
46
water research 44 (2010) 41–57
passageways, establishing a path for the bulk flow of air. As a result, a steady stream of bubbles should be seen in the surfactant solution. Explicitly, the formation of vigorous bubbles in the surfactant solution can be traced to locate specific leaking fibers. In a bubble point test, the excessive gas flow originating from a single or very small defect may not be identified from that coming from the unimpaired part of the membrane, thereby masking the actual bubble point. Another limitation of this test is that the pressure required to reach the bubble point may be impractically high for UF membranes, even with the use of low surface tension fluids. For example, the air– water bubble point for a membrane with a pore size of 20 nm will be in excess of 3500 KPa (Giglia and Krishnan, 2008). Such high bubble point pressure is generally unattainable in practical operation and is likely to lead to membrane compaction and rupture. Therefore, this method is not suitable for UF membrane integrity monitoring. Large-scale operating experience showed (Randles, 1997) that Memcor membrane integrity test, which is based on the bubble point test, was very effective for monitoring membrane integrity of the Memcor CMF process. It is able to detect a single compromised membrane fiber in around 12 million fibers and it offers high sensitivity to guarantee bacterial removal at levels of 6 LRV. So far, the bubble point test is generally employed in conjunction with the PDT rather than as a separate and independent gauge of the membrane integrity. In other words, once a membrane integrity problem is detected, the bubble point test can be applied to identify the compromised fiber(s) following its removal from the rack. Used in this manner, the LRV of the test is limited to that of the PDT.
3.1.5.
Vacuum decay test
The vacuum decay test is a variation of the PDT where a vacuum is applied to the drained side of a fully wetted membrane and the vacuum pressure decay rate is monitored. This method can be used to monitor UF and MF membrane integrity but it is rarely performed in practical operation for membrane drinking water plants. So far, this method, based on ASTM Standard D3923-2 and D6908-3, is used for FILMTEC membranes to detect leaks or confirm the integrity of FILMTEC RO and NF elements after they have been in operation (FILMTEC Membranes). A vacuum decay test is able to identify leaking elements or O-rings within a short time and it is a nondestructive test. This test is useful as a screening procedure and generally it is not intended as a mean of absolute verification of leaks.
3.2.
Binary gas integrity test
Assuming that the gas is completely mixed on both sides of the membrane when two gases permeate through a membrane, the composition of the permeate gas can be calculated from the ratio of diffusive flow rates of the two components and the inlet side composition (Weller and Steiner, 1950). The composition of the permeate gas is independent of membrane thickness, tortuosity, porosity and area. It is also independent of the pressure difference across the membrane but dependent on the pressure ratio. In addition, the permeate composition depends on the feed side
composition and the permeability ratio of these two gases. In order to maintain a constant feed side composition, a constant sweep flow has to be applied. A binary gas integrity test for low-pressure membranes has recently been developed by Giglia and Krishnan (2008). This test uses a binary gas mixture and is based on the gas permeability difference between the two components of a gas mixture through the liquid layer of a wetted membrane. In contrast to the single gas diffusion test, the binary gas test primarily relies on the measurement of downstream gas composition rather than downstream flow rate. The presence of membrane defects results in an elevated concentration of the slower permeating gas in the permeate stream. In an unimpaired membrane, the permeate composition can be predicted from known operating conditions and the transport properties of gases permeating through the liquid layer. A change in gas composition across the membrane indicates the presence of defects or opened pores. The sensitivity of the binary gas test mainly depends on the selectivity (permeabilities ratio) of these two gases through the liquid layer. High selectivity of the two gases results in high detection sensitivity. In addition, the selection of gas concentration in the mixture is influenced by a number of factors, including easiness of composition measurement, gas flow rate through the membrane and economic considerations. Naturally, the permeability of gases through the liquid layer – generally water – is independent of the membrane type. Due to the high selectivity of CO2/C2F6 pair through water of about 1000, a 90/10 CO2/C2F6 molar concentration was selected to test the membrane integrity under a pressure of 345 KPa by Giglia and Krishnan (2008). At this level of concentration, the high permeability of the CO2 enables a convenient flow and composition measurement even for relatively small membrane areas as small as 3 cm2. Evidently, C2F6 is toxic and the storage of C2F6 is not so simple during practical operation. Because the binary gas test has lower sensitivity to membrane porosity, liquid layer thickness and membrane area, integral devices exhibit a relatively narrow range of test values, resulting in a superior defect signal-tonoise ratio. As a result, the binary gas test, which can provide an LRV assurance greater than 6, shows higher defect detection sensitivity than the single gas test. However, this method is conducted off-line and reported just in a lab-scale study.
3.3.
Acoustic sensor test
An acoustic sensor analysis was at first conducted manually by applying an accelerometer (an instrument used to detect vibration) in one or two locations on each membrane module (Adham et al., 1995). Using headphones, an operator listens to vibrations generated by leaking air. This analysis is effectively administered by a skilled and experienced operator and it is to some extent more subjective than other forms of integrity tests. Yet, it cannot monitor membrane integrity continuously. In practical operation, acoustic sensor analysis is usually used to identify the impaired module and the breaches location which have been detected by the PDT. However, this method has the potential to eliminate the subjectivity and to be developed into an on-line and continuous membrane integrity test if it is automated and computerized.
water research 44 (2010) 41–57
Such an automated acoustic system to monitor membrane integrity was described by Laıˆne et al. (1998). Based on the hydrophonic sensor technique, the acoustic integrity monitoring (AIM) technique measures the noise signal (i.e., pressure fluctuation) in a given frequency range (e.g., 280–650 Hz). A distinctive noise signal is created due to a compromised fiber. The main advantage of a hydrophonic sensor is that the membrane integrity is continuously monitored during filtration (on-line sensor) independently of feed water quality. It has been reported that AIM technology is able to ensure more than 6 log removal of viruses. However, experimental results from the full-scale plants showed that the acoustic detection depends significantly on background noise and flow rate. The higher the background noise is, the lower the sensor response is. In the same way, the higher the flow rate is, the higher the noise generated by a compromised fiber is. Laıˆne et al. (1998) designed an AIM prototype to check 28 modules. This prototype includes 28 acoustic sensors, two collectors and one processor. The total cost for a full system was about $7000, i.e., $250 per module showing that AIM technology is economically competitive.
3.4.
Liquid porosimetry technique
The liquid porosimetry technique (CorrTest) proposed by DiLeo and Phillips (DiLeo and Phillips, 1994, 1995; Phillips and DiLeo, 1996), has been used to monitor UF membrane integrity and to characterize UF membranes. This technique uses a pair of mutually immiscible fluids to validate the pore size distribution and particularly the maximum pore size. One of the fluids is employed as a membrane wetting agent and the other is used as an intrusion fluid. A transmembrane pressure is used so that the intrusion fluid can selectively penetrate the pores accessible to a given sized particle such as virus. Another transmembrane pressure is used to make the intrude fluid penetrate nearly all the membrane pores. The ratio of these two permeabilities obtained with each operating pressure is thus the percentage of total flow through the membrane pores accessible to a given size particle. That is to say, the different size particle retention characteristics can be expressed with the permeability ratio of these two liquids. Thus it is possible to obtain the pore size distribution or the maximum pore size when using the liquid porosimetry to characterize UF membranes. For this method, low intrusion pressure is needed due to the low interfacial tension associated with many pairs of immiscible liquids. Gekas and Zhang (1989) have used liquid porosimetry to characterize the entire pore size distribution of UF membranes at pressures less than 8 bars, which is much less than the required pressure when using gas–liquid porosimetry. In other words, to detect a defect with a given size, the required pressure of this method is lower than that of gas–liquid methods (e.g., pressure-based tests). The lower pressure results in less cost requirements and is more feasible during practical operation. As a result, this method is feasible for economic and operational point of views. For this method, the CorrTest value (CTV) of the membrane, the ratio of the intrusion fluid flow rate through all of the membrane pores to the flow rate through those pores
47
penetrated at the test pressure, is calculated at each intrusion pressure (P) using the following equation: Ltotal Lint ðPÞ
CTV ¼ log
(6)
where Ltotal is the total membrane permeability of the intrusion fluid and it is essentially pressure independent. Lint ðPÞ is the membrane permeability measured at each intrusion pressure. Assuming Poiseuille flow through cylindrical pores, the CTV can be rewritten as: ! n rp r4p drp n rp r4p drp Rflow RN
CTV ¼ log R N0
(7)
where rp is the penetrated pore radius, n is the number of membrane pores with rp radius and Rflow is the minimum radius of penetrated pores at the test pressure. As seen in Eq. (7), CTV is dependent only upon the membrane pore size distribution and Rflow . Rflow is related to the intrusion pressure through Washburn’s equation: Rflow ¼
2g cos q P
(8)
where g is the interfacial tension, q is the contact angle and P is the intrusion pressure. By definition, CTV is independent of membrane thickness and porosity, membrane surface area and the fluid properties. This is an important distinction of the CTV compared to the membrane hydraulic permeability which is strongly dependent on the membrane porosity and thickness as well as the pore size distribution. Gadam et al. (1997) found that the liquid porosimetry technique is able to accurately characterize the membrane pore size distribution, independently of module configuration. This technique is able to identify membranes of different molecular weight cut-off and is sensitive enough to capture slight changes of sieving coefficient for the same membrane cut-off with slight variations. Therefore, it is sufficiently sensitive to detect membrane sieving changes resulting from membrane breaches. Additionally, the liquid porosimetry technique is non-destructive and relatively simple to perform, which is convenient for practical operation. However it should be noted that this method is performed off-line.
3.5.
Novel membrane-based sensor
Phattaranawik et al. (2007, 2008) proposed a novel membranebased sensor device for upstream membrane integrity monitoring shown in Fig. 2. The sensor is based on monitoring relative transmembrane pressure, which is created by two unimpaired membranes set in series inside the sensor device that detects deposition from the sample stream onto the first membrane of the sensor. Based on the principle of a transition from ‘‘sustainable flux’’ to ‘‘non-sustainable flux’’ conditions (Field et al., 1995; Howell, 1995) when a contaminated sample passes (Fig. 3), the sensor pressure signals are able to detect either intact or damaged membranes in the upstream membrane filtration process. The sensitivity of this method, based on response time of the membrane sensor for particle detection, depends on
48
water research 44 (2010) 41–57
Membrane plant
Upstream Feed stream
Membranes
Downstream Permeate stream
Membrane-based sensor of the invention Permeate reservoir
a Control unit
PID control designed for novel sensor
Output Computer
Report the integrity status or activate warning alarm
Pressure sensors
Downstream from membrane process
Membrane sensor module Sensor unit
b
Recycle back
Fig. 2 – (a) Installation and (b) schematic diagram of the membrane-based sensor and control unit (Phattaranawik et al., 2008).
3.6.
Particle counting and particle monitoring
The particle concentration in a membrane system with impaired fibers can be estimated by mass-balance equations (Panglisch et al., 1998). In the case of a negligible particle concentration in the permeate of an intact membrane, the particle concentration in the permeate of impaired membrane can be calculated from flux ratio of defect to total membranes, as demonstrated at Eq. (9).
cP ¼
nd vd cF ni vi þ nd vd
(9)
where cP is the particle concentration in the permeate of the membrane system, cF is the particle concentration in the feed, vi is the volume flow through the intact fiber, vd is the volume
Apply the same magnitude of pressure or flux
Sensor (pressure) signal
increasing supply pressures and particle concentrations. Results obtained by Phattaranawik et al. (2008) also showed that both sensitivity and stability of this sensor strongly depends on the membrane sandwich configurations (membrane characteristics) in the sensor and operation mode (pressurized or vacuum). In addition, the membrane sensor has shown to be able to detect bentonite particles with a 0.3 mg L1 concentration in approximately 35 min in the vacuum mode. This membrane-based sensor is reliable, sensitive and low cost. It has potential applications in decentralized systems or in multi-channel monitoring of local conditions in a large plant. A disadvantage of this test is that it is necessary to use first an integrity test to validate the integrity of the sensor itself.
Regime of super-critical flux or non-sustainable condition
Regime of sub-critical flux or sustainable condition Clean permeate from intact membranes
Contaminated permeate by feed from broken membranes
Trans-membrane pressure signal ( )
Time
Fig. 3 – Concept of sustainable conditions for relative transmembrane pressure technique (Phattaranawik et al., 2008).
water research 44 (2010) 41–57
flow through the impaired fiber, ni is the number of the intact fibers and nd is the number of impaired fibers. It can be seen from Eq. (9) that the particle concentration in the permeate strongly depends on the particle concentration in the feed. In general, it is impossible to estimate the total number of particles in a water sample, because there are always particles which are smaller than the particle threshold of a particle counter. Below this threshold, the particles cannot be counted by the particle counter. It should be noted that, if it refers to particle counting, the counted particles are larger than the threshold of the particle counter. A particle counter can count and monitor particles with different sizes in the feed and filtrate using a laser-based light scattering technique. As mentioned above, particle counting is related to a given size or size distribution of particles. The threshold of particle counters will directly affect the sensitivity of this method for membrane integrity monitoring. The sensitivity of particle counters increases with the decreasing threshold. For this reason, particle counters should have a threshold as small as possible. However, the cost of counters will spurt if the threshold of the counter is smaller. Adham et al. (1995) speculated that the 0.5–1 mm range is too sensitive because contaminants in this size range are likely to take place under field conditions and the data of this range could mislead to an operational standpoint. Thus, they only counted particles larger than 1 mm. It should be noted that particle counters are non-specific; i.e., they detect all particles in the appropriate size range. This results in a lower sensitivity of the membrane integrity monitoring. Many factors for on-line particle counting configuration also affect the quality and the accuracy of results. Adham et al. (1995) affirmed that some measures have to be taken when using a more sensitive particle counter. Flow control devices are recommended since they are able to maintain the flow rate through the particle sensor within very strict limits. The ideal flow controller has to be operated without introducing severe mechanical forces (pumps, gears, etc.), pulsations in sample flow rate through the sensor or without conditions that allow the particle size distribution to change by means of flocculation, settling or contamination. The distance of the particle sensor from the sample source or the flow controller should be as short as possible. Tubes made of an inert material should be used to prevent the adhesion of particles. The sensors have to be cleaned and the electronic background noise has to be checked in periodical intervals. In addition, the sensor should be calibrated at least every year. Many studies (Adham et al., 1995; Panglisch et al., 1998; Glucina et al., 1997) were implemented to investigate how sensitive the particle counter has to be and how many membrane modules can be monitored by only one counter. The results about the membrane quantity which can be monitored by one counter are different. Adham et al. (1995) estimated that for a feed particle concentration of 200,000 particles mL1, a maximum membrane area of 18 m2 can be monitored by one particle counter in case of dead-end microfiltration with the Memtec system, while Glucina et al. (1997) estimated the maximum membrane area for a deadend filtration with Aquasource membranes at a particle feed concentration of 200,000 particles mL1 to be 385 m2. It has been demonstrated that the sensitivity of particle counting for
49
membrane integrity monitoring strongly depends on the particle concentration in feed solution and it increases with the rising feed concentration. This conclusion is consistent with the result obtained by Panglisch et al. (1998). That is to say, the number of membrane elements, which can be controlled by just one particle counter, increases with the increasing feed particle concentration. If the feed solution is relatively clean, differences before and after filtration would be beyond the limits of current particle counters and cannot be detected by the particle counter. Under these circumstances, particle counters are not able to detect water quality changes at the levels required to ensure pathogen removal such as Cryptosporidium and Giardia. As the two most frequently used indirect membrane integrity tests, particle counting is more expensive than turbidity monitoring but it has higher detection sensitivity than turbidity monitoring (Adham et al., 1995; Jacangelo et al., 1991). However, it should be noted that current particle counting presents a lack of sensitivity (less than 4 LRV) sufficient for membrane integrity monitoring, which has also been reported by Landsness (2001). Particle counting may not be suitable in some cases, especially in the case of membrane operation in dead-end mode on low turbidity water (high diluted effect). In addition, this method is able to count air bubbles like particles as a result of air entrapment, especially for the membrane systems that use air for backwashing. This may make development of a stable baseline value more difficult for particle counters. Particle counting is also susceptible to counting particles shedding from downstream plumbing (Farahbakhsh et al., 2004). Another problem for particle counters is that it maybe produce unstable output during startup and shutdown. Particle monitoring is similar to particle counting in principle. It provides an index of the water quality. Compared to particle counters which give a direct measurement of particle numbers by size, particle monitoring provides qualitative data based on a relative scale. This method is cheaper than particle counting. It offers less sensitivity for monitoring water quality changes due to membrane failures but is more sensitive than turbidity (Adham et al., 1995). Particle monitoring is seldom used, thus the water industry has limited experience with these devices.
3.7.
Turbidity monitoring
This test is based on the difference of turbidity between the feed water and the filtrate. An intact membrane would be expected to show a 90% reduction in turbidity from feed to filtrate. Turbidity monitoring is less expensive than particle counting but offers lower sensitivity, which is reportedly not adequate to respond to changes in membrane integrity (Hirata and Hashimoto, 1998). For instance, even filtrate of low turbidity –<0.1 NTU– can have significant bacterial contamination (Rajagopalan et al., 2001), limiting the operation of this technique as a separate method for membrane integrity monitoring. Compared to the conventional turbidimeters, laser turbidimeters can improve detection sensitivity in excess of two orders of magnitude over conventional turbidimeters and are able to measure very low turbidity in the range of 0–1 NTU. Since most MF and UF systems produce filtrate water consistently in the range of 0.01–0.05 NTU, the
50
water research 44 (2010) 41–57
laser turbidimeter may be well suited to monitor membrane integrity. However, the results obtained by Colvin et al. (2001) indicated that the laser turbidimeter is somewhat less sensitive to integrity breaches than particle counting. Other experiences using laser turbidimeter have been also reported (Banerjee et al., 2001; Van Hoof et al., 2001; Naismith, 2005). Turbidity monitoring is also subject to some of the same problems as particle counting. For example, if the feed water is relatively clean, the differences in measurements between feed and filtrate water would be beyond the capacity of current turbiditimeters. In addition, it is also subject to air entrapment error.
3.8.
Surrogate challenge tests
As an alternative to on-line indirect methods, surrogate challenge tests are proposed to improve sensitivity of membrane integrity monitoring and to guarantee more efficiently drinking water quality. This method overcomes the low detection sensitivity of conventional indirect methods and the drawbacks of conventional direct integrity testing systems, such as off-line operation and lack of a direct relationship between the measured data and the removal efficiency (log removal). Certain surrogate challenge tests are used to represent the pathogen retention characteristics by membranes. Walsh et al. (2005) employed dissolved organic carbon (DOC) and color measurements to evaluate UF membrane integrity and showed promising results in their ability to identify a significant increase in the concentration of dissolved contaminants within the permeate stream. This study demonstrated the need for robust on-line indirect test methodologies for monitoring membrane integrity in the drinking water industry. Sakaji (2001) reported that maintaining membrane integrity monitoring with challenge tests can validate high removal levels of 5 or 6 LRV. For drinking water treatment, the ultimate integrity test is the natural bacterial challenge test. Unfortunately, this test may result in additional problems, such as membrane biofouling or penetration of bacteria into permeate in case of compromised integrity. Therefore, the key point for surrogate challenge tests is to search for a suitable surrogate and corresponding measurement method. Surrogate challenge tests should be sensitive enough to detect changes in membrane integrity so that regulatory requirements for drinking water can be achieved. The precision and the minimum detection level of the measurement method directly affect the sensitivity of surrogate challenge tests. As for the choice of surrogate, multiple criteria have to be considered, such as well defined size, easily detectable, non-destructive, reasonable price and representative of pathogenic retention properties in drinking water treatment. Monodisperse surrogates are preferred because they can be more accurate to predict membrane retention based on the size exclusion (Causserand et al., 2002). Here, several surrogate challenge tests for monitoring lowpressure membrane integrity are introduced, including the spiked integrity monitoring system-SIM, microbial challenge tests, nanoscale probe challenge tests and magnetic particle challenge tests.
3.8.1.
Spiked integrity monitoring system – SIM
The spiked integrity monitoring system or SIM-system was developed by NORIT Membrane Technology in close co-operation with Water Supply Company North Holland and IWW Rhenish–Westphalian Institute for Water Research, which combines the accuracy of a challenge test with the speed of a pressure test, while keeping the system under test during operation (Van Hoof et al., 2001). For this test, high concentrations of powdered activated carbon (PAC) are spiked in the membrane feed, followed by the monitoring of particulate level in permeate with particle or turbidity monitors. Then, a log removal value can be calculated. It was reported that the SIM system provides increased sensitivity by increasing feed particle concentration (Kruithof et al., 2001). The SIM is an on-line membrane integrity test with high sensitivity. However, it is not suitable from a regulatory point of view, since the test results cannot be linked directly to pathogen removal. The PAC particle size does not remain constant during the test, and the particle size distribution may vary considerably, depending on the manufacturer, the dispersion mode of the PAC and the water quality. In addition, the problems derived from the measurement apparatus, which exist in particle counting test or turbidity monitoring test, also affect the sensitivity of this method. Finally, it should be noted that the PAC particles may occur in the permeate but it cannot be removed with a backwash due to the module capacity.
3.8.2.
Microbial challenge tests
Microbial challenge tests, e.g., bacteriophage and oocyst, spike a given concentration of microorganisms in the feed and measure the particulate levels in the permeate in the case of no chloration. Bacteriophages, a class of viruses that infect bacteria, are often used as microbial surrogates because they avoid the health risks of native pathogen. It can well represent viral pathogen removal by the membrane system because they are close in size, shape, and surface properties. Microbial challenge tests can provide a good sensitivity for fiber break detection and provide some indication of membrane disinfection efficiency. This method is also a basic research tool to allow the log removal calculation for a specific organism. However, microbial monitoring is not practical for routine test in large-scale membrane plants because of several reasons (Gitis et al., 2002). Firstly, the measurement of bacteriophage in the water is based on the plaque forming unit (PFU) count. Because the PFU count is very complicated and needs a long time (12–48 h), it cannot reflect the membrane integrity in real time. In addition, plaque techniques count a virus cluster as a single unit rather than reflecting the actual number of viruses. Secondly, bacteriophage inactivation and transport phenomena are always coupled and it is difficult to distinguish the decrease of bacteriophage which is due to membrane retention or biological inactivation. Thirdly, microbial monitoring may not correctly characterize the membrane integrity since bacterial re-growth may take place in the permeate piping system. So bacteria in the permeate may be over-estimated although there is no membrane failure. Trimboli et al. (2001) described the implementation of a bacillus spore challenge test to measure the integrity of a large MF membrane system. This test is sensitive but expensive. In addition, some researches which use
water research 44 (2010) 41–57
fluorescent-dye-labelled MS2 bacteriophages to evaluate the UF membrane integrity have also been reported (Gitis et al., 2006a,b). Generally, microbial challenge tests are performed simultaneously with pressure-based integrity tests in drinking water treatments to validate and improve membrane integrity monitoring techniques (Brehant et al., 2008).
3.8.3.
Nanoscale probe challenge tests
To efficiently detect nanoscale breaches in membranes, Gitis et al. (2006a,b) proposed applying new monodispersed nanoscale probes – citrate-stabilized (12 nm) or thiol-stabilized (15 nm) gold nanoparticles – to evaluate the UF membrane integrity. Gold nanoparticles are preferred to other metal nanoparticles due to the extremely low background level in water system (usually between 5 and 50 ppt), non-pathogenicity and safe use, high monodispersity, and inexpensive price. The gold nanoparticles challenge test assumes that the membrane retains particles by a size exclusion mechanism regardless of surface chemistry, vitality, pathogenesis, and other characteristics that differentiate organic from inorganic particles. Experimental results showed that breaches in membrane integrity were detected as early as the first appearance of small holes with an average diameter of 20 nm when applying these two nanoscale probes. These results indicate that gold nanoparticle challenge tests are sufficiently sensitive to detect breaches which permit viruses to pass through the membrane. Here, gold nanoparticles were detected using anodic stripping voltammetry (ASV). ASV system showed high sensitivity (low minimum detection level) to the order of a single part per billion, indicating the feasibility to develop the experimental protocol for simple and sensitive on-line detection. But ASV needs a prolonged run time or the addition of some other metal to resolve this problem. Another highly sensitive nanoparticle analysis method based on the laser-induced breakdown detection (LIBD) was found to be sensitive enough to monitor the particle passage through low-pressure membranes (Lipp et al., 2008). This may also provide a suitable on-line monitoring method for UF/MF membrane integrity. However, it is also reported (Lipp et al., 2008; Lohwacharin and Takizawa, 2009) that more significant membrane fouling due to intermediate pore blocking took place with small size of nanoparticles. Nanoparticle challenge tests provide a basis for developing a new on-line membrane integrity monitoring method with high sensitivity to detect virus-size breaches in membranes.
3.8.4.
Magnetic particle challenge tests
Currently on-line membrane integrity tests are generally analyzed by non-specific turbidimeters or particle counters. An alternative method of probing membrane integrity by the use of magnetically susceptible particles was developed to improve both specificity and sensitivity (Moulin, 2008; Rajagopalan et al., 2006; Deluhery and Rajagopalan, 2008). The principle of magnetic particle challenge tests is spiking a certain concentration of magnetic particles in the feed and detecting the material magnetic properties in the filtrate. The particle size used should be greater than the membrane pore size to ensure that there is no particle passing through an intact membrane. In the case of an impaired membrane, particles are detected in the filtrate by an appropriate
51
magnetic sensor, indicating the loss of membrane integrity. Magnetic sensors include superconducting quantum interference device (SQUID) magnetometer, giant magneto-resistive (GMR) sensors, measurement of magnetic permeability and measurement of magnetic susceptibility. Non-magnetic sensors can also be used but it is non-specific. Superparamagnetic particles are preferred for the test due to their high magnetic susceptibility and low density. High magnetic susceptibility makes it more easily detected by magnetic sensors and decreases the external magnetic field strength needed for high capture efficiency if particle capture is needed. Lower particle densities reduce the possibility of particle sedimentation at low flow rates. Here, the particle size and detection sensor have a decisive effect on the sensitivity of this membrane integrity test. For example, nanoscale particles can detect nanometric breaches of the membrane. Recently, some researchers (Rajagopalan et al., 2006; Deluhery and Rajagopalan, 2008) proposed to use a magnetic field to concentrate the particles for detection, which allows the use of smaller and less powerful sensors and thus realizes cost savings or enhanced detection limits of this integrity test. They implemented the membrane integrity test with magnetic particles of 1 mm diameter and the particles in the filtrate were successfully separated and concentrated in the influence of a magnetic field. Moulin (2008) is the first to propose an integrity test based on the magnetic characteristics without a concentration step due to a high sensitivity of the analytical apparatus (LRV >6). Lab-scale results about this method have been reported. On-line magnetic challenge tests for monitoring membrane integrity, with the advantages of detection specificity, high detection sensitivity and on-line operation, is a plausible one for large-scale applications.
4.
Operation of membrane integrity tests
Thus far, the most widely used membrane configuration in drinking water treatment is hollow fiber membrane filtration. The main membrane system suppliers have employed hollow fiber membranes. In a membrane drinking water treatment plant, the detection of membrane system failures requires that an effective membrane integrity test system at an acceptable frequency is used. As the most widely used methods, the PDT and DAF test are considered to be reliable for monitoring membrane system integrity. The PDT and DAF test are generally proposed to be conducted according to the same sequences: (1) detecting compromised rack(s) in the plant; (2) detecting compromised module(s) in the rack; and (3) detecting compromised fiber(s) in the module. Current PDT or DAF test for hollow fiber membranes are typically set to alarm against parameters based on an absolute size removal (4 mm and greater) and a particle log removal value (4–5 LRV). Some companies set alarms against other measured variables, e.g., feed turbidity, high pH in treated water. These criteria reflect the international position but do not satisfy the regulatory requirements for an absolute removal of more than 1 mm. Higher starting pressure for the test would enable smaller defects (e.g., 1 mm) to be detected thereby satisfying the current requirements. However, higher
52
Table 2 – Advantages and disadvantages of different membrane integrity tests. Membrane integrity test Pressure-driven tests PDT
Advantages
(1) Independent of feed water quality (2) Low maintenance (3) Non-destructive (4) Reliable and high sensitive to detect membrane breaches to meet the regulatory requirements, 4.5–5 LRV (5) The most widely used and accepted by both utilities and primacy agencies
Bubble point test
Vacuum decay test
(6) High degree of automation (1), (2) and (3) Are the same as PDT (4) More sensitive than the PDT, even more than 6 LRV (5) Widespread use (6) Ease to conduct and accuracy of test (when measuring water displacement)
(1) Is the same as PDT
(2) Ease of conducting (3) Useful tool for pinpointing compromised fiber(s) and leaks identified by other test methods (4) High sensitivity of 6 LRV for Memcor CMF process (1) and (3) Are the same as PDT (2) Useful as a screening procedure
(3) Ability to test spiral-wound membranes or other systems that cannot be pressurized on the filtrate side of the membranes
Typical module test time
(1) Off-line operation (2) No measurement of filtrate water quality (3) Not sufficiently sensitive to detect nanoscale breaches (4) Need more sensitive pressure sensor when the number of fibers tested is important (5) Potential to yield false–positive results if the membrane is not fully wetted
10 min (5 min for Memsure PDT)
(1), (2), (3) Are the same as PDT (4) Not included as standard equipment for MF and UF systems (5) Sensitive to temperature (6) Requires some additional pipe work when measuring water displacement
15 min (5 min for Memsure DAF)
Adham et al., 1995 Johnson, 1997, 1998 Landsness, 2001 U.S. EPA, 2001, 2003
Jackson, 2001
Adham et al., 1995 Johnson, 1997, 1998 Trimboli et al., 2001 Cheryan, 1998
(1) and (2) Are the same as PDT, and the module to be tested should be removed from the rack (3) Manual application (4) Practically unattainable for UF membranes
Depends on the bubble point
(1) and (2) Are the same as PDT (3) Rarely performed as a mean of absolute verification of a leak in drinking water treatments (4) Should be repeated to confirm its reproducibility
Several minutes
(5) Difficult in removing entrained air after the test has been completed
References
Jacangelo et al., 1997 Jackson, 2001 Randles, 1997
Giglia and Krishnan, 2008
FILMTEC Membranes, troubleshooting: membrane element evaluation
water research 44 (2010) 41–57
DAF test
Disadvantages
Other direct tests Binary gas integrity test
Acoustic sensor test
(1) Non-destructive (2) Higher detection sensitivity than the single gas test, greater than 6 LRV (3) Independent of feed water quality
(1) Simple and non-destructive (2) Economic competitive
Liquid–liquid porosimetry
(3) On-line sensor (4) Independent of feed water quality (5) High sensitivity of 6 LRV of viruses (1) Non-destructive and relatively simple to perform (2) Independent of feed water quality
Indirect tests Particle counting
(1) Continuous and on-line operation (2) Convenient for routine qualitative monitoring (3) Independent of membrane configurations
(4) Widespread use and familiarity in the water industry (5) More sensitive than particle and turbidity monitoring
Particle monitoring
(1), (2), (3) Are the same as particle counting (4) Significantly lower cost than particle counters (5) More sensitive than turbidity monitoring (6) No calibration required
(3) Just in pilot-scale study (4) the storage of gas and the gas toxicity (e.g., C2F6) (1) Depends on the background noise and flow rate (2) No measurement of filtrate water quality
Longer than single gas tests
Giglia and Krishnan, 2008
Direct
Adham et al., 1995 Laıˆne et al., 1998 Jackson, 2001
(1) Off-line operation
Longer than pressurebased tests
(2) No measurement of filtrate water quality (3) No report on practical operations
(1) Low detection sensitivity (<4 LRV) (2) Strongly depends on feed concentration and operating conditions (3) Susceptible to entrapment air as particles and to count particles shedding from downstream plumbing (4) Relatively high cost (5) Produces unstable output during startup and shutdown (6) Periodic cleaning and calibration of the sensor (7) Flow control devices are recommended before the sensor (1), (2), (3) Are the same as particle counting (4) Seldom used in the water industry
DiLeo and Phillips, 1994, 1995 Phillips and DiLeo, 1996 Gekas and Zhang, 1989
Direct
Jacangelo et al., 1991 Adham et al., 1995
Panglisch et al., 1998
water research 44 (2010) 41–57
(3) More feasible than gas–liquid porosimetry to characterize UF membrane
(1) Off-line operation (2) No measurement of filtrate water quality
Landsness, 2001 Glucina et al., 1997 Farahbakhsh et al., 2004 Jackson, 2001
Direct
Adham et al., 1995 Jackson, 2001
53
(continued on next page)
54
Table 2 (continued) Membrane integrity test Turbidity monitoring
Membrane-based sensor
Surrogate challenge tests SIM
(1), (2), (3), (4) Are the same as particle monitoring (5) Widely used in drinking water plants
Disadvantages
Typical module test time
(1), (2), (3) Are the same as particle counting
Direct
(1) Needs an integrity test firstly for the integrity test (2) No further information about membrane pilot operation
35 min
(1) On-line operation (2) High detection sensitivity
(1) High feed concentration (2) Cannot be linked directly to the virus removal (3) The PAC size may vary considerably during the test (4) Subject to the same problems that particle counters or turbidity monitoring (1) Natural bacterial challenge test is destructive (2) Long time for microbial analysis and cannot reflect membrane integrity in time
30 min
(2) Generally performed simultaneously with other direct integrity tests to validate and improve membrane integrity monitoring techniques
(1) On-line operation
(2) High detection sensitivity
Magnetic particle challenge tests
(3) Ability to detect virus-size breaches (1), (2) Are the same as the nanoscale probe challenge test (3) Ability to detect virus-size breaches when using magnetic nanoparticles as surrogates (4) Simple and detection specificity (5) Low cost (6) Independent of feed water quality
(1) Significant membrane fouling due to the pore blocking with the smaller size of nanoparticles (2) Cannot be linked directly to the pathogen removal (3) Just in pilot-scale study (1) Just in pilot-scale study, in progress on industrial scale (2) Need more information on the effect on nanomaterials on health
Colvin et al., 2001 Naismith, 2005 Jackson, 2001 Phattaranawik et al., 2007, 2008
Van Hoof et al., 2001 Kruithof et al., 2001 Jackson, 2001
24–48 h
Brehant et al., 2008 Trimboli et al., 2001
(3) Not continuous and off-line operation Nanoscale probe challenge tests
Banerjee et al., 2001 Van Hoof et al., 2001
(1) Can be operated on-line and continuously (2) Sensitive and low cost
(1) Sensitive
References
Gitis et al., 2002
Direct
Jackson, 2001 Gitis et al., 2006a,b
Lipp et al., 2008
Direct
Lohwacharin and Takizawa, 2009 Moulin 2008 Rajagopalan et al., 2006 Deluhery and Rajagopalan, 2008
water research 44 (2010) 41–57
Microbial challenge tests
Advantages
water research 44 (2010) 41–57
start pressures may cause some membranes to experience higher defect rate than that are currently assumed. For different commercially available membranes, an annual fiber failure rate has been found between 1 and 10 per million fibers (Gijsbertsen et al., 2006). The frequency of the membrane integrity test is important for maintaining specific LRV requirement. There are no common guidelines for the frequency of membrane integrity tests. Membrane integrity test frequencies need to be related to a risk assessment as the first priority and operational feasibility as the second priority. Given an expected number of fiber breaks per year and a LRV target, the frequency of membrane integrity tests can be calculated based on a simple mode or a statistical model (Bennett, 2005). Membrane integrity tests can be implemented once a day, once every 72 h or longer, or after cleaning. It has been summarized that, in most cases, a weekly test is required to ensure LRV more than 4.5 (Pearce, 2007). For Bristol Water, a PDT frequency of once every week to once every other week is recommended to provide a margin of safety for the target of 4 LRV (Bennett, 2005). Generally, a daily test frequency would be beneficial when ascertaining system integrity, which is consistent with the continuous sampling and analysis practice. In general, the data from membrane integrity tests during operation should be kept between 2 and 5 years (Jackson, 2001). In general, a perfect method for low-pressure membrane integrity monitoring should be low cost, simple, on-line and continuous. In addition, it should be reliable and highly sensitive to detect membrane breaches, even for nanoscale breaches. Here, a summarization of current different membrane integrity tests is shown in Table 2.
5.
Conclusions
Membrane integrity tests are very important for membranebased drinking water treatment because they can ensure the quality of treated water, especially in terms of pathogen removal. The criteria of assessing a method for monitoring membrane integrity include sensitivity, continuity, reliability and cost effectiveness. Generally, the sensitivity of membrane integrity monitoring methods is expressed in term of log removal value (LRV) of pathogens for drinking water treatments. Of current different membrane integrity tests, PDT and DAF test are the most frequently used and considered to be simple and reliable. These two tests are able to detect membrane integrity at high levels of 5–6 LRV, even more than 6 LRV for DAF test. However, they are conducted off-line. In contrast, indirect integrity tests, such as particle counting and turbidity monitoring, are performed simply and on-line but have lower sensitivity. They are not able to detect water quality changes at the levels required to ensure pathogen removal. To realize reliable and efficient on-line membrane integrity monitoring, some other methods are proposed, including acoustic sensor method, liquid porosimetry, surrogate challenge tests, etc. All these methods have their own advantages and disadvantages. Membrane integrity tests are specific for the type of membrane used and are dependent of membrane manufacturer and membrane system supplier. In
55
general, membrane suppliers have their own membrane integrity test procedures based on 4–5 LRV. During practical operation, the frequency of membrane integrity tests is very important to achieve the required LRV. So, to better satisfy the regulatory requirements of the drinking water industry, it remains urgent to develop an alternative on-line monitoring technique for quick, accurate, simple, continuous and relatively inexpensive evaluation of the low-pressure membrane (UF and MF) integrity. Based on the above information, surrogate challenge tests with nanometric material are interesting and promising because it can ensure the disinfection efficiency of the UF membrane and make it possible to realize the required objective with the development of more accurate and advanced measurement apparatus. For this purpose, an alternative on-line UF membrane integrity test by using magnetic nanoparticles as surrogates reveals that this challenge test is suitable for largescale drinking water applications. But in this case, more information on the effect of nanometric material on health is needed.
references
Adham, S.S., Jacangelo, J.G., 1994. Assessing the Reliability of Low Pressure Membrane Systems for Microbial Removal. The National Conference on Environmental Engineering, Boulder, CO. Adham, S.S., Jacangelo, J.G., Laıˆne, J.M., 1995. Low-pressure membranes: assessing integrity. Journal AWWA 3, 62–75. Banerjee, A., Lambertson, M., Carlson, K., 1999. Sub-micron Particles in Drinking Water and their Role in Monitoring the Performance of Filtration Processes. The AWWA Water Quality Technology Conference, Tampa FL. Banerjee, A., Lambertson, M., Lozier, J., Colvin, C., 2001. Monitoring membrane integrity using high sensitivity laser turbidimetry. Water Science & Technology: Water Supply 1 (5-6), 273–276. Bennett, A., 2005. Maintaining the integrity of filtration systems. Filtration & Separation 42 (1), 30–33. Bennett, A., 2008. Drinking water: pathogen removal from water – technologies and techniques. Filtration & Separation 45 (10), 14–16. Brehant, A., Glucina, K., Lemoigne, I., Laine, J.M., 2008. Risk Management Approach for Monitoring UF Membrane Integrity and Experimental Validation using MS2-phages. IWA World Water Congress, Austria Vienna. Causserand, C., Aimar, P., Vilani, C., Zambeli, T., 2002. Study of the effects of defect in ultrafiltration membranes on the water flux and the molecular weight cut-off. Desalination 149 (1-3), 485–491. Cheryan, M., 1998. Ultrafiltration and Microfiltration Handbook. Technomic Publishing Company, Lancaster, Pennsylvania. Colvin, C., Brauer, R., Dinatale, N., Seribner, T., 2001. Comparing Laser Turbidimetry with Conventional Methods for Monitoring MF and UF Membrane Integrity. The AWWA Membrane Technology Conference, San Antonio, TX. Crozes, G.F., Sethi, S., Xi, B., Curl, J., Marin˜as, B., 2002. Improving membrane integrity monitoring indirect methods to reduce plant downtime and increase microbial removal credit. Desalination 149 (1-3), 493–497. David, H., Furukawa, P.E., Ch, E., March 2008. NWRI Final Project Report: a Global Perspective of Low Pressure Membranes. USA, California.
56
water research 44 (2010) 41–57
Deluhery, J., Rajagopalan, N., 2008. Use of paramagnetic particles in membrane integrity testing. Journal of Membrane Science 318 (1-2), 176–181. DiLeo, A.J., Phillips, M.W., 1994. Integrity test for membranes. U.S. Patent 5,282,380. DiLeo, A.J., Phillips, M.W., 1995. Process for evaluating solute retention characteristics of membranes. U.S. Patent 5,457,986. Edwards, D., Donn, A., Meadowcroft, C., 2001. Membrane solution to a ‘‘significant risk’’ Cryptosporidium groundwater source. Dasalination 137 (1-3), 193–198. Faber, D., Pearce, G., 2004. ‘HYDRAcap ‘‘Capillary Ultrafiltration Membrane Integrity Verification Plan’. Hydranautics Inc, USA. http://www.membranes.com. Farahbakhsh, K., Smith, D.W., 2004. Estimating air diffusion contribution to pressure decay during membrane integrity tests. Journal of Membrane Science 237 (1-2), 203–212. Farahbakhsh, K., Adham, S.S., Smith, D.W., 2004. Monitoring the integrity of low-pressure water treatment membranes. Journal AWWA 95 (6), 95–107. Field, R.W., Wu, D., Howell, J.A., Gupta, B.B., 1995. Critical flux concept for microfiltration fouling. Journal of Membrane Science 100 (3), 259–272. FILMTEC Membranes, Troubleshooting: Membrane Element Evaluation (http://www.dow.com/PublishedLiterature/dh_ 003d/0901b8038003d390.pdf?filepath¼liquidseps/pdfs/noreg/ 609-02115.pdf&fromPage¼GetDoc). Freeman, S.D.N., Logsdon, G.S., Harris, A.T., Moles, A.D., 1996. Evaluation of Microfiltration Performance with Bacillus Spore, Particle Count and Particle Index Measurements. The AWWA Annual Conference, Toronto, Ontario, pp. 1013-1039. Gadam, S., Philippe, M., Orlando, S., Kuriyel, R., Pearl, S., Zydney, A., 1997. A liquid porosimetry technique for correlating intrinsic protein sieving: application in ultrafiltration processes. Journal of Membrane Science 133 (1), 111–125. Gekas, V., Zhang, W., 1989. Membrane characterization using porosimetry and contact angle measurements: a comparison with experimental UF results. Process Biochemistry 24 (5), 159–166. Giglia, S., Krishnan, M., 2008. High sensitivity binary gas integrity test for membrane filters. Journal of Membrane Science 323 (1), 60–66. Gijsbertsen, A.J., Cornelissen, E.R., Hofman, J.A.M.H., 2006. Fiber failure frequency and causes of hollow fiber integrity loss. Desalination 194 (1-3), 251–258. Gitis, V., Adin, A., Nasser, A., Gun, J., Lev, O., 2002. Fluorescent dye labeled bacteriophages – a new tracer for the investigation of viral transport in porous media: 1. Introduction and characterization. Water Research 36 (17), 4227–4234. Gitis, V., Haught, R.C., Clark, R.M., Gun, J., Lev, O., 2006a. Application of nanoscale probes for the evaluation of the integrity of ultrafiltration membranes. Journal of Membrane Science 276 (1-2), 185–192. Gitis, V., Haught, R.C., Clark, R.M., Gun, J., Lev, O., 2006b. Nanoscale probes for the evaluation of the integrity of ultrafiltration membranes. Journal of Membrane Science 276 (1-2), 199–207. Glucina, K., Do-Quang, Z., Laine, J.M., 1997. Assessment of Particle Counting Method for Hollow Fiber Membrane Integrity. Workshop on Membranes in Drinking Water Production – Technical Innovations and Health Aspects, Italy L’Aquila, pp. 20–23. Hirata, T., Hashimoto, A., 1998. Experimental assessment of the efficacy of microfiltration and ultrafiltration for Cryptosporidium removal. Water Science and Technology 38 (12), 103–107. Howell, J.A., 1995. Sub-critical flux operation of microfiltration. Journal of Membrane Science 107 (1-2), 165–171.
Jacangelo, J., Adham, S., Laine, J.-M., 1997. Membrane Filtration for Microbial Removal, Report No. 90715, American Water Works Association Research Foundation, Denver, CO. Jacangelo, J.G., Laine, J.M., Carns, K.E., Cummings, E.W., Mallevialle, J., 1991. Low-pressure membrane filtration for removing Giardia and microbial indicators. Journal AWWA 83, 97–106. Jackson, C., 2001. Review of the adequacy of existing proposals for membrane integrity monitoring, DWI ref 43/2/159. Johnson, W.T., 1997. Automatic monitoring of membrane integrity in microfiltration systems. Desalination 113 (2-3), 303–307. Johnson, W.T., 1998. Predicting log removal performance of membrane systems using in-situ integrity testing. Filtration & Separation 1 (35), 26–29. Kruithof, J.C., Kamp, P.C., Folmer, H.C., Nederlof, M.M., Van Hoof, S.C.J.M., 2001. Development of a membrane integrity monitoring strategy for the RO/UF Heemskerk drinking water treatment plant. Water Science & Technology: Water Supply 1 (5-6), 261–271. Laıˆne, J.M., Glucina, K., Chamant, M., Simonie, P., 1998. Acoustic sensor: a novel technique for low pressure membrane integrity monitoring. Desalination 119 (1-3), 73–77. Landsness, L.B., 2001. Accepting MF/UF Technology – Making a Final Cut. The American Water Works Association 2001 Membrane Technology Conference, TX. Lipp, P., Mu¨ller, U., Hetzer, B., Wagner, T., 2008. Characterization of Particulate Fouling and Breakthrough During Low Pressure Membrane Filtration. Membranes in Drinking Water Production and Waste Water Treatment, Toulouse. Lohwacharin, J., Takizawa, S., 2009. Effects of nanoparticles on the ultrafiltration of surface water. Journal of Membrane Science 326 (2), 354–362. Moch, I., Paulson, D.J., September 2003. Establishing an Integrity Standard for Membrane Systems. International Desalination Association World Congress, Bahamas. Moulin, P., 2008. Patent Number(s): FR2901607–A1 EP1862791-A2. Naismith, J., 2005. Membrane integrity – direct turbidity measurement of filtrate from MF membrane module at an operating potable water treatment plant. Desalination 179 (1-3), 25–30. Panglisch, S., Deinert, U., Dautzenberg, W., Kiepke, O., Gimbel, R., 1998. Monitoring the integrity of capillary membranes by particle counters. Desalination 119 (1-3), 65–72. Pearce, G., 2007. Water and wastewater filtration: membrane module format. Filtration & Separation 44 (4), 31–33. Phattaranawik, J., Fane, A.G., Wong, F.S., 2007. Detection apparatus and method utilizing membranes and ratio of transmembrane pressures. PCT/SG2007/000130, WO/2007/ 129994. Phattaranawik, J., Fane, A.G., Wong, F.S., 2008. Novel membranebased sensor for online membrane integrity monitoring. Journal of Membrane Science 323 (1), 113–124. Phillips, M.W., DiLeo, A.J., 1996. A validatable porosimetric technique for verifying the integrity of virus retentive membranes. Biologicals 24 (3), 243–253. Rajagopalan, N., Qi, S., Pickowitz, J.P., 2001. Field Evaluation of Ceramic Microfiltration Membrane in Drinking Water Application. The AWWA Membrane Technology Conference, San Antonio, TX. Rajagopalan, N., Rusk, T., Sanford, R., 2006. Methods and systems for membrane testing. US Patent 7, 758. 011. Randles, N., 1997. Large scale operating experience in membrane systems for water and waste water reclamation. Desalination 108 (1-3), 205–211. Sakaji, E.R., 2001. California Surface Water Treatment: Alternative Filtration Technology. CA Dept of Health Service.
water research 44 (2010) 41–57
Trimboli, P., Lozier, J., Johnson, W., 2001. Demonstrating the integrity of a large scale microfiltration plant using a bacillus spore challenge test. Water Science & Technology: Water Supply 1 (5-6), 1–12. U.S. EPA, 2001. Low-Pressure Membrane Filtration for Pathogen Removal: Application Implementation, and Regulatory Issues. EPA 815-c-01–001, U.S. Environmental Protection Agency Office of Water. U.S. EPA, 2003. Membrane Filtration Guidance Manual. EPA 915-d03–008, U.S. Environmental Protection Agency Office of Water. Van Hoof, S.C.J.M., Broens, L., Nahrstedt, A., Panglisch, S., Gimbel, R., Hagmeyer, G., Gimbel, R., 2001. Development of
57
a New Testing System. The AWWA Membrane Technology Conference, San Antonio, TX. Walsh, M.E., Chaulk, M.P., Gagnon, G.A., 2005. Indirect integrity testing on a pilot-scale UF membrane. Journal of Water Supply: Research and Technology – AQUA 54 (2), 105–114. Weller, S., Steiner, W., 1950. Separation of gases by fractional permeation through membranes. Journal of Applied Physics 21, 279–283. Zondervan, E., Zwijinenburg, A., Roffel, B., 2007. Statistical analysis of data from accelerated aging tests of PES UF membranes. Journal of Membrane Science 300 (1-2), 111–116.
water research 44 (2010) 59–68
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Consecutive chemical cleaning of fouled PVC membrane using NaOH and ethanol during ultrafiltration of river water Jia-yu Tian a,*, Zhong-lin Chen a, Yan-ling Yang b, Heng Liang a, Jun Nan a, Gui-bai Li a a
State Key Laboratory of Urban Water Resource and Environment, School of Municipal and Environmental Engineering, Harbin Institute of Technology, No. 73 Huanghe Road, Nangang District, Harbin 150090, P.R. China b Beijing Key Laboratory of Water Quality Science and Water Environment Recovery Engineering, Beijing University of Technology, No. 100 Pingleyuan, Chaoyang District, Beijing 100022, P.R. China
article info
abstract
Article history:
Chemical cleaning of fouled hollow-fiber polyvinyl chloride (PVC) membrane with the
Received 15 April 2009
consecutive use of NaOH and ethanol during ultrafiltration of river water was investigated
Received in revised form
in the study. Results showed that through the chemical cleaning with 1% NaOH for 30 min,
9 August 2009
a negative cleaning efficiency of 14.6% was observed for the PVC membrane. This might
Accepted 30 August 2009
be due to the increase of membrane hydrophobicity, which was reflected by the increase of
Available online 8 September 2009
contact angle from 69.7 to 87.6 . On the other hand, the cleaning efficiency of 85.1% was obtained by the consecutive cleaning with 30 min of 1% NaOH and 30 min of ethanol.
Keywords:
Individual ethanol cleaning could remove 48.5% of the irreversible resistance, indicating
Polyvinyl chloride (PVC) membrane
that NaOH cleaning also made its contribution (36.6%) to the removal of membrane fou-
Membrane fouling
lants. Scanning electronic microscopy (SEM) and atomic force microscopy (AFM) analyses
Chemical cleaning
demonstrated that both NaOH and ethanol were not only able to eliminate the foulants on
Ethanol
membrane surface, but also able to remove the in-pore fouling of the PVC membrane. The
Surface water
synergetic effects for removing membrane foulants were observed between the NaOH and ethanol. Furthermore, ethanol could also restore the hydrophilicity of the membrane by decreasing the contact angle from 87.6 to 71.4 . Considering that ethanol is easy to be used and reclaimed, the consecutive chemical cleaning by alkali and ethanol is recommended for PVC membrane in filtration of surface water. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Low-pressure, hollow-fiber (LPHF) membrane filtration has been recognized as the most promising technology for drinking water treatment in the 21st century, all over the world (Huang et al., 2007). Microfiltration (MF) and ultrafiltration (UF) are able to remove particles and colloids from the raw water almost completely, and considered as the attractive substitutes for the conventional clarification and filtration units. Furthermore, MF and UF exhibit significant advantages
in controlling microorganisms and pathogens (Peter-Varbanets et al., 2009). Recent studies also demonstrate that MF and UF are effective for virus removal when combined with coagulation or adsorption (Fiksdal and Leiknes, 2006; Oh et al., 2007; Shirasaki et al., 2009). Therefore, it is reasonable to consider that MF or UF together with suitable pre-treatment or post-treatment is the new generation of process for advanced drinking water treatment. Although MF and UF are now attracting more and more attention, and being practically applied in a wider geography
* Corresponding author. Tel. þ86 451 86284512; fax: þ86 451 86282100. E-mail address:
[email protected] (J.-y. Tian). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.08.053
60
water research 44 (2010) 59–68
with a larger treatment capacity (Huang et al., 2007; Huang et al., 2008), membrane fouling still remains as one of the major issues regarding the membrane filtration technology (Choi et al., 2008; Fabris et al., 2007; Meng et al., 2009). Three fouling models have been developed to address the fouling mechanisms in various situations, including pore constriction, pore blocking, and cake formation (Lee et al., 2008a; Sun et al., 2008). From the practical point of view, membrane fouling is generally caused by initial internal-pore deposition or pore blocking, followed by gel and cake layer formation (Meng et al., 2007). Many substances in surface water could be potential foulants to membrane, such as iron, manganese, colloids, particles, etc.. Among which, natural organic matter (NOM) had been identified as one of the most important foulants (Lee et al., 2008b). Hydrophobic organic compounds with high aromaticity were traditionally believed to make the major contribution to membrane fouling (Gray et al., 2008). Later researches suggested that hydrophilic organic matter in surface water were the main membrane fouling components (Gray et al., 2008; Lee et al., 2008b). More recently, colloidal NOM was brought forward and considered as the primary membrane foulant (Costa et al., 2006; Lee et al., 2006). To be more complicated, Gray et al. (2004) and Jermann et al. (2007) found that the interaction among various NOM fractions determined the potential of membrane fouling; while Li and Elimelech (2006) observed a significant synergistic effect between inorganic colloids and NOM in the combined fouling of nanofiltration (NF) membrane; both of them are universally presented in natural water. Apart from the foulants in raw water, many other factors may also influence the fouling behavior during membrane filtration process, such as membrane properties (hydrophilicity, roughness and surface charge) and solution chemistry (pH, ion strength and divalent-cation content) (Choi et al., 2008; Costa et al., 2006). To alleviate membrane fouling, some strategies have been applied to immersed membrane systems (Meng et al., 2008; Ng et al., 2006), including (1) increasing back-transportation through intermittent suction; (2) enhancing cross-velocity by proper aeration; (3) backwashing the membrane periodically; (4) operating under the critical flux. Although there are quite a few of the preventing and mitigating approaches, membrane fouling will inevitably take place after a period of filtration operation. Thus, chemical cleaning is necessary to remove the foulants deposited on membrane surface or in membrane pores for the restoration of membrane permeability (Ang et al., 2006). Several literatures had been reported on the topic of chemical cleaning of membrane for filtration of surface water (Ang et al., 2006; Kimura et al., 2004; Lee and Elimelech, 2007; Liang et al., 2008; Yamamura et al., 2007a; Yamamura et al., 2007b; Zondervan and Roffel, 2007). Generally, the cleaning reagents used could be classified into five categories: (1) acids (HCl, oxalic acid, and citric acid), (2) alkalis (NaOH and Na2CO3), (3) metal chelating agents (EDTA), (4) surfactants (SDS), and (5) oxidants (NaClO and H2O2). Although the results showed that the cleaning effectiveness was highly dependent on the feed water qualities and was membrane-specific (in terms of material and pore size), acid and alkali were the most common chemicals
for the removal of inorganic and organic foulants, respectively. Ethanol is often used as a kind of wetting reagent for membrane in laboratory researches (Gray et al., 2008; Lu et al., 2008). In this paper, ethanol in combination with alkali was brought forward and investigated for the chemical cleaning of fouled hollow-fiber membrane during ultrafiltration of river water; the involved mechanisms were also discussed.
2.
Material and methods
2.1.
Experimental set-up
The experimental set-up was schematically shown in Fig. 1, which was mainly composed of a raw water tank, a peristaltic pump, a pressure transducer and a data acquisition system. The raw water was taken from Songhua River, a main drinking water source in Northeast of China. The membrane module was vertically immersed in the raw water tank, which had an effective volume of 1.2 L. The permeate was suck from the tank by a peristaltic pump (BT100-1J, Longer Pump, China), and then recycled back to the raw water tank. The transmembrane pressure (TMP) was monitored by a pressure transducer (PTP708, Tuopo Electric, Foshan, China). A computer together with a data acquisition software was used to record the TMP data in real time.
2.2.
The hollow-fiber PVC membrane module
The hollow-fiber membrane module used in this study was made in the laboratory, as shown in the photo in Fig. 1. The membrane fibers were made of Polyvinyl Chloride (PVC), with the inner and outer diameters of 0.85 mm and 1.45 mm, respectively, which were provided by Suzhou Litree Ultrafiltration Membrane Technology Co. Ltd., China. According to the manufacturer, the average pore size of the membrane was 0.01 mm. Each membrane module contained 10 membrane fibers, with an effective length of 22.0 cm, corresponding to a membrane area of 0.01 m2.
Fig. 1 – Schematic of the experimental set-up (the photo shows the submerged hollow-fiber membrane module used in the experiments).
61
water research 44 (2010) 59–68
2.3.
Membrane fouling and cleaning
The membrane fouling behavior during filtration operation could be characterized by the resistance-in-series model (Fabris et al., 2007). For constant-flux filtration, the resistancein-series model could be expressed as follows (Zheng et al., 2009): Rm þ Rf ¼ Rm þ Rrev þ Rirr ¼
DP hJ
(1)
Where DP is the trans-membrane pressure (TMP, Pa); h is the dynamic viscosity (Pa s); J is the membrane flux (m3/m2$s); Rm is the intrinsic membrane resistance (1/m); Rf is the total fouling resistance, including the physically reversible fouling resistance (Rrev, 1/m) and physically irreversible fouling resistance (Rirr, 1/m). To make the experiments conducted under the same operating conditions, a new membrane module was employed in each run of the experiments. Prior to ultrafiltration of river water, the new membrane was wetted with ethanol (analytical grade) for at least 60 min. It is well known that membrane flux imposes significant influence on the membrane fouling, and the concept of ‘‘critical flux’’ has been widely applied to membrane filtration processes (Guglielmi et al., 2007). When the membrane is operated under sub-critical flux condition, the membrane fouling developed very slowly; however, the rate of membrane fouling would become much faster when a flux higher than the critical flux is employed. This investigation focused on the consecutive chemical cleaning of fouled PVC membrane by NaOH and ethanol. To accelerate the membrane fouling process, the permeate flux was set at a relatively high value of 40 L/m2 h. Before each run of the experiments, the initial membrane resistance (Rm) of the new membrane was determined by filtering deionized (DI) water until the constant TMP was achieved (20 min was shown to be enough). After that, the feed water was changed to the river water (1.2 L) and the fouling experiment was performed for 6.0 h. At the end of the fouling experiment, the fouled membrane was thoroughly wiped with a sponge and rinsed with DI water, to remove the accumulated cake on the membrane surface that causing physically reversible fouling resistance (Rrev) (Yamamura et al., 2007b). Then, the wiped membrane was subjected to filtration of DI water for 20 min, to obtain the irreversible membrane fouling resistance (Rirr). After that, 1% of NaOH, 2% of citric acid, or ethanol was added to the raw water tank and filtration with the chemical solution was conducted for 30 min. Thereafter, the membrane was subjected to filtration with DI water (20 min) again to determine the Rirr after chemical cleaning. Thus, the removal of Rirr through the chemical cleaning could be calculated; and the cleaning efficiency was defined as the removal efficiency of Rirr after chemical cleaning by different reagents in this investigation.
2.4.
Analytical methods
Turbidity was measured by a turbidimeter (TURBO550, WTW, Germany). Total organic carbon (TOC) and dissolved organic carbon (DOC) were determined with a TOC analyzer (TOC-
VCPH, SHIMADZU, Japan). UV absorbance at 254 nm (UV254) was analyzed using a spectrometer (UV754, CANY, China). Concentrations of metals including Al, Fe, Mn, Ca and Mg were determined with an inductively coupled plasma-atomic emission spectroscopy (ICP-AES, Optima 5300DV, PerkinElmer, USA). The surface and cross-section of membrane samples before and after chemical cleaning were observed under a scanning electron microscopy (SEM) (HITACHI S4800 HSD, Japan). For cross-section observation, the membrane fibers were firstly frozen in liquid nitrogen and fractured to obtain the sharp edge. Then, the membrane samples were goldcoated and examined with the SEM. Atomic force microscopy (AFM, Digital Instruments, Veeco, USA) was also employed to determine the surface morphology of the hollow-fiber PVC membrane used in the study. AFM observation was performed under the tapping mode, by using a tip made of etched single crystal silicon. The membrane samples were scanned over a range of 10 mm 10 mm. The obtained data were analyzed using the software of Nanoscope V5.30. The contact angle of the membrane samples was measured with a goniometer (DSA-100, KRUSS, Germany). The standard sessile drop method was employed, with Milli-Q (MQ) water as the reference liquid. Five measurements were performed for each membrane sample with the MQ water drop of 5 mL.
3.
Results and discussion
3.1.
Characteristics of the river water
The raw water qualities of the river water used in this investigation were shown in Table 1. It could be seen that the Songhua River water contained a high content of organic matter, with the concentrations of 8.200 mg/L in terms of TOC and 7.255 mg/L in DOC. On the other hand, the SUVA (UV254/ DOC 100) of the river water was calculated to be a low value of 1.54 on average. The SUVA value represents the aromaticity
Table 1 – Raw water qualities of the river water. Parameters Temperature ( C) pH Turbidity (NTU) Organic carbon (mg/L) UV254 (1/cm) Al (mg/L) Fe (mg/L) Mn (mg/L) Ca (mg/L) Mg (mg/L)
Total concentration
Dissolved fraction
17.2 1.2 7.74 0.16 17.8 6.8 8.200 0.819 – 0.78 0.24 0.57 0.17 0.05 0.01 25.44 0.76 6.85 0.34
– – – 7.255 0.503 0.112 0.022 0.12 0.10 0.08 0.07 0.02 0.01 25.01 0.67 6.66 0.26
Total concentrations of various water parameters were measured using the raw river water; while the corresponding dissolved fractions were determined after pre-filtration through 0.45 mm membrane.
62
water research 44 (2010) 59–68
of dissolved organic matter in water. Thus, it might be considered that there was a certain amount of hydrophilic organic compounds in this surface water. Furthermore, it could be noticed that metal substances including Al, Fe and Mn were also presented in the river water. However, they were mainly in the particulate fraction, implying that the irreversible fouling caused by dissolved metals on the membrane surface and in the membrane pores might be negligible. On the other hand, Ca and Mg in the river water were predominantly in the dissolved form.
(10.9%) of irreversible fouling was obtained by the chemical cleaning with citric acid alone. Based on the cleaning results with single alkali or acid reagent, it might be reasonable to infer that no remarkable recovery of the permeability of fouled PVC membrane could be achieved by the combination of alkali and acid cleaning. As shown in Fig. 2, the cleaning efficiency was still a minus value of 4.5% after 30 min of 1% NaOH cleaning followed by 30 min of 2% citric acid cleaning.
3.3. 3.2. Cleaning of hollow-fiber PVC membrane by alkali and acid As a kind of alkaline reagent, NaOH has always been used for the chemical cleaning of fouled membranes in surface water treatment, and more or less the positive cleaning effect could be obtained depending on the membrane material and type as well as the foulants in feed water (Ang et al., 2006; Kimura et al., 2004; Yamamura et al., 2007a; Yamamura et al., 2007b; Zondervan and Roffel, 2007). However, it could be seen from Fig. 2 that in this investigation, the irreversible resistance of fouled PVC membrane increased after cleaning by 1% NaOH for 30 min, with the negative efficiency amounting to 14.6% on average. The reason accounting for this phenomenon would be discussed in ‘‘3.4’’. Through the chemical cleaning with 2% citric acid for 30 min, a moderate removal of irreversible resistance (10.9%) was observed for the PVC membrane (Fig. 2). It had been identified that metals such as Fe, Mn and Al were able to cause the irreversible membrane fouling, on which acid cleaning was especially effective (Kimura et al., 2004; Yamamura et al., 2007a; Yamamura et al., 2007b). However, the concentrations of dissolved metals in the river water used in this study were shown to be low (Table 1). Furthermore, although some organic membrane foulants (such as carbohydrate) could also be eliminated from fouled membrane through acid cleaning (Yamamura et al., 2007b), it might be assumed that only a small amount of these organic substances was presented in natural river water. As a result, only the moderate removal
Fig. 2 – Effects of chemical cleaning of hollow-fiber PVC membrane by different reagents (A: 30 min of 1% NaOH; B: 30 min of 2% citric acid; C: 30 min of 1% NaOH D 30 min of 2% citric acid. Measurements number n [ 7).
Cleaning of hollow-fiber PVC membrane by ethanol
To restore the permeability of PVC membrane fouled in ultrafiltration of river water, ethanol was employed as a kind of cleaning reagent and evaluated for the chemical cleaning. From Fig. 3, it could be seen that a significant removal of irreversible fouling resistance was achieved after the consecutive chemical cleaning with 30 min of 1% NaOH and 30 min of ethanol, and the cleaning efficiency reached to 85.1% on average. On the other hand, the results showed that cleaning by ethanol alone for 30 min could eliminate 48.5% of irreversible fouling on the membrane, which indicated that the 30 min of 1% NaOH cleaning also made its contribution (36.6% on average) to the restoration of membrane permeability. This seemed to be in contradiction with the experimental results given in ‘‘3.2’’, which would be discussed in the following sections.
3.4. Microscopic analyses of the consecutive membrane cleaning with alkali and ethanol 3.4.1.
SEM observation for the surface of the membrane
In ‘‘3.3’’, it was calculated that alkaline cleaning of the fouled PAC membrane by 1% NaOH for 30 min contributed to 36.6% of the removal of irreversible fouling. However, it was also found from the experimental results that NaOH cleaning exhibited the negative efficiency of 14.6% (Fig. 2). To make this point
Fig. 3 – Effects of chemical cleaning of hollow-fiber PVC membrane by different reagents (D: 30 min of 1% NaOH D 30 min of ethanol; E: 30 min of ethanol. Measurements number n [ 7).
water research 44 (2010) 59–68
clear, membrane samples before and after the chemical cleaning were observed under SEM. As shown in Fig. 4a, the surface of new hollow-fiber PVC membrane was rather clean and smooth. By contrast, the fouled membrane was covered with an irregular and accidented gel layer even after thoroughly wiping with a sponge (Fig. 4b), which could not be eliminated through the physical methods. However, much of the foulants on the membrane surface were removed after chemical cleaning by NaOH solution for 30 min (Fig. 4c). When the ethanol cleaning was further performed, the membrane surface recovered the smooth and flat topography; and almost the entire gel layer disappeared (Fig. 4d).
63
membrane surface restored nearly the same clear-cut observation (Fig. 5d) as that on the new membrane. From the results of SEM and AFM analyses for the membrane surface, it might be inferred that both the alkali and ethanol were able to remove a portion of the irreversible gel layer on the PVC membrane surface, which confirmed the positive effect of NaOH on membrane cleaning suggested in ‘‘3.3’’. The effectiveness of NaOH cleaning for removal of membrane foulants from the fouled membrane has long been recognized (Kimura et al., 2004; Yamamura et al., 2007a; Zhu and Nystro¨m, 1998; Zondervan and Roffel, 2007). However, to the authors’ knowledge, no previous literature is available as for the ethanol cleaning of fouled membrane, which deserves further studies.
3.4.2. Three-dimensional AFM observation for the surface of the membrane
3.4.3.
Three-dimensional AFM image analysis was also employed to investigate the morphological changes of the fouled PVC membrane during consecutive cleaning with alkali and ethanol. As could be seen from Fig. 5a, the pores and lines on the surface of new membrane were clear-cut. In contrast, the surface of fouled membrane was blurry, and no pore and line could be clearly identified even after wiping with sponge (Fig. 5b). However, as chemical cleaning with NaOH was carried out, the surface morphology of the PVC membrane became much clearer, and some lines on the surface might be observed roughly (Fig. 5c). Moreover, after further cleaning by 30 min of ethanol, the pores and lines on the fouled
To evaluate the cleaning effects of alkali and ethanol on inpore fouling formed during ultrafiltration of river water, Cross-sectional SEM images were also taken for the hollowfiber PVC membrane samples before and after the chemical cleaning, as shown in Fig. 6. By comparing the new membrane (Fig. 6a) with the fouled membrane (Fig. 6b), it could be found that the internal membrane fouling was even formed on the supporting material under the active filtration layer of the membrane. The channels on the supporting layer became narrow and obscured near the membrane surface. After cleaning by 1% NaOH for 30 min, a portion of the internal foulants deposited on the wall of the channels was removed
SEM observation for the cross-section of the membrane
Fig. 4 – SEM images of the surface of (a) new membrane, (b) fouled membrane after wiping with sponge, (c) fouled membrane after cleaning by 1% NaOH for 30 min, and (d) fouled membrane after consecutive cleaning by 1% NaOH for 30 min and ethanol for 30 min.
64
water research 44 (2010) 59–68
Fig. 5 – Three-dimensional AFM images of (a) new membrane, (b) fouled membrane after wiping with sponge, (c) fouled membrane after cleaning by 1% NaOH for 30 min, and (d) fouled membrane after consecutive cleaning by 1% NaOH for 30 min and ethanol for 30 min.
(Fig. 6c). After further cleaning with ethanol for 30 min, the cross-sectional observation of the membrane became distinct again (Fig. 6d); and the in-pore fouling was eliminated almost completely.
3.4.4.
AFM section analyses for the membrane
To further evaluate the effect of consecutive alkali and ethanol cleaning on the membrane fouling, AFM section analyses were also employed to characterize the membrane samples before and after the cleaning. This technology had been successfully applied to characterize NOM in the Songhua River (Guo and Ma, 2006), as well as membrane fouling property (Lee et al., 2005). As could be seen from Fig. 7a, the new PVC membrane used in the experiments presented a zigzag feature of the section profile. However, most of the valleys were filled up with foulants after ultrafiltration of river water, which could not be removed by sponge wiping. As a result, the section profile of the fouled membrane turned rather smooth (Fig. 7b). When 30 min of 1% NaOH cleaning was applied to the fouled membrane, the section restored the zigzag shape to a certain extent (Fig. 7c). After the membrane was further cleaned by 30 min of ethanol, it could be noticed that almost all of the foulants deposited on the membrane disappeared, and the
membrane section regained the distinct indented observation again (Fig. 7d). AFM analysis also provided the value of surface roughness for characterizing membrane fouling quantitatively. Two parameters, arithmetical mean deviation of the profile (Ra) and ten-point height of irregularities (Rz) were listed in Table 2. The new PVC membrane used in this investigation exhibited high roughness with Ra of 16.7 nm and Rz of 51.2 nm, respectively. After filtration with the river water, the Ra and Rz were significantly reduced to 4.7 nm and 19.4 nm, respectively, due to the deposition of foulants on the membrane. After cleaning with the NaOH solution, the roughness parameters Ra and Rz increased to 10.9 nm and 34.8 nm, corresponding to the recovery rates of 52% and 49%, respectively. After further cleaning by ethanol, the membrane roughness nearly regained the original state, as proved by the increase of Ra and Rz to 15.3 nm and 48.2 nm, with the respective recovery of 88% and 91%. The recovery rates of membrane roughness after consecutive NaOH and ethanol cleaning were in accordance with the cleaning efficiencies indicated in ‘‘3.3’’. Therefore, it might have been made clear that both the alkali and the ethanol were not only able to remove the
water research 44 (2010) 59–68
65
Fig. 6 – SEM images of the cross-section of (a) new membrane, (b) fouled membrane after wiping with sponge, (c) fouled membrane after cleaning by 1% NaOH for 30 min, and (d) fouled membrane after consecutive cleaning by 1% NaOH for 30 min and ethanol for 30 min.
Fig. 7 – AFM section analyses of (a) new membrane, (b) fouled membrane after wiping with sponge, (c) fouled membrane after cleaning by 1% NaOH for 30 min, and (d) fouled membrane after consecutive cleaning by 1% NaOH for 30 min and ethanol for 30 min.
66
water research 44 (2010) 59–68
Table 2 – Roughness changes of the membrane during the consecutive cleaning by NaOH and ethanol (unit: nm). Membrane samples
Ra
Rz
M1 M2 M3 M4
16.7 4.7 10.9 15.3
51.2 19.4 34.8 48.2
Membrane samples: M1-new membrane; M2-fouled membrane after wiping with sponge; M3-fouled membrane after cleaning by 30 min of 1% NaOH; M4-fouled membrane after consecutive cleaning by 30 min of 1% NaOH and 30 min of ethanol. Ra: arithmetical mean deviation of the profile; Rz: ten-point height of irregularities.
foulants on membrane surface, but also able to eliminate the membrane fouling formed in membrane pores during the ultrafiltration of river water. The synergistic effects were observed between the two reagents for the removal of membrane foulants.
3.5. Contact angle changes of the membrane during the consecutive cleaning with alkali and ethanol It had been recognized by several researchers that NaOH solution could desorbed a large amount of organic matter from the fouled membrane in drinking water treatment, including carbohydrate, protein and humic substances (Kimura et al., 2004; Yamamura et al., 2007a; Yamamura et al., 2007b). According to the results and discussion in ‘‘3.3’’ and ‘‘3.4’’, NaOH solution was also able to remove the surface and in-pore foulants on the PVC membrane used in the experiments. However, this was in contradiction with the negative cleaning efficiency achieved by the alkali solution demonstrated in ‘‘3.2’’.
To understand this point, contact angle was determined for the PVC membrane samples during the consecutive chemical cleaning by NaOH and ethanol. Contact angle is the reflection of the hydrophobic/hydrophilic character of the membrane (Buonomenna et al., 2007). The higher the contact angle is, the more hydrophobic the membrane would be, which leads to the lower membrane permeability if other membrane properties (such as pore size and pore density) are the same. From Fig. 8, it could be seen that the new PVC membrane used in the study had an average contact angle of 69.7 1.2 . After fouling with the river water, the contact angle increased to 73.8 2.3 , implying the hydrophobicity of the membrane increased slightly, which was in coincidence with the results obtained by Nghiem et al. (2008) in nanofiltraion of humic acids. This might be due to the accumulation of hydrophobic organic matter on the membrane. As NaOH cleaning was performed, the contact angle of the PVC membrane increased significantly to 87.6 2.1 . The reason might be that NaOH reacted with or desorbed some hydrophilic groups on the PVC membrane during the cleaning. Further studies would be required to illustrate the detailed mechanisms. However, with the increase of the contact angle, the hydrophobicity of the PVC membrane increased correspondingly, resulting in the decrease of membrane permeability. This might account for the increase of irreversible resistance after the chemical cleaning by NaOH (‘‘3.2’’). After further cleaning with ethanol, the contact angle of the PVC membrane restored to the low value of 71.4 1.7 , which was in comparison with the 69.7 1.2 of new membrane. Therefore, ethanol cleaning was not only able to remove the foulants (‘‘3.3’’ and ‘‘3.4’’), but also able to recover the hydrophilicity of the PVC membrane. As a result, the permeability of the membrane could be regained after the consecutive chemical cleaning by alkali and ethanol.
4.
Conclusions
The effectiveness and mechanism of consecutive cleaning of hollow-fiber PVC membrane with NaOH and ethanol during ultrafiltration of river water were investigated. The following conclusions could be drawn.
Fig. 8 – Contact angle changes of the membrane during the consecutive cleaning by NaOH and ethanol (M1-new membrane; M2-fouled membrane after wiping with sponge; M3-fouled membrane after cleaning by 1% NaOH for 30 min; M4-fouled membrane after consecutive cleaning by 30 min of 1% NaOH and 30 min of ethanol).
(1) Negative cleaning efficiency of 14.6% was observed for the chemical cleaning with 30 min of 1% NaOH; while cleaning with 2% citric acid for 30 min exhibited a moderate capacity for the removal of irreversible resistance (10.9%). (2) As high as 85.1% of the cleaning efficiency was achieved by the consecutive cleaning with 30 min of 1% NaOH and 30 min of ethanol. Individual ethanol cleaning could remove 48.5% of the irreversible membrane fouling. (3) SEM and AFM analyses showed that both NaOH and ethanol were not only able to eliminate the foulants on membrane surface, but also able to remove the in-pore membrane fouling from the PVC membrane. The synergetic effects were observed between the NaOH and ethanol for the removal of membrane foulants. (4) After cleaning with NaOH, the hydrophobicity of PVC membrane increased significantly; after further cleaning
water research 44 (2010) 59–68
with ethanol, the membrane could recover the hydrophibilicity, and restore the permeability. Considering that ethanol is easy to be used and reclaimed, the consecutive chemical cleaning with alkali and ethanol is recommended for PVC membrane in the treatment of surface water.
Acknowledgement This research is jointly supported by the State Key Laboratory of Urban Water Resource and Environment (HIT, Grant No. 2008DX04), the National Natural Science Foundation of China (Grant No. 50678047), the National Science and Technology Supporting Project (Grant No. 2006BAJ08B06), and the Important Items of Science and Technology for the Control and Treatment of Water Pollution (Grant. No. 2008ZX07422-005).
references
Ang, W.S., Lee, S., Elimelech, M., 2006. Chemical and physical aspects of cleaning of organic-fouled reverse osmosis membranes. Journal of Membrane Science 272 (1–2), 198–210. Buonomenna, M.G., Lopez, L.C., Favia, P., d’Agostino, R., Gordano, A., Drioli, E., 2007. New PVDF membranes: the effect of plasma surface modification on retention in nanofiltration of aqueous solution containing organic compounds. Water Research 41 (19), 4309–4316. Choi, Y.H., Kim, H.S., Kweon, J.H., 2008. Role of hydrophobic natural organic matter flocs on the fouling in coagulationmembrane processes. Separation and Purification Technology 62 (3), 529–534. Costa, A.R., Norberta de Pinho, M., Elimelech, M., 2006. Mechanism of colloidal natural organic matter fouling in ultrafiltration. Journal of Membrane Science 281 (1–2), 716–725. Fabris, R., Lee, E.K., Chow, C.W.K., Chen, V., Drikas, M., 2007. Pre-treatments to reduce fouling of low pressure micro-filtration (MF) membranes. Journal of Membrane Science 289 (1–2), 231–240. Fiksdal, L., Leiknes, T.O., 2006. The effect of coagulation with MF/ UF membrane filtration for the removal of virus in drinking water. Journal of Membrane Science 279 (1–2), 364–371. Gray, S.R., Ritchie, C.B., Bolto, B.A., 2004. Effect of fractionated NOM on low pressure membrane flux declines. Water Science and Technology 4 (4), 189–196. Gray, S.R., Ritchie, C.B., Tran, T., Bolto, B.A., Greenwood, P., Busetti, F., Allpike, B., 2008. Effect of membrane character and solution chemistry on microfiltration performance. Water Research 42 (3), 743–753. Guo, J., Ma, J., 2006. AFM study on the sorbed NOM and its fractions isolated from River Songhua. Water Research 40 (10), 1975–1984. Guglielmi, G., Chiarani, D., Judd, S.J., Andreottola, G., 2007. Flux criticality and sustainability in a hollow fibre submerged membrane bioreactor for municipal wastewater treatment. Journal of Membrane Science 289 (1–2), 241–248. Huang, H.O., Lee, N.H., Young, T., Gary, A., Lozier, J.C., Jacangelo, J.G., 2007. Natural organic matter fouling of lowpressure, hollow-fiber membranes: effects of NOM source and hydrodynamic conditions. Water Research 41 (17), 3823–3832.
67
Huang, X.H., Leal, M., Li, Q.L., 2008. Degradation of natural organic matter by TiO2 photocatalytic oxidation and its effect on fouling of low-pressure membranes. Water Research 42 (4–5), 1142–1150. Jermann, D., Pronk, W., Meylan, S., Boller, M., 2007. Interplay of different NOM fouling mechanisms during ultrafiltration for drinking water production. Water Research 41 (8), 1713–1722. Kimura, K., Hane, Y., Watanabe, Y., Any, G., Ohkuma, N., 2004. Irreversible membrane fouling during ultrafiltration of surface water. Water Research 38 (14–15), 3431–3441. Lee, E.K., Chen, V., Fane, A.G., 2008b. Natural organic matter (NOM) fouling in low pressure membrane filtration-effect of membranes and operation modes. Desalination 218 (1–3), 257–270. Lee, N.H., Amy, G., Croue´, J.P., 2006. Low-pressure membrane (MF/UF) associated with allochthonous versus autochthonous natural organic matter. Water Research 40 (12), 2357–2368. Lee, N.H., Amy, G., Croue´, J.P., Buisson, H., 2005. Morphological analyses of natural organic matter (NOM) fouling of lowpressure membranes (MF/UF). Journal of Membrane Science 261 (1–2), 7–16. Lee, S., Elimelech, M., 2007. Salt cleaning of organic-fouled reverse osmosis membranes. Water Research 41 (5), 1134–1142. Lee, S.H., Park, P.K., Kim, J.H., Yeon, K.M., Lee, C.H., 2008a. Analysis of filtration characteristics in submerged microfiltration for drinking water treatment. Water Research 42 (12), 3109–3121. Li, Q.L., Elimelech, M., 2006. Synergistic effects in combined fouling of a loose nanofiltration membrane by colloidal materials and natural organic matter. Journal of Membrane Science 278 (1–2), 72–82. Liang, H., Gong, W.J., Chen, J., Li, G.B., 2008. Cleaning of fouled ultrafiltration (UF) membrane by algae during reservoir water treatment. Desalination 220 (1–3), 267–272. Lu, Y., Ding, Z.W., Liu, L.Y., Wang, Z.J., Ma, R.Y., 2008. The influence of bubble characteristics on the performance of submerged hollow fiber membrane module used in microfiltration. Separation and Purification Technology 61 (1), 89–95. Meng, F.G., Chae, S.R., Drews, A., Kraume, M., Shin, H.S., Yang, F.L., 2009. Recent advances in membrane bioreactors (MBRs): membrane fouling and membrane material. Water Research 43 (6), 1489–1512. Meng, F.G., Yang, F.L., Shi, B.Q., Zhang, H.M., 2008. A comprehensive study on membrane fouling in submerged membrane bioreactors under different aeration intensities. Separation and Purification Technology 59 (1), 91–100. Meng, F.G., Zhang, H.M., Yang, F.L., Liu, L.F., 2007. Characterization of cake layer in submerged membrane bioreactor. Environmental Science and Technology 41 (11), 4065–4070. Ng, H.Y., Tan, T.W., Ong, S.L., 2006. Membrane fouling of submerged membrane bioreactors: impact of mean cell residence time and the contributing factors. Environmental Science and Technology 40 (8), 2706–2713. Nghiem, L.D., Vogel, D., Khan, S., 2008. Characterising humic acid fouling of nanofiltration membranes using bisphenol A as a molecular indicator. Water Research 42 (15), 4049–4058. Oh, H.K., Takizawa, S., Ohgaki, S., Katayama, H., Oguma, K., Yu, M.J., 2007. Removal of organics and viruses using hybrid ceramic MF system without draining of PAC. Desalination 202 (1–3), 191–198. Peter-Varbanets, M., Zurbru¨gg, C., Swartz, C., Pronk, W., 2009. Decentralized systems for potable water and the potential of membrane technology. Water Research 43 (2), 245–265. Shirasaki, N., Matsushita, T., Matsui, Y., Kobuke, M., Ohno, K., 2009. Comparison of removal performance of two surrogates
68
water research 44 (2010) 59–68
for pathogenic waterborne viruses, bacteriophage Qb and MS2, in a coagulation–ceramic microfiltration system. Journal of Membrane Science 326 (2), 564–571. Sun, X.H., Kanani, D.M., Ghosh, R., 2008. Characterization and theoretical analysis of protein fouling of cellulose acetate membrane during constant flux dead-end microfiltration. Journal of Membrane Science 320 (1–2), 372–380. Yamamura, H., Chae, S., Kimura, K., Watanabe, Y., 2007a. Transition in fouling mechanism in microfiltration of a surface water. Water Research 41 (17), 3812–3822. Yamamura, H., Kimura, K., Watanabe, Y., 2007b. Mechanism involved in the evolution of physically irreversible fouling in microfiltration and ultrafiltration membranes used for
drinking water treatment. Environmental Science and Technology 41 (19), 6789–6794. Zheng, X., Ernst, M., Jekel, M., 2009. Identification and quantification of major organic foulants in treated domestic wastewater affecting filterability in dead-end ultrafiltration. Water Research 43 (1), 238–244. Zhu, H., Nystro¨m, M., 1998. Cleaning results characterized by flux, streaming potential and FTIR measurements. Colloids and Surfaces A: Physicochemical and Engineering Aspects 138 (2–3), 309–321. Zondervan, E., Roffel, B., 2007. Evaluation of different cleaning agents used for cleaning ultrafiltration membranes fouled by surface water. Journal of Membrane Science 304 (1–2), 40–49.
water research 44 (2010) 69–76
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Fate and impact of organics in an immersed membrane bioreactor applied to brine denitrification and ion exchange regeneration Ewan J. McAdam a, Mark Pawlett b, Simon J. Judd a,* a b
Centre for Water Science, Cranfield University, Bedfordshire, MK43 0AL, United Kingdom Natural Resources Department, Cranfield University, Bedfordshire, MK43 0AL, United Kingdom
article info
abstract
Article history:
The application of membrane bioreactors (MBRs) to brine denitrification for ion exchange
Received 18 March 2009
regeneration has been studied. The developed culture was capable of complete brine
Received in revised form
denitrification at 50 gNaCl.l1. Denitrification reduced to c.60% and c.70% when salinity was
17 August 2009
respectively increased to 75 and 100 g.l1, presumed to be due to reduced growth rate and
Accepted 30 August 2009
the low imposed solids retention time (10 days). Polysaccharide secretion was not induced
Available online 4 September 2009
by stressed cells following salt shocking, implying that cell lysis did not occur. Fouling propensity, monitored by critical flux, was steady at 12–15 l.m2.h1 during salinity
Keywords:
shocking and after brine recirculation, indicating that the system was stable following
Ion exchange
perturbation. Low molecular weight polysaccharide physically adsorbed onto the nitrate
Brine
selective anion exchange resin during regeneration reducing exchange capacity by c.6.5%
Biological denitrification
when operating up to complete exhaustion. However, based on a breakthrough threshold
Salt
1 the exchange capacity was comparative to that determined when using of 10 mgNO 3 -N.l
Nitrate
freshly produced brine for regeneration. It was concluded that a denitrification MBR was an appropriate technology for IEX spent brine recovery and reuse. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Anion-exchange (aIEX) is the most frequently adopted technology for nitrate (NO 3 ) removal during potable water treatment due to its low cost and operational simplicity. A strong salt (NaCl) solution is used to regenerate the resin resulting in the production of concentrated waste brine containing the target anion, chloride and other oxyanions. This waste stream can comprise 0.8–2.4% of treated product flow (McAdam and Judd, 2008) and its disposal (usually by tankering) constitutes a significant proportion of the process cost. Operation of aIEX in combination with biological nitrate reduction of the waste brine for regenerant recovery presents a more sustainable
alternative by reducing the waste volume, salt (NaCl) consumption and treated product losses. Studies adapting non-halophilic microbial communities from standard activated sludge processes for this application have reported inhibition of denitrification and, in some cases, plasmolysis to be promoted by the elevated salt concentrations (>30gNaCl.l1). More recently, halophilic monocultures Halomonas denitrificans (Cyplik et al., 2007) and Halomonas campisalis (Peyton et al., 2001) have been successfully adapted at laboratory scale for denitrification at high salt concentrations from 30 to 180 g.l1, obviating dilution prior to biotreatment. However, adaptation of halophiles to brine processing is yet to be examined in detail.
* Corresponding author. Tel.: þ44 1234 754842. E-mail address:
[email protected] (S.J. Judd). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.08.048
70
water research 44 (2010) 69–76
Other restrictions to this application include the accumulation of inorganic constituents (e.g. sulphate) due to recirculation, and the impact of organics and microbial carryover from the bioreactor on downstream resin regeneration. In brine re-use trials, elevated sulphate concentrations were not reported to impact upon either resin or biological performance when nitrate selective resins have been used (Clifford and Liu, 1993). However, Bae et al. (2002) reported that microbial associated particulates and organics present in the regenerant fouled anion exchange resins, thus the integration of both sand filtration and GAC were required downstream of the denitrification reactor to nullify the impact. Though little information exists on the impact of residual organics on resin capacity, the application of ‘‘classical’’ biomass separation membrane bioreactor (MBR) technology to this duty has been mooted to provide absolute bacterial rejection and high MW biopolymer retention, promoting a consistent permeate quality (McAdam and Judd, 2008). The current paper assesses the viability of a denitrification MBR for waste aIEX brine treatment and reuse in the regeneration of ion exchange resins, specifically this study will address: the fate of organics during permeate brine recirculation (to simulate re-use); the impact of organics on resin capacity; the influence of salt variation on halophilic treatment performance; and the impact of perturbation on fouling propensity.
2.
Material and methods
2.1.
Experimental rig
To establish a salt tolerant bacterial community, a seed culture was harvested from the anaerobic layer of a coastal sediment at low tide. Following acclimation in batch condi1 tions (50 gNaCl.l1, 500 mgNO 3 -N.l ), a 75 l reactor (Fig. 1) was
seeded at a v/v ratio of 15:1. The influent nitrate concentration 1 was set at 500 mg NO 3 -N.l . During substrate optimisation, ethanol was supplied as the exogenous substrate and dosed at a C:N ratio of between 0.77:1 and 8.5:1 (g.g1); under normal conditions, a C:N of 0.85:1 was used. Reactor temperature was maintained at c.20 C using a thermostatically controlled heating jacket. An impeller mixer was used to ensure complete biomass distribution with the impeller blade sited below the membrane module. The hydraulic and solids residence times (HRT and SRT respectively) were 17.5 h and 10 days respectively. The process was allowed 3 SRTs to acclimatise prior to testing. During recirculation experiments, MBR permeate was collected in a holding tank (T2), supplemented with NO 3 and pumped back to the feed tank (T1). During salt upshocking/down shocking experiments, the NaCl concentrate dosed into T1 was changed to meet the required concentration providing an incremental spike; fluid residence time in T1 was approximately 20 h. A 0.93 m2 out-to-in immersed PVDF hollow-fibre membrane with 0.04 mm nominal pore size was used. Permeate was withdrawn under suction from the membrane using a piston pump (FMI Inc., Syosset, US). To maintain anoxic conditions, nitrogen-enriched air (>99%) was used to scour the membrane. Gas was introduced via a solenoid valve (Zoedale Plc, Bedford, UK) and controlled with a programmable digital relay (Ku¨bler Gmbh, Ludwigshafen, Germany); flow rate was controlled with a 0–50 l min1 needle valve (RS Ltd., Corby, UK). Pressure was monitored using a 0.5 to 0.5 barg calibrated pressure transducer (Gem Sensors, Basingstoke, UK) and data recorded using a 16-bit 0–2.5 V data conversion unit (Pico technology, St. Neots, UK).
2.2.
Anion exchange resin
A commercially available nitrate selective macroporous styrene based anion exchange resin (Purolite A520E,
Concentrate KNO KNO33 NaCl
Exogenous substrate Feed tank
Tap water
Mixer
EtOH
KNO3 T1
Permeate
Waste
Backflush Pressure Transducer
T2
T
Recycle tank N2selective membrane PC
Oxygen
Air
Fig. 1 – Experimental set-up.
Nitrogen
71
water research 44 (2010) 69–76
Llantrisant, UK) was loaded into a 50 mm diameter 1 m glass chromatography column and retained using 25 mm-rated frits at either end; the bed comprised 120 g of resin. After initial rinsing, the A520E resin size ranged from 0.28 to 1.26 mm (d50 0.61 mm). Prior to use, DI water was pumped through the resin bed at 20 bed volumes (Bv).h1 for 30 min, followed by a 30 min 50gNaCl.l1 flush (to ensure saturation) at 5 Bv.h1 and a subsequent DI rinse for 60 min at 5 Bv.h1. The exhaustion flow rate was set to 20 Bv.h1 and the IEX feed contained 30 1 1 1 and 22.6 mg.l1 mgSO2 4 .l , 115 mgCl .l , 150 mgCaCO3.l NO -N. Regeneration comprised a 60 min cycle at 5 Bv.h1 3 1 followed by slow and fast rinses of 5 Bv.h for 30 minutes and 20 Bv.h1 for 10 min respectively.
2.3.
Chemical analysis
2.3.1.
General analysis
Mixed liquor suspended solids (MLSS) and bicarbonate were determined by standard methods. Oxyanion (NO 3 , NO2 ) and chloride concentration was measured using proprietary cell tests (Merck Spectroquant) with spectrophotometric detection. Dissolved organic carbon (DOC) was measured using a Shimadzu TOC-5000A analyser. Ethanol concentration was determined using a commercially available enzymatic method (Boehringer-Mannheim, Roche). Soluble microbial products (SMP) were extracted according to the method described in Judd (2006) and polysaccharide and protein concentration quantified using the phenol–sulphuric acid method (Zhang et al., 1999) and modified Lowry method (Frølund et al., 1995) respectively. Absorbance for polysaccharide and protein determinations was measured using a Jenway 6505 UV/Vis spectrophotometer at UV480nm and UV750nm respectively with D-glucose and bovine serum albumin (BSA) as standards. Particle size distribution was measured with an integrated laser diffractor (Malvern Mastersizer 2000).
Molecular weight fractionation
Serial fractionation was undertaken using an Amicon 8400 series stirred cell, pressurised with N2 (1 barg), and standard UF (Millipore) membranes, size range 10, 30, 50, 100 and 300 kDa. Sample supernatant was pre-filtered using a 1.2mm filter and the subsequent sample split between two 300 kDa membranes to limit concentration polarisation. Concentration polarisation was limited by application of an integrated bar stirrer operated at a constant 100 rpm; the adopted filtrate/ retentate ratio was 0.4.
Nitrate removal efficiency increased from 84.6% to a maximum 99.8% as the carbon to nitrogen ratio (C:N) increased from 0.77 to 0.94 (Fig. 2). Once a C:N ratio of 0.89 had been exceeded, ethanol was detected in the permeate above the limit of detection (>0.5 mg l1). Although the existence of an optimum C:N has been reported previously (McAdam et al., 2007), research studies typically observe low NO 2 -N and NO3 -N effluent concentrations as the optimum C:N is exceeded due to the surplus of available carbon (Chiu and Chung, 2003; McAdam et al., 2007). In this study, on increasing C:N >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 <1.2 mm and 300 kDa and a principal polysaccharide peak of 48.3% below 10 kDa (Fig. 3). Organics between 1.2 mm and 100 kDa were absent in the permeate indicating the molecular weight cut off (MWCO) of the hollow-fibre (and any associated biofilm) was c.50 to 100 kDa.
3.2. Impact of organics accumulation during recirculation After 7 days recirculation, the SMP DOC had increased from an initial concentration of c.170 mgDOC.l1 (c.34 mgDOC.gMLSS1) up to a maximum concentration of 557 mgDOC.l1 (Fig. 4). At steady state, DOC transmission was recorded between 54% and 80% and was attributed to accumulation of low molecular weight (MW) organics (below the membrane MWCO).
Phospholipid fatty acid analysis
Phospholipid fatty acid (PLFA) analysis was used to assess the community structure using the method of Frostega˚rd et al. (1991). Samples were freeze dried prior to analysis. Lipids were extracted from the freeze dried sample using the Bligh and Dyer (1959) ratio of 1:2:0.8 (v/v/v) of chloroform, methanol and citrate buffer. Lipids were then fractionated by solid phase extraction. The phospholipid fraction was derivatised by mild alkaline methanolysis (Dowling et al., 1986). The resultant fatty-acid methyl esters (FAMES) were analyzed by GC-FID (Agilent). Peak identification was undertaken using GC–MS (Agilent).
100
200
90
150
80
100
NO3--N Removal
70
50
Total N Removal EtOH Remaining
60 0.7
0.8
0.9
1.0
1.1
EtOH concentration (mg.l-1)
2.3.3.
Results
3.1. Exogenous and endogenous organics transmission at steady state
Removal (%)
2.3.2.
3.
0 1.2
C:N
Fig. 2 – Optimising C:N ratio during steady state operation. Influent: 500 mgNO-3-N.lL1; 50 gNaCl.lL1.
72
water research 44 (2010) 69–76
100 SMP proteins Permeate DOC Permeate polysaccharides
SMP DOC SMP polysaccharides Permeate proteins
Relative abundance (%)
80
60
40
20
0 <1.2µm-300
300-100
100-50
50-30
30-10
<10
Apparent molecular weight (kDa) Fig. 3 – Molecular weight SMP and permeate fractionation for protein and polysaccharide at steady state.
Critical flux analysis (Jc) was conducted using the flux step method before recirculation and after reaching steady state (Fig. 5). In both cases, Jc was between 12 and 15 l.m2.h1. Interestingly, similar exponential dP/dt trends were obtained for both sets of conditions, evidenced by similar gradients ((dP/dt)/J ) of between 0.23 and 0.27, however, dP/dt measured post-recirculation exhibited lower overall fouling potential. This appears counter-intuitive, based on the presence of accumulated organics and challenges previous reports which link fouling propensity to elevated concentrations of biopolymers in the bulk phase (Judd, 2006; Reid et al., 2006).
3.3.
IEX Resin capacity
To allow comparison with previous aIEX resin studies (Clifford and Liu, 1993; Bae et al., 2002), NO 3 -N breakthrough curves were determined using a 10 mgN.l1 threshold effluent concentration (US regulatory limit). Breakthrough curves (1 to 6) were run to complete exhaustion initially using freshly
produced regenerant (Brinefp, 50gNaCl.l1, Fig. 6(a)). The threshold was reached at c.400 bed volumes (BVs) in the second run which corresponded to a resin capacity of 0.61 eq.l1 or 88% of throughput obtained with the virgin resin during the first run. Subsequent runs 3–6 indicated a near identical trend demonstrating reproducible regeneration efficiency under these conditions. A capacity of c.0.46 eq.l1 has been observed previously by Bae et al. (2002) using the same commercially available resin (A520E); the lower capacity may be explained by the authors’ application of a lower strength regenerant (30gNaCl.l1). Breakthrough curves (1–6) were subsequently generated with fresh resin using biologically treated brine (Brinebt, 50gNaCl.l1) as the regenerant (Fig. 6(b)). Brinebt was sampled from the MBR permeate once steady state had been reached during permeate recirculation. At steady state, the DOC concentration of the brinebt was c.287 mg.l1. Under these conditions, breakthrough occurred at c.390 BVs in the second run, corresponding to 0.58 eq.l1 or 87% of throughput obtained with the virgin resin during the first run. Comparison 10
SMP Permeate Feed
500
Fouling rate, dP/dt (mbar.min-1)
Dissolved organic carbon (DOC) concentration (mg.l-1)
600
400 300 200
100
y = 0.01e0.27x
0
2
4
6
8
10
12
14
Time (days)
Fig. 4 – Impact of permeate recirculation to the main feed tank on dissolved organic carbon concentration (DOC) in the feed, permeate and SMP.
R2 = 0.99
1
0.1
0.01
0
y = 0.04e0.23x R2 = 0.99
Post-recirculation Pre-recirculation
0
5
10
15
20
25
Flux, J (l.m-2.h-1)
Fig. 5 – Critical flux analysis (Jc) before, during and after permeate recirculation to the main feed tank. Specific gas demand per unit membrane area (SGDm), 0.39 m3.mL2.hL1.
73
IEX permeate NO3--N Concentration (mg.l-1)
a
100
40000
80
32000
60
24000
Transmission (%)
25 Run 2 Run 3 Run 4 Run 5
20
15
10
Protein Polysaccharides DOC Chloride
20 0
5
16000
40
2
0
6
4
8000
8
0
Chloride concentration (mg.l-1)
IEX permeate nitrate concentration (mg.l-1)
water research 44 (2010) 69–76
Bed volumes (Bv) 0
0
100
200
300
400
500
600
Bed volumes (Bv)
Fig. 7 – Adsorption of regenerant organics (protein, polysaccharide and DOC) by anion exchange resin.
b 25
120 g resin bed was estimated at 26.4 mg (0.22 mgDOC.gResin). During exhaustion/regeneration cycles (Fig. 6(b)), the adsorptive mechanism of the polysaccharides was evaluated by regenerating the resin with a 50/50 fresh brine/biological regenerant (Run 5) and 100% fresh brine (Run 6). The similarity of the subsequent exhaustion curves suggested low polysaccharide exchange potential (i.e. reversibility).
Run 2 Run 3
20
Area 1
Run 4 Run 5 (50:50)
15
Run 6 (0:100) Run 4a (Fresh brine)
10
5
0
3.4. 0
100
200
300
400
500
600
Bed volumes (Bv)
Fig. 6 – Breakthrough curves observed from runs 2–6 using: (a) freshly produced brine; and (b) biologically treated brine. L1 ; Influent concentration: NO-3-N 22.6 mg.lL1; SO24 30 mg.l L1 L1 and HCO3 150 mg.l . Cl 115 mg.l
with Run 4 (a) using brinefp (Fig. 6(b)) demonstrated a loss in capacity (Area 1) indicating the extent of interference created by the biologically derived organics. Integration of the area between the brinefp and brinebt exhaustion curves recorded a capacity loss of 59 meq.l1 or c.6.5% of the estimated exhaustive resin capacity. Bae et al. (2002) observed significant capacity losses when using permeate from an upflow sludge blanket reactor (USBR) for regeneration unless subsequent treatment steps were incorporated. However, in this study subsequent regenerations using brinebt displayed a similar reproducibility indicating that the resin had reached a maximum organic capacity at the end of the first regeneration cycle. The uptake of brinebt organics by the resin was quantified using a virgin salt saturated resin (Fig. 7). Brine regenerant was assumed to exit the column once chloride transmission reached 100% (assuming chloride uptake to be zero at saturation). Chloride and protein transmission reached 100% simultaneously between 1 and 1.5 BVs indicating protein adsorption to be negligible. Polysaccharide and DOC transmission were recorded at c.15% and c.90% respectively up to 4 BVs, where a rapid increase in transmission in the interval between 4 and 7 BVs was observed. At 7 BVs, polysaccharide and DOC transmission reached 100%, indicating saturation of the resin with polysaccharide. Total adsorbed DOC on the
Salt shocking
To reflect the significant salt variations occurring in brine regenerant waste, the regenerant was initially upshocked to 75 gNaCl.l1 which was subsequently further increased to 100 gNaCl.l1 after 7 days. Following the initial upshock (75 gNaCl l1), nitrate removal decreased from 99.7 to 60.1% (Table 1) demonstrating a decrease in the specific biomass denitrification capacity. Protein release was also recorded with an increase in bulk phase concentration from c.30 to c.50 mg.l1 and from c.15 to c.30 mg.l1 following salt upshock to 75 and 100 gNaCl.l1 respectively (Fig. 8). A transition in floc structure also occurred; at steady state (50 gNaCl.l1), a floc size distribution ranging 60–800 mm was measured, however, following upshocking to 75 gNaCl.l1, a bi-modal distribution was recorded with the dominant peak ranging 0.2–5 mm, indicating floc breakage into primary particles (Wile´n et al., 2003). After 7 days at 100gNaCl.l1, the system was downshocked to 50 gNaCl.l1; sampling 24 h after downshocking demonstrated near complete denitrification recovery to 98.4%.
Table 1 – Treatment performance during salt spiking. NaCl (g.l1)
50 75 100 50b
NO-3-N
Ethanola (mg.l1)
Recovery Time
C:N
N/a 24 h 7d 24 h 7d
0.92 0.87 0.91 0.97 0.99
99.7 60.1 58.6 73.6 73.3
4.1 75.1 113 176.7 209
24 h
0.94
98.4
10.7
a Permeate concentration. b Salt downshock.
Reml. (%)
water research 44 (2010) 69–76
60
50 gNaCl.L-1
75 gNaCl.L-1
600
100 gNaCl.L-1
50
500
40
400
30
300
20
200
10 0
100
Polysaccharides Proteins DOC
5
0
10
15
20
25
30
DOC Concentration (mg.l-1)
Protein and polysaccharide (mg.l-1)
74
0
Time (days)
Fouling rate, dP/dt (mbar.min-1)
10 75 g.L-1 (7 d)
50
75
100
C16:0 C16:1 C17:0 Exhibit 1a C19:0cy
13.87 24.95 2.45 48.9 7.07
13.04 27.12 1.76 50.47 5.24
13.62 26.06 1.73 51.2 4.77
Total (%) Cyc/cis
97.2 0.145
97.6 0.110
97.4 0.093
corresponding to salt concentration (Fig. 10). Analysis of variance of the principal components (PC) confirmed significant differences of P < 0.001 and P < 0.01 for principal components PC1 and PC2 respectively. This distinction indicates abrupt changes in phenotypic profile between step changes in salinity.
4.
Discussion
4.1.
MBR Fouling
High polysaccharide transmission of c.81.5% was observed during steady state due to the production of low MW biopolymers and corresponded to a mean DOC removal of c.44%. Low MW biopolymers are generally associated with substrate metabolism and biomass growth (Barker et al., 2000) and are produced in all MBR applications. Using LC-OCD, Zhang et al. (2006) observed 99.8% high MW (c.250 kDa) and 93.6% low MW (5 to 250 kDa) biopolymer rejection when using a 0.2 mm flat sheet membrane in an MBR and cited
2.0 50 gNaCl.L-1
1.5
100 g.L-1 (24 h)
1
75 gNaCl.L-1 100 gNaCl.L-1
1.0
PC2 (8.62%)
Although the volume of particulate and colloidal material had apparently increased, critical flux analysis conducted before and after each salt increment (Fig. 9) indicated that fouling propensity remained stable as demonstrated by the similar dP/dt trends obtained. In addition, Jc was consistently recorded at c.12 l.m2.h1 and is comparable to that recorded during steady-state recirculation. This contradicts a previous non-halophilic MBR study where upon exposure to a 5 g.l1 chloride residual (0.83% NaCl) both protein and polysaccharide were released causing permeability decline (flat sheet, 0.4 mm) which was correlated to the elevated SMP polysaccharide concentration (Reid et al., 2006); the absence of elevated concentrations of secreted polysaccharide in this current study may in part explain this disparity. Twenty PLFA fatty acid methyl esters (FAMES), identified by MS, principally comprised normal saturates and terminally branched saturates. Trans-monoenoic fatty acid concentrations were below the limit of detection. Dominant FAMES were C16:0, C16:1, C17:0, C18, C18:1u9c and C19:0cy at 50, 75 and 100 gNaCl.l1 and accounted for c.95% of PLFAs detected (Table 2). Similar elution profiles (and the absence of transmonoenoic fatty acids) were observed previously for a range of moderately and extremely halophilic bacterium (Aston and Peyton, 2007; Yakimov et al., 2001). Principal component analysis (PCA) showed three discrete data groupings
75 g.L-1 (24 h)
NaCl (g.l1) Concentration
a Exhibit 1 – Comprises C18:0 and C18:1u9c.
Fig. 8 – Impact of salt upshock on biologically derived organics measured in the SMP.
50 g.L-1
Table 2 – Major constituents of PLFA analysis (%).
0.5 0.0 -0.5 -1.0
0.1
-1.5 0.01
-2.0 -2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
PC 1 (85.37%)
0.001 0
5
10
15
20
Flux, J (l.m-2.h-1)
Fig. 9 – Critical flux analysis (Jc) before and after each increase in salt concentration. Specific gas demand per unit membrane area (SGDm), 0.39 m3.mL2.hL1.
25
Fig. 10 – First and second principal components (PCs) derived from phospholipid fatty-acid profiles originating from biomass samples at the three salinities. Mean and standard deviation plotted. Percentage variation accounted for by PC shown in parenthesis on each axis.
water research 44 (2010) 69–76
polysaccharide as the major foulant. The authors suggested this behaviour to be a common trait of fouled MF membranes; improved retention of low MW biopolymers (and higher dP/dt) in their investigation may arise from more significant internal deposition created by the larger pore size. In this study, concentration (accumulation) of low MW biopolymers in the bulk phase by permeate recirculation did not increase fouling propensity. This indicates that: (1) low MW biopolymers asserted poor aggregation potential upon recirculation and thus were not filtered; and (2) biopolymers exhibited limited binding potential to the membrane surface and any biofilm present. This contradicts previous experiences with polysaccharides (Zhang et al., 2006; Frank and Belfort, 2003), however, past research has typically focused on high MW polysaccharides (100 to 1600 kDa) which possess more structural and functional complexity than those of lower MW biopolymers (48.3% below 10 kDa) as in this study; higher MW structures may thus concentrate at the membrane surface by both size exclusion and surface adhesion (Frank and Belfort, 2003). Fouling propensity was not greatly increased by salt shocking. The characteristic response of non-halophilic micro-organisms exposed to salt upshock is to undergo plasmolysis due to a loss in turgor pressure (Reid et al., 2006). This induces the release of soluble cellular components through the cell membrane (Laspidou and Rittmann, 2002) and in some instances the subsequent release of cell wall components such as acid mucopolysaccharides, resulting in high concentrations of proteins and polysaccharides in the bulk phase (Reid et al., 2006, Zhang et al., 2006). In this study, only protein was released, implying that cell lysis did not occur. Halophilic bacteria possess modified highly negatively charged proteins on the external cell wall to mediate osmotic shifts (Petrovic et al., 1999); the protein release observed may therefore have been an adjustment in cell wall composition (Russell, 1989). In addition, cell wall modification may have initiated the floc destabilisation observed upon upshocking causing indirect release of extracellular protein from the floc matrix as postulated by Reid et al. (2006). The absence of secreted polysaccharide, structure and size distribution of the organics produced by halophilic bacteria and the lower membrane pore size adopted in this investigation (0.04 mm, potentially limiting internal deposition) may explain the disparity in organics rejection and fouling compared to previous literature findings (Reid et al., 2006; Zhang et al., 2006).
4.2.
75
decrease in nitrate removal from 99.7% to between 58.6% and 73.6% may therefore be due to microbial restructuring, however, lower specific bacterial growth rates have been observed at high salt concentrations. Peyton et al. (2001) established that Halomonas campisalis could effectively denitrify at 180 gNaCl.l1 (Peyton et al., 2001), though the maximum specific growth rate for the monoculture was identified at c.30gNaCl.l1 (Aston and Peyton, 2007). In this study, upon down shocking to 50gNaCl.l1 denitrification capacity recovered to 98.4% within 24 h. This demonstrates that whilst transition in community structure occurred following salt upshocking, an effective residual halotolerant denitrifying community remained following perturbation; extension of SRT (>10 days) may be sufficient to offset the lower growth rates observed at high salt concentrations.
4.3.
Resin operation
It has been suggested that polysaccharides do not normally deposit easily onto aIEX resin due to impeded diffusion (by size exclusion) and low contact times (Cornelissen et al., 2008). In this study, polysaccharides contacted the resin during regeneration rather than exhaustion, thus increasing contact time by a factor of four. Adsorption of exopolysaccharides to anionic resins is intuitive as their structure is principally polyanionic due to the number of uronic acid or ketal linked pyruvate groups contained within the long chain high MW (500–2000 kDa) structures (Sutherland, 2001). However, based on the low affinity shown for desorption of polysaccharides in this study, it appears that the dominant adsorption mechanism associated with the lower MW polysaccharides present in the brine is physical rather than exchange based. DOC uptake could not be quantified during exhaustion runs due to competition effects with the influent DOC. Therefore, based on physical data, the theoretical charge density (approximated by normalising lost resin capacity with DOC uptake, Fig. 7) was c.3.9 104 meq.gDOC1; this negligible result further demonstrates that adsorption was not exchange based and indicates that the adsorbed organics exhibited a charge closer to neutrality. Kim and Symons (1991) postulated that physical adsorption was more likely to occur at the resin skeleton. After the first regeneration with brinebt, physical adsorption reached a maximum, presumably due to the limited number of adsorption sites available.
Microbial community and treatment performance
5. PLFA profiles were dominated by a small range of fatty acids that are common amongst halophiles (Aston and Peyton, 2007; Yakimov et al., 2001). The trans/cis ratio was consistently < 0.15 at all three salt concentrations indicating that the cell membrane remained stable – a ratio above 0.25 indicating instability (Aston and Peyton, 2007) – further suggesting that plasmolysis did not occur. Transition of PLFA profiles at each salt increment indicated reordering of the membrane lipid composition for osmoregulation (Russell, 1989; Pflu¨ger and Muller, 2004) and microbial community restructuring (Forney et al., 2001) as described previously for salt upshocking of denitrifying halophiles (Yoshie et al., 2006). The
Conclusions
A study of the denitrification of high salinity ion exchange brine regenerant and the impact of accumulation on process performance has demonstrated recycling for ion exchange regeneration to be viable. 1. Whilst recirculation generated high concentrations of low MW organics, their impact on membrane permeability was negligible. 2. Although salt upshock induced protein release, the permeability decline was minimal, contrary to previous studies based on non-halophilic communities.
76
water research 44 (2010) 69–76
3. Nitrate removal of c.99.7% was observed at steady-state (50 gNaCl.l1); at salt concentrations above 50 gNaCl.l1 nitrate removal decreased and the community profile was modified, though this could be countered by adoption of a higher SRT to offset the lower growth rates. 4. Adsorption of the low MW organics generated during denitrification onto the resin structure resulted in minimal loss in resin capacity, implying long-term operation using recovered brine is possible. 5. Under halophilic conditions, addition of exogenous substrate must be controlled to minimise breakthrough and to support complete denitrification (limiting the preferential formation of nitrite). 6. The efficacy of the denitrification MBR process is closely related to membrane rejection and the structural and functional attributes of the resultant organics; both the process operational determinants and the bacterial community generated may influence performance.
Acknowledgements The authors would like to thank the Engineering and Physical Sciences Research Council (EPSRC), Anglian Water, Severn Trent Water and Yorkshire Water for their financial support.
references
Aston, J.E., Peyton, B.M., 2007. Response of Halomonas campisalis to saline stress: changes in growth kinetics, compatible solute production and membrane phospholipid fatty acid composition. FEMS Microbiol. Lett. 274, 196–203. Bae, B.-U., Jung, Y.-H., Han, W.-W., Shin, H.-S., 2002. Improved brine recycling during nitrate removal using ion exchange. Water Res. 36, 3330–3340. Barker, D.J., Salvi, S.M.L., Langenhoff, A.A.M., Stuckey, D.C., 2000. Soluble microbial products in ABR treating low-strength wastewater. J. Environ. Eng 126, 239–249. Bligh, E.G., Dyer, W.J., 1959. A rapid method of total lipid extraction and purification. Can. J. Biochem. Physiol 37, 911–917. Chiu, Y.-C., Chung, M.-S., 2003. Determination of optimal COD/ nitrate ratio for biological denitrification. International biodeterioration biodegradation 51, 43–49. Clifford, D., Liu, X., 1993. Ion exchange for nitrate removal. J. Am. Water Works Assoc. 85, 135–143. Cornelissen, E.R., Moreau, N., Siegers, W.G., Abrahamse, A.J., Rietveld, L.C., Grefte, A., Dignum, M., Amy, G., Wessels, L.P., 2008. Selection of anionic exchange resins for removal of natural organic matter (NOM) fractions. Water Res. 42, 413–423. Cyplik, P., Grajek, W., Marecik, R., Kro´liczak, P., Dembczyn´ski, R., 2007. Application of a membrane bioreactor to denitrification of brine. Desalination 207, 134–143. Dowling, N.J.E., Widdel, F., White, D.C., 1986. Phospholipid esterlinked fatty acid biomarkers of acetate-oxidising sulphatereducers and other sulphide-forming bacteria. J. Gen. Microbiol. 132, 1815–1825. Fawehinmi, F., (2006). Anaerobic MBR treatment of a low strength municipal wastewater. PhD thesis, Cranfield University, UK.
Forney, L.J., Liu, W.-T., Guckert, J.B., Kumagai, Y., Namkung, E., Nishihara, T., Larson, R.J., 2001. Structure of microbial communities in activated sludge: potential implications for assessing the biodegradability of chemicals. Ecotoxicol. Environ. Saf 49, 40–53. Frank, B.P., Belfort, G., 2003. Polysaccharides and sticky membrane surfaces: critical ionic effects. J. Membr. Sci. 212, 205–212. Frølund, B., Griebe, T., Nielsen, P.H., 1995. Enzymatic activity in the activated-sludge floc matrix. Appl. Microbiol. Biotechnol 43, 755–761. ˚ ., Tunlid, A., Ba˚a˚th, E., 1991. Microbial biomass Frostega˚rd, A measured as total lipid phosphate in soils of different organic content. J. Microbiol. Methods 14, 151–163. Judd, S.J., 2006. The MBR Book: Principles and Applications in Water and Wastewater Treatment. Elsevier Science, Amsterdam. Kim, P.H.-S., Symons, J.M., 1991. Using anion exchange resins to remove THM precursors. J. Am. Water Works Assoc. 83, 61–68. Laspidou, C.S., Rittmann, B.E., 2002. A unified theory for extracellular polymeric substances, soluble microbial products, and active and inert biomass. Water Res. 36, 2711–2720. McAdam, E.J., Judd, S.J., Cartmell, E., Jefferson, B., 2007. Influence of substrate on fouling in anoxic immersed membrane bioreactors. Water Res. 41, 3859–3867. McAdam, E.J., Judd, S.J., 2008. Biological treatment of ionexchange brine regenerant for re-use: a review. Sep. Purif. Technol 62, 264–272. Petrovic, U., Gunde-Cimerman, N., Plemenitas, A., 1999. Salt stress affects sterol biosynthesis in the halophilic black yeast Hortaea werneckii. FEMS Microbiol. Lett. 180, 325–330. Peyton, B.M., Mormile, M.R., Peterson, J.N., 2001. Nitrate reduction with Halomonas Campisalis: kinetics of denitrification at pH9 and 12.5% NaCl. Water Res. 35, 4237–4242. Pflu¨ger, K., Mu¨ller, V., 2004. Transport of compatible solutes in extremophiles. J. Bioenerg. Biomembr 36, 17–24. Reid, E., Liu, X., Judd, S.J., 2006. Effect of high salinity on activated sludge characteristics and membrane permeability in an immersed membrane bioreactor. J. Membr. Sci. 283, 164–171. Russell, N.J., 1989. Adaptive modifications in membranes of halotolerant and halophilic microorganisms. J. Bioenerg. Biomembr 21, 93–113. Sutherland, I.W., 2001. Biofilm exopolysaccharides: a strong and sticky framework. Microbiology 147, 3–9. Wile´n, B.-M., Jin, B., Lant, P., 2003. The influence of key chemical constituents in activated sludge on surface and flocculating properties. Water Res. 37, 2127–2139. Yakimov, M.M., Giuliano, L., Chernikova, T.N., Gentile, G., Abraham, W.-R., Lu¨nsdorf, H., Timmis, K.N., Golyshin, P.N., 2001. Alcalilimnicola halodurans gen. nov., sp. nov., an alkaliphilic, moderately halophilic and extremely halotolerant bacterium, isolated from sediments of soda depositing Lake Natron, East Africa Rift Valley. Int. J. Syst. Evol. Microbiol. 51, 2133–2143. Yoshie, S., Ogawa, T., Makino, H., Hirosawa, H., Tsuneda, S., Hirata, A., 2006. Characteristics of bacteria showing high denitrification activity in saline wastewater. Lett. Applied Microbiol. 42, 277–283. Zhang, X., Bishop, P.L., Kinkle, B.K., 1999. Comparison of extraction methods for quantifying extracellular polymers in biofilms. Water Sci. Technol 39, 211–218. Zhang, J., Chua, H.C., Zhou, J., Fane, A.G., 2006. Factors affecting the membrane performance in submerged membrane bioreactors. J. Membr. Sci. 284, 54–66.
water research 44 (2010) 77–84
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Combined Bayesian statistics and load duration curve method for bacteria nonpoint source loading estimation Jian Shen, Yuan Zhao* Virginia Institute of Marine Science, College of William and Mary, 1208 Greate Road, P.O. Box 1346, Gloucester Point, VA, 23062, USA
article info
abstract
Article history:
Nonpoint source load estimation is an essential part of the development of the bacterial
Received 30 May 2009
total maximum daily load (TMDL) mandated by the Clean Water Act. However, the
Received in revised form
currently widely used watershed-receiving water modeling approach is usually associated
1 September 2009
with a high level of uncertainty and requires long-term observational data and intensive
Accepted 1 September 2009
training effort. The load duration curve (LDC) method recommended by the EPA provides
Available online 6 September 2009
a simpler way to estimate bacteria loading. This method, however, does not take into consideration the specific fate and transport mechanisms of the pollutant and cannot
Keywords:
address the uncertainty. In this study, a Bayesian statistical approach is applied to the
Bacteria
Escherichia coli TMDL development of a stream on the Eastern Shore of Virginia to inversely
Nonpoint source
estimate watershed bacteria loads from the in-stream monitoring data. The mechanism of
Bayesian statistics
bacteria transport is incorporated. The effects of temperature, bottom slope, and flow on
Load duration curve
allowable and existing load calculations are discussed. The uncertainties associated with
Water quality
load estimation are also fully described. Our method combines the merits of LDC, mech-
Total maximum daily load
anistic modeling, and Bayesian statistics, while overcoming some of the shortcomings associated with these methods. It is a cost-effective tool for bacteria TMDL development and can be modified and applied to multi-segment streams as well. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
In the United States, a total of 95,292 miles of streams and 3653 square miles of estuaries were impaired by pathogens, the number one cause of impairment for the assessed streams and estuaries (USEPA, 2008). Fecal indicator bacteria, e.g., fecal coliform, enterococci, and Escherichia coli (E. coli), are commonly used to indicate fecal pollution and the possible presence of pathogens. Total maximum daily load (TMDL) for impaired waterbodies is mandated by Section 303(d) of the Clean Water Act for establishing bacteria loading strategy (USEPA, 2001). The dominant sources of fecal bacteria in waterbodies in the Chesapeake Bay region are non-point sources such as livestock and wildlife feces from adjacent watersheds (Shen
et al., 2006). To estimate the nonpoint source loads from the watershed, a watershed model is widely used. Upon receiving the estimated loads from the watershed model, a water quality model can be applied to simulate bacteria concentrations in the receiving waterbody. The calibration of the watershed model and the receiving water model are often conducted through an iterative trial-and-error method and long-term observational data are highly warranted. However, as field monitoring data are limited in most situations, uncertainties associated with model simulation are unknown. On the other hand, even with sufficient monitoring data, given the highly variable bacteria concentrations driven by episodic events (e.g., storms and wildlife migration), it is difficult to calibrate these events without knowing the source variations
* Corresponding author. Tel.: þ1 (804) 832 3533; fax: þ1 (804) 684 7899. E-mail address:
[email protected] (Y. Zhao). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.002
78
water research 44 (2010) 77–84
(Sisson et al., 2008). As a result, the watershed-receiving water modeling approach is often associated with a high level of uncertainty (Wu et al., 2006). In addition, the current widely used mechanistic watershed models, such as Hydrologic Simulation Program-Fortran (HSPF), Generalized Watershed Loading Function (GWLF), and Soil and Water Assessment Tool (SWAT), are complex models that require a great amount of input data and training effort. Therefore, how to accurately and efficiently estimate nonpoint source load remains a highly challenging topic. Besides the watershed-receiving water modeling approach, the load duration curve (LDC) method recommended by the EPA (USEPA, 2007) provides a simpler way to estimate watershed loading in bacteria TMDL development (e.g., FDEP, 2004; USEPA, 2007). The major steps of this method include: (1) generating a flow duration (cumulative frequency) curve based on available historical flow data; (2) calculating TMDL by multiplying the numerical water quality target with flows, and plotting against the cumulative frequency; (3) estimating the existing instantaneous loads by multiplying the observed ambient water quality data with the averaged flows on the sampling dates, and plotting against the corresponding cumulative frequency on the LDC; and (4) calculating the margin of safety, load allocation and reduction for different flow regimes based on the LDC. This method is simple and straightforward, and particularly useful in addressing the essential role played by flows in determining load capacity. It does, however, have an intrinsic shortcoming as it does not take the specific fate and transport mechanisms of the pollutant into consideration. If factors other than flow significantly affect the loading capacity, the LDC method becomes insufficient. For example, it cannot address the effect resulting from fecal bacteria decay, which is modified significantly by temperature. In addition, similar to mechanistic models, the LDC method cannot address the uncertainty associated with loading estimation. In this paper, we present a novel, LDC-like method that uses the Bayesian statistical approaches to inversely estimate watershed bacteria loads from the in-stream monitoring data. The inverse modeling approach, which has wide applications to environmental problems, provides an efficient way for load estimation by directly incorporating observational data into model simulations (Shen et. al., 2006; Shen, 2006; Barth and Hill, 2005). The Bayesian statistical method has been applied in environmental modeling (Dowd and Meyer, 2003; Huang and McBean, 2007) and watershed load estimation (Liu et al., 2008; Shen and Zhao, 2009) because of its capability to manage the uncertainties of unknown parameters. Our method was applied to the E. coli TMDL development of the Sandy Bottom Branch (SBB) and its tributary, Unnamed Tributary to Sandy Bottom Branch (UTSBB), on the Eastern Shore of Virginia. The effects of temperature, bottom slope, and flow on allowable and existing load calculations were considered. The uncertainties associated with load estimation were also fully described. A LDC was used to calculate the required load reductions for different flow regimes. This method combines the merits of LDC, mechanistic, and statistical methods, while overcoming some of the shortcomings associated with them.
2.
Methods
2.1.
Study site
The SBB and its tributary, UTSBB, are located in Accomack County of Virginia’s Eastern Shore (Fig. 1). The total length of the stream is about 4200 m and the mean width is 2 m. The watershed area is about 6.9 km2 and dominated by forest and agriculture. According to the field measurements conducted by the Virginia Department of Environmental Quality (VADEQ) from its monitoring station at the outlet of SBB (Fig. 1), the E. coli concentrations exceeded the VA water quality criterion of a maximum of 126 counts/100 ml for non-tidal recreational waters. The bacteria source tracking data suggested that the contamination was mainly contributed by non-point sources from livestock and wildlife.
2.2.
Distributed-source model and inverse modeling
Since the stream is narrow and the watershed associated with the headwater is very small, the bacteria loads can be assumed to be discharged laterally into the system. Assuming the bacteria is fully mixed laterally and vertically, in a steady state the E. coli concentration C in the stream can be described by the distributed-source model (Chapra, 1997) as follows: u
dC ¼ kC þ L dx
(1)
where L is the diffuse source load, x is the distance measured from the headwater, k is the first-order bacteria decay rate, and u is the mean cross-sectional velocity, which is calculated as flow Q divided by cross-sectional area A. To account for the temperature effect, the bacteria decay rate k was treated as a function of temperature (Thomann and Mueller, 1987): k ¼ k0 qT20
(2)
where k0 is the decay constant at 20 C, T is temperature, and q ¼ 1.07. When the bacteria loading from the headwater is minimal and therefore can be neglected, the solution for Eq. (1) is: Ci ¼ f ðLi ; k0 ; Ti ; Qi ; Ai ; xÞ ¼
A Li k qTi -20 xQi i 1e 0 Ti 20 k0 q
(3)
where subscript i denotes different monitoring times. Eq. (3) is a forward model, which calculates Ci from Li. Its inverse form can be expressed as: Li ; k0 ¼ f 1 ðCi ; Ti ; Qi ; Ai ; xÞ
(4)
Thus the problem of estimating the load and decay constant can be transformed into an inverse model (Shen et al., 2006). It estimates k0 and Li jointly based on the observed in-stream E. coli concentrations, temperatures, simulated flows, crosssectional areas, and distance from the headwater. The Li obtained from the above equations is the daily unit load (counts$m3$day1). The daily load (counts$day1) can be computed as the daily unit load times the stream volume. To account for the change of cross-sectional area with respect to the change of flow, the Manning’s equations is used:
water research 44 (2010) 77–84
79
Fig. 1 – Location of the SBB and UTSBB watershed.
u¼
1 2=3 pffiffiffi s R n
(5)
where n is the Manning coefficient, R is the hydraulic radius, and s is the slope of the channel bed. Assuming the channel is rectangular, Eq. (5) can be written as: Q¼
2=3 pffiffiffi 1 WH s WH n W þ 2H
(6)
where W is the width of the channel and H is the depth. For natural rivers, the value of n varies approximately between 0.025 and 0.1. Throughout our study, n was fixed at 0.035 for simplicity as the bottom materials are similar in the river. For TMDL and existing load calculations of SBB and UTSBB, the s was also fixed at 1.0 105, a reasonable value for a coastal plain stream. For a given flow generated from LSPC and mean stream width of 2 m, the water depth can be obtained using an iteration method from Eq. (6). The cross-sectional areas corresponding to different flows thus can be computed and used as the input for Eq. (3).
2.3.
Bayesian parameter estimation
The Bayesian parameter estimation approach was used to compute the loadings Li and decay constant k0. The unknown parameters were treated as random variables with certain distributions and derived from known information (Ci, x, Qi, Ai, and Ti). The Bayes’ theorem can be written as: pðqjCÞ ¼
pðCjqÞ pðqÞ pðCÞ
(7)
In this equation, pðqjCÞ is called the posterior distribution and represents the probability of the model parameter (q) values given the observed data (C ). In our case, the parameter q refers to loads Li and decay constant k0, i.e., q¼{L1, L2,., Ln, K0}. p(C ) is the expected value of the likelihood function over the parameter distributions as a normalizing constant. pðCjqÞ is
the probability density function (likelihood function), which describes the mechanistic and statistical relationships between the predictors and response variables. p (q) is the prior, the probability density function over all values of q prior to the observed data. Because the bacteria concentration can be considered to follow a log-normal distribution, the modeled bacteria concentration with the consideration of random measurement error is given by: lnðCi Þ ¼ ln Ci þ 3i 0; s2 A k qTi 20 xQi i ¼ lnðLi Þ ln k0 qTi 20 þ ln 1 e 0 þ 3i 0; s2
(8)
where Ci is the observed concentrations and 3i is the error term following a N (0, s2) distribution. The likelihood function can be expressed as: n Y
1 2 1 pffiffiffiffiffiffiffiffiffiffiffi e2s2 ðlnCi lnCi Þ 2 2ps i¼1
(9)
where n ¼ 45, the number of observations at the DEQ water quality monitoring station. The software WinBUGS (Spiegelhalter et al., 2003), which implements Markov Chain Monte Carlo (MCMC) using Gibbs sampling, was used for the Bayesian estimation. The major steps in Bayesian modeling using the MCMC sampling can be found in Malve and Qian (2006). The variance s2 was assumed to follow a standard non-informative diffuse inverse-Gamma distribution (1.0 103, 1.0 103). The prior densities of ln(Li) and k0 were assumed to follow uninformative uniform distributions within the ranges of (0, ln(Lm)) and (0, km). The parameter Lm is the estimated maximum load (1.0 1012 counts/day/m3) based on experience. km ¼ 0.8/ day and (0, km) is a reasonable range for the decay constant in fresh water (USEPA, 2001; Mancini, 1978; Thomann and Mueller, 1987).
80
water research 44 (2010) 77–84
1/1/1985 to 1/1/1994 Observed flow
Modeled flow over the same period
10
Flow (m3/s)
1
0.1
0.01
0.001
0
10
20
30
40
50
60
70
80
90
100
Percent of time indicated flows are equaled or exceeded
Fig. 2 – 10-year accumulated daily stream flow comparison between model simulation and the reference flow station USGS 01484800.
2.4.
Flow simulation
As there is no USGS gauge station in our study site, the watershed model developed in the nearby Onancock Creek, VA, was adopted to simulate the watershed flow for this study. The model was calibrated using USGS flow station 01484800 in Guy Creek near Nassawadox, which is located approximately 53 km south of our study site. The Loading Simulation Program Cþþ (LSPC; Wang, 2005; Shen et al., 2005) was used to simulate surface runoff and subsurface flow from different land uses. The model is essentially a re-coded Cþþ version of selected Hydrological Simulation Program FORTRAN (HSPF) (Bicknell et al., 1996; Shen et al., 2005). The model is driven by hourly precipitation and simulates hourly surface runoff and subsurface flows. Both PWATER and IWATER modules were used to simulate discharges from pervious and impervious land uses. All the hydrology parameters for each land use were calibrated based on the flow observations in Guy Creek and Onancock Creek (Wang, 2005). Since the land uses and soil characteristics of both watersheds are very similar, the calibrated model parameters from Onancock Creek watershed are applicable to our study site. The simulated daily fresh water discharge by LSPC during 1999–2007 was used to generate the flow duration curve.
2.5. Calculation of existing load, TMDL, and load reduction The observed bacteria concentrations and temperatures from 2002 to 2007, the simulated flows, and the corresponding crosssectional areas were input to WinBUGS. The means, medians, and 2.5% and 97.5% quantiles of the predicted ln(unit load) were exponentiated and multiplied by the flow-adjusted stream volume (Eq. 6) to obtain the existing daily loads and their corresponding 95% credible intervals. The water quality criterion of 126 E. coli counts/100 ml, the temperature of 1 C, and the flows over the entire flow duration together with the
corresponding cross-sectional areas were directly substituted into Eq. (3) to calculate the maximum allowable unit loads, which were subsequently multiplied by the flow-adjusted stream volume to generate the maximum allowable loads. Here, the low temperature of 1 C was used for decay rate calculation as a conservative estimate. For TMDL purposes, the flow duration was divided into three regimes, representing low (> 70th percentile), median (30–70th percentile), and high (< 30th percentile) flows. Then, the TMDL for each flow regime was set as the maximum allowable load corresponding to the median flow of the regime, i.e., the 85th, 50th, and 15th percentile flows (5.7 102, 9.4 102, and 2.0 101 cms), respectively. The mean value of the maximum existing load in each flow regime was used to represent the existing load for the regime. The difference between the TMDL and existing load was used to calculate the load reduction for each regime.
3.
Results
Fig. 2 shows the 10-year daily stream flow frequency comparison between the LSPC model result and field data collected at the USGS station. It can be seen that LSPC has reasonably reproduced the observed flow over a 10-year period. The flow rates of the SBB and UTSBB system ranged from 5.7 105 to 4.8 cms, and 50% of them were smaller than 5.1 102 cms. When the observed E. coli concentrations and temperatures, and the simulated flow rates at the same days of the field observations were input into WinBUGS using Eq. (4), the model converged after 15,000 runs. The Monte Carlo error associated with each prediction, an estimate of the difference between the mean of the sampled values and the true posterior mean, is very small (less than 5% of the standard deviation), suggesting good convergence and high accuracy (Spiegelhalter et al., 2003). The estimated k0 has a mean of 0.5 day1 with a standard deviation of 0.2 day1. This mean k0 can be used for the subsequent load calculations. With the
water research 44 (2010) 77–84
Modeled (Count/100ml)
10000
1000
100
10 10
100
1000
10000
Observed (Count/100ml) Fig. 3 – Scatter plot of the observed and modeled in-stream E. Coli concentrations. The straight line denotes the 1:1 ratio.
temperatures, the simulated flow rates, and the posterior means of k0 and daily unit loads, the mean in-stream bacteria concentrations were predicted by WinBUGS as well. Fig. 3 shows the scatter plot of the observed vs. simulated in-stream concentrations. It can be seen that the comparison is very good with the R2 value close to 1.0. Fig. 4 illustrates the posterior mean, median, and 2.5% and 97.5% quartiles (95% credible intervals) of the daily unit load, corresponding to each field observation. From 2002 to 2007, the mean unit load ranged from 1.4 105 to 2.9 107 count$day1$m3. There was no apparent temporal trend of load but the several most recent loads seemed very stable. The existing mean loads together with their 95% credible intervals, TMDLs, and load reductions in each flow regime are illustrated in the LDC of Fig. 5. The E. coli impairment existed in all flow regimes, and mostly occurred under a low flow condition. The TMDL, existing load, and load reduction of each flow regime are listed in Table 1.
4.
Discussion
4.1.
Incorporating uncertainty in load estimation
Uncertainty is a fundamental characteristic of all scientific activities; therefore environmental decisions should be made in ways that reflect the uncertainty (Ellison, 1996). In TMDL development, both mechanistic modeling and the LDC methods can only provide point estimations for loadings, while the uncertainties associated with them are substantial but unknown. To address the uncertainty, the common way is to include a margin of safety, which is usually 5% or 10% of the maximum allowable load. However, this is arbitrary and lacks a scientific base. In comparison, by solving the mechanistic equation (i.e., Eq. 3) with the inverse modeling technique and Bayesian statistical methods, we are able to estimate watershed
81
bacteria loads from the in-stream monitoring data, thus save a great amount of effort from the tedious yet subjective model calibration process. Most importantly, we are able to obtain both point and interval estimations for loads. Taking the first unit load estimation in Fig. 4 as an example, it can be concluded that the point estimate of the unit load is 1.4 105 count$day1$m3 (mean value), there is a 0.5 probability that the unit load is greater than 1.3 105 count$day1$m3 (median), and that 95% of the potential values of the unit load will fall within the range of 9.1 104 to 4.0 105 count$day1$m3 (95% credible interval). From management point of view, this is clear language that can be easily understood by decision-makers and the public. Compared with the traditional modeling approaches, this method provides decision-makers and stakeholders with an explicit basis for their decision on the load reduction of a TMDL. For example, if large uncertainty is associated with the existing load, caution should be exercised and more observational data or complex transport model may be needed to lower the uncertainty. In fact, the load reduction scenario shown in Fig. 5 is just one of the many ways to quantify the load reduction. If being conservative is the major concern during a TMDL development, the lowest TMDL and the upper 95% credible level of the highest existing load within each flow regime can be selected as the maximum and existing loads, respectively, and higher load reductions will be needed.
4.2. Effects of flow, temperature, and stream channel slope on bacteria TMDL In general, a high flow washes more pollutants off the watershed and results in a higher load. The effect of flow on the pollutant load (e.g., sediment, nutrients, and bacteria) has been widely recognized and incorporated into most load estimation methods (e.g., Sheeder and Evans 2004, USEPA, 2007, Shen et al., 2005; Haith and Shoemaker, 1987). The LDC method recommended by the EPA, for example, divides flow into several intervals and calculates load reduction within each interval (USEPA, 2007). It only focuses on flow while ignoring the effects of other factors, such as temperature and channel slope, because of its inability to incorporate specific fate and transport mechanisms. Temperature plays an important role in determining the fate of bacteria by affecting the decay rate (Crane and Moore, 1986; Collins and Rutherford, 2004). From Eq. (2), a 10 C temperature increase will result in a doubling of the decay rate. Channel bottom slope affects the TMDLs mainly through affecting water depth, and thus the cross-sectional area and stream volume (Eq. 6). The effects of temperature and slope on loading capacity of the SBB and UTSBB system can be further illustrated by the TMDL response surface in Fig. 6. Using the E. coli water quality criterion of 126 counts/100 ml, the decay constant k0 of 0.5 estimated by WinBUGS, the stream length of 4200 m, and the mean stream width of 2 m, the TMDLs corresponding to temperature ranging from 0 to 30 C and slope ranging from 1 105 to 5 103 under the 50 percentile flow rate were computed. It can be seen that higher load is allowed under higher temperature and smaller bottom slope. Any TMDL calculation ignoring this fact will result in inaccurate estimations. For example, with a slope of 1 105, the maximum allowable load increases from 1 1010
82
water research 44 (2010) 77–84
1.E+07
1.E+06
1.E+05 Median Mean
1.E+04
7/23/02 9/24/02 10/2/02 11/19/02 1/21/03 3/18/03 9/24/03 10/27/03 11/18/03 1/20/04 4/6/04 5/25/04 7/20/04 9/22/04 10/27/04 11/9/04 1/26/05 3/29/05 4/26/05 5/18/05 7/19/05 9/20/05 10/19/05 11/17/05 1/17/06 3/21/06 4/27/06 5/23/06 7/18/06 9/19/06 10/18/06 11/15/06 1/23/07 3/27/07 4/10/07 5/8/07 5/22/07 5/29/07 6/19/07 7/16/07 7/24/07 8/27/07 9/11/07 10/31/07 11/15/07
Unit Load (Count/Day/m3 )
1.E+08
Date Fig. 4 – Predicted mean and median of the load per unit volume. The error bars denote the 2.5% and 97.5% quartiles.
counts$day1 at 0 C to 2 1010 counts$day1 at 32 C; at 32 C, the maximum allowable load decreases from 2 1010 counts$day1 at a slope of 1 105 to 1 1010 counts$day1 at a slope of 5 103. The TMDL surface calculated by multiplying the 50 percentile flow and water quality criterion (LDC method) is shown in Fig. 6 as well. It can be seen that the LDC TMDLs are always lower than those calculated from our method within the parameter ranges used, reflecting its conservative characteristic. The ignorance of these factors may result in a high relative error. While being conservative is necessary for environmental protection purposes, to underestimate the TMDLs to such a degree may also involve much more unnecessary costs on loading reduction practices. Therefore, environmental managers should take all these factors into consideration and make corresponding decisions for implementation strategies.
4.3.
Application of the method to multi-segment streams
The transport model used in this study (Eq. 3) assumes the bacteria loading from the headwater is minimal and can be neglected. This is justified as the SBB and UTSBB system is relatively short and its topography does not change significantly from upstream to downstream. In the case of long distance with varying width, the multiple segment method proposed by Liu et al. (2008) can be used. The river can be divided into several segments and a multi-segment transport model can be used: k0j q
Cj;i ¼ Cj1;i e
-
Tj;i 20 Aj;i xj Q j;i
A
T 20 j;i Lj;i k0j q j;i xj Q j;i 1e þ k0j qTj;i 20
! (10)
where j and i refer to river segment and monitoring time respectively. Cj-1,i is the bacteria concentration in the
1.0E+13
E. coli (Count/Day)
1.0E+12
1.0E+11
1.0E+10
1.0E+09 TMDL Existing Load
1.0E+08
0
10
20
30
40 50 60 70 Flow Duration Interval (%)
80
90
100
Fig. 5 – TMDL, existing load, and load reductions for different flow regimes. The circles and error bars indicate the means and 95% credible intervals of the loads. The dash line denotes the TMDL, and arrows the load reduction of each regime.
water research 44 (2010) 77–84
Table 1 – TMDLs, existing loads, and required reductions for each flow regime. Load unit in count/day. Flow regime
>70% (Low) 30–70% (Median) <30% (High) 7.30 109 1.54 1011 95.3%
TMDL Existing load Required reduction (%)
x 10
1.19 1010 3.03 1011 96.1%
2.59 1010 5.40 1011 95.2%
83
estimate the existing loads. The TMDL and loading reduction was calculated on a LDC. This approach not only takes bacteria fate and transport mechanism into consideration but also addresses the uncertainty associated with loading estimation, which is important to environmental management. It provides researchers and managers a cost-effective tool for bacteria TMDL development. Compared with the LDC method that the EPA recommended, the approach provides better estimations. If needed, it can be applied to multi-segment rivers for load estimation as well.
10
1.9
Acknowledgements
1.8
TMDL
1.7 1.6
The funding of this study is supported by the Virginia Department of Environmental Quality. We thank Jennifer Howell and David Lazarus of VADEQ for discussions and support. This is contribution number 3040 of the Virginia Institute of Marine Science.
1.5 1.4 1.3 1.2 1.1 1 6
references 4
x 10
30
-3
Slope
20
2
10 0
0
Temperature (C)
Fig. 6 – Response surface of TMDL to temperature and channel slope. The TMDL calculated by the traditional LDC method is also shown (unfilled blue grids at the bottom).
upstream segment j-1. In this way, the bacteria load contributed by the upstream of segment j is incorporated. The likelihood function is written as: m Y n Y i¼1
1 1 pffiffiffiffiffiffiffiffiffiffiffi e 2s2 2 2ps j¼1
lnCj;i lnCj;i
2 (11)
The values of bacteria decay constant and unit loading within each segment, together with the associated uncertainty, can be estimated by WinBUGS. The existing loads of the entire river can then be calculated as the sum of all the segment loads. Similarly, the information can be displayed on a LDC and the reduction for each flow interval calculated. This approach can be applied to other water quality constituents as well. For example, Liu et al. (2008) used the multi-segment method to estimate nutrients and biological oxygen demand loadings and obtained satisfactory results. This method can also be extended to an estuary with bi-directional flows and mixing processes. Because no analytical solution can be obtained for bacteria transport in estuaries, an alternative approach using a finite difference method can be used (Shen and Zhao, 2009).
5.
Summary and conclusions
We applied a modified LDC method to the bacteria TMDL development of SBB and UTSBB. A distributed-source model was used to simulate the transport of bacteria in the stream. An inversed Bayesian modeling approach was used to
Barth, G., Hill, M., 2005. Numerical methods for improving sensitivity analysis and parameter estimation of virus transport simulated using sorptive-reactive processes. Journal of Contaminant Hydrology 76 (3–4), 251–277. Bicknell, B.R., Imhoff, J.C., Kittle, J., Donigian, A.S., Johansen, R.C., 1996. Hydrological Simulation Program – FORTRAN, User’s Manual for Release 11. U.S. Environmental Protection Agency, Environmental Research Laboratory, Athens, GA, p. 284. Chapra, S.C., 1997. Surface Water-Quality Modeling. McGraw-Hill, New York, USA, pp. 844. Collins, R., Rutherford, K., 2004. Modelling bacterial water quality in streams draining pastoral land. Water Research 38 (3), 700–712. Crane, S.R., Moore, J.A., 1986. Modelling enteric bacterial die-off: a review. Water, Air, and Soil Pollution 27 (3–4), 411–439. Dowd, M., Meyer, R., 2003. A Bayesian approach to the ecosystem inverse problem. Ecological Modelling 168 (1–2), 39–55. Ellison, A.M., 1996. An introduction to Bayesian inference for ecological research and environmental decision-making. Ecological Applications 6 (4), 1036–1046. Florida Department of Environmental Protection (FDEP), 2004. Fecal and total coliform TMDL for Delaney Creek. http://www. dep.state.fl.us/water/tmdl/docs/tmdls/final/gp2/ DelaneyCreekColiformsTMDL.pdf (accessed Dec 2008). Haith, D.A., Shoemaker, L.L., 1987. Generalized watershed loading functions for stream-flow nutrients. Water Resource Bulletin 23 (3), 471–478. Huang, J.J., McBean, E.A., 2007. Using Bayesian statistics to estimate the coefficients of a two-component second-order chlorine bulk decay model for a water distribution system. Water Research 41 (2), 287–294. Liu, Y., Yang, P., Hu, C., Guo, H., 2008. Water quality modeling for load reduction under uncertainty: a Bayesian approach. Water Research 42 (13), 3305–3314. Malve, O., Qian, S.S., 2006. Estimating nutrients and chlorophyll a relationships in Finnish lakes. Environmental Science and Technology 40 (24), 7848–7853. Mancini, J.L., 1978. Numerical estimates of coliform mortality rates under various conditions. Journal of Water Pollution Control Federation 50 (11), 2477–2484. Sheeder, S.A., Evans, B.M., 2004. Estimating nutrient and sediment threshold criteria for biological impairment in
84
water research 44 (2010) 77–84
Pennsylvania watersheds. Journal of the American Water Resources Association 40 (4), 881–888. Shen, J., Zhao, Y., 2009. A Bayesian approach for estimating bacterial nonpoint source loading in an estuary with limited observations. Journal of Environmental Science and Health (A) 44 (14), doi:10.1080/10934520903263553. Shen, J., 2006. Optimal estimation of parameters for an estuarine eutrophication model. Ecological Modeling 191 (3–4), 521–537. Shen, J., Jia, J., Sisson, M., 2006. Inverse estimation of nonpoint sources of fecal coliform for establishing allowable load for Wye River, Maryland. Water Research 40 (18), 3333–3342. Shen, J., Sun, S., Wang, T., 2005. Development of the fecal coliform total maximum daily load using loading simulation program Cþþ and tidal prism model in estuary shellfish growing areas: a case study in the Nassawadox coastal embayment, Virginia. Journal of Environmental Science and Health (A) 40 (9), 1791–1807. Sisson, G.M., Jin, Z., Currey, L., Shen, J., Jia, J., 2008. Developing a cost-effective methodology to manage fecal coliform loading in shellfish harvesting areas of upper Chesapeake Bay, Maryland. Proceedings of the 10th International Conference on Estuarine and Coastal Modeling, November 3–7, 2007, Newport, Rhode Island, USA.
Spiegelhalter, D., Thomas, A., Best, N., Lunn, D., 2003. WinBUGS User Manual, Version 1.4. http://www.mrc-bsu.cam.ac.uk/ bugs/winbugs/manual14.pdf (accessed Dec 2008). Thomann, R.V., Mueller, J.A., 1987. Principles of Surface Water Quality Modeling and Control. Harper and Row Publishers, New York, USA, p. 644. United States Environmental Protection Agency (USEPA), 2001. Protocol for developing pathogen TMDLs. http://www.epa.gov/ owow/tmdl/pathogen_all.pdf (accessed Dec 2008). USEPA, 2007. Fecal Coliform TMDL for Horseshoe Creek. Georgia, Atlanta. http://www.epa.gov/Region4/water/tmdl/florida/ documents/2horseshoe_FC_EPA.pdf (accessed Dec 2008). USEPA Assessment Total Maximum Daily Load (TMDL), 2008. Tracking and implementation system (ATTAINS). http:// iaspub.epa.gov/waters10/attains_nation_cy.control#causes (accessed Dec 2008). Wang, T., 2005. Hypoxia in shallow coastal waters: a case study in Onancock Creek, Virginia. Master thesis, Virginia Institute of Marine Science, College of William and Mary, Gloucester Point, VA, p. 129. Wu, J., Zou, Rui, Yu., S.L., 2006. Uncertainty analysis for coupled watershed and water quality modeling systems. Journal of Water Resources Planning and Management 132 (5), 351–361.
water research 44 (2010) 85–96
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Nitrate removal and biofilm characteristics in methanotrophic membrane biofilm reactors with various gas supply regimes Oskar Modin a,b,*, Kensuke Fukushi b, Fumiyuki Nakajima c, Kazuo Yamamoto c a
Department of Urban Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan Integrated Research System for Sustainability Science (IR3S), The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8654, Japan c Environmental Science Center, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan b
article info
abstract
Article history:
Aerobic methanotrophs can contribute to nitrate removal from contaminated waters,
Received 4 June 2009
wastewaters, or landfill leachate by assimilatory reduction and by producing soluble
Received in revised form
organics that can be utilized by coexisting denitrifiers. The goal of this study was to
31 August 2009
investigate nitrate removal and biofilm characteristics in membrane biofilm reactors
Accepted 1 September 2009
(MBfR) with various supply regimes of oxygen and methane gas. Three MBfR configurations
Available online 8 September 2009
were developed and they achieved significantly higher nitrate removal efficiencies in terms of methane utilization (values ranging from 0.25 to 0.36 mol N mol1 CH4) than have
Keywords:
previously been observed with suspended cultures. The biofilm characteristics were
Biofilm
investigated in two MBfRs with varying modes of oxygen supply. The biofilms differed in
Denitrification
structure, but both were dominated by Type I methanotrophs growing close to the
Membrane
membrane surface. Detection of the nitrite reductase genes, nirS and nirK, suggested
Methane
genetic potential for denitrification was present in the mixed culture biofilms.
Methanotrophs
ª 2009 Elsevier Ltd. All rights reserved.
Nitrate
1.
Introduction
Denitrification with methane would potentially be an attractive nitrate removal method for waters, wastewaters, and landfill leachate requiring the addition of an external electron donor. Wastewater treatment plants and landfills where methane is generated onsite would be particularly well-suited for this technology. The literature describes two microbial processes capable of carrying out denitrification with methane. The first, anaerobic methane oxidation coupled to denitrification, is as
of yet poorly understood and appears to be accomplished by slow-growing microorganisms (Raghoebarsing et al., 2006; Ettwig et al., 2008). The other, aerobic methane oxidation coupled to denitrification (AME-D), has been observed in several laboratory studies (Rhee and Fuhs, 1978; Meschner and Hamer, 1985; Werner and Kayser, 1991; Amaral et al., 1995; Costa et al., 2000; Eisentraeger et al., 2001) and is accomplished by aerobic methanotrophs oxidizing methane and releasing soluble organic compounds that are utilized by coexisting denitrifiers. This study focuses on the latter of these two processes.
* Correspondence to: Oskar Modin, Integrated Research System for Sustainability Science (IR3S), The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8654, Japan. Tel.: þ81 90 66341728; fax: þ81 3 58418533. E-mail addresses:
[email protected] (O. Modin),
[email protected] (K. Fukushi),
[email protected] (F. Nakajima),
[email protected] (K. Yamamoto). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.009
86
water research 44 (2010) 85–96
Two major challenges must be overcome for the AME-D process to be feasible in engineering applications (Modin et al., 2007): The efficiency of nitrate removal must be improved. Oxygen inhibits denitrification; however, it must be added to the process to allow methane oxidation. Thus, the bioreactor should provide both aerobic spaces for methane oxidation and anoxic spaces for denitrification. Methane must be delivered to the microorganisms without losses to the atmosphere and without flammable mixtures with oxygen. Methane is a greenhouse gas so losses to the atmosphere are not acceptable, furthermore, mixtures of methane and oxygen may pose a safety hazard for this type of reactor. Methane mixtures with air are flammable between 5% and 15% methane concentration. A membrane biofilm reactor (MBfR) is a bioreactor configuration that potentially meets the above challenges. In an MBfR, a gaseous substrate is supplied from the interior of a membrane to a biofilm growing on the membrane surface. The biofilm consumes the supplied gas preventing losses to the bulk liquid. MBfRs are previously described for delivery of oxygen to nitrifying (Brindle et al., 1998; Downing and Nerenberg, 2008) and nitrogen removing biofilms treating wastewater (Timberlake et al., 1988; Hibiya et al., 2003; Semmens et al., 2003). MBfRs have also been described for delivery of hydrogen to biofilms reducing oxidized contaminants such as nitrate, perchlorate, chromate, selenate, and bromate (Lee and Rittmann, 2000; Nerenberg and Rittmann, 2004). Clapp et al. (1999) investigated TCE removal by an MBfR fed with methane and oxygen from the interior of a membrane and inoculated with the methanotroph Methylosinus trichosporium OB3b (Clapp et al., 1999). Hypothetically, the biofilm-mode of growth provides conditions that favor a more efficient AME-D process. In a biofilm, microorganisms live in close proximity to each other, which may enhance carbon transfer from aerobic methanotrophs to coexisting microbes. Furthermore, biofilms tend to develop anoxic regions in which denitrification would be the favored respiration route. An MBfR with both methane and oxygen supplied from the interior of a silicone tube has previously been shown to achieve nitrate removal with higher efficiency than a suspended growth reactor. The nitrate removal efficiencies in terms of methane utilization were 0.45 and 1 CH4 for the tested MBfR and suspended 0.11 mol NO 3 mol growth reactor, respectively (Modin et al., 2008). The gas supply regime used in that study, however, required gaseous mixtures of methane and oxygen. Such gas mixtures may be flammable and pose a safety concern for this type of bioreactor. The first goal of this study was to investigate nitrate removal in MBfR configurations with various supply regimes for methane and oxygen. The second goal was to investigate microproperties (dissolved oxygen concentration profiles, microbial structure, and genetic potential) of the methanotrophic biofilms that developed under two different gas supply regimes. Two experiments with different reactor setups were carried out. Nitrate removal efficiency was measured in Experiment 1 and microscale biofilm characteristics were investigated in Experiment 2.
2.
Materials and methods
2.1.
Growth medium and culture
Experiments were conducted with the nitrate minerals salts (NMS) medium described in Modin et al. (2008). Methane was supplied as the sole source of carbon and nitrate was the sole source of nitrogen. A highly enriched aerobic methaneoxidizing culture originating from activated sludge was used as inoculum for the experiments.
2.2.
Experimental procedures
2.2.1.
Experiment 1
The laboratory MBfR setup used in Experiment 1 consisted of a tubular membrane (silicone rubber tubing with an outside diameter of 1.0 mm, wall thickness of 0.25 mm, and approximately 30 cm long) placed centrally within a clear acrylic pipe with an inside diameter of 25 mm. NMS medium was circulated through the pipe from a 600 ml reservoir. Both ends of the silicone tubing membrane were connected to a gas reservoir where methane and/or oxygen and helium were pressurized to a total pressure of approximately 150 kPa. The experiment was carried out in room temperature which was about 25 C. The biofilm thickness was measured every 2–7 days. Four runs with varying gas supply regimes were carried out (Fig. 1): In run 1, methane and oxygen were mixed within the membrane and the bulk liquid was anoxic (Fig. 1A). Run 2 used two separate intertwined membrane tubes for supply of methane and oxygen. The bulk liquid was anoxic (Fig. 1B). In runs 3 and 4 only methane was supplied from the membrane and the bulk liquid was aerobic (Fig. 1C). Each run was started with clean silicone tubing as membrane and 500 ml of NMS medium in the liquid reservoir. The NMS medium was circulated through the membrane module at a flow rate of 0.83 ml s1, which resulted in a flow velocity of 0.17 cm s1 and a Reynolds number of approximately 48. The reactor was inoculated with the enrichment
Fig. 1 – Supply regimes of methane and oxygen used in Experiment 1. A: methane and oxygen are mixed within the membrane tube. B: methane and oxygen are supplied through separate intertwined membrane tubes. C: only methane is supplied from the membrane but the bulk liquid is aerobic.
water research 44 (2010) 85–96
culture. After 4 days the inoculation culture was replaced with 500 ml of fresh NMS medium. The reactor was operated as a batch and the liquid medium was typically replaced every 7 days. Since the nitrate-N concentration of the medium was about 200 mgN l1 and the medium was replaced every 7 days, the bulk liquid nitrate-N concentration in the reactor was typically decreasing from 200 to about 150 mg NO3-N l1 in each feed cycle. Because of the high nitrate concentrations, the biofilms were not likely limited by nitrate availability during the runs. The partial pressures of methane and oxygen in the gas reservoir were measured every 2–3 days. The flux of methane through the membrane could be calculated using the ideal gas law and the difference in partial pressure between the start and the end of each 2–3 day period. In run 1, methane and oxygen were mixed in the same 1100 ml reservoir; in run 2, two separate 600 ml reservoirs were used for supply of methane and oxygen. In runs 3 and 4, only methane was supplied from a 600-ml reservoir, while the liquid reservoir was aerated and its headspace open to the atmosphere. In run 3 the bulk liquid was aerated by air-sparging (100 ml min1) and the measured DO concentration ranged from 4.7 mg l1 to 8.2 mg l1. In run 4 it was aerated by stirring only and the measured DO concentration ranged from 2.5 mg l1 to 7.8 mg l1. In run 1 and 2, the liquid reservoirs were sealed from the atmosphere and oxygen was evacuated by sparging with He gas every 2–3 days. The maximum oxygen partial pressure of the liquid reservoir headspace never exceeded 1.1 kPa. Runs 1 and 2 lasted for 37 days and measurements were carried out in 5 batch feed cycles. Runs 3 and 4 lasted for 35 day and measurements were carried out in 4 batch feed cycles.
2.2.2.
Experiment 2
In Experiment 2, a flat-sheet silicone membrane was used to facilitate analysis of microscale properties in the biofilms. Two runs were conducted in parallel: In run A, methane and oxygen were supplied from the membrane and the bulk liquid was anoxic. In run B, only methane was supplied from the membrane and the bulk liquid was aerobic.
approximately 10.4 ml h1 to both reactors. The width of the flow channel was 2 cm, the height of liquid flow 1 cm, so the flow velocity over the membrane was 0.16 cm s1 and the Reynolds number 36. The length of the module was 29 cm. In the headspace above the liquid in the channel, a gas flow of either air or N2 was provided to keep the bulk liquid either aerobic or anoxic. Sampling ports were drilled into the top of the flow channel to allow for microsensor measurements of dissolved oxygen (DO) concentrations in the biofilm. The two reactors (A and B) were operated in parallel under continuous-flow conditions. Reactor A had an aerated liquid phase and only methane was supplied from the membrane (at a partial pressure of about 50 kPa, helium was provided as the inert gas). Reactor B had an anoxic liquid phase with both methane and oxygen supplied from the membrane (partial pressure was about 50 kPa of each gas). To accurately calculate the hydraulic retention time (HRT) of each reactor, the effluent volume was measured on each sampling occasion (usually every 2 days). Based on these measurements the HRT for reactors A and B were 58.3 and 55.3 h, respectively. The reactors were operated for 53 days. The influent nitrate concentration was stepwise increased throughout the run: from about 20 mg NO3-N l1 in the beginning to 80 mg NO3N l1 in the end. Microsensor measurements of the DO concentration within the biofilm were carried out on two occasions: the first on day 38 and the second on day 52.
2.2.3.
Post-harvest preparations (Experiment 2)
After 53 days of operation the biofilms in Experiment 2 were harvested. The silicone membranes with attached biofilms were excised from the reactors and cut into several segments. For microscopic observation and microbial analysis, some biofilm segments were prepared according to Okabe et al. (1999). The biofilm was fixed in a freshly prepared solution of 4% paraformaldehyde (PFA) dissolved in PBS (1X, pH 7.2), for 8 h. Then, the biofilm was washed three times in PBS and covered with an embedding medium (OCT compound 4583, Sakura Finetek). The OCT compound was allowed to infiltrate the biofilm overnight, before it was frozen at 70 C. The frozen biofilm segments were sliced at 20 C using a cryostat/microtome (Leica CM3050).
2.3. The laboratory MBfR setup consisted of two pieces, the upper partially liquid-filled flow channel and the lower gasfilled box. The pieces were separated by a 0.5-mm thick silicone rubber membrane placed on top of the gas-filled box. The top of the gas box gave the membrane a solid support to rest on, and many small perforations in the box allowed the silicone membrane to be in contact with the gas mixture inside. The gas-filled box (2 2 29 cm) was connected to a gas reservoir with a total volume of about 1.5 litre kept at approximately atmospheric pressure, which was necessary to avoid bulging of the silicone sheet. NMS medium was recirculated through the flow channel from a 500-ml liquid reservoir placed on a magnetic stirrer. Either air or N2 gas was sparged through the bulk liquid to provide either aerobic or anoxic conditions. The total liquid volume in the system was 540 ml. A peristaltic pump kept the recirculating liquid at a flow rate of 0.33 ml s1. The influent flow rate was
87
Analytical methods
Methane and oxygen were measured by gas chromatography (8A, Shimadzu) with a thermal conductivity detector. The gases were separated on a Molecular Sieve 5A column with helium as the carrier gas. Nitrate and nitrite were analyzed using ion chromatography on a Metrohm 761 Compact IC. Dissolved organic carbon (DOC) concentration was measured by filtering the sample through 0.45-mm PTFE filter and analysis with a Shimadzu TOC-5000A, total organic carbon analyzer. Volatile solids (VS) mass was measured according to Standard Methods (APHA et al., 1998). In Experiment 1, the biofilm thickness was measured nondestructively using a method developed by Freitas dos Santos and Livingston (1995). The clear pipe containing the silicone tubing with attached biofilm was placed between a light source and a lens. An image of the biofilm-covered silicone tube was projected on a screen 5 m away from the
88
water research 44 (2010) 85–96
lens. The biofilm thickness could be calculated based on the diameter measured on the screen. The accuracy of this method was approximately 20 mm. Biofilm thickness was measured in a room outside the laboratory, so the reactors were shut down for approximately 20 min every time biofilm thickness was measured. Biofilm density was measured at the end of the runs by removing all the biofilm from a section of the membrane by vortexing, measuring the VS mass, and dividing by the biofilm volume. The bulk liquid DO concentrations were measured with a membrane electrode (Broadley James Corp.). Microscale DO concentration profiles were measured using a Unisense OX10 dissolved oxygen microsensor, which has a tip diameter less than 10 mm. The microsensor was manually controlled with a micromanipulator and measurements were typically taken every 10–50 mm.
2.4.
Microbiological analysis
2.4.1.
Fluorescence in-situ hybridization (FISH)
In Experiment 2, the harvested biofilms from both reactors were prepared as either 10 mm thick cross-sections or 20 mm thick slices cut parallel to the membrane. The slices were placed onto glass slides and air dried, followed by a series of dehydration steps in 50, 80, and 100% ethanol (3 min each). The biofilm slices were hybridized with FISH probes targeting the 16S rRNA of type I and II methanotrophs (Mg84 þ Mg705 and Ma450) (Eller et al., 2001) and all bacteria (EUB338 I, II, III) (Amann et al., 1990; Daims et al., 1999), as well as with DAPI. A concentration of 5 ng ml1 was used for both the FISH probes and DAPI. The samples were observed using an epifluorescence microscope (Olympus BX51).
2.4.2.
PCR, cloning, and sequencing
DNA was extracted from the biofilms in Experiment 2 using the FastDNA Spin Kit for Soil and the FastPrep instrument (QBiogene, Inc., CA, USA) following the manufacturer’s instructions. PCR amplification of the nitrite reductase genes (nirS and nirK ) was carried using the cd3aF-R3cd (nirS, Throback et al., 2004) and F1aCu-R3Cu (nirK, Hallin and Lindgren, 1999) primer sets. Products of the expected length were confirmed with gel electrophoresis. The PCR products were purified using QIAquick PCR purification kit and cloned using QIAGEN PCR cloning kit. Colonies with inserts were again PCR amplified and the products were sequenced by Macrogen.
3.
Results
3.1.
Experiment 1
The average intramembrane pressures of methane and (when applicable) oxygen, and the nitrate and nitrite concentrations are shown in Fig. 2. Fig. 2 also shows nitrate removal efficiency, which is defined as the molar ratio between the amount of TIN (total inorganic nitrogen, nitrate þ nitrite) removed and the amount of methane that passed through the membrane (DTIN/DCH4) for each batch feed cycle. In run 1 and 2, the methane and oxygen partial pressures were increased after 23 days of operation. This led to an increase in the gas
flux through the membranes and a temporary drop in the nitrate removal rate, which resulted in a sudden drop in the TIN/CH4 ratio. In runs 3 and 4, only the methane pressure was increased (since oxygen was supplied from the bulk liquid). In all four runs, the nitrate removal efficiency was high in the first batch feed cycle, but stabilized at a lower value for subsequent cycles. The TIN removal rates were calculated for each feed cycle and ranged from 0.53 to 6.59 gTIN m2 d1 in run 1, 0.30 to 1.05 gTIN m2 d1 in run 2, 1.82 to 2.75 gTIN m2 d1 in run 3, and 1.56 to 2.89 gTIN m2 d1 in run 4. The 30-cm long membrane modules were operated under upflow conditions and biofilm thickness was measured in runs 1, 3, and 4 (Fig. 3). In run 1 the biofilm was clearly thicker on the lower 10 cm of the membrane tube where it reached about 600 mm. The middle and upper section of the membrane reached a biofilm thickness of 300 mm. For run 4, a certain difference between the lower and upper part of the membrane was also observed with the lower part reaching a biofilm thickness of 550 mm and the upper part reaching about 400 mm. This phenomenon was likely due to biomass detaching from the upper part of the membrane settling down and reattaching on the lower part. Run 3, however, had an evenly distributed biofilm reaching about 550 mm. Due to the uneven surface caused by the twisting of the membranes in run 2, biofilm thickness could not be quantitatively measured in this run. However, biofilm was mostly associated with the methane-permeating tubes or the interface between tubes, suggesting methane was the limiting substrate in run 2. It is also possible that differences in intramembrane gas pressures caused the varying biofilm thickness observed on the membrane tubes. The top and bottom of the membrane tubes were connected to the gas reservoirs. Back diffusion of He, N2, or CO2 may have lowered the partial pressure of CH4 (and O2) in the centre section of the membrane tubes. Lower partial pressure would lead to lower flux of the gas through the membrane (Rishell et al., 2004) leading to less biofilm growth. At the end of each run, the biofilm was removed from the membranes and the VS mass was measured. The mass of VS attached to the membranes were: 45.3 mg (run 1); 28.4 mg (run 2); 41.6 mg (run 3); and 27.8 mg (run 4). The average biofilm densities could be calculated for runs 1, 3, and 4 based on the biofilm thicknesses measured on the last day of the runs. The densities were: 78.6 g l1 (run 1), 48.9 g l1 (run 3), and 39.8 g l1 (run 4). In all runs some loose biomass could also be observed in the membrane modules. In runs 1 and 2, the amount of loose biomass appeared to be fairly small. The amount in run 1 appeared less than in run 2 and was not quantified; in run 2 it was 3.5 mg. However, in runs 3 and 4 significant amounts of loose biomass were observed. In run 3 it was 28.7 mg and in run 4 it was 47.7 mg. The loose biomass indicates sloughing took place during the runs, particularly in runs 3 and 4. However, we did not observe any dramatic sloughing events leaving parts of the membrane completely bare, which occurred with the methanotrophic biofilms described by Rishell et al. (2004). The DOC concentrations in all runs of Experiment 1 were low, typically below 5 mg/l. The ranges of DOC production rates were 0.15–0.46 gDOC m2 d1 in run 1, 0.05–0.15 in run 2, 0.16–0.33 in run 3, and 0.23–0.30 in run 4. The highest
89
PCH4=37 kPa
PO2=46 kPa
PO2=78 kPa
Run 2 200
0.8
150
0.6
100
0.4
50
0.2 0
10
20
30
40
0
1.2 1.0
PCH4=21 kPa
PCH4=36 kPa
PO2=43 kPa
PO2=77 kPa
150
0.6
100
0.4
50
0.2 0.0
0
10
Time (Day)
200
1.0 0.8
150
0.6
100
0.4 0.2
50
0.0
0
0
10
20
30
40
TIN/CH4 (molar ratio)
Run 4
PCH4=42 kPa
PCH4=26 kPa
20
30
40
0
Time (Day)
NO3-N, NO2-N (mg/l)
TIN/CH4 (molar ratio)
Run 3 1.2
200
0.8
1.2
PCH4=41 kPa
PCH4=26 kPa
1.0
200
0.8
150
0.6
100
0.4
50
0.2 0.0
Time (Day)
0
10
20
30
40
0
NO3-N, NO2-N (mg/l)
0.0
TIN/CH4 (molar ratio)
1.0
PCH4=24 kPa
NO3-N, NO2-N (mg/l)
TIN/CH4 (molar ratio)
Run 1 1.2
NO3-N, NO2-N (mg/l)
water research 44 (2010) 85–96
Time (Day)
Fig. 2 – Average pressures of methane ðPCH4 Þ and oxygen ðPo2 Þ, and concentrations of nitrate (-) and nitrite (,) in Experiment 1. The nitrate removal efficiencies, defined as the molar ratio of the amounts of total inorganic nitrogen (TIN) removed and methane passing through the membrane, for each 7-day batch feed cycle are also shown (line).
production rate was observed in the first feed cycles of all runs except run 4 where it was observed in the third feed cycle.
3.2.
Experiment 2
3.2.1.
Nitrate removal
In Experiment 2, the MBfRs were operated in continuous mode with a stepwise increasing influent nitrate concentration. The influent and effluent nitrate and nitrite concentrations are shown in Fig. 4. The effluent nitrate concentration dropped to zero in both reactors after approximately 7 days of operation with an influent nitrate concentration of around 20 mgN l1. After 11 days, the influent nitrate concentration was increased to about 35 mgN l1. This resulted in the effluent nitrate concentration in reactor A increasing to around 20 mgN l1. In reactor B, however, the nitrate concentration remained at zero. The effluent nitrite concentration increased significantly, particularly in reactor B, reaching 20 mgN l1. From day 33, the influent nitrate concentration was again increased, to nearly 80 mgN l1. This caused a significant increase in the effluent nitrate concentration from reactor A, though the nitrite concentration in this reactor eventually decreased to almost zero. In reactor B, an increase in the effluent nitrate concentration could initially be seen as a result of the high influent concentration, however, the effluent concentration finally stabilized at a fairly low value, less than 5 mgN l1. The nitrite concentration in reactor B, on the other hand, increased significantly, to over 50 mgN l1. The TIN removal rates in each reactor were calculated from the shaded periods indicated in Fig. 4, representing periods in which the effluent concentrations were judged to have reached steady state. The rates were calculated as the difference in influent and effluent TIN concentrations divided by the HRT. For reactor A the
removal rates were 0.74, 0.34, and 0.29 gN m2 d1 for the 1st, 2nd, and 3rd periods, respectively. For reactor B, the rates were 0.78, 0.56, and 0.76 gN m2 d1. With the experimental setup used in Experiment 2, methane consumption rates could not be quantified, so the nitrate removal efficiencies could not be determined. The effluent DOC concentrations were declining from about 5 mg/l in the beginning of the run to about 3 mg/l in the end of the run for reactor A, and from 4 mg/l to 2 mg/l for reactor B. This corresponds to a reduction in DOC production rate from 0.19 to 0.11 gDOC m2 d1 in reactor A, and from 0.16 to 0.08 in reactor B. At the end of the runs the area-based biofilm densities were similar in the two reactors, 24.6 gVS m2 in A and 22.3 gVS m2 in B.
3.2.2.
Microscale properties
The DO concentration in the biofilms was measured by lowering the microsensor through one of the sampling ports located on top of the membrane modules. Measurements were taken in sampling ports located 12, 17, or 22 cm from the inlet to the module, which was 29 cm long. When measurements were taken, the liquid flow was temporarily shut off since the vibrations from the flowing water caused highly oscillating readings. The microsensor was lowered through the sampling port into the liquid until the tip was located a small distance above the biofilm surface. From thereon, the microsensor position was controlled with a manually operated micromanipulator. For reactor A, which had an aerobic bulk liquid, it was lowered into the biofilm until the DO concentration reached zero. For reactor B, which had oxygen supplied from the membrane, the microsensor was lowered until the DO concentration ceased increasing or the slope of profile suddenly decreased significantly, which was assumed to occur when the sensor tip pressed against the membrane
90
water research 44 (2010) 85–96
Run 1
Biofilm thickness (µm)
800
5cm 10cm 15cm
700 600
20cm 25cm
500 400 300 200 100 0
0
10
20
30
40
30
40
Time (Day) Run 3
Biofilm thickness (µm)
800
5cm 10cm 15cm
700 600
20cm 25cm
500 400 300 200 100 0 0
10
20
Time (Day) Run 4
Biofilm thickness (µm)
800
5cm 10cm 15cm
700 600
20cm 25cm
Fig. 4 – Influent and effluent nitrate and nitrite concentrations in reactor A and B. The shaded regions indicate time periods for which the TIN removal rates were calculated.
500 400 300 200 100 0
0
10
20
30
40
Time (Day)
Fig. 3 – Biofilm thickness for runs 1, 3, and 4. Each data point shows the average of three measurement points taken approximately 5, 10, 15, 20, or 25 cm from the bottom of the membrane module. The error bars show the standard deviations.
surface. Several profiles were taken at each sampling ports, but in different randomly selected locations. DO concentration profiles were measured on day 38 and 52 of the runs. The average concentrations for reactor A and B are shown in Fig. 5. For reactor A the average concentrations are based on 6 profiles on day 38 and 7 profiles on day 52. The DO concentration drops from about 6 to 7 mg/l in the bulk liquid to zero inside the biofilm. For reactor B, the average concentrations on day 38 are based on 5 profiles. On day 52, the profiles looked significantly different depending on which sampling port was used. Thus, the average concentrations from the sampling ports located 12 or 17 cm from the reactor inlet are plotted separately and are based on 3 profiles each. The DO concentration is high near the membrane surface, which is located at 0 mm, and lower further out into the biofilm. The elongated profile from one of the sampling ports on
day 52 may have been caused by a sloughing event at that location leading to a significantly lower biofilm density at the time of sampling. It should be noted that the measured DO concentration profiles may differ from the actual concentrations during normal operation, since the liquid flow was shut off during measurements. Shutting off the liquid flow may have two effects on the concentration profiles. First, it may increase the mass transfer resistance of nutrients from the bulk liquid into the biofilm. Casey et al. (2000) compiled data showing that the diffusion boundary layer thickness between the bulk liquid and biofilm, which can be correlated to the mass transfer coefficient, increases significantly from about 100 mm at liquid velocities over 8 cm s1 to between 800 and 1200 mm at liquid velocities of about 0–1 cm s1. Second, during liquid flow, advection of oxygen not consumed by the biofilm in an upstream location may affect the DO concentration profile in a downstream location. Shanahan and Semmens (2006) observed this effect when they studied DO concentration profiles in a nitrifying biofilm attached on a membrane fed with air. The first effect, an increased mass transfer resistance, probably affected reactor A more than reactor B. In reactor B, both major substrates, oxygen and methane, were supplied from the membrane, so an increased mass transfer coefficient from the bulk liquid did not likely affect the oxygen consumption significantly. In reactor A, oxygen was supplied from the bulk liquid, so the measured DO profiles could have
water research 44 (2010) 85–96
7.0
DO (mg/l)
Japan and assigned accession numbers AB489846 – AB489849. A BLAST search showed that all four sequences were most closely related to uncultured bacteria (Table 1).
Reactor A
6.0
Membrane Membrane pos ion positition unk n unknown
5.0 4.0 3.0 2.0 1.0
0.0 -500
0
500
1000
1500
Distance ( m) 16.0
DO (mg/l)
91
Reactor B
14.0 12.0 10.0 8.0 6.0 4.0 2.0 0.0
0
200
400
600
800
1000
1200
Distance ( m)
Fig. 5 – Top: Average DO concentration profiles for reactor A from day 38 (-) and day 52 (B). Bottom: Average DO concentration profiles for reactor B from day 38 (-) and day 52. On day 52, the concentration profiles were plotted separately for sampling ports 12 cm (B) and 17 cm (x) from the inlet of the reactor. The error bars show standard deviations.
underestimated the DO availability in the biofilm during normal operation. The second effect, advection of substrate, may have affected both reactors. Methane (reactor A and B) and oxygen (reactor B) supplied from the membrane but escaping consumption by the biofilm near the inlet of the reactors may have been consumed further downstream, creating a layer of methanotrophs living near the biofilm– liquid interface. The structure of the biofilms in reactors A and B can be seen in Figs. 6 and 7. Fig. 6 shows horizontal slices taken from various depths of the biofilms; Fig. 7 shows cross-sections. The slices and cross-section were taken from biofilm samples cut from near the middle of the reactors. None of the biofilm samples hybridized with the FISH probes for Type II methanotrophs. However, Type I methanotrophs appeared to dominate the bottom part of the biofilms in both reactor. In reactor A, with only methane supplied from the membrane a dense methanotrophic layer formed on the membrane surface whereas in reactor B, with methane and oxygen mixed in the membrane the methanotrophs formed a more porous structure with clusters separated by voids. In the biofilm from both reactors, some methanotrophs can also be seen growing a considerable distance from the membrane surface (this can be seen primarily in reactor A, Fig. 7). This may be caused by methane escaping the biofilm and becoming available in the bulk liquid, as described in the paragraph above. The nitrite reductase gene nirS was successfully amplified from both reactors whereas nirK was only amplified from reactor B. For nirS, one sequence was obtained from each reactor. For nirK, two sequences were obtained from reactor B. The sequences have been submitted to the DNA Database of
4.
Discussion
4.1.
Nitrate removal efficiency
The goal of Experiment 1 was to determine whether MBfRs with various gas supply regimes could achieve high nitrate removal efficiencies compared to suspended growth reactors. The nitrate removal efficiencies (i.e the TIN/CH4 ratios) based on the total removal of TIN and CH4 for each of the four runs were 0.34, 0.25, 0.36, and 0.27 mol TIN mol1 CH4. Suspended growth experiments with the same enrichment culture have achieved efficiencies ranging from 0.067 to 0.11 mol TIN mol1 CH4 under various headspace conditions. A description of the suspended culture experiments is provided in Modin et al. (2008). In brief, 1.1-litre bottles containing 250-ml NMS medium inoculated with the enrichment culture were operated with mixtures of methane, oxygen, and helium in the headspace. The consumption of nitrate and methane were monitored. The best nitrate removal efficiency in terms of methane utilization observed under these conditions was 0.11 mol TIN mol1 CH4 (Modin et al., 2008). Werner and Kayser (1991) also studied aerobic methane oxidation coupled to denitrification in suspended batch cultures and measured 1 CH4. a ratio of 0.083 mol NO 3 mol The results obtained in Experiment 1 show that MBfRs can achieve nitrate removal coupled to aerobic methane oxidation with a higher efficiency than suspended cultures. The results also show that methane and oxygen do not have to be mixed within the membrane lumen (Fig. 1A) to achieve high nitrate removal efficiency, since MBfRs with methane and oxygen supplied through separate intertwined membrane tubes (Fig. 1B) or only methane supplied through the membrane and an aerobic bulk liquid (Fig. 1C) achieved similarly high efficiencies. In this study, nitrate and nitrite concentrations were measured, but the amount of nitrogen leaving the reactor in gaseous form or being assimilated into biomass was not quantified. However, some conclusions can be drawn based on the observed TIN/CH4 ratios. Leak and Dalton (1986a) studied the growth yields of the aerobic methanotroph Methylococcus capsulatus and found carbon conversion efficiencies (i.e. the fraction of converted methane-carbon assimilated into biomass) with nitrate as N source ranging from 0.312 to 0.433. The elemental content of microbial biomass can be represented by empirical formulas such as C4H8O2N (Leak and Dalton, 1986b) or C5H7O2N, thus it is reasonable to expect the molar N/C ratio in the biomass to be around 0.20–0.25 mol N mol1 C. Multiplying the carbon conversion efficiency with the N/C ratio results in expected nitrate removal efficiencies based on assimilation to be between 0.062 1 CH4. These values are close to those and 0.11 mol NO 3 mol observed with suspended cultures suggesting assimilation was responsible for most of the nitrate removal in such reactor systems. However, the efficiencies observed in the
92
water research 44 (2010) 85–96
Fig. 6 – Images showing biofilm slices cut parallel to the membrane surface from reactor A (left) and reactor B (right), stained with DAPI (blue), EUBmix (green), and probes for Type I methanotrophs (red). Scale bars indicate 200 mm. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
MBfRs were significantly higher, indicating denitrification played a role in the biofilms. It was experimentally confirmed that the enrichment culture used in this study was unable to utilize methane under anoxic conditions, so anaerobic coupling between methane oxidation and denitrification in the MBfR biofilms was not responsible for the high efficiencies observed. Aerobic methanotrophs, which according to the FISH images appeared to dominate the biofilms, are incapable of denitrification (Knowles, 2005). Thus, the most likely explanation for the high nitrate removal efficiencies is coexistence of aerobic methanotrophs with denitrifiers utilizing various organic compounds released by the methanotrophs, a phenomenon that has been observed in previous studies (e.g. Amaral et al., 1995; Eisentraeger et al., 2001; Waki et al., 2002). Variations in nitrate removal efficiency could be observed between different feed cycles. With the exception of the first feed cycle, the nitrate removal efficiencies in runs 3 and 4 (Experiment 1) remained fairly stable between 0.19 and
0.31 mol TIN mol1 CH4 despite a change in the methane pressure in the third feed cycle. Also in runs 1 and 2, the efficiency appeared stable between 0.32 and 0.48 mol TIN mol1 CH4 in feed cycles 2 and 3. However, when the gas pressures were increased before the last two feed cycles the efficiencies were significantly affected. In runs 1 and 2, in which oxygen was also fed from the interior of the membranes, the nitrate removal efficiency dropped significantly as a result of the increased gas pressures (Fig. 2). In runs 3 and 4, in which oxygen was supplied with the bulk liquid and thus was kept relatively constant, the increased methane pressure did not have a significant effect on the nitrate removal efficiency. This indicates that increased flux of oxygen in runs 1 and 2 reduced the anoxic biofilms spaces and temporarily inhibited nitrate removal, thus further pointing to denitrification as a sink of nitrate. In the final feed cycle the nitrate removal efficiency recovered in runs 1 and 2 possibly as a result of an increased biofilm thickness (Fig. 3) allowing for the development of new anoxic zones.
water research 44 (2010) 85–96
93
Fig. 7 – Images of biofilm cross-sections from reactors A and B, stained with DAPI (blue), EUBmix (green), and probes for Type I methanotrophs (red). The white line shows the approximate position of the membrane. Scale bars indicate 200 mm. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
The nitrate removal efficiency was the highest in the beginning of all four runs (Experiment 1). The reasons for this behavior are not entirely clear. The nitrate concentrations in the reactors were high throughout the experiment (150–200 mgN l1), so mass transfer limitation caused by increased biofilm thickness in the later phase of the runs was not likely an issue limiting nitrate removal. A possible explanation is an increased release of organics by the methanotrophs in the early phases of biofilm establishment. Although the bulk liquid DOC concentrations measured in all the experimental runs were low (<5 mg/l), the DOC production rate was high in the first feed cycle for all 4 runs in Experiment 1 (in run 4 it was also high in period 3) correlating with the high nitrate removal efficiency. Moreover, the DOC concentrations were the highest in the beginning of Experiment 2 in both reactor A (w5 mg/l, which later decreased to w3 mg/l) and reactor B (w4 mg/l, which later decreased to w2 mg/l). This suggests that the methanotrophic biofilms in both experiments released more organic carbon in the initial phase of biofilm establishment than in the later mature phase, and correlates with a higher nitrate removal efficiency observed in the early phase of Experiment 1.
4.2.
Nitrate removal rate
In Experiment 2, both reactors could at the start of the experiment remove the entire incoming nitrate load. When the influent concentration was subsequently increased, the two reactors showed different behavior in terms of nitrate removal. Reactor A, with aerobic bulk liquid, removed nitrate with an over time declining removal rate. This could have been caused by an over time increasing biofilm thickness, leading to higher diffusion resistance for methane and oxygen in the biofilm, resulting in less methanotrophic activity. Reactor B, with mixed gas supplied from the membrane, removed almost all
the nitrate at continuously increasing influent loads. However, high concentrations of nitrite were produced, so the TIN removal rate remained fairly constant. Since aerobic methanotrophs are not known for dissimilatory nitrate reduction (Ren et al., 2000), the high nitrite production observed in Experiment 2 could either be explained by assimilatory nitrate reduction by methanotrophs or the development of other microbes able to reduce nitrate to nitrite. Since aerobic methanotrophs were active in all experimental runs, but high nitrite concentrations were only observed in run B of Experiment 2, the second explanation appears more likely. The DO concentration profiles showed that in reactor B an anoxic zone existed, starting 100–200 mm from the membrane surface. For the reactor A profiles, the exact position of the membrane is not known. However, the cross-sectional images suggest oxygen penetrated almost the entire biofilm since a dense methanotrophic (aerobic) layer is located near the membrane (Figs. 6 and 7). Thus, the DO concentration profiles suggest that larger anoxic zones existed in the reactor B biofilm than in reactor A. This supports the theory that non-methanotrophs using nitrate as electron acceptor were responsible for the high nitrite production seen in reactor B and also correlates with the higher TIN removal rates in reactor B. The reactors in Experiment 1 were operated under a much higher nitrate concentration (>150 mgN l1) than would be expected in a real application. It is possible that mass transfer limitations could affect the nitrate removal rate at lower concentrations. Assuming zero-order kinetics, the penetration depth of nitrate into a flat biofilm can be estimated using equation (1) (Stewart, 2003).
a¼
0:5 2$D$S0 q$X
(1)
94
water research 44 (2010) 85–96
Table 1 – Accession numbers to obtained nirK and nirS sequences and nearest matches of a BLAST search. Reactor, target
Acc. No.
Nearest match, (Acc. No.), Nucleotide identities
B, nirK
AB489846
B, nirK
AB489847
A, nirS
AB489848
B, nirS
AB489849
Uncultured bacterium from acetate-degrading sludge, (AB162313), 95% Uncultured bacterium from activated sludge, (DQ182171), 92% Uncultured bacterium from arable soil, (AY583407), 87% Uncultured bacterium from activated sludge, (AY583443), 89%
where a is the penetration depth (m), D is the diffusion coefficient of nitrate in the biofilm (m2 d1), S0 is the nitrate concentration at the biofilm–liquid interface (mgN l1), q is the specific nitrate consumption rate (mgN g1VS d1), and X is the biomass density (gVS l1). The diffusion coefficient of nitrate in the biofilm was assumed to be 0.8 times that in water (Rittmann and McCarty, 2001), which equaled 0.8 1.64 104 ¼ 1.31 104 m2 d1 (Lide, 2008). Assuming a biofilm density of 78.6 g l1, which was the highest density measured in Experiment 1, we can calculate the penetration depth for various S0 and q. The specific nitrate removal rates could be calculated using the mass of VS and the TIN removal rates measured at the end of the experiments. In Experiment 1, the rates were 37.5, 54.6, 59.5, and 68.2 mgN g1VS d1 for runs 1, 2, 3, and 4, respectively. In Experiment 2, the rates were 11.7 and 33.9 mgN g1VS d1 for reactor A and B, respectively. For the highest specific consumption rate of 68.2 mgN g1VS d1 the penetration depth at an S0 of 150 mg l1 is 2712 mm, whereas at an S0 of 10 mg l1, it is 700 mm. This rough estimation suggests that only in thick biofilms (>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
<0.5 K
Mn
Mw
r
Normal times
SF L90-4 L90-15 M1300 L85 M350
4.3 5.2 4.9 4.5 5.1 4.7
17.0 17.0 17.1 16.6 17.3 16.0
32.1 31.3 31.6 31.7 31.6 31.4
46.6 46.5 46.4 47.2 46.0 47.9
750 766 771 740 773 739
1574 1558 1581 1427 1547 1446
2.10 2.04 2.05 1.93 2.00 1.96
Dry season
SF L90-4 L90-15 M1300 L85 M350
5.4 3.5 5.0 4.1 5.5 4.4
15.1 14.2 14.8 14.6 15.1 14.7
29.0 29.8 29.0 29.7 28.6 30.1
50.5 52.5 51.2 51.6 50.8 50.8
744 736 671 696 756 707
1627 1640 1246 1392 1709 1454
2.19 2.23 1.86 2.00 2.26 2.06
Rainy season
SF L90-4 L90-15 M1300 L85 M350
1.7 1.8 1.5 1.9 2.1 1.9
13.2 13.3 13.0 13.5 13.5 13.3
31.6 31.4 31.0 31.5 31.1 31.2
53.5 53.5 54.5 53.1 53.3 53.6
613 602 616 622 628 619
979 984 1002 1046 1078 1033
1.60 1.63 1.63 1.68 1.72 1.67
Mn: Number-average molecular weight Mw: Weight-average molecular weight
factor. P17 AOC and NOX AOC were calculated by the following equations, and AOC was calculated by the sum of P17 AOC and NOX AOC. P17ðCFU=mLÞ 1000 mL=L 4:1 106 ðCFU=mgCÞ
NOX AOCðmg=LÞ ¼
NOXðCFU=mLÞ 1000 mL=L 1:2 107 ðCFU=mgCÞ
AOCðmg=LÞ ¼ P17 AOC þ NOX AOC
2.6.
Chlorine demand and decay rate
Chlorine demand and decay rate were estimated for the sand filtered water and the UV treated water taken from each UV
Turbidity (NTU), Rain fall (mm), Chl.a (µg/L)
250
5 Rainy season
200 Dry seaon
Turbidity Rain fall Chl.a BOD TOC Sampling
4
Noraml times
150
3
100
2
50
1
0 '05
BOD, TOC (mg/L)
P17 AOCðmg=LÞ ¼
reactor. Chlorine decay rate was measured with the procedure proposed by Powell et al. (2000). The freshly cleaned glassware was filled with distilled water and sodium hypochlorite solution was added to make 10 mg/L of free chlorine solution and left for 24 h. It was then emptied, rinsed thoroughly with ultrapure water and left to dry. 2 L volumetric flask was filled with ultrapure water and the sample water. Chlorine was added to 1w 2 mg/L and left for 15 min to ensure homogeneity. The sample water was decanted into eleven 125 mL brown glass bottles without headspace and sealed with teflon lined caps. All the bottles were stored in the incubator, at 4 and 15 C. The chlorine concentration was measured with time. Initial chlorine concentration was defined as the chlorine concentration when the same amount of chlorine was added to 2 L of ultrapure water. Chlorine concentration was measured by the DPD colorimetric method using Hach pocket
0
Oct. Nov. Dec.'06 Jan. Feb. Mar. Apr. May Jun Jul. Aug. Sep. Oct. Nov. Dec.
Month Fig. 2 – Hydraulic characteristics and the raw water quality change by season.
water research 44 (2010) 115–122
450
Normal times
Dry season
Rainy season
400 350
AOC (µg/L)
300 250 200 150 100 50 0
SF
L90-4
L90-15
M1300
L85
M350
Fig. 3 – Seasonal AOC concentration before and after UV treatment in each system.
chlorine colorimeters (pocket Hachs). All samples were taken and analyzed in triplicate. The decay rate constants were estimated by the first-order chlorine decay model (Jadas-He´cart et al., 1992). In this study, two chlorine decay rates were used, i.e. rapid chlorine decay rate (K1) for the first 4 h and slow chlorine decay rate (K2) after 4 h considering the retention time in the clearwell of the WTP.
3.
Results and discussions
3.1.
Effects of UV treatment on DOM properties
DOC, UV254, SUVA254, hydrophilic/hydrophobic ratio and apparent molecular weight before and after UV treatment were investigated. The DOC concentration of the sand filtered water was less than 1 mg/L in normal times but increased to 1.5 mg/L in dry and rainy seasons.
0
20
Pre-UV NOX AOC (µg/L) 40 60 80
100
120
400
120
300
80 L90-4 (P17) L90-15 (P17) M1300 (P17) L85 (P17) M350 (P17)
200
60
equal value line L90-4 (NOX) L90-15 (NOX) M1300 (NOX) L85 (NOX) M350 (NOX)
100
0
0
100 200 300 Pre-UV P17 AOC (µg/L)
40
20
400
Fig. 4 – P17 AOC and NOX AOC before and after UV treatment.
0
Post-UV NOX AOC (µg/L)
Post-UV P17 AOC (µg/L)
100
119
There were little change in DOC, UV254, and SUVA254 after UV treatment throughout all seasons (Table 2). Only in the L90 system, with low pressure lamp for TOC reduction, a little decrease of UV254 and SUVA254 were observed with the UV dosage of 150 mJ/cm2. The reductions might be caused by high energy of short wavelength at 185 nm and high dosage of 150 mJ/cm2. With dosage of 40 mJ/cm2, which is usually applied for drinking water disinfection process in WTP, all UV systems had no effect on DOC, UV254, and SUVA254. The ratio of hydrophilic and hydrophobic fractions was calculated from DOC concentration of each fraction. The fraction of hydrophilic DOC was relatively high throughout all seasons with the range of 62w81 %, but hydrophobic fraction was increased in rainy season (Fig. 1). The source water from the Han river, has been known to have relatively higher concentration of hydrophilic organic fraction (Oh et al., 2003; Kim et al., 2007; Jeong et al., 2007). The ratio of hydrophilic and hydrophobic fractions can be changed in water treatment process. The hydrophilic fraction tends to be increased in treated water as humic material with high SUVA value and high hydrophobic organic carbon is removed easily by coagulation process (White et al., 1997). It was also reported, from the previous studies with the Han river as the source water, that hydrophilic fraction in the settled water was increased (Oh et al., 2003; Kim et al., 2007). However, there was not significant difference in the ratio of hydrophilic and hydrophobic fractions before and after UV treatment throughout all seasons. Shaw et al. (2000) also reported that there was little or no statistical evidence that hydrophilic and hydrophobic ratios were altered by UV treatment. Distribution of apparent molecular weight was measured by HPLC-SEC system with UV and TOC detectors. The molecular weight of most DOM (over 95 %) was less than 2 kDa (Table 3), and especially DOM fraction between 0.3 and 0.4 kDa was dominant throughout all seasons. While the distribution of apparent molecular weight was not changed before and after UV treatment in normal times and rainy season, there was increase in low molecular weight fraction around 0.3 kDa after UV treatment in dry season (data was not shown). Number-averaged molecular weights (Mn) and weightaveraged molecular weight (Mw) have also shown that average molecular weight in the post-UV treated water was decreased in dry season (Table 3). This suggested that DOM structure in dry season is broken down more easily by UV radiation than those in other seasons. DOM structure might be related with the origin of DOM of each season. In aquatic system, the origin of DOM can be categorized as allochthonous DOM entering from the terrestrial watershed, and autochthonous DOM derived from biota (e.g., algae, bacteria) growing in the water body (Aiken and Cotsaris, 1995). In Korea, during the rainy season in late summer with lots of heavy rain, DOC increases because heavy rain washes large amount of organic carbon from the watershed into river while in dry season, algal biomass and BOD increases (Fig. 2). This allochthonous DOM in rainy season is known to be relatively refractory DOM with high SUVA, high molecule weight, and hydrophobic properties. In contrast, the autochthonous DOM is relatively labile, and consists of low SUVA, low molecular weight, and hydrophilic DOM (Wetzel, 1983; Kitis et al., 2002). Ma et al. (2001) reported that hydrophilic fraction was
120
water research 44 (2010) 115–122
Table 4 – Chlorine demands and chlorine decay rate before and after UV treatment, K1 : rapid decay rate(< 4 h), K2 : slow decay rate(> 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 (<5 dry tonnes solids produced per day), where high-capital options to enhance degradability, such as sonication or thermal hydrolysis are not available (Barr et al., 2008). To address these limitations in smaller plants, an anaerobic option should (Batstone et al., 2008a):
(a) Improve biogas production to offset energy demand (b) Increase solids destruction to reduce the volume of sludge requiring ultimate disposal (c) Increase hydrolysis rates to allow reduced digester size and capital cost and (d) Achieve pathogen free stabilised solids to expand reuse options. Temperature phased anaerobic digestion (TPAD) may allow enhanced degradability and biogas production, as well as pathogen destruction, at a relatively low capital cost. TPAD consists of a pre-treatment stage operated under thermophilic temperature (50–70 C) and short hydraulic retention times (HRT), followed by a main stage operated at lower mesophilic temperature with a longer retention time. Pathogen destruction and hydrolytic and acidogenic conditions can be further optimised in the pre-treatment process. In the following main
* Corresponding author: Tel.: þ61 7 3346 9051; fax: þ61 7 3365 4726. E-mail address:
[email protected] (D.J. Batstone). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.005
124
water research 44 (2010) 123–130
stage, a longer retention time and a neutral pH favour methanogenesis for maximum conversion of organic components to methane. There have been a number of studies evaluating TPAD systems. Han et al. (1997) tested the effect of different solids retention times for TPAD system (55 C and 35 C) compared with conventional single-stage mesophilic (35 C) digestion on primary sludge and waste activated sludge. They showed that the optimal solids retention time of across both stages of a TPAD system ranged from 11 to 17 days, with volatile solids (VS) destruction up to double in TPAD system compared to single-stage anaerobic digestion. Skiadas et al. (2005) found a VS destruction with TPAD system (70 C, 2 day HRT and 55 C) of 55% and 43% for primary and secondary sludge respectively, higher than 43% and 6% achieved in the singlestage thermophilic (55 C) anaerobic digestion. Watts et al. (2006) reported that lower thermophilic temperatures (47 C and 54 C, 2 day HRT) treating waste activated sludge did not offer higher VS destruction over single-stage mesophilic (37 C) anaerobic digestion. When the thermophilic temperature was increased to 60 C, VS destruction was improved to 35%, compared with 24% in single-stage mesophilic anaerobic digestion. They also observed increased gas production consistent with the increased VS destruction. These studies indicate enhanced treatment performance for TPAD systems as compared to single-stage thermophilic or mesophilic systems. However, rigorous analysis is missing, as there is no direct parallel comparison of mesophilic–mesophilic and thermophilic–mesophilic TPAD. There is also little analysis of which conditions (temperature and pH) can optimise eventual hydrolytic conversion. Finally, it has not been established whether enhanced performance is due to increased hydrolysis in the pre-treatment stage, increased overall degradability, or a conditioning process (such as a physical breakdown of sludge similar to that achieved during thermal hydrolysis and sonication), that allows better performance in the main stage. This paper addresses these limitations on a particular feed (primary sludge) by operating parallel thermophilic–mesophilic and mesophilic–mesophilic TPAD systems, and detailed analysis of the pre-treatment process.
2.
Materials and methods
2.1.
Substrate
The substrate used in this study was primary sludge collected from a large wastewater treatment plant in Brisbane, Australia. The feed was screened with a 3 mm sieve and diluted with tap water to a total solids (TS) concentration of 2– 3%. Feed batches were prepared at intervals of 1–2 months and stored at below 4 C. Regular analysis was performed to determine the characteristics and consistency of the feed material. The average characteristics of the primary sludge feed are shown in Table 1.
Table 1 – Characteristics of the primary sludge used in this study. Measure 1
TS (g L ) VS (g L1) pH COD (g L1) VFA (g COD L1) TKN (g N L1) 1 NHþ 4 -N (g L )
26.9 2.9a 20.7 2.0 5–6.5 30.2 3.2 0.6 0.2 1.3 0.6 0.09 0.02
a Indicates standard deviation across 5 different feed materials used in the study over 6 months.
(HRT 2 days) for pre-treatment and a 4.0L reactor (HRT 13–14 days approx) as main methanogenic stage. The thermophilic pre-treatment (TP) system and mesophilic pre-treatment (MP) system were operated identically, except for the pre-treatment stage, which was either 50–65 C (TP1), or 35 C (MP1). The temperature in the pre-treatment stages was maintained with temperature controlled water jackets, while temperature in the main methanogenic stages was maintained using submersed electrical heating elements. All reactors were continually mixed using magnetic stirrer bars. Gas production volumes and pH were recorded from each reactor and recorded online by a process logic control system.
2.3.
Start-up and operation
Each reactor was inoculated from a full-scale anaerobic digester (35 1 C) in Brisbane, Australia. Reactors were fed at intervals of 4 hours (6 times daily). During feed events, approximately 50 mL of feed was pumped through the system simultaneously using multi-head peristaltic pumps located between the feed reservoirs and pre-treatment stages; pretreatment stages and methanogenic stages; and methanogenic stages and the waste effluent drums. The systems were operated for over 6 months. During this time the temperature of TP1 was altered to create different operating periods:
Period 1: 50 C (117 days) Period 2: 60 C (20 days) Period 3: 65 C (32 days) Period 4: 65 C, pH 4.5 by dosing of 1 M HCl (14 days).
The TP system had been operated for 64 days before the MP system commenced operation. The temperature of MP1, TP2 and MP2 were held constant at 35 C during all periods. After Period 4 the acid dosing was stopped and the pH in TP1 returned to its natural level of 6.8. Only data from Day 75 was used in comparative analysis (i.e., after stabilisation of both digesters).
2.4. 2.2.
Primary sludge
Analysis
Laboratory scale reactor systems
Two identical TPAD systems, as shown in Fig. 1 were used throughout the study. Each system contained a 0.6L reactor
Gas production was measured using tipping bucket gas meters and continuously logged. Gas meters were regularly recalibrated and switched between reactors to prevent
125
water research 44 (2010) 123–130
Gas to exhaust
Gas meter F
Gas meter PLC
F pH
pH
Water jacket temperature control
Heating coil
Feed reservior
Effluent drum Pretreatment 0.6L TP1 = 50-65°C MP1 = 35°C
Feed pump
Main Digester 4L TP2=35°C MP2=35°C Digester pump
Effluent pump
Fig. 1 – Schematic diagram of TPAD systems.
systematic errors. Biogas composition (H2, CH4, CO2) was analysed by a Perkin Elmer loop injection gas chromatography (GC), as described by Tait et al. (2009). The pH in each reactor was measured daily with a calibrated glass body probe (TPS, Brisbane, Australia). Liquid samples were collected from each reactor three times per week. Analyses were performed for TS, VS, volatile fatty acid (VFA), chemical oxygen demand (COD), total Kjeldahl nitrogen (TKN) and ammonium–nitrogen (NHþ 4 -N). Analytical methods were as for Standard Methods (APHA, 1998). COD was measured on Merck Method for total (TCOD) and soluble fractions (SCOD), using an SQ 118 Photometer (Merck, Germany). For measurement of SCOD, VFA and NHþ 4N, the liquid samples were centrifuged at 2500g for 20 min and filtered through a syringe filter (0.22 mm PES membrane) prior to analysis. VFA concentrations were measured by GC (Agilent, FID with polar capillary column). NHþ 4 -N and TKN were measured using a Lachat Quik-Chem 8000 Flow Injection Analyser (Lachat Instrument, Milwaukee).
2.5.
Calculation of VS destruction
VS destruction was calculated using the Van Kleeck equation, which assumes that the amount of fixed solids is conserved during digestion (Switzenbaum et al., 2003). It can be expressed as VS destruction% ¼
VSfraci VSfrac0 VSfraci VSfraci VSfrac0
(1)
Where VSfraci and VSfrac0 are volatile fractions (VS/TS) in the influent and effluent solids. VS destruction was also calculated based on the gas flow, expressed as VS destructiongas % ¼
CODCH4 =fraci VSi
Where CODCH4 is daily CH4 production as g COD d1
(2)
fraci is COD/VS ratio of influent, measured as 1.47 0.02 (95% confidence in mean over 197 measurements) VSi is volatile solids loading rate as g VS d1.
2.6.
Mathematical analysis
2.6.1.
Model implementation
The IWA Anaerobic Digestion Model No. 1 (ADM1) (Batstone et al., 2002a) was used. The reference Aquasim 2.1d version was used (Reichert, 1994) with inputs as described below. Initial conditions were based on a steady state, adjusted for measured initial conditions (organic solids, organic acids, ammonia, TKN, etc).
2.6.2.
Model inputs
Defining inputs well is important to achieve reliable model predictions. In this case, inputs were divided into particulate inerts, carbohydrates, proteins, lipids, organic acids and ammonia, based on a modified form of the COST ASM1-ADM1 interface (Nopens et al., 2009). The main difference is that the inert fraction was mapped in terms of an overall degradability parameter. Other fractions were based on VS, COD, TKN, organic acids, and NHþ 4 -N measurements as in the standard interface. There were 170 input changes over 180 days used in the model.
2.6.3.
Parameter estimation and analysis
Estimation of parameter value and confidence in value are critical to assess difference between two systems. The main parameters compared were degradability extent ( fd) and apparent first order hydrolysis rate coefficient (khyd) (Pavlostathis and Giraldo-Gomez, 1991), based on the method of Batstone et al., (2003, 2008b) used to estimate parameter confidence regions for a two-parameter system. A 95% confidence limit was used, with appropriate F-values for 2 parameters and the number of degrees of freedom (approx. 158, F ¼ 2.996). A modified version of Aquasim 2.1d was used to determine the parameter surfaces. Gas flow was used as
126
water research 44 (2010) 123–130
a measured variable, with sum of squared errors (c2) as an objective function.
3.
Results
3.1.
Overall performance of TPAD systems
There are three methods of calculating VS destruction; Van Kleeck VS destruction, which uses inlet mineral solids as a reference, mass balance VS destruction, which uses inlet organic solids as a reference, and gas flow VS destruction, which uses inlet flow and COD as references. In this study, Van Kleeck VS destruction was consistent with gas flow VS destruction, but was higher than mass balance VS destruction. This may indicate incomplete mixing in the methanogenic stage. Therefore, Van Kleeck and gas flow VS destruction were used as main indicators of performance. Mineral solids, NHþ 4 -N, and other analyses supported the use of Van Kleeck and gas flow VS destruction measures. The TP and MP systems both achieved VS destruction greater than 38% (Fig. 2, top), the required value specified in the 40 Code of Federal Regulations (CFR) Part 503 regulations to minimize vector attraction (US EPA, 1994). Statistical analysis (student t-test, a ¼ 0.05) confirmed that VS destruction in 100
the TP system was significantly greater than that in the MP system from Day 75 to Day 183. However, varying the thermophilic pre-treatment temperature from 50 C to 65 C did not have a significant impact on VS destruction. Additionally, VS destruction was not improved under acidic pre-treatment conditions. A summary of the average VS destruction during each period is shown in Fig. 3. Thermophilic pre-treatment enhanced VS destruction, resulting in higher conversion of organic components to methane. This was reflected in the approximately 25% higher methane production from the TP system compared to the MP system, as shown in Fig. 2 (bottom). The increase was confirmed as a statistically significant improvement by the student t-test analysis (a ¼ 0.05). In both systems, the methane production from pre-treatment stage was negligible compared to that in the methanogenic stage. Methane accounted for 72% and 69% of biogas composition in TP2 and MP2 respectively with carbon dioxide being the other major component during all operating periods. The methane production increase was not observed when the thermophilic pre-treatment temperature was increased to 60 C and 65 C. Methane production results were reflected in apparent VS destruction. Methane production in both systems was lowest during Period 4, it is not clear if this was due to variations in the feed or operational conditions.
% VS destruction in TP system % VS destruction in MP system
VS destruction (%)
80
60
40
20
0
Methane production in TP system Methane production in MP system
-1
Methane production (L day )
3
2
1
0
Period 1 0
20
40
60
80
100
Period 2 120
140
Period 3 160
Period 4 180
200
Time in operation (days)
Fig. 2 – VS destruction (top) and daily methane production (bottom) during each period in the TP system and MP system (% VS destruction is based on the primary sludge feed characteristics and Van Kleeck equation).
127
100
VS destruction in TP system VS destruction in MP system Apparent VS destruction on methane flow in TP system Apparent VS destruction on methane flow in MP system
Hydrolysis rate
80
VS destruction (%)
Degrades faster
water research 44 (2010) 123–130
60
40
0.5
TP system
0.4 0.3 0.2 0.1
20
0
Period 1
Period 2
Period 3
Period 4
Fig. 3 – Average VS destruction and apparent VS destruction on methane flow during each period in the TP system and MP system (Error bars are 95% confidence in mean VS destruction and methane production). Fig. 4 shows the biogas production for a 24 h period from TP2 and MP2, indicating the increase in performance was across the feed cycle. During each feed event, the methane production declined with the substrate consumed in both methanogenic stages, and TP2 demonstrated a faster response to feed than MP2.
3.2.
Model based analysis
Fig. 5 shows the 95% confidence regions for degradability ( fd) (x-axis) and apparent hydrolysis rate (khyd) ( y-axis) in both systems using complete gas flow over 180 days as objective functions. In the TP system, khyd values were between 0.20– 0.51 d1, with fd of 0.56–0.64. In the MP system, the confidence region was right-unbounded in fd, indicating that a degradability upper limit could not be determined. Therefore, there was statistical overlap between the two fd values, but hydrolysis was significantly faster in the TP system.
3.3.
Pre-treatment mechanism
SCOD in TP1 was higher than that formed in MP1 for all periods, increased with temperature increase, and dropped 4.5
Biogas production in TP2 Biogas production in MP2
4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0
MP system
0.5
0.55
0.6
0.65
0.7
0.75
0.8
Degradability
0
Biogas production rate (L day-1)
0.6
0
4
8
12
16
20
24
Time in operation (h) Fig. 4 – Biogas production for a 24 h period from TP2 and MP2.
0.85
0.9
0.95
1
Degrades more
Fig. 5 – Confidence regions for khyd (dL1) and fd (degradable fraction) using gas flow as an objective function in the TP system (172 measurements) and MP system (108 measurements).
under acidic conditions (pH 4.5). It is also important to note that only 5.5% of the organic material was solubilised during thermophilic pre-treatment, while the final release was considerably more. Organic acids did not follow this trend, as organic acid concentrations were lower in TP1 as compared to MP1. This indicates that the material is being solubilised to a greater extent at thermophilic conditions, but not subsequently converted to organic acids. The main organic acid produced in TP1 was acetate, while propionate was the main VFA produced in MP1, as shown in Table 2. Other VFAs were also measured (iso-butyrate, butyrate, iso-valerate, valerate and hexanoate), at significantly lower levels than acetate and propionate. Increasing the thermophilic pre-treatment temperature from 50 C to 60 C resulted in an increase in acetate concentration. However, acetate did not increase further with temperature increase to 65 C and dropped under acidic conditions. Propionate concentration was lower in TP1 than MP1. Propionate did not appear influenced by temperature, and dropped significantly under acidic conditions. The total VFA concentration dropped by approximately half with pH decrease, indicating that low pH may be responsible for inhibition of fermentation or hydrolysis. Although VFA concentrations were high in both TP1 and MP1, in the methanogenic stages (TP2 and MP2) the concentrations were very low (<100 mg COD L1). is another key intermediate released from NHþ 4 -N fermentation of protein or other nitrogenous organic compounds. Generally, but especially during periods 1–3, NHþ 4 -N concentration was higher in TP1 than in MP1 (Fig. 6), indicating enhanced protein fermentation under thermophilic conditions. Again, this was not influenced by temperature. NHþ 4 -N release decreased significantly under acidic conditions, which was consistent with SCOD and VFA concentration, indicating the low pH has a negative impact on fermentative activity. The final concentration of NHþ 4 -N in TP2 and MP2 was similar at each period, indicating the thermophilic pre-treatment does not substantially influence protein degradation extent.
128
water research 44 (2010) 123–130
Table 2 – Summary of solubilisation performance in TP1 and MP1 according to VFAs. Acetate (mg COD L1) Period 1 Period 2 Period 3 Period 4
TP1 MP1 TP1 MP1 TP1 MP1 TP1 MP1
990 730 1250 900 1240 540 670 400
Propionate (mg COD L1)
(230)a (210) (90) (80) (70) (150) (100) (70)
560 1130 700 1680 650 1630 200 1380
Total VFA (mg COD L1)
(90) (380) (70) (220) (50) (140) (70) (50)
2470 2500 2870 3300 2800 2970 1300 2600
SCOD (mg COD L1)
(400) (540) (170) (360) (140) (220) (230) (50)
3030 2630 3520 3120 4220 3270 3510 3380
(500) (370) (170) (210) (560) (360) (230) (140)
a Indicates standard deviation across different measurements over each period.
4.
Discussion
temperature, suggesting that the temperature may be selected to optimise pathogen destruction rather than VS destruction.
4.1.
Overall performance of TPAD systems
4.2.
Primary sludge is a large and unstabilised stream common in wastewater treatment plants. An increase in VS destruction in real plants translates to better sludge dewaterability (Novak, 2006) and lower overall costs of disposal, as this is generally weight based. A corresponding higher methane production in the TP system can be used to produce heat and power for the whole treatment process and offset the higher energy demand required by thermophilic temperature. Since most of the heat used is excess heat from cogeneration, the heating demand required at 55 C can be provided from methane production. It is important that VS destruction and methane production were not improved with increased thermophilic pre-treatment
2000
Model analysis of TPAD systems
The estimates of apparent khyd based on the gas flow showed a greater hydrolysis rate during thermophilic pre-treatment, which was 67% higher than during mesophilic pre-treatment. However, the degradability extent in the TP system was not increased compared to the MP system, indicating that thermophilic pre-treatment influences speed of degradation rather than extent of degradation. This is similar to other lower impact pre-treatment methods such as sonication, which alter physical properties of the substrate to enhance hydrolysis rates (Tiehm et al., 2001). In contrast, high-intensity pre-treatment methods such as thermal hydrolysis increase both rate and extent (Neyens and Baeyens, 2003).
TKN in TP1
TKN in TP2
NH4+-N in TP1
NH4+-N in TP2
1500
N concentration (mg L-1)
1000
500
0 2000
TKN in MP1
TKN in MP2
NH4+-N in MP1
NH4 -N in MP2
+
1500 1000 500 0
Period 1
Period 2
Period 3
Period 4
Fig. 6 – Concentrations of TKN and NHD 4 -N during each period in the TP system (top) and MP system (Error bars are 95% confidence in mean TKN and NHD 4 -N).
water research 44 (2010) 123–130
There are a wide range of hydrolysis constants reported in the literature for primary sludges (e.g., 0.2–0.6 d1 in the review of Batstone et al., 2002b). However, the best comparison is probably to that of Siegrist et al. (2002) who reported a hydrolysis rate of 0.25 d1 at mesophilic conditions compared to 0.4 d1 at thermophilic conditions. This was for a thermophilic methanogenic reactor. Our results indicate comparative performance can be obtained simply by conditioning, rather than operating the main digester at thermophilic conditions. Since hydrolysis rate rather than extent is increased, the increase in performance can be accomplished by either adding a thermophilic pre-treatment stage or increasing the main digester size. If considering the footprint and capital investment of anaerobic digestion process, the addition of a thermophilic pre-treatment stage will benefit the design due to the smaller process vessels compared to mesophilic pre-treatment process or conventional mesophilic digestion. In terms of solids destruction, a larger main digester could achieve the same performance as adding a thermophilic pre-treatment stage. However, the thermophilic pre-treatment process enables pathogen destruction to achieve the ultimate solids hygienisation required for land application and agricultural use (Sung and Santha, 2003).
4.3.
Pre-treatment mechanisms
Analysis of the pre-treatment reactors as assessed by NHþ 4 -N and SCOD confirmed hydrolysis (solubilisation) in TP1 was improved compared to MP1, however this did not translate to increased conversion to organic acids. There was a considerable component of SCOD which could not be attributed to organic acids. Digestion intermediates such as glucose, pyruvate, succinate, lactate, and ethanol (Elefsiniotis and Oldham, 1994) were not detected in TP1. However, anaerobic organisms are able to directly take up and utilise partially hydrolysed organics including oligosaccharides and long-chain fatty acids (Lynd et al., 2002). Significantly, all hydrolysates produced during the TPAD processes were biologically degradable, exhibited by the significant reduction in SCOD concentrations in the effluent of the methanogenic stages (<500 mg L1) compared to the pretreatment effluents and raw sludge feed (1000–2000 mg L1). Variations in the biological processes occurring in TP1 and MP1 (5.5% solubilisation compared to 5.1% solubilisation) were minimal compared to the increased methane production and VS destruction (25% and 20%) observed between thermophilic–mesophilic TPAD and mesophilic–mesophilic TPAD. It is clear that key mechanisms active during thermophilic pretreatment affected biological availability of the substrate during downstream processes. However, the specific nature of these mechanisms is not clear. From a biological perspective, possible mechanisms include stimulated growth of the microbial population or production of extracellular hydrolytic enzymes which are then passed downstream into the methanogenic reactors. Increased microbial concentrations or enzyme activities could explain the increases in apparent hydrolysis rates, without an increase in sludge degradability (as determined from model simulations).
129
From a non-biological perspective, increased disintegration of the sludge may have reduced particle size and increased the surface area available to the microbial community. Hydrolysis is a surface process and rates may be improved by increasing the surface area of feed particles (McAllister et al., 1994; Lynd et al., 2002). Furthermore, effluent from TP1 may have contained increased colloidal substrates that are readily degradable, but not measured as SCOD. Further investigations into these mechanisms are required. As a final note, decreased pH did not enhance hydrolysis. Control of pH in full scale sludge fed systems is inherently difficult to manipulate due to buffering from NHþ 4 -N release.
5.
Conclusion
The following conclusions can be drawn from this study: Thermophilic–mesophilic TPAD achieved 20% and 25% higher VS destruction and methane production respectively, compared to mesophilic–mesophilic TPAD. Increasing thermophilic pre-treatment temperature from 50–65 C had no further impacts. Higher SCOD was produced during thermophilic pre-treatment over mesophilic pre-treatment, and further increased by increasing the thermophilic pre-treatment temperature from 50 C to 65 C. Higher NHþ 4 -N was released during thermophilic pre-treatment, but did not increase at increased temperatures. Both SCOD and NHþ 4 -N decreased under acidic pre-treatment conditions (pH 4.5). Model based analysis indicated that the improved performance was due to an increased hydrolysis rate (0.1 0.05 d1 to 0.3 0.15 d1), rather than overall degradability.
Acknowledgements This work was funded by the Queensland State Government, under the Smart State Research-Industry Partnerships Program (RIPP), Meat and Livestock Australia, and Environmental Biotechnology Cooperative Research Centre (EBCRC), Australia as P23 ‘‘Small-medium scale organic solids stabilization’’. Huoqing Ge and Paul Jensen are recipients of an EBCRC postgraduate scholarship and postdoctoral award, respectively. We thank Beatrice Keller, and the AWMC Analytical Services Laboratory for conducting organic acid and nitrogen analysis.
reference
APHA, 1998. Standard Methods for the Examination of Water and Wastewater, twentieth ed. American Public Health Association, Washington, DC, USA. Barr, K.G., Solley, D.O., Starrenburg, D.J., Lewis, R.G., 2008. Evaluation, selection and initial performance of a large scale centralised biosolids facility at Oxley Creek Water Reclamation Plant, Brisbane. Water Science and Technology 57 (10), 1579–1586.
130
water research 44 (2010) 123–130
Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S., Pavlostathis, S.G., Rozzi, A., Sanders, W.T.M., Siegrist, H., Vavilin, V., 2002a. Anaerobic Digestion Model No. 1 (ADM1), IWA Task Group for Mathematical Modelling of Anaerobic Digestion Processes. IWA Publishing, London, UK. Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhny, S., Pavlostathis, S.G., Rozzi, A., Sanders, W.T.M., Siegrist, H., Vavilin, V., 2002b. Anaerobic Digestion Model No. 1 (ADM1) Scientific and Technical Report No. 13, IWA Task Group for Mathematical Modelling of Anaerobic Digestion Processes. IWA Publishing, London, UK. Batstone, D.J., Pind, P.F., Angelidaki, I., 2003. Kinetics of thermophilic, anaerobic oxidation of straight and branched chain butyrate and valerate. Biotechnology and Bioengineering 84 (2), 195–204. Batstone, D.J., Darvodelsky, P., Keller, J., (2008a). Trends in biosolids handling technologies: economics and environmental factors. In: AWA biosolids specialty conference IV, Adelaide, Australia. Batstone, D.J., Tait, S., Starrenburg, D., 2008b. Estimation of hydrolysis parameters in full-scale anaerobic digesters. Biotechnology and Bioengineering 102 (5), 1513–1520. Elefsiniotis, P., Oldham, K.W., 1994. Influence of pH on the acidphase anaerobic digestion of primary sludge. Journal of Chemical Technology and Biotechnology 60 (1), 89–96. Gossett, J.M., Belser, R.L., 1982. Anaerobic-digestion of waster activated sludge. Journal of the Environmental Engineering Division-Asce 108 (6), 1101–1120. Han, Y., Sung, S., Dague, R.R., 1997. Temperature-phased anaerobic digestion of wastewater sludges. Water Science and Technology 36 (6-7), 367–374. Lynd, L.R., Weimer, P.J., van Zyl, W.H., Pretorius, I.S., 2002. Microbial cellulose utilization: fundamentals and biotechnology. Microbiology and Molecular Biology Reviews 66 (3), 506–577. McAllister, T.A., Bae, H.D., Jones, G.A., Cheng, K.J., 1994. Microbial attachment and feed digestion in the rumen. Journal of Animal Science 72 (11), 3004–3018. Neyens, E., Baeyens, J., 2003. A review of thermal sludge pretreatment processes to improve dewaterability. Journal of Hazardous Materials 98 (1–3), 51–67.
Nopens, I., Batstone, D.J., Copp, J.B., Jeppsson, U., Volcke, E., Alex, J., Vanrolleghem, P.A., 2009. An ASM/ADM model interface for dynamic plant-wide simulation. Water Research 43 (7), 1913–1923. Novak, J.T., 2006. Dewatering of sewage sludge. Drying Technology 24 (10), 1257–1262. Pavlostathis, S.G., Giraldo-Gomez, E., 1991. Kinetics of anaerobic treatment. Water Science and Technology 24 (8), 35–59. Reichert, P., 1994. Aquasim – a tool for simulation and dataanalysis of aquatic systems. Water Science and Technology 30 (2), 21–30. Siegrist, H., Vogt, D., Garcia-Heras, J.L., Gujer, W., 2002. Mathematical model for meso- and thermophilic anaerobic sewage sludge digestion. Environmental Science & Technology 36 (5), 1113–1123. Skiadas, I.V., Gavala, H.N., Lu, J., Ahring, B.K., 2005. Thermal pretreatment of primary and secondary sludge at 70 C prior to anaerobic digestion. Water Science and Technology 52 (1–2), 161–166. Sung, S., Santha, H., 2003. Performance of temperature-phased anaerobic digestion (TPAD) system treating dairy cattle wastes. Water Research 37 (7), 1628–1636. Switzenbaum, M.S., Farrell, J.B., Pincince, A.B., 2003. Relationship between the Van Kleeck and mass-balance calculation of volatile solids loss. Water Environment Research 75 (6), 377–380. Tait, S., Tamis, J., Edgerton, B., Batstone, D.J., 2009. Anaerobic digestion of spent bedding from deep litter piggery housing. Bioresource Technology 100 (7), 2210–2218. Tiehm, A., Nickel, K., Zellhorn, M., Neis, U., 2001. Ultrasonic waste activated sludge disintegration for improving anaerobic stabilization. Water Research 35 (8), 2003–2009. US Environmental Protection Agency, 1994. A Plain English Guide to the EPA Part 503 Biosolids Rule. In: Environmental Protection Agency 8322/R-93-003, September Washington, DC, USA. Watts, S., Hamilton, G., Keller, J., 2006. Two-stage thermophilic– mesophilic anaerobic digestion of waste activated sludge from a biological nutrient removal plant. Water Science and Technology 53 (8), 149–157.
water research 44 (2010) 131–140
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Influence of nanoscale zero-valent iron on geochemical properties of groundwater and vinyl chloride degradation: A field case study Yu-Ting Wei a, Shian-Chee Wu a, Chih-Ming Chou b, Choi-Hong Che a, Shin-Mu Tsai b, Hsing-Lung Lien b,* a b
Graduate Institute of Environmental Engineering, National Taiwan University, Taipei, Taiwan, ROC Department of Civil and Environmental Engineering, National University of Kaohsiung, 811 Kaohsiung, Taiwan, ROC
article info
abstract
Article history:
A 200 m2 pilot-scale field test successfully demonstrated the use of nanoscale zero-valent iron
Received 17 March 2009
(NZVI) for effective remediation of groundwater contaminated with chlorinated organic
Received in revised form
compounds in Taiwan within six months. Both commercially available and on-site synthe-
3 September 2009
sized NZVI were used. A well-defined monitoring program allowing to collect three-dimen-
Accepted 3 September 2009
sional spatial data from 13 nested multi-level monitoring wells was conducted to monitor
Published online 11 September 2009
geochemical parameters in groundwater. The degradation efficiency of vinyl chloride (VC) determined at most of monitoring wells was 50–99%. It was found that the injection of NZVI
Keywords:
caused a significant change in total iron, total solid (TS) and suspended solid (SS) concen-
Nanoparticles
trations in groundwater. Total iron concentration showed a moderate and weak correlation
Groundwater remediation
with SS and TS, respectively, suggesting that SS may be used to indicate the NZVI distribution
Nanotechnology
in groundwater. A decrease in oxidation–reduction potential (ORP) values from about 100 to
Chlorinated organic compound
400 mV after NZVI injection was observed. This revealed that NZVI is an effective means of
Taiwan
achieving highly reducing conditions in the subsurface environment. Both VC degradation efficiency and ORP showed a correlative tendency as an increase in VC degradation efficiency corresponded to a decrease of ORP. This is in agreement with the previous studies suggesting that ORP can serve as an indicator for the NZVI reactivity. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Using nanoscale zero-valent iron (NZVI) for groundwater remediation is one of the most promising environmental nanotechnologies up to date. It has been demonstrated that, with its high specific surface area and special molecular conformation (Zhang, 2003), NZVI is a reactive reagent for the treatment of various contaminants in aqueous systems
including chlorinated organics, heavy metals, perchlorate and nitrate (Cao et al., 2005; Elliott et al., 2008; Giasuddin et al., 2007; Lien et al., 2007; Liu et al., 2007; Ponder et al., 2000; Sohn et al., 2006; Wang and Zhang, 1997; Yuan and Lien, 2006). Additionally, bimetallic iron nanoparticles prepared by coating catalytic metals such as palladium or nickel on NZVI exhibited an enhancing effect to accelerate the reactivity and reduce the formation of toxic byproduct during the dechlorination
* Corresponding author. Tel.: þ886 7591 9221; fax: þ886 7591 9376. E-mail address:
[email protected] (H.-L. Lien). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.012
132
water research 44 (2010) 131–140
Fig. 1 – Layout of the pilot study area and injection and monitoring wells.
reaction (Choi et al., 2009; Feng and Lim, 2005; He and Zhao, 2005; Lien and Zhang, 2001, 2005, 2007; Nutt et al., 2005, 2006; Zhang et al., 1998). While laboratory studies have indicated the superior capability of NZVI, the feasibility for implementation and application of NZVI in groundwater remediation needs to be further tested in the field. An increasing number of field tests including pilot- and full-scale remediation sites have been implemented using NZVI (Glazier et al., 2003; Lien et al., 2006; Mace´ et al., 2006). The United States Environmental Protection Agency (USEPA) has reported a summary of a total of 26 sites using or testing nanoparticles for remediation in the U.S. and Canada (USEPA, 2008). Several pilot tests have been conducted in other countries including Czech Republic, Germany, Italy and Slovakia (Mace´ et al., 2006). The cost of NZVI was on the order of 50 USD/kg as of mid-2004 and is expected to decrease as supply increases (Elliott et al., 2009). Though NZVI is considered the most widely used nanoparticle in groundwater remediation, our current understanding of its risk to human or ecological health is still limited (Tratnyek and Johnson, 2006; USEPA, 2008). The assessment of the impact of nanomaterials on the environment and human health has attracted worldwide attention (e.g., Nel et al., 2006). Taiwan has long been known as an island with dynamic and flexible industries. The manufacturing industry including the production of chemicals, petrochemicals, advanced electrical equipment, and electronics accounted for 27.5% of the gross domestic product (GDP) in 2007 (Government Information
Office, 2008). To protect the groundwater resource, Taiwanese government promulgated The Soil and Groundwater Pollution Remediation Act in 2000. The groundwater pollution control standards of chlorinated hydrocarbons vary from 0.02 to 8.5 mg/L for non-drinking water source protection areas. The acceptable level of vinyl chloride in groundwater is regulated at 0.02 mg/L (Environmental Protection Administration, Taiwan, 2006). However, limited technologies such as pump-and-treat and chemical oxidation are currently available for groundwater remediation in Taiwan. As a result, the development of suitable remedial technologies is certainly a need. In this paper, we report the first field test for groundwater remediation using the NZVI technology in Taiwan. The study focuses on a pilot-scale field demonstration of injecting NZVI, either on-site synthesized or commercially available, to degrade a variety of chlorinated organic contaminants in the groundwater. A well-defined monitoring program was established by measuring the iron concentration and groundwater geochemical parameters including dissolved oxygen (DO), oxidation–reduction potential (ORP), pH and electric conductivity (EC) from 13 nested multi-level monitoring wells installed on a 200 m2 plot. Total solid (TS), suspended solid (SS), chloride and chlorinated organic compound concentrations in the groundwater were also determined. The monitoring program allows this study to collect three-dimensional spatial data that offer a systematic analysis to better understand the influence of NZVI on the groundwater geochemistry and the impact of geochemical properties on the effectiveness
133
water research 44 (2010) 131–140
below ground surface (m bgs) (Fig. 1 inset). The pumping test indicated that the hydraulic conductivity is 0.275 cm/sec, the transmissivity is 4349 m2/day, specific yield is 0.18 and the anisotropy is 14.3. The linear velocity of the groundwater is 28.5 cm/day. The natural gradient at the site is approximately 0.0012 m/m. Three injection wells and 13 nested multi-level monitoring wells were installed within the 200 m2 plot (Fig. 1). A background well was located upgradient of the injection wells while four multi-level monitoring wells were installed in the downstream direction of each injection well. The positions of the four nested monitoring wells are approximately one, two, three, and five meters away from the injection well. The injection wells are all eighteen-meter deep with fifteen-meter screens. The nested monitoring wells are consisted of three separate wells approximately six, twelve and eighteen-meter deep with three-meter screens (Fig. 1). Groundwater samples can therefore be collected from three different depths: upper, middle and bottom layers.
2.2.
of NZVI. To the best of our knowledge, this study presents the first three-dimensional spatial data for the in situ field test using the NZVI technology.
2.
Experimental methods
2.1.
Site selection and test area description
The testing site was selected at an active industrial complex including petrochemical plants and vinyl chloride monomer (VCM) manufacturing plants in Kaohsiung, Taiwan. The NZVI pilot test was conducted in a 10 m 20 m area located south of the VCM plant in the downstream groundwater direction (Fig. 1 inset) where the groundwater was contaminated with high concentrations of chlorinated organic compounds. The contaminated plume containing vinyl chloride (VC, 620– 4562 mg/L), trichloroethylene (TCE, 53–682 mg/L), 1,1-dichloroethylene (42–134 mg/L), cis-1,2-dichloroethylene (27–1151 mg/L), and dichloroethanes (55–270 mg/L) was measured from an upgradient monitoring well near the plant. In this paper, VC was selected as a target compound because of its high concentration. At the testing site, an unconfined aquifer, composed of medium to coarse sand and few silt, lies approximately 4–18 m
Two types of NZVI were used in this study. A commercially available nanoiron solution was purchased from Lehigh Nanotech, LLC and an on-site synthesized NZVI was prepared using a borohydride reduction. Prior to the field test, laboratory feasibility experiments for two kinds of NZVI were conducted to treat contaminated groundwater obtained from the testing site. Bench-scale batch tests indicated that both NZVI showed a better performance for TCE degradation in the TCE-spiked deionized water than in the TCE-contaminated groundwater (Fig. S-1). Nevertheless, palladized NZVI (0.5–1 wt%) exhibited a rapid and effective degradation of TCE regardless of the types of iron and water samples (Figs. S-2 and S-3). The on-site synthesized NZVI was prepared by slowly adding ferrous sulfate solution into sodium borohydride solution (>98.5%, Beckman Coulter, Inc.) containing a nonionic surfactant (Taiwan NJC Corp., industrial grade) in a 1000-L tank. After the reaction was complete, palladium acetate dissolved in methanol was mixed with the NZVI suspension to 1000
Total iron concentration (mg/L)
Fig. 2 – SEM images of (a) on-site synthesized NZVI and (b) commercial NZVI.
Preparation of NZVI
3M-1-1 (upper layer) 3M-1-2 (middle layer) 3M-1-3 (bottom layer) 100
10
1 0
20
40
60
80
100
120
Time (day) Fig. 3 – Total iron concentration measured at different depths in the monitoring well (3M-1) during the period of on-site synthesized NZVI injection.
134
water research 44 (2010) 131–140
a 0.0
50
100
150
0.5
150
150
200
250
1.0
250
150
250
200 300
1.5
2.0
3.0
50 250 100 150 200 250 300 350 200 1.0
300
350
300
2.5
Flow direction
150
200
200
Y-direction (m)
200
100
100
250
350
200
250
250
200
200
150
150
150
100
100
1.5
250
300
300
50
2.0
2.5
3.0
X-direction (m)
b
0.0 60
40
50 0.5
30
40 30
30
Y-direction (m)
1.0
30 1.5
30
40
30
50
30 40
60
2.0
30
50 40
2.5
3.0
Flow direction
50 60 70 80 90 100 110 120 130 140 360
30
30
30
1.0
1.5
2.0
2.5
3.0
X-direction (m)
Fig. 4 – Total iron concentration distribution in the upper layer during the period of on-sited synthesized NZVI injection at (a) Day 8 and (b) Day 66. (Refer to Fig. 1 for the corresponding position of the well locations.)
reach the ratio of palladium to iron at 0.5 wt%. The NZVI mixture was then pumped into a storage tank for the injection later. The cost of on-site synthesized NZVI was approximately 200 USD/kg. It should be noted that both types of NZVI exhibited a similar behavior in terms of their influence on pH, ORP and contaminant degradation efficiency in groundwater. Therefore, data presented in this paper are primarily based on the results of the on-site synthesized NZVI study.
X-ray (EDX) at 10 kV and a transmission electron microscope (TEM) (JEOL JSM-1200EX II). A surface area analyzer (Beckman Coulter SA3100) was used to determine surface areas of NZVI. The particle size was analyzed by a particle size analyzer (Brookhaven 90Plus, Brookhaven Instruments Co.). This instrument analyzes particle sizes ranging from <1 nm to 6 mm. Diluted samples were placed in plastic cells for analysis and elapsed time was set at 30–60 s.
2.3.
2.4.
Characterization of NZVI
Morphological and elemental analyses of NZVI were performed by a scanning electron microscope (SEM) (Hitachi S-4300, Hitachi Science Systems, Ltd.) equipped with energy-dispersive
Injection of NZVI
In the first period of the field test, the commercial nanoiron solution was applied and monitored for three months. A total amount of 2250 L diluted commercial nanoiron solution
water research 44 (2010) 131–140
135
NZVI was conducted in the second period of the trial and monitored for another three months. A total amount of 8500 L suspension containing 20 kg NZVI was injected via gravity. The first 1000 L was injected using the injection well IW-3 while the rest of NZVI suspension (7500 L) was injected using the injection well IW-1 after ten days. The 7500 L of on-site synthesized NZVI was prepared and stored for a week before deployment. Prior to the second NZVI injection, a measurement of VOC and total iron concentration was conducted to establish a new baseline for comparison with the post-injection result. It was found that the first NZVI injection resulted in a slight increase in total iron concentration from 10 3 to 16 9 mg/L.
2.5.
Fig. 5 – SME image of soil samples with the agglomeration of spherical iron particles.
containing 40 kg NZVI with palladium (1 wt%) was injected into the aquifer using the injection well IW-2 via gravity. The injection rate was about 1200 L/h. Five months after the end of the three-month monitoring program, the on-site synthesized
a 18000
Injection well 2M-3-1 3M-3-1 5M-3-1 5M-3-2 5M-3-3
TS concentration (mg/L)
16000 14000 12000 10000 8000 6000 4000 2000 0
0
10
20
30
40
50
Time (day)
b
Injection well 2M-3-1 3M-3-1 5M-3-1 5M-3-2 5M-3-3
SS concentration (mg/L)
6000
Field sampling and analyses
The groundwater samples were collected using a peristaltic pump at various time intervals. A low-flow sampling technique minimizing losses of volatiles and gases was used to take samples for organic analytes. Filtered and acidified samples were analyzed by an inductively coupled plasmaoptical emission spectrometry (ICP-OES, PerkinElmer Optima 2000DV, PerkinElmer Inc.) for total iron concentrations. Triplicate analysis was performed for each sample and data were accepted only if relative standard deviation was less than 1%. Geochemical parameters of groundwater including pH, DO and ORP were measured in the field using an YSI 650 MDS6600 probe (V2-4 Sonde, YSI Inc.). Concentrations of TS and SS were determined based on the USEPA method 160.2 (USEPA, 1999). One duplicate field sample was collected for every ten samples obtained. Duplicate samples were collected at the same time as the primary samples. Volatile organic compounds were measured by GC/MS (Angilent 6890/5973 with a DB-624 capillary column) using a purge and trap sampling equipment (OI Analytical. Model 4560). Methane, ethane and ethene were determined from the headspace of serum vials containing water samples after equilibration. Concentrations of hydrocarbons were measured by a HP4890 GC-FID equipped with a GS-GASPRO capillary column (J&W, 30 m 0.32 mm). Analysis was generally performed in triplicate with relative differences less than 15%. Soil samples were obtained after 7 days of the commercial NZVI injection. The testing site was cored using 5 cm i.d. core barrels to examine the iron distribution in the aquifer. The cores were taken at the testing site near the monitoring wells (1M-2, 2M-2 and 3M-2) and the injection well (IW-2) using the Geoprobe technique. A SEM-EDX was used to identify the NZVI in the soil samples.
4000
3.
Results and discussion
2000
3.1. Characterization of on-site synthesized and commercial NZVI 0
0
10
20
30
40
Time (day)
Fig. 6 – The concentration changes of (a) TS and (b) SS observed in the injection and monitoring wells.
50
Fig. 2 shows the SEM images for the on-site synthesized and commercial NZVI, respectively. The synthesized NZVI showed a particle size in nanoscale range (Fig. S-4) while the commercial NZVI had a spherical shape and relatively large particle size. Based on the particle size analysis, the on-site
a
0.0
-400
-450
-350
-350
0.5
-400
Y-direction (m)
1.0
-400
-350
-250 -250
Flow direction
-450
-300
-300
-550
-500
-300 -250
-200
-200
1.5
-200
-200
-550 -500 -450 -400 -350 -300 -250 -200
2.0
2.5 -200 3.0
b
1.0
1.5
0.0
2.0
-450
2.5
3.0 -400
-450
Flow direction
-400 0.5
-300
-300
-300
1.0
-350
-350
-350
-350
Y-direction (m)
-400
-400
-300 -300
-300
1.5
-300 2.0 -300
-300 -600 -550 -500 -450
2.5
3.0
c
1.0
-300
1.5
2.0
0.0
-240 -300
0.5
-280
-260
Y-direction (m)
-280
Flow direction
-300
-300
-300 -260
-260
3.0
-280
-280
1.0
2.5
-320
-280
-320
-320 -320 -300
1.5
-280
2.5
3.0
-300
-400 -380 -360 -340 -320 -300 -280 -220
2.0
-300
-300
-280 1.0
1.5
2.0
2.5
3.0
X-direction (m)
Fig. 7 – ORP changes in the upper layer during the period of on-sited synthesized NZVI injection at (a) Day 8, (b) Day 40 and (c) Day 66. (Refer to Fig. 1 for the corresponding position of the well locations.)
137
water research 44 (2010) 131–140
700
Upper layer IW-1 1M-1-1 2M-1-1 3M-1-1 5M-1-1
VC Concentration (µg/L)
600 500 400 300 200 100 0
0
VC Concentration (µg/L)
b 8000
10
20
30 40 Time (days)
50
Bottom layer
60
70
IW-1 1M-1-3 2M-1-3 3M-1-3 5M-1-3
6000
4000
2000
0
0
10
20
30 40 Time (days)
50
60
70
Fig. 8 – Degradation of VC in the (a) upper layer and (b) bottom layer during the NZVI injection.
synthesized NZVI had the particle size in the range of 80–120 nm whereas the commercial NZVI had the average particle size between 450 and 550 nm. The elemental analysis conducted by SEM-EDX revealed that the iron content accounted for nearly 100 wt% for commercial NZVI. The specific surface area of on-site synthesized NZVI was 29.3 m2/ g, similar to those found in literatures (Lien and Zhang, 2001; Liu et al., 2005; Ponder et al., 2000). However, the specific surface area of commercial NZVI was only approximately 4.61 m2/g. Detail characterization of synthesized NZVI by the borohydride reduction can be found elsewhere (e.g., Nurmi et al., 2005).
3.2.
Total iron concentration in groundwater and soil
Fig. 3 shows the total iron concentration change at different depths in the monitoring well (3M-1) during the period of onsite synthesized NZVI injection. As shown in Fig. 3, it is clear that the iron concentration found in the upper layer was significantly higher than that found in the middle and bottom layers. The iron concentration decreased in the order: upper layer > middle layer > bottom layer. It is believed that with the injection via gravity, much of the NZVI first seeped through channels in the unsaturated zone, causing NZVI to accumulate more in the upper layer as compared to that in the lower layer.
It should be pointed out that a significant iron concentration drop in the upper layer was observed at Day 66. This is likely because a heavy storm occurred 3 days before field sampling. Iron in the upper layer may be washed downstream or may penetrate into the bottom layer. A concentration distribution of total iron in the upper layer determined from monitoring wells located down gradient from the injection well during the on-site synthesized NZVI injection is illustrated in Fig. 4. The injected NZVI influenced the geochemistry of the aquifer significantly. A noticeable increase in the total iron concentration at the whole testing site was observed after NZVI was injected. The iron concentration was in the range of 40–370 mg/L. In general, the iron concentration showed a tendency of decrease towards downstream. Fig. 4b shows the iron concentration distribution at the testing site after 66 days of NZVI injection. A dramatic decrease in iron concentrations with time was found. This suggested that iron either was consumed through the oxidative corrosion or transported through the groundwater flow. It should be pointed out that the total iron concentration measured in the injection well was about 2500 mg/L after 7 days of the commercial NZVI injection. It is about 7 times greater than the on-site synthesized NZVI injection. This suggested that the on-site synthesized NZVI has a better mobility than the commercial NZVI, which is at least partly due to the size effect of particles. Fig. 5 shows the SEM image of the soil sample obtained from the injection well after 7 days of the commercial NZVI injection. Based on the SEM-EDX analysis, the agglomeration of spherical particles was determined to have the major composition of iron (81.7 wt%) suggesting they were originally from the commercial NZVI. Analysis of the concentration profile of total iron in the soil indicated that the soil iron concentration was all greater than 30 g/kg within the depth of 1–6 m. The soil iron showed the highest concentration in the vicinity of the injection well (up to 90 g/kg), which is consistent with the groundwater data. In addition, it was found that a large portion of commercial NZVI was trapped in the depth of about 3.6 m that is in a good agreement with the groundwater samples indicating the upper layer contained the highest total iron concentration among three layers.
18000 16000
Solid concentration (mg/L)
a
SS TS
14000 12000 10000
R2 = 0.28
8000 6000
R2 = 0.63
4000 2000 0
0
100
200
300
Total iron concentration (mg/L) Fig. 9 – Correlation of total iron concentrations with TS and SS concentrations.
138
a
water research 44 (2010) 131–140
0 -100
ORP (mV)
-200 -300 -400 -500 -600
0
10
20
30
40
50
60
70
Time (day)
VC degradation efficiency (%)
b
100
80
60
40
3.5.
0
10
20
30
40
50
60
70
Time (day) Fig. 10 – Trades of (a) ORP and (b) VC degradation efficiency at the testing site.
Electric conductivity (EC) and chloride
The EC of groundwater is dependent on the depth of the aquifer. The average EC values measured in upper, middle and bottom layer were 1.5, 8.5 and 14.5 ms/cm, respectively. A similar trend of the chloride concentration was observed. Chloride concentrations, ranging of 550–6190 mg/L, generally increased with increasing the aquifer depth. The unusual high value of chloride concentrations and EC detected in the bottom layer may be attributed to the seawater intrusion because the testing site is located near the coast.
3.4.
ORP and pH
20
0
3.3.
Day 8 measured in the upper, middle and bottom layer at the monitoring well 5M-3 was 1303, 3641 and 9401.5 mg/L, respectively. This is in a good correlation with the EC results. The impact of NZVI on the TS concentration in the bottom layer is negligible. This is consistent with the result shown in Fig. 3 that only small amounts of NZVI were able to reach the bottom layer. Furthermore, it was found that the change of TS concentrations is minor at the furthest monitoring distance of 5 m, which may reflect the maximum mobile distance of NZVI. It is worthy of mention that although the groundwater possesses a high ionic strength in the testing site that may cause the agglomeration of NZVI and limit its movement (Phenrat et al., 2007; Saleh et al., 2008), this study has demonstrated the capability of NZVI for effective remediation of chlorinated organic solvents. Fig. 6b illustrates the change of the SS concentration with time along the downstream path after the injection of NZVI. Compared to the TS concentration, a similar trend in the change of SS concentrations was observed. The highest concentration of SS was about 5000 mg/L measured at 2 m downstream from the injection well. Analogous to TS, no significant change of the SS concentration was found at the furthest monitoring well 5M-3.
Total solid (TS) and suspended solid (SS)
The injection of NZVI influenced the TS and SS concentration in the groundwater. A significant increase of SS and TS concentrations was observed within an effective distance of 3 m in the upper layer after the NZVI injection. As shown in Fig. 6a, the TS concentration in the upper layer peaked at Day 8 and then gradually decreased. The highest TS concentration was measured at 2 m downstream from the injection well suggesting the NZVI migration occurred. It should be noted that the TS concentration in the bottom layer was higher than that in the upper layer. For example, the TS concentration of
In a groundwater environment, the dissolved oxygen concentration is usually very low (e.g., <1 mg/L in this study); therefore, the predominant electron receptor is water: Fe0 þ 2H2 O/Fe2þ þ H2 þ 2OH
(1)
According to the above reaction, the iron oxidation reaction should produce a characteristic increase in solution pH and a concomitant decline in the redox potential. Redox labile contaminants such as chlorinated organics can also serve as possible electron acceptors. However, these contaminants are generally present at low concentrations in the environment and typically do not strongly influence the pH or ORP profiles observed. As it has been observed in field tests of the NZVI technology, the pH and ORP profiles at given monitoring locations over time may serve as an indicator for the NZVI reactivity and to track the migration path of the nanoparticles (Elliott and Zhang, 2001; Glazier et al., 2003; Zhang, 2003). In this study, it was found that the ORP decreased significantly from about 100 to 400 mV at the central area of the testing site after NZVI injection. The ORP distribution shown in Fig. 7 reveals that a strong reducing condition was established at the beginning while ORP values slowly increased during the course of the testing period. This corresponds to the results of the NZVI distribution shown in Fig. 4 suggesting the NZVI gradually migrated towards downstream. Overall, the data from this study suggested that NZVI is an effective means of achieving highly reducing conditions in the subsurface environment. The impact of NZVI on groundwater pH is minor. No significant increase of pH was found in all monitoring wells where the groundwater pH was maintained under near neutral conditions (pH 6–7) except for the injection well. At the injection well, the pH value increased to 8.5 indicating the oxidation of NZVI took
water research 44 (2010) 131–140
place and caused the release of hydroxyl ions (Gillham and O’Hannesin, 1994).
3.6.
Effectiveness of NZVI for VC degradation
The concentration profiles of VC also exhibited a dependence on the aquifer depth at the testing site. The VC concentration in the upper layer was in the range of 31.2–624.5 mg/L while VC in the bottom layer can reach up to a concentration greater than 6245 mg/L. Fig. 8 shows the VC concentration steadily decreased as the test date progressed. In general, the degradation efficiency was greater than 90% in both upper and middle layers but was about 20–85% in the bottom layer (Fig. 8). The degradation efficiency determined at most of the monitoring wells was 50–99%. The lowest degradation efficiency was about 20%, which was found in the bottom layer of the monitoring well 5M-1. Hydrocarbons including methane, ethylene and ethane were observed. High concentrations (up to 20 mg/L) of methane and ethylene were determined while the ethane concentration remained in trace amounts (at few mg/L). The formation of hydrocarbons may be resulted from the microbial activity since strongly reducing conditions are beneficial to the anaerobic bioremediation (Mace´ et al., 2006; USEPA, 2008). However, the cause for high hydrocarbon concentrations is still unclear at the current stage.
3.7.
Correlation of TS and SS with the total iron
A challenge with evaluating the effectiveness of NZVI injection is to monitor the NZVI distribution in the aquifer. ORP and dissolved iron have been used as indicators (USEPA, 2008). In this study, geochemical parameters including ORP, pH, TS, SS and total iron concentrations were systematically monitored. It was found that the injection of NZVI resulted in a significant impact on ORP, TS, SS and total iron concentration in the groundwater. As shown in Fig. 9, the correlation analysis indicated a moderate dependence (R2 ¼ 0.63) of SS on the total iron concentration. Further, a relatively weak correlation of TS with total iron concentration was found (R2 ¼ 0.28) whereas there was no correlation between ORP and total iron concentration (R2 ¼ 0.08, Fig. S-5). This suggests that SS may have potential to serve as a proper indicator to locate the NZVI distribution in the subsurface. The SS, TS, and total iron concentration represent the ‘‘primary’’ parameter that directly reflects to the iron content of injected NZVI. However, ORP can be considered as a ‘‘secondary’’ parameter that represents the reducing conditions caused by the redox reaction of iron and water after the post injection of NZVI. In other words, SS, TS and the total iron concentration can indicate specific groundwater conditions at a specific time while ORP reveals a bulk condition of groundwater within a period of time. As a result, it is reasonable to observe no correlation between ORP and the total iron concentration in this study.
3.8. Correlation of ORP with the VC degradation efficiency To facilitate the evaluation process of the NZVI effectiveness, there is a need to find a simple yet reliable geochemical
139
parameter capable of serving as a performance indicator. ORP has been extensively tested for this purpose (Elliott et al., 2008; Glazier et al., 2003; Mace´ et al., 2006). Fig. 10 summarizes all the data of the VC degradation efficiency and ORP measured in various time intervals from upper, middle and bottom layers during the injection of on-site synthesized NZVI. It was found that both ORP and VC degradation efficiency showed a correlative tendency as an increase in VC degradation efficiency corresponded to a decrease of ORP (Fig. 10). This is in agreement with the previous studies suggesting that ORP can act as a threshold for indicating whether or not the VC degradation takes place.
4.
Conclusions
This pilot-scale field study demonstrated a successful application of nanoscale zero-valent iron (NZVI), either on-site synthesized or commercially available, for remediation of groundwater contaminated with chlorinated organic compounds. A total amount of 60 kg palladized NZVI was injected into the groundwater via gravity at a 10 m 20 m testing site. Based on the results of this study, the following conclusions can be drawn: The testing site is located near the coast where the deep aquifer may be intruded by seawater. As a result, it was found that the electric conductivity, chloride, VC and TS concentration are all dependent on the aquifer depth. The VC degradation efficiency determined at most of the monitoring wells was 50–99%. It was greater than 90% in both upper and middle layers but was about 20–85% in the bottom layer. High concentrations (up to 20 mg/L) of methane and ethylene were detected. However, the cause of which is still unclear at the current stage. It is likely that enhanced bioremediation was involved at the testing site because of its strongly reducing conditions. NZVI is mobile in the aquifer. The effective travel distance is at least 3 m according to the SS and TS analysis. Correlation analysis showed a moderate dependence of the SS concentration on the total iron concentration. This suggested that SS may be used to indicate the ZVI distribution in groundwater. An increase in VC degradation efficiency corresponded to a decrease of ORP values, which is in agreement with the previous studies suggesting that ORP can serve as a proper indicator for the NZVI reactivity.
Acknowledgements The authors would like to thank the National Science Council (NSC), Taiwan, R.O.C. for the financial support under Grant no. NSC 95-2221-E-002-162-MY2 and NSC 95-2221-E390-014-MY2. We would also like to thank Mr. De-Huang Huang of Chinese Petroleum Corporation for his on-site assistance.
140
water research 44 (2010) 131–140
Appendix. Supplementary data Supplementary data associated with this article can be found in the online version, at doi:10.1016/j.watres.2009.09.012
references
Cao, J., Elliott, D., Zhang, W.-X., 2005. Perchlorate reduction by nanoscale iron particles. Journal of Nanoparticle Research 7 (4–5), 499–506. Choi, H., Agarwal, S., Al-Abed, S.R., 2009. Adsorption and simultaneous dechlorination of PCBs on GAC/Fe/Pd: mechanistic aspects and reactive capping barrier concept. Environmental Science and Technology 43 (2), 488–493. Elliott, D., Zhang, W.-X., 2001. Field assessment of nanoscale biometallic particles for groundwater treatment. Environmental Science and Technology 35 (24), 4922–4926. Elliott, D., Lien, H.-L., Zhang, W.-X., 2008. Zerovalent iron nanoparticles for treatment of ground water contaminated by hexachlorocyclohexanes. Journal of Environmental Quality 37, 2192–2201. Elliott, D., Lien, H.-L., Zhang, W.-X., 2009. Degradation of lindane by zero-valent iron nanoparticles. Journal of Environmental Engineering 135 (5), 317–324. Environmental Protection Administration, Taiwan, 2006. Soil and groundwater pollution control site preliminary assessment regulations. http://law.epa.gov.tw/en/laws/278076101.html. Feng, J., Lim, T., 2005. Pathways and kinetics of carbon tetrachloride and chloroform reductions by nano-scale Fe and Fe/Ni particles: comparison with commercial micro-scale Fe and Zn. Chemosphere 59 (9), 1267–1277. Giasuddin, A.B.M., Kanel, S.R., Choi, H., 2007. Adsorption of humic acid onto nanoscale zerovalent iron and its effect on arsenic removal. Environmental Science and Technology 41 (6), 2022–2027. Gillham, R.W., O’Hannesin, S.F., 1994. Enhanced degradation of halogenated aliphatics by zero-valent iron. Ground Water 32 (6), 958–967. Glazier, R., Venkatakrishnan, R., Gheorghiu, F., Walata, L., Nash, R., Zhang, W.-X., 2003. Nanotechnology takes root. Civil Engineering 73 (5), 64–69. Government Information Office, the Republic of China, 2008. The 2008 edition of the Republic of China yearbook. http://www. gio.gov.tw/taiwan-website/5-gp/yearbook/contents.html. He, F., Zhao, D., 2005. Preparation and characterization of a new class of starch-stabilized bimetallic nanoparticles for degradation of chlorinated hydrocarbons in water. Environmental Science and Technology 39 (9), 3314–3320. Lien, H.-L., Zhang, W.-X., 2001. Nanoscale iron particles for complete reduction of chlorinated ethenes. Colloids and Surfaces A: Physicochemical and Engineering Aspects 191 (1–2), 97–105. Lien, H.-L., Zhang, W.-X., 2005. Hydrodechlorination of chlorinated ethanes by nanoscale Pd/Fe bimetallic particles. Journal of Environmental Engineering 131 (1), 4–10. Lien, H.-L., Elliott, D.W., Sun, Y.-P., Zhang, W.-X., 2006. Recent progress in zero-valent iron nanoparticles for groundwater remediation. Journal of Environmental Engineering and Management 16 (6), 371–380. Lien, H.-L., Jhuo, Y.-S., Chen, L.-H., 2007. Effect of heavy metals on dechlorination of carbon tetrachloride by iron nanoparticles. Environmental Engineering Science 24 (1), 21–30.
Lien, H.-L., Zhang, W.-X., 2007. Nanoscale Pd/Fe bimetallic particles: catalytic effects of palladium on hydrodechlorination. Applied Catalysis B: Environmental 77 (1–2), 110–116. Liu, Y., Majetich, S.A., Tilton, R.D., Sholl, D.S., Lowry, G.V., 2005. TCE dechlorination rates, pathways, and efficiency of nanoscale iron particles with different properties. Environmental Science and Technology 39 (5), 1338–1345. Liu, Y., Phenrat, T., Lowry, G.V., 2007. Effect of TCE concentration and dissolved groundwater solutes on NZVI-promoted TCE dechlorination and H2 evolution. Environmental Science and Technology 41 (22), 7881–7887. Mace´, C., Desrocher, S., Gheorghiu, F., Kane, A., Pupeza, M., Cernik, M., Kvapil, P., Venkatakrishnan, R., Zhang, W.-X., 2006. Nanotechnology and groundwater remediation: a step forward in technology understanding. Remediation Journal 16 (2), 23–33. Nel, A., Xia, T., Ma¨dler, L., Li, N., 2006. Toxic potential of materials at the nanolevel. Science 311 (5761), 622–627. Nurmi, J.T., Tratnyek, P.G., Sarathy, V., Baer, D.R., Amonette, J.E., 2005. Characterization and properties of metallic iron nanoparticles: spectroscopy, electrochemistry, and kinetics. Environmental Science and Technology 39 (5), 1221–1230. Nutt, M.O., Hughes, J.B., Wong, M.S., 2005. Designing Pd-on-Au bimetallic nanoparticles for trichloroethylene hydrodechlorination. Environmental Science and Technology 39 (5), 1346–1353. Nutt, M.O., Heck, K.N., Alvarez, P., Wong, M.S., 2006. Improved Pdon-Au bimetallic nanoparticle catalysts for aqueous-phase trichloroethylene hydrodechlorination. Applied Catalysis B: Environmental 69, 115–125. Phenrat, T., Saleh, N., Sirk, K., Tilton, R.D., Lowry, G.V., 2007. Aggregation and sedimentation of aqueous nanoiron dispersions. Environmental Science and Technology 41 (1), 284–290. Ponder, S.M., Darab, J.G., Mallouk, T.E., 2000. Remediation of Cr(VI) and Pb(II) aqueous solution using supported nanoscale zero-valent iron. Environmental Science and Technology 34 (12), 2564–2569. Saleh, N., Kim, H.-J., Phenrat, T., Matyjaszewski, K., Tilton, R.D., Lowry, G.V., 2008. Ionic strength and composition affect the mobility of surface-modified Fe0 nanoparticles in watersaturated sand columns. Environmental Science and Technology 42 (9), 3349–3355. Sohn, K., Kang, S.W., Ahn, S., Woo, M., Yang, S.-K., 2006. Fe(0) nanoparticles for nitrate reduction: stability, reactivity, and transformation. Environmental Science and Technology 40 (17), 5514–5519. Tratnyek, P.G., Johnson, R.L., 2006. Nanotechnologies for environmental cleanup. Nano Today 1 (2), 44–48. United States Environmental Protection Agency (USEPA), (2008) Nanotechnology for site remediation fact sheet, EPA 542-F-08–009. United States Environmental Protection Agency (USEPA), 1999. EPA method 160.2 (gravimetric, dried at 103–105 C). http:// www.epa.gov/region09/qa/datatables.html. Wang, C., Zhang, W.-X., 1997. Synthesizing nanoscale iron particles for rapid and complete dechlorination of TCE and PCBs. Environmental Science and Technology 31 (7), 2154–2156. Yuan, C., Lien, H.-L., 2006. Removal of arsenate from aqueous solution using nanoscale iron particles. Water Quality Research Journal of Canada 41 (2), 210–215. Zhang, W.-X., Wang, C.B., Lien, H.-L., 1998. Treatment of chlorinated organic contaminants with nanoscale bimetallic particles. Catalysis Today 40, 387–395. Zhang, W.-X., 2003. Nanoscale iron particles for environmental remediation: an overview. Journal of Nanoparticle Research 5 (3–4), 323–332.
water research 44 (2010) 141–150
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Citizen monitoring: Testing hypotheses about the interactive influences of eutrophication and mussel invasion on a cyanobacterial toxin in lakes Orlando Sarnelle a,*, Jamie Morrison a, Rajreni Kaul a, Geoffrey Horst a, Howard Wandell a, Ralph Bednarz b a b
Department of Fisheries and Wildlife, Michigan State University, 13 Natural Resources Building, East Lansing, MI 48824, Michigan, USA Department of Environmental Quality, Lansing, Michigan, USA
article info
abstract
Article history:
An existing volunteer monitoring network in the state of Michigan was exploited to
Received 22 May 2009
conduct a statewide survey of the cyanobacterial toxin, microcystin, and to test hypotheses
Received in revised form
about the interactive influences of eutrophication and dreissenid mussel invasion. A total
2 September 2009
of 77 lakes were sampled by citizen volunteers for microcystin, total phosphorus (TP) and
Accepted 4 September 2009
chlorophyll a. Microcystin was measured in depth-integrated samples collected from the
Available online 11 September 2009
euphotic zone as well as in surface-water samples collected along the shoreline. Average microcystin in samples collected by volunteers was not different from samples collected
Keywords:
side-by-side by professionals. Euphotic-zone microcystin was positively related to TP in
Microcystin
lakes without dreissenids (uninvaded) but not in lakes with dreissenids (invaded).
Microcystis
Regression-tree analysis indicated that euphotic-zone microcystin was eight times higher
Cyanobacteria
in the presence of dreissenids for lakes with TP between 5 and 10 mg L1. In contrast,
Total phosphorus
euphotic-zone microcystin was almost identical in invaded and uninvaded lakes with TP
Recreational exposure
between 10 and 26 mg L1. Across all lakes, microcystin concentrations at the surface were
Dreissena polymorpha
on average more than double, and in some cases an order-of-magnitude greater than,
Dreissena bugensis
concentrations in the euphotic-zone. Given these results, it seems prudent to include dreissenid invasion status in forecasting models for microcystin, and to include shoreline sampling in monitoring programs aimed at assessing recreational exposure to cyanobacterial toxins. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Cyanobacteria are a major source of water-quality problems, in part because many species produce toxins that contaminate drinking water and degrade the recreational value of freshwaters. Given the importance of cyanobacteria to water quality, major research efforts have been directed at predicting the influence of environmental factors on the biomass of
bloom-forming nuisance cyanobacteria (hereafter referred to as cyanobacteria). Most notably, this body of work has shown that the biomass of cyanobacteria increases faster with nutrient enrichment than that of phytoplankton biomass in general, such that the percentage of total phytoplankton biomass comprised by cyanobacteria can reach 100% during the summer in lakes in which total phosphorus concentrations exceed w100 mg L1 (Downing et al., 2001; Kalff, 2002;
* Corresponding author. Tel.: þ1 517 353 4819; fax: þ1 517 432 1699. E-mail address:
[email protected] (O. Sarnelle). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.014
142
water research 44 (2010) 141–150
Trimbee and Prepas, 1987). More recently, it has been shown that at extremely high enrichment levels (TP >w500 mg L1), dominance by cyanobacteria may decline again with further enrichment (Jensen et al., 1994). Recent advances in analytical methodology have enabled more widespread monitoring of particular cyanobacterial toxins, most notably the class of peptides known as microcystins. Microcystins are intracellular compounds produced by several cyanobacterial taxa (most notably Microcystis aeruginosa) that act as potent liver toxins when ingested by terrestrial animals and humans (Chorus and Bartram, 1999). Given the relative ease of microcystin measurement, studies have accumulated documenting the distribution and abundance of microcystins in a variety of habitats and geographical regions (Jacoby and Kann, 2007; Kotak and Zurawell, 2007). Despite this increased attention, research into the influence of environmental factors on microcystin concentrations across natural systems remains relatively underdeveloped (Graham et al., 2004). A few studies have correlated microcystin concentrations with environmental factors across lakes, with the general consensus that concentrations increase with nutrient enrichment in all but the most enriched lakes (Giani et al., 2005; Graham et al., 2004; Kotak et al., 2000). This conclusion is not surprising given the existing base of knowledge about the effects of eutrophication on cyanobacteria in general (Kalff, 2002). The most comprehensive studies of environmental influences on microcystin concentrations have relied on sampling methods that involve concentrating the toxin by collecting particles from lake water using sieves, with the mesh sizes employed being as large as 64 mm (Giani et al., 2005; Graham et al., 2004; Kotak et al., 2000). Concentrating samples via sieving enables detection of very low environmental concentrations, but at the potential cost of underestimating concentrations since microcystin may be associated with particles that pass through the sieve. We have found microcystin concentrations as high as 5 mg L1 in the filtrate from a 35 mm mesh sieve when Microcystis aeruginosa, a colonyforming cyanobacterium, was dominant (unpublished data). A recent study estimated that sieving with 53 mm mesh resulted in an average 37% underestimation of total microcystin concentrations in lakes (Graham and Jones, 2007). From the perspective of protecting public health, underestimating toxin levels when they are high would seem to be more of an issue for a monitoring program than not detecting toxin when levels are very low. Thus, we decided to depart from common practice in our survey by measuring microcystin concentrations in unaltered water samples. In an era of restricted budgets for environmental monitoring, some agencies have begun to rely on citizen-volunteer networks to provide data on large numbers of habitats, including lakes (Bruhn and Soranno, 2005). These networks enlist interested citizens as field workers who collect samples for the analysis of relatively simple parameters such as total phosphorus (TP) and chlorophyll a. Studies have shown that sampling by citizens can provide high-quality data for such parameters when the laboratory analyses are performed by professionals (Canfield et al., 2002; Obrecht et al., 1998). We took advantage of an existing citizen-monitoring network in the state of Michigan (the Cooperative Lakes Monitoring Program, CLMP) to conduct a state-wide survey of microcystin
concentrations in 2006. The CLMP has been in existence since 1974 and the current program includes monitoring of more than 100 lakes for water clarity (Secchi Disk depth), TP and chlorophyll a. By coordinating with the CLMP, we were able to collect microcystin data from 77 lakes (a total of 378 samples) on a small budget, along with parallel data on TP and chlorophyll a. In this paper, we present analyses aimed at validating microcystin data from samples collected by citizen-volunteers, followed by use of that data to examine a set of hypotheses about the interacting influences of total phosphorus and dreissenid-mussel invasion on microcystin concentrations. The ongoing invasion of North American lakes by dreissenid mussels (Dreissena polymorpha, the zebra mussel and Dreissena bugensis, the quagga mussel) has been implicated as the cause of a widely-observed increase in a particular species of toxigenic cyanobacteria, Microcystis aeruginosa (Knoll et al., 2008; Raikow et al., 2004; Sarnelle et al., 2005; Vanderploeg et al., 2001). Curiously, promotion of M. aeruginosa by dreissenids appears to be limited to lakes with low-moderate TP concentrations (Knoll et al., 2008; Raikow et al., 2004; Sarnelle et al., 2005). Thus, there is concern that lakes not normally at risk for high concentrations of cyanobacterial toxins may become so after dreissenid invasion. One survey of lakes in Michigan has also suggested that dreissenid invasion alters the well-established positive influence of TP on cyanobacterial dominance (Raikow et al., 2004). Although previous studies have investigated the interactive influences of eutrophication and dreissenid invasion on the dominance of cyanobacteria (Raikow et al., 2004), and the separate influences of eutrophication (Giani et al., 2005; Graham et al., 2004; Kotak et al., 2000) and dreissenid invasion (Knoll et al., 2008) on microcystin concentrations, no studies have examined the interactive influences of eutrophication and dreissenid invasion on microcystin concentrations. Based on existing studies, we constructed testable hypotheses about how microcystin should respond to the interacting influences of TP and dreissenid invasion. First, we expected that microcystin concentrations would increase with TP, given that our survey was limited to lakes with TP levels below 200 mg L1 (Giani et al., 2005; Graham et al., 2004; Kotak et al., 2000). However, based on the response of cyanobacteria to dreissenid invasion and TP in a previous survey (Raikow et al., 2004), we also expected that microcystin would increase with TP at a faster rate in uninvaded lakes than in invaded lakes. Finally, we expected that microcystin concentrations would be elevated in invaded lakes, but only if those lakes have relatively low nutrients (Knoll et al., 2008; Raikow et al., 2004). In addition to examining the influences of dreissenid invasion and TP, we also examined how microcystin levels measured at the surface near the shore compare with concentrations in offshore samples collected from the entire euphotic zone. Surface sampling near the shore targets toxin levels at the point of contact with swimmers and terrestrial animals, whereas depth-integrated sampling through the mixed layer represents the typical method used by limnologists to assess phytoplankton species composition (Sarnelle, 1993), a major driver of variation in microcystin concentrations across lakes (Kotak and Zurawell, 2007). Given that cyanobacteria can regulate their buoyancy and form scums at the surface in calm weather (Reynolds, 1984), we hypothesized that
water research 44 (2010) 141–150
microcystin concentrations should be generally higher at the surface than in depth-integrated samples from the mixed layer. We are not aware of any broad-scale systematic comparison of this kind for cyanobacterial toxins (but see Johnston and Jacoby, 2003 for an example within one lake).
2.
Materials and methods
2.1.
Sample collection and processing
Samples were collected by citizen-volunteers in conjunction with the CLMP, which is administered by the Michigan Department of Environmental Quality (MDEQ) in partnership with Michigan Lake and Stream Associations, Inc., Michigan State University (MSU) and the Huron River Watershed Council under the Michigan Clean Water Corps (MiCorps) program. A one-day training session was held in April of 2006 to instruct volunteers in sample collection/processing and to distribute sampling containers, shipping supplies and instruction manuals. Each of 77 lakes (Supplementary Table 1, Fig. 1) was sampled on a single occasion within the period 25 August–29 September, 2006. Water samples from the euphotic zone (defined as 2X the Secchi Depth, (Koenings and Edmundson, 1991)) were collected from a single site over the deepest basin of the lake with a depth-integrating sampler. Samples for TP analysis were taken at the same location but collected by submerging a 250 ml polyethylene sample bottle w0.3 m below the surface. Surface-water samples were collected from four sites along the north, south, east and west shorelines. Surface water was collected by orienting the sample bottle such that the bottle opening was partially submerged and air could escape the bottle without bubbling during filling. Volunteers were instructed to take shoreline samples where water depth was w0.6 m and to avoid disturbing the sediment in the vicinity of the sample-collection site. Samples destined for microcystin
143
and chlorophyll a analysis were stored in brown 250 ml polyethylene bottles in the field. Volunteers were instructed to store samples on ice in the field and process (chlorophyll a) and freeze samples at their home immediately after completion of field collection. For chlorophyll a, volunteers filtered 50 ml of lake water through a 0.45 mm membrane filter (Millipore, MF) using a syringe-filter holder assembly. For microcystin, one subsample from each of the five 250 ml bottles was dispensed into a matching set of five 60 ml polyethylene bottles. All samples were kept frozen until transported or shipped to the laboratory for analysis. Volunteers transported euphotic-zone samples for chlorophyll a and TP to a designated MDEQ district office. Only samples that were frozen upon delivery at the MDEQ office were accepted for analysis. These samples were then kept frozen by MDEQ until analyzed. Chlorophyll a from shoreline samples and all microcystin samples were shipped to MSU within 8 days after collection in styrofoam-insulated boxes via Express Mail (U. S. Postal Service). Upon arrival at MSU, samples were transferred immediately to a lab freezer and kept frozen until analyzed. Notes were also recorded about shipping and arrival date and condition of samples. In most cases, water samples were still frozen upon arrival at MSU. Volunteers were asked to note presence/absence of dreissenid mussels in their lake on a data sheet. Generally speaking, CLMP participants are experienced in the detection of dreissenid presence through participation in other state monitoring programs. We checked volunteer information, where possible, against a database of dreissenid monitoring maintained by Michigan Sea Grant (www.miseagrant.umich. edu/ais/lakes.html). A total of 33 lakes in our survey were listed in Sea Grant’s database. Zebra mussel characterizations by citizen-volunteers matched the database in 30 of 33 cases (91% accuracy), indicating that volunteer characterizations were generally reliable. Of the 3 mismatches, one lake was listed as uninvaded by volunteers but as recently invaded (in
Fig. 1 – Map of Michigan showing surveyed lakes. Solid symbols- invaded by dreissenid mussels, open symbols- not invaded by dreissenid mussels.
144
water research 44 (2010) 141–150
2005) by Sea Grant (Pickerel Lake, Kalkaska County). This lake was scored as invaded in our analyses. Two lakes (Gilletts Lake, Jackson County and Hubbard Lake, Alcona County) were listed as invaded by volunteers but uninvaded by Sea Grant. However, Sea Grant has no record of invasion monitoring for these two lakes since 1997. We assumed that they were invaded at some point between 1997 and 2006. For lakes that were not listed in the Sea Grant database, we assumed that volunteer characterizations were correct. No attempt to distinguish between D. polymorpha and D. bugensis was made.
2.2.
Sample analysis
Chlorophyll a from the euphotic zone and TP were analyzed by MDEQ. Shoreline chlorophyll a and all microcystin analyses were conducted by MSU. Total phosphorus in whole-water samples was measured via automated sulfuric acid-mercuric oxide digestion followed by standard colorimetry for phosphate (U. S. Environmental Protection Agency, 1983). For three lakes, no summer TP data for 2006 was available, so we estimated TP from mean TP in 2004 and 2005 and an empirical relationship between averaged 2004–2005 TP and 2006 TP for all CLMP lakes ( y ¼ 2.0 þ 0.82x, R2 ¼ 0.63, n ¼ 170, p < 0.0001). Chlorophyll a in euphotic-zone samples was measured at MDEQ via extraction of filters in 90% acetone followed by fluorometric measurement of extracted chlorophyll a (APHA, 1998). Chlorophyll a in shoreline samples was measured at MSU via extraction of filters in 95% ethanol followed by fluorometric measurement of extracted chlorophyll a (Welschmeyer, 1994). We found generally good correspondence in chlorophyll a for parallel samples analyzed by MDEQ versus MSU (unpublished data). Microcystin was measured on whole-water samples via Enzyme-Linked ImmunoSorbent Assay (ELISA) using a commercial kit (Envirologix, Inc.) according to manufacturer’s instructions for high sensitivity. The detection limit for microcystin in lake water using our ELISA protocol was w0.02 mg L1. When microcystin concentration was above 1 mg L1, the sample was diluted with deionized water and re-analyzed.
2.3.
Methodological assessments
Data quality for microcystin was assessed via comparisons of data from samples collected by volunteers versus MSU personnel. Samples were either collected, processed and shipped to MSU by volunteers or collected by MSU personnel along side volunteers and brought back to the laboratory for processing on the day of collection. This comparison was conducted for 10 lakes, with five microcystin samples per lake. The MDEQ annually conducts side-by-side sampling with volunteers and performs data assessments for TP and chlorophyll a in samples collected by volunteers as part of the CLMP Quality Assurance Project Plan (Bednarz, 2007). These assessments have shown that data from volunteer-collected samples are highly comparable to data from MDEQ-collected samples (Bednarz and Wandell, unpublished data). We also examined the influence of three procedures on whole-water microcystin concentrations using water samples collected from a eutrophic pond with high microcystin concentrations located on the MSU campus. The procedures
investigated were: storage time in insulated shipping containers, heating of samples to release toxin from cells, and extraction in methanol of dried samples. Given that microcystin is a relatively stable, intracellular toxin (Chorus and Bartram, 1999), we expected that incomplete release of toxin from cells would be the most likely source of bias in our data. A standard protocol for microcystin measurement in water samples calls for freezing and thawing samples 2–3 times before analysis (Graham and Jones, 2007; Harada et al., 1999), but we are aware of no published examination of the efficacy of this technique. Preliminary tests indicated that this technique was not significantly superior to our standard protocol of one freeze-thaw cycle. To examine the effects of storage in insulated shipping containers, we collected water from the MSU pond in midsummer and dispensed replicate aliquots into 35 60-ml sample bottles. The shipping containers and sample bottles used for this test were identical to those used by volunteers during the survey. Each of six shipping containers was packed the same as by volunteers: five frozen sample bottles at the bottom, two frozen ‘‘blue ice’’ surrounding the samples, and styrofoam peanuts on top of the ice packs such that the container was completely full. A small temperature recorder (Hobo Model U10, Onset Computer Corporation) was placed next to the samples within each container. Containers were placed in a room without air conditioning to simulate summer transport conditions. Containers were opened after 1, 2, 3, 4, 7 and 8 days and bottles were removed and immediately frozen. Results were compared to bottles filled and frozen on the day of collection (day 0 samples). To examine the effect of heating on microcystin concentrations, we selected 42 volunteer-collected samples of widely-varying toxin concentrations and simultaneously analyzed replicate aliquots for microcystin with our standard protocol (direct analysis of untreated water) versus immersing the samples in a boiling water bath for 30 min before analysis (Metcalf and Codd, 2000). To examine the effect of methanol extraction on microcystin concentrations, we selected 12 volunteer-collected samples and simultaneously analyzed replicate aliquots for microcystin with our standard protocol versus evaporating 25 ml of the sample to dryness and then extracting the residue in 75% methanol.
2.4.
Data analysis
Relationships among TP, chlorophyll a and microcystin were examined via linear regression after data were normalized by log transformation. The interactive influences of log TP and dreissenid invasion on log microcystin were tested statistically with a least-squares general linear model having invasion status as a categorical variable and log TP as a covariate. To examine the hypothesis that dreissenid invasion only increases microcystin concentrations in low-nutrient lakes, we ran regression-tree analysis (Systat 9.0) with log microcystin as the dependent variable and invasion status and log TP as independent variables. Regression trees subdivide a data set into increasingly homogenous subsets that maximize the reduction in error produced by the partitioning (De’ath and Fabricius, 2000; Knapp and Sarnelle, 2008). This procedure was used as an objective means of defining the ‘‘low nutrient’’
145
water research 44 (2010) 141–150
5
Microcystin (µg L-1)
category of lakes. We estimated error reduction via least squares and used a criterion of 5% for the minimum improvement in the proportion of variance explained to terminate splitting. When the regression tree split the data into subsets based on invasion status, we tested the significance of the difference in microcystin between lake subsets with a t-test.
4 3 2 1 0 0
3.
Results
3.1.
Methodological assessments
1
2
3
4
5
6
7
8
Days in shipping container
Microcystin data from samples collected and handled by volunteers were highly comparable to data from samples collected and handled by MSU personnel (means not significantly different by paired t-test, p > 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 (<5 mg L1) and very low microcystin levels (untransformed mean ¼ 0.05 mg L1) from the remainder of the data set. The second split separated out 5 lakes (2 invaded) with the highest TP ( 26 mg L1) and a high average microcystin (untransformed mean ¼ 0.32 mg L1). The third split divided the remaining 62 lakes based on a TP criterion of 10 mg L1, with the 21 lower-TP lakes (TP < 10 mg L1) being further subdivided into invaded (untransformed mean microcystin ¼ 0.80 mg L1) and uninvaded (untransformed mean microcystin ¼ 0.10 mg L1). The difference in microcystin concentrations between these last two subsets was significant at p < 0.02 (Fig. 8, t-test on logtransformed data, n ¼ 21). In contrast, microcystin concentrations were not statistically different between invaded (untransformed mean ¼ 0.15 mg L1) and uninvaded (untransformed mean ¼ 0.19 mg L1) lakes with TP 10 mg L1 (but < 26 mg L1), despite a larger sample size (Fig. 8, t-test, p > 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 (<15 mg L1). Given that microcystin appears to be the most common cyanobacterial toxin encountered in temperate freshwaters (Boyer, 2007; Chorus and Bartram, 1999), monitoring programs for toxins should take into account the role of dreissenid invasion in elevating microcystin concentrations in low-nutrient lakes. Dreissenid invasion is limited to waters with sufficient alkalinity, so this accounting can be ignored in lakes with pH or calcium concentrations below established thresholds (Ramcharan et al., 1992). We monitored microcystin concentrations at the surface near shore since this is the most likely point of contact for recreational users of lakes (none of the lakes in the survey provide drinking water). We predicted that shoreline concentrations would be higher than concentrations in depth-integrated samples from the euphotic zone because of the buoyancy-regulating abilities of Microcystis and other cyanobacterial taxa potentially capable of producing microcystin. In addition, zebra mussel impacts on Microcystis are likely to be strongest in the littoral zone because contact between the phytoplankton and benthic suspension feeders is greatest in shallow water. Our data clearly support this hypothesis (Fig. 9), with average shoreline concentration (0.62 mg L1) being more than double the average for euphotic-zone samples (0.27 mg L1). In two cases, average shoreline levels were an order of magnitude higher than euphotic-zone levels (Fig. 9), suggesting that measuring the latter may not be the most appropriate sampling strategy for assessing risks from recreational exposure. The close correspondence in microcystin levels between samples collected by volunteers and those collected by professionals (Fig. 2), the efficacy of the method we employed for shipping samples, and our ability to use the data to test hypotheses with high statistical power all suggest that sample collection by citizen volunteers is a valid, valuable and costeffective approach for monitoring microcystin in freshwaters. Given that microcystin was positively related to chlorophyll a in surface-water samples (Fig. 6), monitoring costs might be reduced by routinely determining chlorophyll a at the surface, followed by microcystin analysis on selected samples with high chlorophyll a.
149
The primary limitation of our survey was that we were only able to collect and analyze samples from a single date in each lake. Given that interannual and seasonal variation in lake wide cyanobacterial abundance, and day-to-day variation in scum formation, are likely very high, we suggest that future monitoring included multiple samplings in each lake. A second limitation stems from the fact that we had no control over the lakes selected for sampling. This resulted in there being an unintended difference in TP between invaded and uninvaded lakes (Table 1). To account for this potential source of bias, we used regression-tree analysis to compare microcystin in invaded and uninvaded lakes having the same TP (Fig. 7). Given that TP had an overall positive influence on microcystin, lower TP in invaded lakes was a conservative source of bias with respect to the general conclusion that dreissenid invasion leads to increased microcystin.
5.
Conclusions
1. The positive influence of eutrophication (as indexed by TP) on microcystin concentrations appears to be stronger in lakes that lack dreissenid mussels (uninvaded) relative to lakes in which dreissenids have become established (invaded). A positive relationship between log TP and log microcystin was found for uninvaded lakes but not invaded lakes. 2. For lakes with TP between 5 and 10 mg L1, euphotic-zone microcystin concentrations were on average eight times higher in invaded lakes than uninvaded lakes. The highest levels of microcystin in near shore surface waters were found in invaded lakes with low TP. 3. Across all 75 lakes, microcystin concentrations in surfacewater samples were on average more than double concentrations in euphotic-zone samples. 4. Sample collection by citizen volunteers is an effective method for monitoring microcystin.
Acknowledgements Funding for this research was provided by a grant from the Michigan Department of Environmental Quality, Clean Water Fund, Clean Michigan Initiative.
Appendix. Supplementary information Supplementary data associated with this article can be found, in the online version, at doi: 10.1016/j.watres.2009.09.014.
references
American Public Health Association, 1998. Standard Methods for the Examination of Water and Wastewater, 20 ed. American Public Health Association, American Water Works Association, Water Environment Federation, Washington, D. C.
150
water research 44 (2010) 141–150
Bednarz, R.L., 2007. Cooperative Lakes Monitoring Program Quality Assurance Project Plan. Michigan Department of Environmental Quality, Lansing, Michigan. Boyer, G.L., 2007. The occurrence of cyanobacterial toxins in New York lakes: lessons from the MERHAB-Lower Great Lakes program. Lake and Reservoir Management 23, 153–160. Bruhn, L.C., Soranno, P.A., 2005. Long term (1974–2001) volunteer monitoring of water clarity trends in Michigan Lakes and their relation to ecoregion and land use/cover. Lake and Reservoir Management 21, 10–23. Canfield, D.E., Brown, C.D., Bachmann, R.W., Hoyer, M.V., 2002. Volunteer lake monitoring: testing the reliability of data collected by the Florida LAKEWATCH program. Lake and Reservoir Management 18, 1–9. Chorus, I., Bartram, J. (Eds.), 1999. Toxic Cyanobacteria in Water: A Guide to Their Public Health Consequences, Monitoring and Management. E & FN Spon, London, UK. De’ath, G., Fabricius, K.E., 2000. Classification and regression trees: a powerful yet simple technique for ecological data analysis. Ecology 81, 3178–3192. Downing, J.A., Watson, S.B., McCauley, E., 2001. Predicting cyanobacteria dominance in lakes. Canadian Journal of Fisheries and Aquatic Sciences 58, 1905–1908. Giani, A., Bird, D.F., Prairie, Y.T., Lawrence, J.F., 2005. Empirical study of cyanobacterial toxicity along a trophic gradient of lakes. Canadian Journal of Fisheries and Aquatic Sciences 62, 2100–2109. Graham, J.L., Jones, J.R., 2007. Microcystin concentrations in physical size class separations of natural phytoplankton communities. Lake and Reservoir Management 23, 161–168. Graham, J.L., Jones, J.R., Jones, S.B., Downing, J.A., Clevenger, T.E., 2004. Environmental factors influencing microcystin distribution and concentration in the Midwestern United States. Water Research 38, 4395–4404. Harada, K., Kondo, F., Lawton, L., 1999. Laboratory analysis of cyanotoxins. Pages 369–405. In: Chorus, I., Bartram, J. (Eds.), Toxic Cyanobacteria in Water: A Guide to Their Public Health Consequences, Monitoring and Management. E & FN Spon, London, UK. Idrisi, N., Mills, E.L., Rudstam, L.G., Stewart, D.J., 2001. Impact of zebra mussels (Dreissena polymorpha) on the pelagic lower trophic levels of Oneida Lake, New York. Canadian Journal of Fisheries and Aquatic Sciences 58, 1430–1441. Jacoby, J.M., Kann, J., 2007. The occurrence and response to toxic cyanobacteria in the Pacific Northwest, North America. Lake and Reservoir Management 23, 123–143. Jensen, J.P., Jeppesen, E., Olrik, K., Kristensen, P., 1994. Impact of nutrients and physical factors on the shift from cyanobacterial to chlorophyte dominance in shallow Danish lakes. Canadian Journal of Fisheries and Aquatic Sciences 51, 1692–1699. Johnston, B.R., Jacoby, J.M., 2003. Cyanobacterial toxicity and migration in a mesotrophic lake in western Washington, USA. Hydrobiologia 495, 79–91. Kalff, J., 2002. Limnology: Inland Water Ecosystems. Prentice-Hall. Knapp, R.A., Sarnelle, O., 2008. Recovery after local extinction: factors affecting re-establishment of alpine lake zooplankton after removal of nonnative fish. Ecological Applications 18, 1850–1859. Knoll, L.B., Coauthors, 2008. Invasive zebra mussels (Dreissena polymorpha) increase cyanobacterial toxin concentrations in low-nutrient lakes. Canadian Journal of Fisheries and Aquatic Sciences 65, 448–455.
Koenings, J.P., Edmundson, J.A., 1991. Secchi disk and photometer estimates of light regimes in Alaskan lakes – effects of yellow color and turbidity. Limnology and Oceanography 36, 91–105. Kotak, B.G., Lam, A.K.-Y., Prepas, E.E., Hrudey, S.E., 2000. Role of chemical and physical variables in regulating microcystin-LR concentration in phytoplankton of eutrophic lakes. Canadian Journal of Fisheries and Aquatic Sciences 57, 1584–1593. Kotak, B.G., Zurawell, R.W., 2007. Cyanobacterial toxins in Canadian freshwaters: a review. Lake and Reservoir Management 23, 109–122. Metcalf, J.S., Codd, G.A., 2000. Microwave oven and boiling waterbath extraction of hepatotoxins from cyanobacterial cells. FEMS Microbiology Letters 184, 241–246. Nicholls, K.H., Heintsch, L., Carney, E., 2002. Univariate step-trend and multivariate assessments of the apparent effects of P loading reductions and zebra mussels on the phytoplankton of the Bay of Quinte, Lake Ontario. Journal of Great Lakes Research 28, 15–31. Obrecht, D.V., Milanick, M., Perkins, B.D., Ready, D., Jones, J.R., 1998. Evaluation of data generated from lake samples collected by volunteers. Lake and Reservoir Management 14, 21–27. Raikow, D.E., Sarnelle, O., Wilson, A.E., Hamilton, S.K., 2004. Dominance of the noxious cyanobacterium Microcystis aeruginosa in low nutrient lakes is associated with exotic zebra mussels. Limnology and Oceanography 49, 482–487. Ramcharan, C., Padilla, D.K., Dodson, S.I., 1992. Models to predict the potential occurrence and density of the zebra mussel, Dreissena polymorpha. Canadian Journal of Fisheries and Aquatic Sciences 49, 2611–2620. Reynolds, C.S., 1984. The Ecology of Freshwater Phytoplankton. Cambridge University, Cambridge. Sarnelle, O., 1993. Herbivore effects on phytoplankton succession in a eutrophic lake. Ecological Monographs 63, 129–149. Sarnelle, O., Wilson, A.E., Hamilton, S.K., Knoll, L.B., Raikow, D.E., 2005. Complex interactions between the zebra mussel, Dreissena polymorpha, and the harmful phytoplankter, Microcystis aeruginosa. Limnology and Oceanography 50, 896–904. Smith, T.E., Stevenson, R.J., Caraco, N.F., Cole, J.J., 1998. Changes in phytoplankton community structure during the zebra mussel (Dreissena polymorpha) invasion of the Hudson River (New York). Journal of Plankton Research 20, 1567–1579. Trimbee, A.M., Prepas, E.E., 1987. Evaluation of total phosphorus as a predictor of relative biomass of blue-green algae with an emphasis on Alberta lakes. Canadian Journal of Fisheries and Aquatic Sciences 44, 1337–1342. U.S. Environmental Protection Agency. Methods for chemical analysis of water and wastes. U.S. Environmental Protection Agency, Cincinnati, Ohio. Vanderploeg, H.A., Coauthors, 2001. Zebra mussel (Dreissena polymorpha) selective filtration promoted toxic Microcystis blooms in Saginaw Bay (Lake Huron) and Lake Erie. Canadian Journal of Fisheries and Aquatic Sciences 58, 1208–1221. Watson, S.B., McCauley, E., Downing, J.A., 1997. Patterns in phytoplankton taxonomic composition across temperate lakes of differing nutrient status. Limnology and Oceanography 42, 487–495. Welschmeyer, N.A., 1994. Fluorometric analysis of chlorophyll a in the presence of chlorophyll b and phaeopigments. Limnology and Oceanography 39, 1985–1992.
water research 44 (2010) 151–158
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Biotreatment and bioassessment of heavy metal removal by sulphate reducing bacteria in fixed bed reactors C. Cruz Viggi a,*, F. Pagnanelli a, A. Cibati a, D. Uccelletti b, C. Palleschi b, L. Toro a a b
Department of Chemistry, Sapienza University of Rome, P.le Aldo Moro 5, 00185 Rome, Italy Department of Developmental and Cell Biology, Sapienza University of Rome, P.le A. Moro 5, 00185 Rome, Italy
article info
abstract
Article history:
In this work a batch-optimised mixture (w/w %: 6% leaves, 9% compost, 3% Fe(0), 30% silica
Received 3 June 2009
sand, 30% perlite, 22% limestone) was investigated in a continuous fixed bed column
Received in revised form
reactor for the treatment of synthetic acid-mine drainage (AMD). A column reactor was
3 September 2009
inoculated with sulphate-reducing bacteria and fed with a solution containing sulphate
Accepted 4 September 2009
and heavy metals (As(V), Cd, Cr(VI), Cu and Zn). At steady state, sulphate abatement was
Published online 10 September 2009
50 10%, while metals were totally removed. A degradation rate constant (k) of 0.015 0.001 h1 for sulphate removal was determined from column data by assuming
Keywords:
a first order degradation rate. Reduction of AMD toxicity was assessed by using the
Heavy metals
nematode Caenorhabditis elegans as a test organism. A lethality assay was performed with
Sulphate reducing bacteria
the toxicants before and after the treatment, showing that only 5% of the animals were still
Fixed bed reactors
alive after 48 h in presence of the contaminants, while the percentage increased to 73%
Biotreatment
when the nematodes were exposed to the solution eluted from the column.
Bioassessment
ª 2009 Elsevier Ltd. All rights reserved.
Caenorhabditis elegans
1.
Introduction
Anthropogenic release of heavy metals in the environment is mainly related to wastewaters discharged from industrial and mining activities. In particular acid-mine drainage (AMD) is one of the worst environmental problems caused by the natural oxidation of sulphide minerals under the combined action of oxygen and water (Costa et al., 2008; Neculita and Zagury, 2008; Pagnanelli et al., 2008a). Addition of alkaline agents is generally used to promote metal precipitation in AMD (Johnson and Hallberg, 2005; Akcil and Koldas, 2006). This treatment is expensive, produces large amounts of waste sludges and effluents with concentrations of sulphates and metals that do not meet environmental standards (Akcil and Koldas, 2006).
The precipitation of metals with hydrogen sulphide produced by sulphate reducing bacteria (SRB) has been proposed as an alternative process (Foucher et al., 2001) using off-line sulphidogenic bioreactors (Gonc¸alves et al., 2007) and permeable reactive barriers (PRB) (Waybrant et al., 1998; Benner et al., 2002; Komnitsas et al., 2003; Bartzas et al., 2005; Komnitsas et al., 2006; Costa et al., 2008). Mixtures used in biological PRB are generally a mix of organic materials to sustain SRB growth and activity, gravel to improve barrier permeability, and limestone to increase pH (Table 1). PRB design requires the complete hydrogeological characterisation of the site and laboratory tests for evaluating the degradation rate of contaminants and determining the barrier thickness (Gavaskar, 1999).
* Corresponding author. Fax: þ39 06 490631. E-mail addresses:
[email protected] (C. Cruz Viggi),
[email protected] (F. Pagnanelli), alessio.cibati@ uniroma1.it (A. Cibati),
[email protected] (D. Uccelletti),
[email protected] (C. Palleschi),
[email protected] (L. Toro). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.013
152
water research 44 (2010) 151–158
Table 1 – Reactive mixtures used in lab-scale studies and in full-scale permeable reactive barriers for treatment of AMD and heavy metal’s contaminated wastewater. Abatement
SO24 reduction rate (mg L1 d1)
Reference
Batch
SO24 w 100% Fe 99% Ni w 100% Cd 99%
22.4–157.5
Waybrant et al. (1998)
Batch
SO24 97% Ni 72 % Zn 88%
35.6–156.2
Cocos et al. (2002)
Column
SO24 20-60%
48–76.8
Column
–
Gibert et al. (2003)
Column
Fe 99% Zn 55% Cd 80% Cu 97% SO24 18-27%
16.3–42.2
Gibert et al., (2004)
PRB Ontario (Canada) 1995
SO24 60% Fe 85%
10.6–15.3
Benner et al. (2002)
PRB Aznalco`llar (Spagna) 1998
Metals 90%
–
Carrera et al. (2001)
PRB Vancouver (Canada) 2000 PRB Northumberland (UK) 2003 Batch
Metals 80%
4.6
Ludwig et al. (2002)
Column
SO24 50% Metals 99%
Composition Municipal compost Sawdust Manure Cellulose Sediments with SRB Silica sand Limestone Wood chips 3% Composted leaves 30% Chicken manure 20% Silica sand 5% Sediments with SRB 37% Limestone 2% Urea 3% Pyrite Silica sand Leaves Chips Sawdust Biological sludges Sediments with SRB Limestone 50% Compost 45% Sediments with SRB 5% Compost limestone Sheep manure Gravel 50% Municipal compost 20% Soil 20% Wood chips 9% Limestone 1% Module 1 Limestone 50% Municipal compost 30% Sludges 20% Module 2 Limestone 50% Municipal compost 50% Module 3 Iron (0) 1% Limestone 66% Municipal compost 33% Composted leaves 15% Gravel 84% Limestone 1% Manure and straw 25% Municipal compost 25% Limestone 50% 6% leaves, 9% compost, 3% Fe(0), 30% silica sand, 30% perlite, 22% limestone. 6% leaves, 9% compost, 3% Fe(0), 30% silica sand, 30% perlite, 22% limestone.
Type
Laboratory tests consist of a preliminary selection in batch reactors of the optimal mixture for PRB filling, followed by a kinetic characterisation in fixed bed columns operating in fluidodynamic conditions similar to those existing in real PRBs. The advantages of installing PRB should be evaluated also considering the attenuation of AMD toxicity after the
SO24 67% Fe 95% Al 87% SO24 83%
–
Waybrant et al. (2002)
Jarvis et al. (2006)
590
Present study
1550
Present study
treatment. Toxicity tests on column effluents (as simulation of full scale permeable reactive barriers) can give preliminary information about the potential impact of treated effluents in the environment (Pagnanelli et al., 2008b). Bioassessment of AMD toxicity was investigated using different test-organism systems such as microorganisms (Gray and O’Neill, 1997), daphnids (Kim et al., 2006; Lopes et al., 2006;
water research 44 (2010) 151–158
Yim et al., 2006), shrimp, fish (Janssens de Bisthoven et al., 2006) and plants (Arambasic et al., 1995; Cole et al., 2001). Nemathodes can be also used for ecological risk assessment in soil (Freeman et al., 1999; Peredney and Williams, 2000) and water (Freeman et al., 1998; Ura et al., 2002). Among nemathodes Caenorhabditis elegans (C. elegans), a free-living nematode abundant in soil ecosystems (Sohlenius, 1980), is an optimal candidate for biomonitoring because of its short lifespan (approximately 3 weeks at 16 C under optimal conditions) and ease of manipulation. In addition, it is a simple multicellular eukaryote, whose developmental process and behaviour can easily be monitored under the microscope. This work aimed to investigate the performances of a batchoptimised mixture (Pagnanelli et al., 2009) in continuous fixed bed column reactor inoculated with SRB. In particular column data well resembling fluidodynamic conditions inside real PRBs will allow to validate the treatment efficacy and obtain kinetic parameters for the preliminary estimate of barrier thickness. Novelty aspects are the comparison of equilibrium and kinetic characteristics of sulphate and metal abatement obtained in batch and column reactors, the isolation of bioactive mechanism strictly related to SRB activity from the contribution of sorbing properties of organic materials used for column filling, and the bioassessment of the column treatment by ecotoxicological tests using C. elegans.
2.
Materials and methods
2.1.
Sulphate reducing bacteria (SRB)
SRB inoculum was kindly furnished by the research group of Professor Groudev (Department of Engineering Geoecology, University of Mining and Geology, Sofia, Bulgaria), who collected it in the Curilo mine district located near Sophia (Groudev et al., 2001). This area is located near the uranium deposit ‘‘C’’ which has been contaminated with radioactive elements (uranium, radium, thorium) and toxic heavy metals (copper, zinc, cadmium).
2.2.
Biomass cultivation and maintenance
Inoculated bacteria were cultivated in closed flasks (previously filled with N2) using standard procedures for SRB reported in the literature (Postgate, 1984). C Medium, was used for bacterial growth and acclimatization: KH2PO4 0.5 g L1; NH4Cl 1 g L1; Na2SO4 4.5 g L1; CaCl2*6H2O 0.06 g L1; MgSO4*7H2O 0.06 g L1; sodium lactate 6 g L1; yeast extract 1 g L1; FeSO4*7H2O 0.004 g L1; sodium citrate*2H2O 0.3 g L1. Usually 20 mL of bacteria inoculum were added to 80 mL of C Medium (inoculum volume was 20% of total volume). All flasks were incubated at 37 C under shaking conditions.
2.3.
Reactive mixture
Leaf samples were collected in October/November 2006 at the base of oak trees (major percentage), walnut trees, red maples, cherry trees, pear trees and horse chestnuts. Leaf dimensions were reduced to an average diameter of 1.8 mm by manual grinding and stored dry in a stove at 80 C.
153
Limestone (average diameter of dried particles was 5 mm) and silica sand (average diameter of dried particles 0.77 mm) samples were taken from Italian gravel pits. Commercial samples were used for compost (universal fertilized soil for gardening, Verdemix-CERMEC compost), perlite (Perlite expanded Agri 30; Isoperl) and zero-valent iron (Connelly GPM-Iron aggregate ETI CC-1004; Connelly, Chicago, Illinois). The solid reactive mixture had the following composition (w/w %): 6% leaves, 9% compost, 3% Fe(0), 30% silica sand, 30% perlite, 22% limestone.
2.4.
Fixed bed reactor
Column tests were performed in a fixed bed reactor (height 1 m; diameter 0.2 m; volume, V ¼ 6.65 103 m3) made of Plexiglas with 10 equally distant outputs along the axial length, numbered from the bottom to the top of the column. The column was packed with perlite and silica sand on the bottom (10 cm length) followed by reactive mixture (80 cm) (paragraph 2.3) and topped with perlite and silica sand (10 cm) (column pore volume V0 ¼ 1.5 L). SRB were inoculated in the core of the column (outputs 4, 5 and 6). The column was continuously fed (F ¼ 0.5 mL min1) with a solution containing sulphate (31 mM) and heavy metals (Cd 0.1 mM, Cr(VI) 0.1 mM, Cu 0.1 mM, Zn 0.1 mM and As(V) 27 mM). The pH and Eh values of the influent solution were 5.5 0.1 and 200 10 mV, respectively. Samples from five different outputs (1, 4, 5, 6 and 9) were analyzed for pH, Eh and residual amounts of sulphates and metals. Column experiments were run for over 18 months.
2.5.
Sorption isotherms
Sorption isotherms of As, Cd, Cr(VI), Cu and Zn onto compost and leaves were determined by batch equilibrium experiments. Tests were performed by using suspensions of compost (0.4 g) and leaves (0.2 g) in 40 mL (vi) of metal solution with different initial metal concentrations (0.01–1 mM). Metalbearing suspensions were kept under magnetic stirring until the equilibrium conditions were reached after 2 h. pH was kept constant at 6 by HNO3 and/or NaOH additions. Solidliquid separation was performed by centrifugation (5 min at 4000 rpm) and the equilibrium metal concentration in liquid phase (cf, mmol L1) was determined by an inductively coupled plasma spectrophotometer (Varian Vista-MPX CCD Simultaneous ICP-OES). For each condition a blank test without biomass was also performed to determine the initial metal concentration (ci, mmol L1). Metal concentration in solid phase (q) was then determined by the material balance q¼
ci vi cf vf m
(1)
where vf (L) is the final suspension volume and m (g) is the sorbent mass.
2.6.
Analytical determination of sulphates
Sulphate concentrations were determined by a turbidimetric method: 2.5 mL of sample to be analyzed were placed in a testtube and mixed with 500 mL of a solution of glycerine and
154
water research 44 (2010) 151–158
250 mL of a solution of sodium chloride. Glycerine solution was prepared mixing pure glycerine and distilled water (1:1). NaCl solution was prepared mixing 200 g of NaCl þ 40 mL of concentrated HCl diluted to 1 L by distilled water. After the addition of glycerine and NaCl solutions, each sample was added for 300 ml of a solution of BaCl2 (prepared with 90 g of BaCl2*2H2O in 1 L of distilled water). After two minute shaking, samples were analyzed by spectrophotometer at 390 nm. Instrument calibration was performed by standard solutions of anhydrous Na2SO4 (dried for 1 h in an oven at 110 C).
2.7.
Preparation of nematode growth medium plates
The C. elegans strain SEK-1 (kindly provided by the C. elegans Genetics Center) was used for the experiment and maintained on nematode growth medium (NGM) supplemented with Escherichia coli (E. coli) strain OP50 (Brenner, 1974). The NGM plates containing the contaminants before and after the column treatment were prepared under sterile conditions and serially diluted or not in sterile distilled water as indicated.
2.8.
Determination of lifespan
For determination of the lifespan, approximately 45 animals raised to the L4 stage were transferred into four fresh NGM plates (15 animals per plate) with or without the contaminants before and after the column and grown at 16 C. The concentrations of contaminants, before and after the column treatment, in the plates were as indicated in the text. After placement on test plates, the viability of the adult animals was checked under a stereomicroscope every day. When the nematodes stopped moving, they were checked for viability by tapping the plates and by gently touching them with a platinum picker. The tested nematodes were not transferred into new plates during the experiment, because frequent transfer of animals clearly shortened their lifespan (data not shown). It was also noted that the amount of the OP50 food source in the test plates was sufficient for maintaining 15 adult animals. The experiments were independently repeated at least three times.
2.9.
Statistical analysis for the lifespans
The software package JMP IN5.1.2 J (SAS Institute Inc.) was used for statistical analyses. The means and standard errors of the lifespans for animals in each group were calculated and the significant differences in the lifespans in each nematode population was analyzed by the Kaplan-Meier method.
3.
Results and discussion
3.1.
Column experiments
The reactive mixture used for column filling was chosen according to preliminary batch tests performed by varying the nature and composition of organic sources (compost, composted sheep manure, olive pomace, and leaves) and the amounts of limestone (used to ensure neutral pH values), silica sand and perlite (added to increase column permeability), and zero valent iron (which consume oxygen
dissolved in water, and can generate hydrogen used as electron donor by SRB) (Pagnanelli et al., 2009). These preliminary batch tests showed that the selected solid mixture (w/w %: 6% leaves, 9% compost, 3% Fe(0), 30% silica sand, 30% perlite, 22% limestone) was the best among the tested ones in terms of sulphate abatement (83 3% in 22 days) and optimal growth conditions (pH: 7.8 0.1, Eh: 410 5 mV) (Pagnanelli et al., 2009). The efficiency of this reactive mixture as column filling for SRB growth in column reactor was monitored during time for pH, Eh and residual amounts of sulphates and metals in five different outputs of the column, representing the low (output 1) and the upper (output 9) zones filled with inorganic components and the reactive central core (outputs 4, 5 and 6) filled with the solid reactive mixture and inoculated with SRB. The volume of treated effluent is expressed as bed volume BV ¼ V/ V0, where V is the treated volume and V0 is the pore volume of the column. Bacterial growth in the central part of the column was clearly evidenced by the gradual change of colour of the filling due to the formation of a black precipitate of FeS. The effects of SRB activity were also evident by observing Eh and pH values along the column length (Table 2). In fact Eh passed from oxidant values (output 1) to reducing ones (outputs 4, 5, 6 and 9) due to SRB activity. As for pH profiles (Table 2), it is evident the role of inorganic components (perlite and silica sand) in raising pH from 5.5 (feed) to about 7, thus creating favourable conditions for SRB growth. Alkalinity generation due to SRB activity determined a further increase of pH from 7 (output 1) to about 8.5 (inoculated zone, outputs 4, 5, 6 and 9). Steady state was reached for sulphate abatement after the treatment of about 20 bed volumes with a final average abatement of 50 10%, which is comparable to what is reported in the literature for similar column apparatus (Table 1). On the other side 83% sulphate removal was observed in batch reactor experiments. The difference observed between batch and column performances could be due to the different fluidodynamic conditions of the two reactor configurations. In particular in batch reactors mixing conditions determined a mechanical degradation of the solid components which probably favours both sorption phenomena and SRB activity due to an increased release of organics in solution. This showed how important is performing tests in columns in order to obtain realistic estimates of treatment performance. As for metals the complete removal was obtained for all investigated metals as evidenced by the average abatements reported in Table 2 (calculated for Bv from 20 to 100). The profiles of average abatement along the column length were reported in Fig. 1 for sulphate and metals. Steady state profile of sulphate abatement evidenced the effect of SRB activity in the central core of the bed reactor, and the capability of the system in sustaining an active mechanism of removal even for long treatment time (about 8 months) without extra carbon source than the biological matrices in the reactive mixture. As for metals a complete abatement was observed even in the central core of the column. The observed profiles did not allow to distinguish different phenomena occurring in metal removal such as bioprecipitation and biosorption. Nevertheless
155
water research 44 (2010) 151–158
Table 2 – Average values of sulphate and metals abatement (%), pH and Eh of the different column outputs (calculated for Bv from 20 to about 100). Output
Eh (mV)
pH
1 4 5 6 9
6.7 0.4 8.2 0.6 8.4 0.3 8.4 0.3 8.8 0.2
Abatement (%)
230 50 310 50 310 50 320 50 320 50
SO2 4
As
Cd
Cr(VI)
Cu
Zn
52 30 5 40 5 40 5 50 10
50 20 98 3 99 2 99 2 99 1
30 20 97 2 99 1 99 1 99 1
20 10 98 2 99 1 99 1 99 1
70 20 96 4 97 3 98 2 98 2
30 20 97 3 98 2 99 1 99 1
previous tests of cadmium removal by SRB growing on the reactive mixture in batch reactors evidenced a predominant contribution of sorption processes with respect to bioreduction (94% against 6%) (Pagnanelli et al., 2009). In order to isolate the contributions of bioprecipitation and adsorption in heavy metals removal, the adsorption capacities of the filling materials were determined. A quantitative evaluation of maximum uptake capacity for the different metals can be obtained by regressing adsorption data (q versus Cf) using the Langmuir model: q¼
qmax KCf 1 þ KCf
(2)
where q (mmol g1) and Cf (mmol l1) are the equilibrium metal concentrations in the solid and liquid phase, and qmax (mmol g1) and K (l mmol1) are the adjustable parameters related to the maximum specific metal uptake and the metal affinity constants (Table 3). Experimental data of sorption isotherms are reported as supplemental material. qmax values were used in order to calculate the maximum sorption capacity of the column filling material, by assuming that biological matrices gave a predominant contribution on the sorbing performance of the mixture (Pagnanelli et al., 2009). The maximum sorption capacities of the column towards the different heavy metals were compared with the amount of metals removed in the column after the treatment of 100 bed volumes (Table 4), showing that, after 8 month treatment, the adsorption capacity of the column was saturated for Cd and Zn, but not for As, Cr(VI) and Cu. By this comparison the significant sorbing capacities of organic matrices generally used as PRB filling were confirmed. In addition it was evidenced that, for treatment times exceeding the estimated breakthrough time, active metal removal still occurs.
3.2.
Kinetic analysis
Experimental data from batch and column reactors were used in order to determine sulphate removal rates in different fluidodynamic conditions. Under closed anoxic systems, rates of sulphate bioreduction can be determined by sulphate-removal rates (Waybrant et al., 1998). These rates are considered approximate because other processes, such as adsorption of sulphate to mixture components, biosorption of sulphate to biomass surface and acclimation periods can also affect the calculated rates of sulphate-reduction. Experimental data were modelled assuming a first-order degradation rate for sulphate removal: dCt ¼ kCt dt
(3)
Therefore, Ct ¼ C0 ekt
(4)
Or, (5)
lnðCt =C0 Þ ¼ kt
where C is the concentration of sulphate, k the reaction rate constant and C0 the initial concentration of sulphate. Then, ln(C/C0) was plotted against time, and the slope of the regression line represented the reaction rate constant k (supplemental material). A degradation rate constant of 0.0030 0.0005 h1 (R2 ¼ 0.9514) was obtained for batch tests with solid substrates. The same kinetic model was used for column data reporting the effluent concentration along the column length versus the residence time of each output. The residence times (s) were obtained assuming plug-flow conditions by the equation:
Sulphate
Abatement (%)
100
Cd
90 80
Cr(VI)
70
Zn
Cu
60 50
As
Table 3 – Adjustable parameters of the Langmuir isotherm for leaves and compost sorption tests. Leaves
40 30
Compost
qmax K qmax K (mmoli g1) (L mmol 1) (mmoli g1) (L mmol 1)
20 10 0 1
2
3
4
5
6
7
8
9
Column outputs
Fig. 1 – Sulphate and metals profiles along the column length.
As 0.040 0.004 Cd 0.0148 0.0008 Cr(VI) 0.16 0.06 Cu 0.13 0.05 Zn 0.024 0.005
62 70 20 32 42 42
0.0086 0.0005 0.0171 0.0007 0.062 0.005 0.032 0.003 0.0190 0.0004
11 2 18 3 2 0.5 10 3 11 1
156
water research 44 (2010) 151–158
Table 4 – Comparison between the amount of metals removed in the column (mmol) and the maximum sorption capacity of the system (mmol). Total metal removed after 100 BV
Maximum sorption capacity
5.4 7.6 23.8 6.9 11.6
6.9 5.2 32.9 23.1 6.8
As Cd Cr(VI) Cu Zn
s¼
LAe F
(6)
where F is the volumetric flow rate, L is the linear distance from the feed input and the effluent output, A is the crosssectional area of the column and e is the porosity. For sulphate, a degradation rate constant (k) of 0.015 0.001 h1 (R2 ¼ 0.9347) was obtained (supplemental material), denoting a significant improvement of SRB activity working in fluidodynamic conditions similar to those existing in real PRB.
On the base of these degradation rate constants, the sulphate removal rate (mg L1 d1) was calculated and compared with literature data (Table 1). For batch tests a sulphate-removal rate of 590 20 mg L1 d1 was obtained, much higher than other literature results obtained using similar solid mixtures (Table 1). Even for the column, a sulphate reduction rate of 1550 50 mg L1 d1 was obtained, significantly higher than those reported for continuous systems filled with similar complex organic substrates (Table 1). The kinetic analysis was also performed for Cd and Zn, being the adsorption capacity of column saturated towards them. Degradation rate constants (k) for Cd and Zn were 0.16 0.02 h1 and 0.11 0.01 h1, respectively, one order magnitude larger than for sulphate abatement. Residence times (s) required to reduce concentration below the specific the maximum contaminant level (MCL) (250 mg L1 for sulphate, 0.005 mg L1 for Cd and 3 mg L1 for Zn, according to Italian regulation) were calculated for the different contaminants: ssulphate ¼ 8 d, sCd ¼ 2.2 d and sZn ¼ 0.52 d. It is evident that, for the synthetic AMD here considered, the barrier thickness must be designed using the sulphate residence time, that is the highest, among the calculated ones.
3.3.
% Live Animals
A 100 1 100
80
1 10 not dil
60 40 20 0 0
24
48
Time (h)
% Live Animals
B 100 1 100
80
1 10 not dil
60 40 20 0 0
24
48
Bioassessment with C. elegans
To further support the effectiveness of the column in reducing the toxicants concentration from a biological point of view, the nematode C. elegans was used as a test organism. Firstly the sensitivity of the nematodes to the single heavy metals included in the influent was analyzed, utilizing a wild type strain and a mutant strain defective in the P38 signalling cascade, SEK-1 (Kim et al., 2002). It was found that SEK-1 mutant was more sensitive to all the metals analyzed respect to the wild type already after 24 h from the exposure (Data not reported). The sensitivity to heavy metals of SEK-1 was in agreement with recent data reporting that cadmium exposure affects the expression of several genes previously implicated in the nematode immune response (Cui et al., 2007). A lethality assay was then performed with different dilutions of the toxicants before and after the treatment utilizing the C. elegans SEK-1 strain. As shown in Fig. 2 only 5% of the animals were still live after 48 h in presence of the contaminants at the concentration used for the column, while the percentage increased up to 30% when the animals were exposed at the same contaminant solution but diluted 1:10 (panel A). When the toxicants were eluted from the column and assayed at the same dilutions as before, a significant higher percentage of living animals after 48 h was observed: 67% and 73% respectively (panel B).
Time (d) Fig. 2 – Mean survivors of C. elegans SEK-1 animals grown on different dilutions of the contaminants at 16 8C. (A) Lifespan of animals in presence of the contaminant before the column treatment; (B) Lifespan of animals in presence of the contaminant after the column treatment. Animals seeded on plates containing the concentrated contaminants are in white bars, in the 1:10 diluted contaminants in grey bars, in the 1:100 diluted contaminants in black bars.
4.
Conclusions
Experimental data of sulphate bioreduction and metal bioprecipitation obtained in batch and column reactors inoculated with sulphate reducing bacteria denoted the effect of the different fluidodynamic conditions on pollutants removal. In particular steady state sulphate removal in column reactor was about 50% against 83% in batch reactors. Nevertheless preliminary kinetic analysis denoted significant
water research 44 (2010) 151–158
improvement of sulphate degradation rates in column reactor working in fluidodynamic conditions similar to those existing in full-scale PRB (a preliminary estimate of 0.8 m barrier thickness was obtained by assuming a groundwater velocity of 0.1 m/d (Mayer et al., 2006). Independent biosorption tests confirmed the significant contribution in metal removal of sorption onto the organic components of the column filling, but also denoted the presence of an active removal process due to the presence of bacteria once sorption capacity have reached saturation. Finally bioassessment of column performances by C. elegans denoted that bioavailable toxicity was substantially removed by the experimented treatment.
Appendix Supplementary information Supplementary data associated with this article can be found, in the online version, at doi: 10.1016/j.watres.2009.09.013.
references
Akcil, A., Koldas, S., 2006. Acid mine drainage (AMD): causes, treatment and case studies. Journal of Cleaner Production 14, 1139–1145. Arambasic, M.B., Bjelic, S., Subakov, G., 1995. Acute toxicity of heavy metals (copper, lead, zinc), phenol and sodium on Allium cepa l., Lepidium sativum l. and Daphnia magna St.: comparative investigations and the practical applications. Water Research 29, 497–503. Bartzas, G., Komnitsas, K., Paspaliaris, I., 2005. Laboratory evaluation of Fe0 barriers to treat acidic leachates. Minerals Engineering 19 (5), 505–514. Benner, S.G., Blowes, D.W., Ptacek, C.J., Mayer, K.U., 2002. Rates of sulfate reduction and metal sulfide precipitation in a permeable reactive barrier. Applied Geochemistry 17, 301–320. Brenner, S., 1974. The genetics of Caenorhabditis elegans. Genetics 77, 71–94. Carrera, J., Alcolea, A., Bolzicco, J., Knudby, C., Ayora, C., 2001. An experimental geochemical barrier at Aznalcollar. In: Thornton, Oswald (Ed.), Proc. 3rd Internat. Conf. Groundwater Quality Sheffield UK, 18–21 June, pp. 407–409. Cole, M.B., Arnold, D.E., Watten, B.J., 2001. Physiological and behavioural responses of stonefly nymphs to enhanced limestone treatment of acid mine drainage. Water Research 35, 625–632. Cocos, I.A., Zagury, G.J., Clement, B., Samson, R., 2002. Multiple factor design for reactive mixture selection for use in reactive walls in mine drainage treatment. Water Research 32, 167–177. Costa, M.C., Martins, M., Jesus, C., Duarte, J.C., 2008. Treatment of acid mine drainage by sulphate-reducing bacteria using low cost matrices. Water Air Soil Pollution 189, 149–162. Cui, Y., McBride, S.J., Boyd, W.A., Alper, S., Freedman, J.H., 2007. Toxicogenomic analysis of Caenorhabditis elegans reveals novel genes and pathways involved in the resistance to cadmium toxicity. Genome Biology 8, R122. Foucher, S., Battaglia-Brunet, F., Ignatiadis, I., Morinet, D., 2001. Treatment by sulfate-reducing bacteria of chessy acid-mine drainage and metals recovery. Chemical Engineering Science 56, 1639–1645.
157
Freeman, M.N., Marse, T.J., Williams, P.L., 1998. Establishment of quality assurance procedures for aquatic toxicity testing with the nematode Caenorhabditis elegans. In: Little, E.E., Delonay, A. J., Greenberg, B.M. (Eds.), Environmental Toxicology and Risk Assessment, vol. 7. American Society for Testing and Materials, Philadelphia, pp. 45–59. vol. ASTM STP 1333. Freeman, M.N., Peredney, C.L., Williams, P.L., 1999. A soil bioassay using the nematode Caenorhabditis elegans. In: Henshel, D.S., Black, M.C., Harrass, M.C. (Eds.), Environmental Toxicology and Risk Assessment, vol. 8. Americal Society for Testing and Materials, Philadelphia, pp. 305–318. vol. ASTM STP 1364. Gavaskar, A.R., 1999. Design and construction techniques for permeable reactive barriers. Journal of Hazardous Materials 68, 41–77. Gibert, O., Pablo, J., Cortina, J.L., Ayora, C., 2003. Evaluation of municipal compost/limestone/iron mixtures as filling material for permeable reactive barriers for in-situ acid mine drainage treatment. Journal Chemical Technology Biotechnology 78, 489–496. Gibert, O., de Pablo, J., Cortina, J.L., Ayora, C., 2004. Chemical characterization of natural organic substrates for biological mitigation of acid mine drainage. Water Research 38, 4186–4196. Gonc¸alves, M.M.M., da Costa, A.C.A., Leite, S.G.F., Sant’Anna, G.L., 2007. Heavy metal removal from synthetic wastewaters in an anaerobic bioreactor using stillage from ethanol distilleries as a carbon source. Chemosphere 69, 1815–1820. Gray, N.F., O’Neill, C., 1997. Acid mine-drainage toxicity testings. Environmental Geochemistry and Health 19, 165–171. Groudev, S.N., Spasova, I.I., Georgiev, P.S., 2001. In situ bioremediation of soils contaminated with radioactive elements and toxic heavy metals. International Journal of Mineral Processing 62, 301–308. Janssens de Bisthoven, L., Gerhardt, A., Guhr, K., Soares, A.M.V., 2006. Behavioral changes and acute toxicity to the freshwater shrimp Atyaephyra desmaresti Millet (Decapoda Natantia) from exposure to acid mine drainage. Ecotoxicology 15, 215–227. Jarvis, A.P., Moustafa, M., Orme, P.H.A., Younger, P.L., 2006. Effective remediation of grossly polluted acidic, and metalrich, spoil heap drainage using a novel, low-cost, permeable reactive barrier in Northumberland, UK. Environmental Pollution 143, 261–268. Johnson, D.B., Hallberg, K.B., 2005. Acid mine drainage remediation options: a review. Science of Total Environment 338, 3–14. Kim, D.H., Feinbaum, R., Alloing, G., Emerson, F.E., Garsin, D.A., Inoue, H., Tanaka-Hino, M., Hisamoto, N., Hatsumoto, K., Tan, M.-W., Ausubel, F.H., 2002. A conserved p38 MAP kinase pathway in Caenorhabditis elegans innate immunity. Science 297, 623–626. Kim, K.T., Lee, Y.G., Kim, S.D., 2006. Combined toxicity of copper and phenol derivatives to Daphnia magna: effect of complexation reaction. Environment International 32, 487–492. Komnitsas, K., Bartzas, G., Paspaliaris, I., 2003. Efficiency of limestone and red mud barriers: laboratory column studies. Minerals Engineering 17 (2), 183–194. Komnitsas, K., Bartzas, G., Paspaliaris, I., 2006. Inorganic contaminant fate assessment in zero-valent iron treatment walls. Environmental Forensics 7 (3), 207–217. Lopes, I., Baird, D.J., Ribeiro, R., 2006. Genetic adaptation to metal stress by natural populations of Daphnia longispina. Ecotoxicology and Environmental Safety 63, 275–285. Ludwig, R.D., McGregor, R.G., Blowes, D.W., Benner, S.G., Mountjoy, K., 2002. A permeable reactive barrier for treatment of heavy metals. Ground Water 40, 59–66. Mayer, K.U., Benner, S.G., Blowes, D.W., 2006. Process-based reactive transport modeling of a permeable reactive barrier for
158
water research 44 (2010) 151–158
the treatment of mine drainage. Journal of Contaminant Hydrology 85, 195–211. Neculita, C., Zagury, G.J., 2008. Biological treatment of highly contaminated acid mine drainage in batch reactors: long-term treatment and reactive mixture characterization. Journal of Hazardous Materials 157, 358–366. Pagnanelli, F., De Michelis, I., Di Tommaso, M., Ferella, F., Toro, L., Veglio`, F., 2008a. Treatment of acid mine drainage by a combined chemical/biological column apparatus: mechanisms of heavy metal removal. In: Sa´nchez, Mikel L. (Ed.), Chapter 1 in the Book Causes and Effects of Heavy Metal Pollution. Nova Science Publishers, Inc. (ISBN: 978-1-60456-900-1). Pagnanelli, F., De Michelis, I., Di Muzio, S., Ferella, F., Veglio`, F., 2008b. Bioassessment of a combined chemical-biological treatment for synthetic acid mine drainage. Journal of Hazardous Materials 159 (2–3), 567–573. Pagnanelli, F., Cruz Viggi, C., Mainelli, S., Toro, L., 2009. Assessment of solid reactive mixtures for the development of biological permeable reactive barriers. Journal of Hazardous Materials 170, 998–1005. Peredney, C.L., Williams, P.L., 2000. Utility of Caenorhabditis elegans for assessing heavy metal contamination in artificial soil.
Archives of Environmental Contamination and Toxicology 39, 113–118. Postgate, J.R., 1984. The Sulphate-Reducing Bacteria, second ed. Cambridge University Press. Sohlenius, B., 1980. Abundance, biomass and contribution to energy flow by soil nematodes in terrestrial ecosystems. Journal of Nematology 11, 186–194. Ura, T., Kai, S., Sakata, T., Iguchi, K., Arizono, K., 2002. Aquatic acute toxicity testing using the nematode Caenorhabditis elegans. Journal of Health Science 48, 583–586. Waybrant, K.R., Blowes, D.W., Ptacek, C.J., 1998. Selection of reactive mixtures for use in permeable reactive walls for treatment of mine drainage. Environmental Science and Technology 32, 1972–1979. Waybrant, K.R., Ptacek, C.J., Blowes, D.W., 2002. Treatment of mine drainage using permeable reactive barriers: column experiments. Environmental Science and Technology 36, 1349–1356. Yim, J.H., Kim, K.W., Kim, S.D., 2006. Effect of hardness on acute toxicity of metal mixtures using Daphnia magna: prediction of acid mine drainage toxicity. Journal of Hazardous Materials 138, 16–21.
water research 44 (2010) 159–166
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Modeling of slow sand filtration for disinfection of secondary clarifier effluent K. Langenbach a,*, P. Kuschk a, H. Horn b, M. Ka¨stner a a b
Helmholtz Centre for Environmental Research – UFZ, Department of Environmental Biotechnology ,Permoserstr. 15, 04318 Leipzig, Germany Institute of Water Quality Control, Technical University Mu¨nchen, Am Coulombwall, 85748 Garching, Germany
article info
abstract
Article history:
Due to increasing water scarcity, appropriate technologies are needed for disinfection of
Received 9 March 2009
wastewater to enable safe reuse. Research on hygienisation of secondary effluent using
Received in revised form
slow sand filters is very limited but promising with removal of fecal indicator bacteria of
24 August 2009
>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, <50%) to more than 90% for HHCB and AHTN by sorption onto sludge particles ( Bester, 2004; Simonich et al., 2002; Zhang et al., 2008; Zeng et al., 2007). The removal of MK in this study was comparable to those values (38–55%) reported by Yang and Metcalfe (2006). The SMCs removals by other modified versions of AS (i.e., MLE, SBR and A2O) obtained in this study were somewhat higher than those of the conventional AS process. The removals of HHCB and AHTN in MLE, which uses separate anoxic and aerobic reactors, and recycles both sludge and nitrified liquor to the anoxic reactor, were 65% and 64%, respectively. SBR operated by sequencing aeration and settling periods had the highest removals for HHCB and AHTN compared to other biological treatment processes: 72% and 69%, respectively. Only one previous study investigated the removal efficiency of SMCs in SBR and it also reported high removal efficiencies of 87% for HHCB and 83% for AHTN
(Berset et al., 2004). The A2O process, which is composed of anaerobic/anoxic/oxide processes, showed similar removals for HHCB and AHTN as those of the MLE process. The Densadeg (DE) process uses a high-rate solids contact clarifier which combines optimized flocculation, internal and external sludge recirculation, and plate settling in two conjoined vessels to obtain excellent effluent quality. As a chemical coagulation treatment, the DE process had lower removal (37% for HHCB and AHTN and 36% for MK) than the biological treatment processes. In filtration processes followed by secondary treatment, the removal in biofiltration (BF) was 3–8 times higher than that of sand filtration (SF), indicating that biological filtration has higher SMCs removal than physical filtration treatment processes, although this result was obtained from only one case study. Chlorination (Cl) and Ultraviolet disinfection (UV) processes as an additional disinfection treatment showed low removal of below 10%. Cl and UV treatments prior to effluent discharge into surface water have normally been used with high efficiency to remove viruses and bacteria from sewage (Reiner et al., 2007). However, such disinfection processes are not so effective in SMCs removal. To summarize, the HHCB and AHTN removals of modified biological treatment processes (i.e., MLE, SBR and A2O) were slightly higher than the conventional AS process, while the DE chemical treatment was lower than the biological treatment processes. The BF process followed by secondary treatment showed relatively higher removal efficiency than SF. Cl and UV disinfection were ineffective processes for SMCs removal.
3.2.
Synthetic musk compounds (SMCs) in surface water
The HHCB and AHTN concentrations in surface water were measured twice in October, 2007 and April, 2008 in Korea. In the first survey, HHCB and AHTN were detected in all stream and river water samples at a concentration range from 0.26 to 13.92 mg/L (mean, 4.91 mg/L) and 0.05 to 2.80 mg/L (mean, 0.98 mg/L), respectively (Table 3). The study concentrations were comparable with the surface water monitoring data measured in Berlin water highly polluted by continuous STP discharges (Heberer, 2002). That study reported HHCB concentrations ranging from 0.02 to 12.47 mg/L (mean, 2.49 mg/L) and AHTN concentrations from 0.03 to 5.8 mg/L (mean, 1.34 mg/L). Other reported monitoring data for HHCB and AHTN in surface water (mostly, below 1 mg/L) were
water research 44 (2010) 214–222
219
Fig. 3 – The removals of HHCB, AHTN and MK in the sewage treatment processes.
lower than those in this study (Bester, 2005; Dsikowitzky et al., 2002; Rimkus, 1999; Zhang et al., 2008). In the second survey, two more NMCs (i.e., MX and MK) were monitored and MX was not detected. HHCB was the predominant compound, followed by AHTN and MK, in all surface water samples. The concentrations ranged from 0.10 to 2.72 mg/L (mean, 0.68 mg/L) for HHCB, 0.03 to 0.52 mg/L (mean, 0.14 mg/L) for AHTN, and ND to 0.42 mg/L (mean, 0.11 mg/L) for MK. The concentration variation of SMCs in surface waters is shown in Table 3. The HHCB concentrations in surface waters were approximately 4–5 times higher than those of AHTN in all sampling sites (HHCB/AHTN ratio, Table 3). Those of HHCB/ MK in second survey a little varied according to the sampling points (3.69–7.13, Table 3). The total HHCB and AHTN concentrations in the first survey (October) were higher than those in second survey (April). Yang and Met} ller et al. (2006) reported that in influent calfe (2006) and Mu STP samples, the HHCB and AHTN concentrations in cold seasons were relatively higher than those in warm seasons. These studies indicate that the usage volume and discharge of SMCs in household can cause seasonal variation of SMC concentrations in water. There is not enough evidence to match with previous researches because the samples in this study were collected in spring and fall not showing wide temperature variation. Therefore, further continuous research is needed to elucidate the seasonal variation of SMCs concentration in surface water.
As the main source of SMCs in surface water is indirect input via STP effluent, urban areas have shown higher concentrations than rural areas (Bester, 2005; Dsikowitzky et al., 2002; Zhang et al., 2008). The similar result was observed in this study and the SMC concentrations in urban areas were generally higher than those in rural areas (Table 3). Especially, the sampling points near the STPs were shown relatively high SMCs levels. The high concentration in R-1 was attributed to effluent discharge from 3 STPs (C, D-1 and D-2) that are located within 1 km (see Fig. 1). SMCs concentration at S-7 point was also affected by effluent discharge from B STP located within 500 m and especially, the effluent discharge of B STP (0.347 m3/s) accounted for 85% of the flow rate of S-7 point (0.405 m3/s). However, even though S-1 point was located down the effluent discharging point of A STP, the level of SMCs was low. The sampling location of S-1 is far from the A STP (approximately, 10 km) and the discharging SMCs concentration in A STP is relatively low compared to those of other STPs (Table 3). Moreover, there is little mixing between upstream and downstream in this sampling point because of stagnant flow in ordinary days caused by floodgate (see Fig. 1). Zhang et al. (2008) also reported that SMC concentrations in surface water depend on the distance of the sampling sites to the pollution sources. Therefore, S-1 point has the lowest SMCs occurrence (Table 3). In case of S-2 and S-8 points, sewage gathering systems along these two streams were under construction during sampling period. Therefore, it is
220
water research 44 (2010) 214–222
Table 3 – Concentrations of synthetic musk compounds (SMCs) in rivers and streams (mg/L). Area
S-1 S-2 S-3 S-4 S-5 S-6 S-7 S-8 S-9 S-10 R-1 R-2 R-3 R-4
Rural Urban Urban Urban Urban Urban Urban Urban Rural Urban Urban Suburban Rural Suburban
STP (upstream)
Yes No No No No No Yes No No No Yes – No No
First survey (October)
Second survey (April)
HHCB
AHTN
Sum (Rank)
HHCB/ AHTN
HHCB
AHTN
MX
MK
Sum (Rank)
0.37 9.24 6.11 0.42 3.60 1.98 7.43 13.92 0.26 NAa 5.74 NA NA NA
0.07 1.77 1.16 0.09 0.74 0.44 1.49 2.80 0.06 NA 1.13 NA NA NA
0.44 (9) 11.01 (2) 7.27 (4) 0.52 (8) 4.34 (6) 2.42 (7) 8.92 (3) 16.72 (1) 0.32 (10) – 6.88 (5) – – –
5.15 5.22 5.26 4.53 4.89 4.52 4.98 4.96 4.36 – 5.07 – – –
0.10 1.26 0.25 0.11 0.49 0.50 1.07 2.72 NA 0.65 1.14 0.16 0.13 0.21
0.03 0.30 0.05 0.03 0.09 0.10 0.21 0.52 NA 0.12 0.23 0.04 0.03 0.04
NDb ND ND ND ND ND ND ND NA ND ND ND ND ND
0.13 1.80 0.30 0.14 0.67 0.69 1.57 3.66 – 0.91 1.54 0.20 0.15 0.25
4.91
0.98
5.88
4.89
0.68
0.14
–
0.11
0.92
(13) (2) (8) (12) (7) (6) (3) (1) (5) (4) (10) (11) (9)
HHCB/ AHTN
HHCB/ MK
4.08 4.18 5.26 4.27 5.34 4.91 5.17 5.20 – 5.42 4.89 4.16 4.98 5.20
– 5.25 – – 6.13 5.56 3.69 6.48 – 4.64 7.13 – – –
4.85
5.55
S, Stream water; R, River water a NA ¼ Not analyzed. b ND ¼ below LOD.
conjectured that the presence of non-point sources such as intermittent or continuous direct input of untreated residential sewage into streams due to the lack of sewage gathering systems affected the higher SMCs concentrations in these two points. Generally, the SMC concentrations in surface water in this study were related with the urbanization of sampling sites such as indirect input via STP effluent and direct input from non-point sources therefore, the samples collected near urban areas exhibited high concentrations, followed by suburban areas and rural areas (urban > suburban > rural, Table 3 – Rank). The SMC concentrations in surface waters showed concentration variation
over the two survey times. It implies that continuous monitoring is needed according to season.
3.3. The fate of synthetic musk compounds (SMCs) in the water environment The ratios of each SMC in water samples have normally been used to identify their fates in the water system (Dsikowitzky et al., 2002). Heberer (2002) reported that the HHCB/AHTN ratios were similar in the Berlin water system [i.e., sewage (n ¼ 35) and surface water (n ¼ 98)]. The HHCB/AHTN ratios in the influent and effluent samples of STPs in this study were
Fig. 4 – Comparison of measured and estimated AHTN (A) and MK (B) concentrations from measured HHCB concentrations using Eqs. (1) and (2). [Insets: linear regression equation between HHCB and AHTN (Eq. (1)), and HHCB and MK (Eq. (2)); symbols: measured values of HHCB, AHTN and MK.
water research 44 (2010) 214–222
similar with those in surface waters (see Section 3.1.1 and Table 3). The relative distribution profiles of the 3 SMCs (MX was excluded because this compound was only detected in 4 influent samples) were quite similar in all influent and effluent sewage samples (68.3 5.3% and 67.9 2.8% for HHCB; 15.0 1.1% and 15.1 2.4% for AHTN; 15.6 5.3% and 17.0 2.4% for MK, respectively), revealing their similar fates in the sewage treatment processes. Therefore, we investigated the R2 value among these 3 SMCs to examine their interrelations and obtained very significant, positive values among these 3 SMCs (R2 ¼ 0.955, p < 0.001 for HHCB and AHTN; R2 ¼ 0.729, p < 0.001 for HHCB and MK; R2 ¼ 0.803, p < 0.001 for AHTN and MK). This result indicates other SMC concentrations can be predicted based on the measured values for one SMC concentration. Therefore, we performed linear regression analysis using each concentration level of the 3 SMC concentrations in all influent and effluent sewage samples and obtained the following linear regression equations (Eqs. (1) and (2)):
221
continuous monitoring is needed to identify any possible seasonal variation. A close relationship was observed between AHTN and HHCB and the concentration of either one in a water system can be predicted from that of the other; however, the prediction capability for MK concentration was lower due to its relatively lower Kow value.
Acknowledgements This work was supported by the Korea Research Foundation Grant funded by the Korean Government (MOEHRD, Basic Research Promotion Fund) (KRF-2008-331-D00280).
Appendix. Supplementary data
CAHTN ¼ 0:229 CHHCB 0:021
(1)
Supplementary data associated with this article can be found in online version at doi:10.1016/j.watres.2009.08.049.
CMK ¼ 0:284 CHHCB 0:085
(2)
references
CHHCB; measured HHCB concentration, CAHTN; measured AHTN concentration, CMK; measured MK concentration. To evaluate the prediction capability of these equations, we compared the measured SMCs values in the interim STPs effluent samples and surface water samples. Fig. 4 shows the comparison results between the measured concentrations of the 3 SMCs and those estimated using Eqs. (1) and (2). The slope lines show the estimated values of AHTN and MK based on the HHCB concentrations and the symbols are measured values of HHCB, AHTN and MK. As shown in Fig. 4A, the measured values of AHTN fit very well with the estimated values, indicating the similar fate of these compounds in water environments (i.e., STP and surface waters) and confirming that the AHTN concentration can be predicted from the HHCB concentration. The comparison result of MK showed a slightly lower prediction capability than that of AHTN (Fig. 4B), which may have been due to the low Kow value of MK (4.3) compared to the similar values for HHCB (5.9) and AHTN (5.7) (Simonich et al., 2002). These results revealed the similar fate of HHCB and AHTN (generally, a tendency to partition between aqueous and solid matrices) but the slightly different fate of MK in the water environment.
4.
Conclusion
SMC concentrations were investigated in ten STPs and 14 surface water sites. The total SMC concentrations in the STPs were varied according to the input source and treatment volume of each STP. Biological treatment processes showed higher removal for SMCs than chemical treatment, filtration and disinfection processes. However, the removal efficiencies of these treatment processes need to be further compared with more extensive case studies. The presence of SMCs in surface water was mainly attributed to household sewage discharge and concentration variation of SMCs in surface water was observed for two times investigation. However,
Balk, F., Ford, R.A., 1999. Environmental risk assessment for the polycyclic musks, AHTN and HHCB in the EU: I. Fate and exposure assessment. Toxicology Letters 111 (1–2), 57–79. Berset, J.D., Kupper, T., Etter, R., Tarradellas, J., 2004. Considerations about the enantioselective transformation of polycyclic musks in wastewater, treated wastewater and sewage sludge and analysis of their fate in a sequencing batch reactor plant. Chemosphere 57 (8), 987–996. Bester, K., 2004. Retention characteristics and balance assessment for two polycyclic musk fragrances (HHCB and AHTN) in a typical German sewage treatment plant. Chemosphere 57 (8), 863–870. Bester, K., 2005. Polycyclic musks in the Ruhr catchment area – transport, discharges of waste water and transformations of HHCB, AHTN and HHCB-lactone. Journal of Environmental Monitoring 7, 43–51. Bester, K., 2009. Analysis of musk fragrances in environmental samples. Journal of Chromatography A 1216, 470–480. Chou, Y.-J., Dietrich, D.R., 1999. Toxicity of nitromusks in early lifestages of South African clawed frog (Xenopus laevis) and zebrafish (Danio rerio). Toxicology Letters 111 (1–2), 17–25. Clara, M., Strenn, B., Gans, O., Martinez, E., Kreuzinger, N., Kroiss, H., 2005. Removal of selected pharmaceuticals, fragrances and endocrine disrupting compounds in a membrane bioreactor and conventional wastewater treatment plants. Water Research 39 (19), 4797–4807. Dsikowitzky, L., Schwarzbauer, J., Littke, R., 2002. Distribution of polycyclic musks in water and particulate matter of the Lippe River (Germany). Organic Geochemistry 33 (12), 1747–1758. Heberer, T., 2002. Occurrence, fate and assessment of polycyclic musk residues in the aquatic environment of urban areas – a review. Acta Hydrochimica et Hydrobiologica 30 (5–6), 227–243. Horii, Y., Reiner, J.L., Loganathan, B.G., Kumar, K.S., Sajwan, K., Kannan, K., 2007. Occurrence and fate of polycyclic musks in wastewater treatment plants in Kentucky and Georgia, USA. Chemosphere 68 (11), 2011–2020. Kupper, T., Berset, J.D., Etter-Holzer, R., Furrer, T., Tarradellas, J., 2004. Concentrations and specific loads of polycyclic musks in sewage sludge originating from a monitoring network in Switzerland. Chemosphere 54 (8), 1111–1120.
222
water research 44 (2010) 214–222
Lishman, L., Amyrh, S.A., Sarafin, K., Kleywegt, S., Toito, J., Peart, T., Lee, B., Servos, M., Beland, M., Seto, P., 2006. Occurrence and reductions of pharmaceuticals and personal care products and estrogens by municipal wastewater treatment plants in Ontario, Canada. Science of the Total Environment 367 (2–3), 544–558. Ministry of Environment (MoE), Korea, 2006. The Survey on the Distribution Amount of Chemical Compounds. Mitjans, D., Ventura, F., 2004. Determination of musks and other fragrance compounds at ng/L levels using CLSA (closed loop stripping analysis) and GC/MS detection. Water Science and Technology 50 (5), 119–123. } ller, J., Bo¨hmer, W., Litz, N.T., 2006. Occurrence of polycyclic Mu musks in sewage sludge and their behaviour in soils and plants. Part 1: behaviour of polycyclic musks in sewage sludge of different treatment plants in summer and winter. Journal of Soils and Sediments 6 (4), 231–235. } ttmann, W., 2008. Organophosphates and Quednow, K., Pu synthetic musk fragrances in freshwater streams in Hessen/ Germany. Clean 36 (1), 70–77. Reiner, J.L., Berset, J.D., Kannan, K., 2007. Mass flow of polycyclic musks in two wastewater treatment plants. Archives of Environmental Contamination and Toxicology 52 (4), 451–457.
Rimkus, G.G., 1999. Polycyclic musk fragrances in the aquatic environment. Toxicology Letters 111 (1–2), 37–56. Simonich, S.L., Federle, T.W., Eckhoff, W.S., Rottiers, A., Webb, S., Sabaliunas, D., Wolf, W., 2002. Removal of fragrance materials during U.S. and European wastewater treatment. Environmental Science and Technology 36 (13), 2839–2847. Tas, J.W., Balk, F., Ford, R.A., van de Plassche, E.J., 1997. Environmental risk assessment of musk ketone and musk xylene in the Netherlands in accordance with the EU-TGD. Chemosphere 35 (12), 2973–3002. Yang, J.-J., Metcalfe, C.D., 2006. Fate of synthetic musks in a domestic wastewater treatment plant and in an agricultural field amended with biosolids. Science of the Total Environment 363 (1–3), 149–165. Zeng, X., Sheng, G., Gui, H., Chen, D., Shao, W., Fu, J., 2007. Preliminary study on the occurrence and distribution of polycyclic musks in a wastewater treatment plant in Guandong, China. Chemosphere 69 (8), 1305–1311. Zhang, X., Yao, Y., Zeng, Z., Qian, G., Guo, Y., Wu, M., Sheng, G., Fu, J., 2008. Synthetic musks in the aquatic environment and personal care products in Shanghai, China. Chemosphere 72 (10), 1553–1558.
water research 44 (2010) 223–231
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Characterization of the colloidal organic matter from the Amazonian basin by asymmetrical flow field-flow fractionation and size exclusion chromatography Enrica Alasonati a, Vera I. Slaveykova a,*, Herve´ Gallard b, Jean-Philippe Croue´ b, Marc F. Benedetti c,* Environmental Biophysical Chemistry Group (EBC), Environmental Engineering Institute, E´cole Polytechnique Fe´de´rale de Lausanne, Station 2, CH-1015 LAUSANNE, Switzerland b Universite´ de Poitiers, Ecole Supe´rieure d’Inge´nieurs de Poitiers Laboratoire de Chimie et Microbiologie de l’eau UMR CNRS 6008 40, avenue du Recteur Pineau, 86 022 POITIERS Cedex, France c Laboratoire de Ge´ochimie des Eaux – Baˆtiment Lamarck Universite´ Paris Diderot-Paris7-IPGP UMR 7154, case courrier 7052, 75205 PARIS Cedex 13, France a
article info
abstract
Article history:
Colloidal organic matter (COM) collected in small and large watercourse tributaries of the
Received 31 May 2009
Negro River (Brazil) were fractionated and characterized by an asymmetrical flow field-flow
Received in revised form
fractionation (AFlFFF) coupled to UV and seven angle laser light scattering (LS) detectors,
31 August 2009
and by size exclusion chromatography (SEC) coupled with a UV detection. Number and
Accepted 3 September 2009
weight average molar masses, weight average gyration radiuses RGw and polydispersity
Published online 15 September 2009
indexes were obtained for each sample in two separate runs under conditions optimized for lower and larger size fractions. The results demonstrate the existence of a decrease of
Keywords:
size of the colloidal matter when passing from first order streams to higher order rivers. No
Colloids
significant changes were found in size distributions of samples collected during the low
Size distribution
and high flow stages at the same site. The influence of selected pre-treatments such as
Size exclusion chromatography
filtration and reverse osmosis pre-concentration on the size and molar mass distributions
Flow field-flow fractionation
was also studied.
Organic matter
ª 2009 Elsevier Ltd. All rights reserved.
Amazon
1.
Introduction
The Amazon River basin encompasses seven million square kilometers and host to over half of the planet’s remaining rainforests. This region is a huge source of carbon with relevant role in the global cycles (Ometto, et al., 2005). The basin is drained by the Amazon River and its major tributaries, the Negro River, the Madeira River and the Solimo˜es River and supplies the 20% of the global water discharged into the
oceans (Molinier, et al., 1997) as well as 7% of the global flux of terrestrial dissolved organic carbon to the oceans and 70% of this flux is in the colloidal size range (<5000 g mol1; 1 g mol1 ¼ 1 Da). Hence, the understanding of Amazonian system through the study of aquatic geochemistry and sitespecific soil/aquatic organic matter is of fundamental importance in global carbon cycles investigation. Several works on the characterization of organic matter (OM) in the Amazon basin exist, (e.g. Ertel, et al., 1986; Hedges,
* Corresponding authors. E-mail addresses: [email protected] (V.I. Slaveykova), [email protected] (M.F. Benedetti). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.010
224
water research 44 (2010) 223–231
et al., 1994; Krusche, et al., 2002; Moreira-Turcq, et al., 2003; Guo and Macdonald, 2006). There is still a debate on the respective part of the different processes that control OM concentrations and quality (structure, degradability and chemical reactivity) along the Amazon Basin. Hedges, et al., (2000) have built a chromatographic model to account for the change of the OM composition from the soils to the Amazon mainstream. Selective sorption of OM onto minerals is a key process to account for the differences observed in the different size fractions of the total organic carbon of the rivers. A major perspective of these previous works dedicated to the cycling of OM was that additional studies of the composition and reactivity of OM size fractions are required as well as laboratory information that can help us to understand the partitioning processes occurring in soils or riparian zones (Amon and Benner, 1996; Hedges, et al., 2000). Among the different OM constituents, colloidal fraction plays utmost role in the transport, speciation and bioavailability of trace elements. Due to the small size and large specific surface area, colloids, in general and colloidal organic matter (COM) are important mediator in partitioning of metals between truly dissolved and particulate phases e.g. by acting as ‘‘colloidal pumps’’, thus influencing the transport and removal of trace elements (Santschi, et al., 2002; Wilkinson and Lead, 2007) organic carbon cycling and micronutrient bioavailability (e.g. Town and Filella, 2002 and references therein; Schmit and Wells, 2002). The size of COM influences the proton and metal binding, the organic pollutant partitioning, the coagulation efficiency, the adsorption onto minerals, practically all environmental aspects of COM compounds (Cabaniss, et al., 2000; Tipping, 2005). Therefore, accurate colloidal organic matter molar mass and size characterization are necessary. In the geographical region of interest in the present work, the Amazonian Basin, changes in size distribution of colloidally associated Ca in several rivers were studied using AFlFFF coupled to ICP-MS (Dahlqvist, et al., 2004). FlFFF coupled to UV or ICP-AES detectors was also used to investigate particulate and colloidal binding of Al and Fe onto organic matter in samples collected in the Negro and Amazon Rivers (Benedetti, et al., 2002; Benedetti, et al., 2003). Nonetheless, an important limitation of these previous works is that the COM characteristics were obtained for a material pre-concentrated by tangential flow ultrafiltration and not on bulk water. The use
of a pre-concentration step prevents a full extrapolation of the obtained data to in situ conditions (Benedetti, et al., 2002, 2003) therefore, to better understand changes at the field scale data obtained on bulk material is highly sought. Furthermore, a series of recent papers emphasizes the importance of the studied area and the fact that samples are representative for the processes occurring in the Negro River watershed (do Nascimento, et al., 2004; Balan, et al., 2005; Bardy, et al., 2007; Fritsch, et al., 2007; Bardy, et al., 2008; Do Nascimento, et al., 2008; Bardy, et al., 2009; Fritsch, et al., 2009). By exploring the capabilities of the AFlFFF coupled with UV and seven angle laser light scattering detectors, the present study aims to investigate role of spatial and temporal variability of the fluvial system on the size and molar mass of colloidal material. The emphasis is on the size and molar mass distributions and their average values of the colloidal organic matter collected in small and large watercourse tributaries of the Negro River. SEC coupled to UV detection was used as a complementary characterization tool. Colloids originating from podzol water were compared to those from river in order to study the difference between aquatic and soil solution carbon issued from podzolization zones that cover 50% of the Rio Negro watershed or 7% of the 2 millions km2 of the Amazon watershed (Radambrazil, 1972–1978). The study of the small order stream is particularly interesting because they are more sensitive to environmental changes due to their small size and their low power of dilution. The influence of selected pre-treatments as filtration and reverse osmosis pre-concentration on the size and molar mass distributions was also studied.
2.
Material and methods
2.1.
Sampling sites and methods for sample collection
Colloidal samples were collected in the Amazon River basin (Brazil) during two separated cruises. The first one took place in October–November 2004 during to a falling stage period, while the second cruise in May 2005 corresponded to a high-flow period. The samples name, dissolved organic carbon concentration (DOC) and specific UV-absorbance (SUVA) are listed in Table 1 together with the temperature
Table 1 – NOM samples with respective localization, level of water in the sampling period (L [ low and H [ high), temperature and pH. DOC: dissolved organic carbon measured on the samples after GFF filtration (i.e. 0.7 m cut-off). Sample/Type O16/Podzol soil solution M12/Podzol soil solution O7/creek named: Igarape Disjuncta O15/creek named: Igarape Bonito O5/river named: Rio Jau O1/river named: Rio Negro @ Paricatuba M1/river named: Rio Negro @ Paricatuba
Water level
T ( C)
pH
DOC (m1/mg C)
SUVA (mg C L1)
L H L
30 – –
3.5 3.4 3.6
75.7 59.5 57.8
4.76 4.84 4.74
L
29
3.6
57.6
4.97
L
33
5.5
7.2
3.83
L
31
5.3
7.1
4.92
H
–
4.4
11.6
4.53
water research 44 (2010) 223–231
225
Fig. 1 – Pedology and sampling site locations in the Amazon basin (adapted from do Nascimento et al., 2008).
and pH measured at the sampling point. The location is shown in Fig. 1 together with information on the nature of the soils and vegetation. Samples O7, O15 and O16 were taken from the same soil catena: O16 is a Podzol soil solution collected 100 cm under the surface level in a trench made in the soil profile. O7 was collected on the creek Disjuncta, draining a small soil depression; O15 on the creek Bonito draining a large soil depression, after the confluence with the creek Disjuncta. Sample O5 was collected on the Jau River, which collects the smaller creeks and is a major tributary of the Negro River. Sample O1 was collected in the Negro River, close to the city of Paricatuba (Amazonia, Br). Samples M1 and M12 were collected at the same locations as samples O1 and O16, respectively, in May 2005, during the high-flow period. All samples were filtered through 0.7 mm GFF filters, unless mentioned otherwise in the text. The influence of selected sample pre-treatments, size and molar mass distributions of unprocessed sample (raw water), filtered through 0.7 mm filter and concentrated on site by reverse osmosis were compared for the soil solution sample (O16).
2.2.
Total organic carbon determination
Total organic carbon (TOC) was determined using a Shimadzu TOC analyzer that utilized a high temperature oxidation procedure prior to IR detection of CO2. The detection limit was 0.1 mg L1 and the precision ranged from 2 to 3%. SUVA
corresponds to the ratio between the absorbance of the sample measured at 254 nm and its DOC concentration value.
2.3.
Size exclusion chromatography set-up
High pressure size exclusion chromatography (HP-SEC) fractionation was performed in France with a chromatographic system coupled with an UV absorbance detector at the wavelength of 260 nm, using a REPROSIL 200 SEC column. The carrier solution was a 10 m M sodium acetate solution (pH ¼ 7.0) and the flow rate was 1 mL min1. COM solutions for HP-SEC determination had a 10 mg L1 DOC concentration. Polystyrene sulfonate (PSS) standards of molar mass 1400, 4300, 6800, 13,000 and 17,000 Da were used as standards. The obtained calibration equation was: log M ¼ ð16:728 tr Þ=ð2:255Þ
(1)
where tr is the retention time and M is the molar mass of the PSS. All samples were analyzed by SEC immediately after the return in France (less than a couple of weeks) in order to minimize the effect of sampling storage since in situ measurements were not possible on board due to the remoteness of the field sites.
2.4.
Asymmetrical flow field-flow fractionation set-up
Colloidal samples were fractionated and characterized with an asymmetrical flow field-flow fractionation system (AF2000
226
water research 44 (2010) 223–231
Focus, Postnova Analytics) coupled to seven angle laser light scattering (LS; Postnova Analytics) and UV/VIS spectrophotometer a year after the SEC measurements were done. The software AF2000 Control (Postnova Analytics) was used to control the AFlFFF system and detectors as well as for LS data treatment. Measured channel dimensions are: tip-to-tip length, 27.5 mm; inlet triangle breadth and length, 20 and 34 mm, respectively; outlet triangle breadth and length, 5 and 10 mm. Amazonian COM samples were fractionated using two set of working parameters in order to analyze both the lower and the higher size fractions. Two regenerated cellulose membrane were employed with cut-off of 1 kDa and 10 kDa. When 1 kDa cut-off membranes were used, the focusing conditions were: focusing time tfoc ¼ 14 min; inlet flow rate Vin ¼ 0.09 mL min1; crossflow rate focus flow rate Vfoc ¼ 4 mL min1; 1 Vc ¼ 3.09 mL min ; outlet flow rate Vout ¼ 1 mL min1; elution conditions: Vc ¼ 3 mL min1 and Vout ¼ 1 mL min1. When 10 kDa cut-off membranes were used, the focusing conditions were: focusing time tfoc ¼ 14 min; inlet flow rate Vin ¼ 0.09 mL min1; focus flow rate Vfoc ¼ 5 mL min1; crossflow rate Vc ¼ 4.09 mL min1; outlet flow rate Vout ¼ 1 mL min1; elution conditions: Vc ¼ 0.25 mL min1 and Vout ¼ 1 mL min1. The 10 mM NaNO3 (pH ¼ 5.7) used as carrier solution was filtered on a 0.1 mm pore size teflon filter (Postnova Analytics). COM samples were centrifuged 10 min at 5000 rpm and filtered on 0.22-mm polyvinylidene fluoride (PVDF; Millipore) filters, before injection in the channel.
2.5. Asymmetrical flow field-flow fractionation data handling The higher size colloidal fraction characteristics such as the radius of gyration and the molar mass were obtained from the slope and intercept of the Zimm plot, according to Rayleigh scattering theory (Wyatt, 1993), from both LS and UV signal by using AF2000 software (Postnova Analytics). The signal provided by UV detector and the COM absorbance coefficient (e) calculated from the Lambert–Beer law using absorbance and total organic carbon measurements performed on COM samples were used for data processing. Number-average (Mn) and weight average (Mw) molar masses, polydispersity (PDI ) as well as the weight average radius of gyration (RGw) were determined from the respective distributions obtained from light scattering measurement, according to: P Mi ci Mw ¼ P ci P ci Mi M Mn ¼ P ici
(2)
(3)
Mi
P ci Ri RGw ¼ P ci
(4)
PDI ¼ Mw =Mn
(5)
where ci, Mi, and Ri are the sample concentration, the molar mass, and the radius of gyration for a slice i of the corresponding distributions.
For the smaller size fraction, the molar mass and distributions were obtained from the UV detector signal by calibration, injecting 100 mL of 50 mg L1 polystyrene sulfonate (PSS) (Postnova Analytics) dissolved in carrier. The molar mass of PSS standards were respectively 1360, 2220, 6530, 10,600 and 15,200 Da. The following relationship between molar mass, M and the retention time, tr was obtained: log M ¼ ðlog tr 0:480Þ=ð0:449Þ
(6)
from the PSS calibration and used to calculate colloidal fraction molar masses. Since the molar mass is related in a nonlinear way to retention time, UV signals were converted in distributions by multiplying the UV signal with the ratio dtr/dM. The hydrodynamic radius of the colloids Rh, corresponding to the radius of the sphere of equivalent hydrodynamic behavior, was calculated from Eq. (7) in accordance to the FFF theory (Wahlund, 2000): Rh ¼
tr Vo kT t0 w2 Vc ph
(7)
where tr is the retention time, t0 is the eluent average residence time, calculated according to the equation developed for trapezoidal channels (e.g. Litzen and Wahlund, 1991), w the total channel thickness, Vc the crossflow rate, h the eluent viscosity, Vo the void volume, k the Boltzmann constant and T is the absolute temperature. The spacer nominal thickness of 350 mm decreased to 302 mm when measured from the retention time of BSA using tabulated values of its diffusion coefficient. The effective spacer thickness yielded the value of 0.98 mL for the channel void volume.
3.
Results and discussion
3.1. Characterization of colloidal organic matter low-size fraction Fractionation of the filtered (i.e. GFF-0.7 m) samples O1, O5, O7, O15 and O16 yielded a single peak corresponding to weight and number average molar masses in the range 1126–2230 and 519–1142 Da respectively (Fig. 2A, Table 2). The UV signal of sample O16, the Podzol water collected on the top of the soil catena, was higher than in the signal in the river system samples (samples O1, O5, O7, O15), in agreement with higher carbon content of the sample. There is a slight shift toward lower values of the retention time corresponding to the O16 peak, compared to the O15, which resulted in a lower weight average hydrodynamic radius (Table 2). This is probably due to an overloading effect caused by the concentration of the O16 sample (i.e. 76 mg C L1), which is more than 30% higher than O15 and O7. Indeed, the particle effective volume is recognized to influence the retention of the eluting particles or molecules in normal mode FlFFF. The increase in the sample load induces peak distortion and variation in elution profiles (overloading). In the case of particles the retention was noticed to decrease in presence of overloading in the channel (Schimpf, et al., 2000), thus shifting the peak toward lower retention times.
227
water research 44 (2010) 223–231
compared to size exclusion chromatography (Pelekani, et al., 1999; Perdue, et al., 2003; Jackson, et al., 2005). The small changes in the low molar mass have been found to affect Mn more than Mw (Zhou, et al., 2000), while larger molar mass species affect more the Mw than the Mn. Furthermore, the obtained values for the molar masses were consistent with those obtained for model humic substances. For example, weight average and number-average molar masses of 2300 and 1428 Da, and 847 Da and 627 Da were obtained for SRHA and SRFA, respectively. Furthermore, the obtained values were in general agreement with those found in literature by employing several separation and detection techniques (Janos, 2003; Abbt-Braun, et al., 2004; McDonald, et al., 2004). For example, Mw values in the range 2090–6590 Da and 950–2800 Da and Mn in the range 1220–2370 Da and 540–1790 Da were found respectively for humic and fulvic-like fractions (Perdue, et al., 2003; Reemtsma and These, 2003). The measured values are also consistent with data measured using techniques, as the adsorption onto minerals, based on a fractionation principle completely different from field-flow fractionation (Hur and Schlautman, 2003).
3.2. Characterization of larger size COM fraction by AFlFFF–UV–LS
Fig. 2 – Comparison of AFlFFF and HP-SEC data. (A) Fractograms, corresponding to the injection of 100 mL of samples O1, O5, O7, O15 and O16, obtained by AFlFFF. Fractionation conditions – Membrane: RC 1 kDa cut-off. Vc [ 3 mL minL1. (B) Chromatograms, corresponding to the injection of 10 mg C LL1 of samples O16, O15, O5 and O1, obtained by SEC.
Size exclusion chromatography was applied to separate the O5, O1, O15 and O16 samples giving rise to a large peak corresponding to weight and number average molar masses in the range 803–1513 and 655–1016 Da, respectively (Fig. 2B, Table 3). A second smaller peak corresponding to 200–300 Da molar mass species was also found in all samples. Despite of the first SEC peak, comparison of the molar mass distributions and averages values, as well as polydispersity indexes obtained with the two fractionation techniques AFlFFF–UV and SEC–UV, for samples O1, O5, O15 and O16, were consistent and confirmed the same trends. The number average molar mass obtained by both techniques was very close suggesting that long term storage did not significally change the properties of the COM, with average variations of 15%, while the weight average molar mass obtained by AFlFFF measurements was around 1.3–1.6 times higher than that found by SEC, with consequent increase of the polydispersity index. The analytical window of asymmetrical flow field-flow fractionation is often shifted toward higher class of size, when
Samples O16, M12, O7 and O15, which have a fairly high carbon concentration (Table 1) allowing LS detection, were characterized by AFlFFF–UV–LS, under conditions optimized to separate larger size fractions. In contrast to a single peak in the UV signal, the light scattering shows the presence of two not well resolved peaks for all samples (Fig. 3A and B). The first peak corresponds to the species with molar mass in the range from 200 to 2800 103 Da and significant UV-absorbing
Table 2 – Weight average molar mass (±13%), number average molar mass (±3%), polydispersity index (±10%) and weight average hydrodynamic radius (±13%) of colloidal organic matter (COM) samples determined by AFlFFF–UV detection. Averaged values were calculated using the signal acquired in the region of interest corresponding to the retention times between 1 and 10 min. The crossflow rate was 3 mL minL1. O states for October and M for May. Sample/type O16/Podzol soil solution O16 unprocessed/Podzol soil solution O16 osmosis/Podzol soil solution M12/Podzol soil solution O7/creek named: Igarape Disjuncta O15/creek named: Igarape Bonito O5/river named: Rio Jau O1/river named: Rio Negro @ Paricatuba M1/river named: Rio Negro @ Paricatuba
Mw (Da)
Mn (Da)
PDI
Rh (nm)
1918 1920
922 957
2.0 2.0
0.6 0.6
1592
806
2.0
0.6
2230 2120
1142 1132
2.0 1.9
1.5 1.5
2137
1073
2.0
1.5
1126
519
2.2
1.1
1466
635
2.3
1.3
1544
638
2.4
1.3
228
water research 44 (2010) 223–231
Table 3 – Weight and number average molar mass and polydispersity index of COM samples determined by HP-SEC–UV detection. Averaged values were calculated using the signal acquired in the region of interest corresponding to the retention times between 7 and 11 min, corresponding to the range of molar mass 20605– 347 Da. Sample/Type O16/Podzol soil solution O15/creek named: Igarape Bonito O5/river named: Rio Jau O1/river named: Rio Negro @ Paricatuba
Mw (Da)
Mn (Da)
PDI
1513 1438
1016 961
1.5 1.5
803 980
655 740
1.2 1.3
properties, while the second corresponds to the very low amount of a large COM aggregates or colloids dominated by mineral phases such as iron oxides or clays (Benedetti et al., 2003; Allard et al., 2004) Average molar masses, radius of gyration, RG values and the polydispersity index were calculated only in the region corresponding to the UV signal (Table 4). The values of molar mass are similar for all samples and corresponded to the ranges of 344–467 103 Da for the weight average molar mass and 248 308 103 Da for the number-
average. The weight average gyration radius was about 46% lower in O7 and O15 samples. The colloidal material isolated from Podzol water (O16 and M12) exhibited higher size and molar mass compared to the river water samples: gyration radius and molar mass average values are w29% higher than O15/O7 average values (Table 4).
3.3. Influence of filtration and osmosis pre-concentration on the COM size distribution The Podzol water O16 contains rich quantity of dissolved organic carbon (76 mg L1) and dissolved organic nitrogen (1.58 mg N L1). To verify the influence on the size distribution of commonly used sample pre-treatments, three samples of O16 were analyzed by AFlFFF–UV system, using the 1 kDa cutoff membrane: (i) unprocessed sample (raw water), (ii) filtered through 0.7–mm filter and (iii) concentrated by reverse osmosis (RO). No difference in size distribution (Fig. 4) and in average molar mass and polydispersity values (Table 2) can be detected after filtration on 0.7 mm filters. In contrast the preconcentrated sample shows a slight shift of the mass distribution towards smaller masses. The weight average molar mass decreased from w1919 to 1592 Da and the number average from w940 to 806 Da. It is worth noting that there is probably an overloading effect due to the increased concentration after reverse osmosis, as previously discussed. Thus the slight shift of molar mass is to be attributed to the fractionation technique and not to the concentration method, and can be overcome diluting the sample prior fractionation. It can be concluded that filtration and osmosis pre-concentration do not influence significantly the size distribution of the pretreated samples.
3.4. Temporal and spatial variability of COM size characteristics
Fig. 3 – Fractograms of 1 mL of the COM samples O16, M12, O7 and O15. (A) UV signal at 254 nm, (B) LS 908 detectors signals versus elution time. aFlFFF conditions – Membrane: RC 10 kDa cut-off. Vc [ 0.25 mL minL1. Calculated hydrodynamic radiuses are given on the top x-axes.
The fractionation of the O1, M1, O16 and M12 samples provided information about the temporal variability of the COM size distribution in the Amazon River fluvial system. No difference was found in the size distributions of samples from Rio Negro (Paricatuba) collected in October and May (O1 and M1), representative of the large river system (Fig. 5). The nature of very large basin like the Negro River watershed generates very homogeneous size fractions of COM because the contribution of different type of processes (sorption, degradation etc.) controlling the COM in the water and occurring in the different type of head-waters and soils distributed over the whole basin are integrated at the outlet of the watershed like in sample O1 or M1. This is confirmed by the number and weight average molar mass values and the polydispersity (Table 2), even if a slight shift towards higher mass values and higher polydispersity can be noticed for the sample acquired during the high-flow period (M1). This trend is more pronounced when comparing the Podzol water samples, O16 and M12, representative for groundwater colloids. Sample M12 collected in May during the higher water level has higher weight and number average molar mass of 2230 and 1142 Da respectively, by contrast with the 1918 and 922 Da determined for sample O16. It is possible to observe a shift of the mass distribution toward larger values in
water research 44 (2010) 223–231
Table 4 – Weight average molar mass (±6%), number average molar mass (±8%), polydispersity index (±5%) and weight average gyration radius (±12%) of COM samples determined by AFlFFF–UV–LS detection. Averaged values were calculated using the signal acquired in the region of interest 0–4 min. The crossflow rate was 0.25 mL minL1. Sample/Type O16/Podzol soil solution M12/Podzol soil solution O7/creek named: Igarape Disjuncta O15/creek named: Igarape Bonito
Mw (103Da)
Mn (103Da)
PDI
RGw (nm)
467 386 344
308 273 248
1.5 1.4 1.4
42 40 23
464
287
1.6
17
UV response (Fig. 5B). Similarly, the characterization of the larger class of size by AFlFFF–UV–LS showed that the weight average molar mass of the sample M12 is slightly higher than the O16 (Table 4). Samples O7, O15 and O16 were collected on the same soil catena (Fig. 1); O15 comes from the Creek Bonito after the confluence with the Creek Disjuncta corresponding to water sample O7. The O5 and O1 were collected on the Jau River and Negro River, respectively. Comparing the sampling points O7, O15, O5 and O1, an evolution of the size properties of the colloidal material from small to large watercourses is revealed. There is a clear difference between COM size properties in small and large watercourses. Creek Bonito (O15) shows the similar Mw and size distributions (Fig. 2A) as well as polydispersity characteristics (Table 2) with its small affluent Creek Disjuncta (O7) and the podzol soil solution sample (O16). In contrast the molar mass distribution shifts toward lower values in the large watercourses as the Jau River and the Negro River. This trend was found using both fractionation systems
Fig. 4 – Fractograms obtained for the injection of 100 mL of the raw, filtered or pre-concentrated O16 sample. AFlFFF conditions – Membrane: RC with 1 kDa cut-off. Vc [ 3 mL minL1. Since the molar masses are related to retention time in a non-linear way, UV signals were converted to distributions multiplying the UV signal by the ratio dtr/dM.
229
AFlFFF and SEC (Tables 2 and 3). The weight and number average molar mass of the Jau and Negro rivers COM were approximately 40% lower than those found in the small streams (samples O15 and O7). When comparing the larger rivers (Negro and Jau rivers) between them, it seems that the Jau river (O5) is composed of lower sized COM than the Negro River (O1), in opposition to the trend discussed above but in agreement with the integrative nature of larger rivers as discussed previously. In previous works on the Amazon Basin COM from the Amazon and Negro Rivers were characterized with flow-FFF with UV absorbance (Benedetti, et al., 2002; Benedetti, et al., 2003). Samples collected on the Negro River, at the same location than sample O1 in October 1996 corresponding to a falling stage period; have number and weight average molar masses Mn and Mw of 1230 and 2350 Da, values higher than those found in the present study. Furthermore in 1996 the number and weight - average molar masses of the COM isolated from the Amazon River, 890 and 1030 Da respectively, were found to be very close to those obtained from the Negro River, which is an affluent of the former (Benedetti, et al., 2002). This supports the hypothesis on the decreasing of the COM size when passing from small to large watercourses. The differential fate of organic matter in the studied system proposed here and based on the role of size and reactivity are in agreement with the recent finding of (Bardy, et al., 2008; Do Nascimento, et al., 2008; Bardy, et al., 2009; Fritsch, et al., 2009). They demonstrated that the accumulation of organic matter in the Bh soil horizon corresponded to the trapping of highly aromatic humic acid like substances (Bardy, et al., 2008) while the fulvic acid like substances (i.e. smaller and less reactive on a molecular weight basis) would migrate deeper in the soil and reach the rivers (Bardy, et al., 2008; Fritsch, et al., 2009). At the margin and bottom part of the podzolic area, organic compounds of low aromaticity and with abundant oxygen groups may accumulate (Bardy, et al., 2008; Bardy, et al., 2009) and than be released by the quick recharge of the groundwater into the rivers (Do Nascimento, et al., 2008). The more aromatic humic acids would not be affected by these recharge sequences (Bardy, et al., 2008; Do Nascimento, et al., 2008). The changes in size fraction reported here, are also associated with differences in the overall chemical reactivity of the isolated fractions. Indeed, it was shown by potentiometric titrations that the large molecular weight fractions as found in the Negro River were carrying more reactive sites than the smaller molecular weight fractions isolated in the same samples or in other Amazon rivers samples (Benedetti et al., 2002, and Korshin et al, submitted). These properties will prompt a different behavior within the river with respect to interactions with suspended particles or to degradation and will affect the fate of the different COM fractions. In the small watersheds and in the soils solution the larger, more reactive (Benedetti et al., 2002, and Korshin et al, submitted) and hydrophobic organic matter fractions (Gadel et al., 2000; Bardy et al., 2008) could undergo preferential sorption onto positively charged minerals in soils and/or could also react with negatively charged clays via nitrogen rich fraction with amino groups (eNH2) (Croue´ et al., 2003). This scenario based on combined size distribution and reactivity is in agreement with
230
water research 44 (2010) 223–231
substances and for natural colloids from the same geographic location. Sample pre-treatments with ultrafiltration and reverse osmosis were found to not influence significantly the size characteristics of the samples.
Acknowledgements Warm thanks are extended to the Swiss National Science Foundation PP002–102640 for providing funding directly related to the present work. This work was also financed by the IRD-INSU-SNRS program ECCO. The authors would also like to acknowledge for support from the Hibam project (IRD/ CNPq/ANEEL).
references
Fig. 5 – Seasonal variability of COM illustrated on the fractograms obtained after the injection of 100 mL of samples M1 and O1 in panel A and of sample M12 and O16 in panel B versus the elution time. AFlFFF conditions – Membrane: RC with 1 kDa cut-off. Spacer: 500 mm. Vc [ 3 mL minL1. The absorbance signal height variation is consistent with measured differences in the content of dissolved organic carbon (DOC).
the regional chromatographic model developed by Hedges et al. (2000) that was vague as to how organic matter may interact with minerals. The characterization by size and molar mass of colloidal matter in natural freshwater is a first step to better understand the general dynamic of carbon in water system.
4.
Conclusions
The present study explores the capabilities of size exclusion chromatography and asymmetrical flow field-flow fractionation to characterize COM size and molar mass distributions. The fractionation of the samples collected in different periods of the year provided information about the temporal variability of the COM size distribution in the Amazon River fluvial system. No important differences were found in size distributions, whereas average molar mass was found to be weakly shifted towards higher values in samples acquired during the high-flow period. On the other hand the spatial variability of COM size characteristics revealed that COM average molar masses shift toward values up to 40% lower, when passing from small to large watercourses. Size and mass values were in accordance to those found in the literature for humic
Abbt-Braun, G., Lankes, U., Frimmel, F.H., 2004. Structural characterization of aquatic humic substances – the need for a multiple method approach. Aquatic Sciences 66, 151–170. Allard, T., Menguy, N., Salomon, J., Calligaro, T., Weber, T., Calas, G., Benedetti, M.F., 2004. Revealing the forms of iron in river-borne material from major tropical rivers of the Amazon Basin (Brazil). Geochimica Et Cosmochimica Acta 68, 3079–3094. Amon, R.M.W., Benner, R., 1996. Bacterial utilization of different size classes of dissolved organic matter. Limnology and Oceanography 41, 41–51. Balan, E., Allard, T., Fritsch, E., SE`lo, M., FalguE¨res, C., Chabaux, F., Pierret, M.-C., Calas, G., 2005. Formation and evolution of lateritic profiles in the middle Amazon basin: insights from radiation-induced defects in kaolinite. Geochimica Et Cosmochimica Acta 69, 2193–2204. Bardy, M., Bonhomme, C., Fritsch, E., Maquet, J., Hajjar, R., Allard, T., Derenne, S., Calas, G., 2007. Al speciation in tropical podzols of the upper Amazon Basin: a solid-state 27Al MAS and MQMAS NMR study. Geochimica Et Cosmochimica Acta 71, 3211–3222. Bardy, M., Derenne, S., Fritsch, E., 2009. Evolution of lipid abundance and molecular composition during the podzolisation of laterites in the upper Amazon basin. Biogeochemistry 92, 95–118. Bardy, M., Fritsch, E., Derenne, S., Allard, T., do Nascimento, N.R., Bueno, G.T., 2008. Micromorphology and spectroscopic characteristics of organic matter in waterlogged podzols of the upper Amazon basin. Geoderma 145, 222–230. Benedetti, M., Ranville, J., Ponthieu, M., Pinheiro, J., 2002. Fieldflow fractionation characterization and binding properties of particulate and colloidal organic matter from the Rio Amazon and Rio Negro. Organic Geochemistry 33, 269–279. Benedetti, M.F., Ranville, J.F., Allard, T., Bednar, A.J., Menguy, N., 2003. The iron status in colloidal matter from the Rio Negro, Brasil. Colloids and Surfaces a-Physicochemical and Engineering Aspects 217, 1–9. 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 and Technology 34, 1103–1109. Croue´, J-P, Benedetti, M.F., Violleau, D., Chin, Y.P., Leenheer, J.A., 2003. Characterization and Copper Binding of Humic and non Humic Organic Matter isolated from the South Platte River : evidence of the presence of nitrogeneous binding site. Environmental Science and Technology 37, 328–336.
water research 44 (2010) 223–231
Dahlqvist, R., Benedetti, M.F., Andersson, K., Turner, D., Larsson, T., Stolpe, B., Ingri, J., 2004. Association of calcium with colloidal particles and speciation of calcium in the Kalix and Amazon rivers. Geochimica Et Cosmochimica Acta 68, 4059–4075. do Nascimento, N.R., Bueno, G.T., Fritsch, E., Herbillon, A.J., Allard, T., Melfi, A.J., Astolfo, R., Boucher, H., Li, Y., 2004. Podzolization as a deferralitization process: a study of an Acrisol–Podzol sequence derived from Palaeozoic sandstones in the northern upper Amazon Basin. European Journal of Soil Science 55, 523–538. do Nascimento, N.R., Fritsch, E., Bueno, G.T., Bardy, M., Grimaldi, C., Melfi, A.J., 2008. Podzolization as a deferralitization process: dynamics and chemistry of ground and surface waters in an Acrisol–Podzol sequence of the upper Amazon Basin. European Journal of Soil Science 59, 911–924. Ertel, J.R., Hedges, J.I., Devol, A.H., Richey, J.E., Ribeiro, M.D.G., 1986. Dissolved humic substances of the Amazon River system. Limnology and Oceanography 31, 739–754. Fritsch, E., Allard, T., Benedetti, M.F., Bardy, M., do Nascimento, N. R., Li, Y., Calas, G., 2009. Organic complexation and translocation of ferric iron in podzols of the Negro River watershed. Separation of secondary Fe species from Al species. Geochimica Et Cosmochimica Acta 73, 1813–1825. Fritsch, E., Herbillon, A.J., Do Nascimento, N.R., Grimaldi, M., Melfi, A.J., 2007. From Plinthic Acrisols to Plinthosols and Gleysols: iron and groundwater dynamics in the tertiary sediments of the upper Amazon basin. European Journal of Soil Science 58, 989–1006. Gadel, F., Serve, L., Benedetti, M.F., Da Cunha, L.C., Blazi, J.L., 2000. Biogeochemical characteristics of organic matter in the particulate and colloidal fractions downstream of the rio Negro and Solimoes rivers confluence. Agronomie 20, 477–490. Guo, L.D., Macdonald, R.W., 2006. Source and transport of terrigenous organic matter in the upper Yukon River: evidence from isotope (delta C-13, Delta C-14, and delta N-15) composition of dissolved, colloidal, and particulate phases. Global Biogeochemical Cycles 20. Hedges, J.I., Cowie, G.L., Richey, J.E., Quay, P.D., Benner, R., Strom, M., Forsberg, B.R., 1994. Origins and processing of organic-matter in the Amazon River as indicated by carbohydrates and amino-acids. Limnology and Oceanography 39, 743–761. Hedges, J.I., Mayorga, E., Tsamakis, E., McClain, M.E., Aufdenkampe, A., Quay, P., Richey, J.E., Benner, R., Opsahl, S., Black, B., Pimentel, T., Quintanilla, J., Maurice, L., 2000. Organic matter in Bolivian tributaries of the Amazon River: a comparison to the lower mainstream. Limnology and Oceanography 45, 1449–1466. Hur, J., Schlautman, M.A., 2003. Molecular weight fractionation of humic substances by adsorption onto minerals. Journal of Colloid and Interface Science 264, 313–321. Jackson, B.P., Ranville, J.F., Bertsch, P.M., Sowder, A.G., 2005. Characterization of colloidal and humic-bound Ni and U in the ‘‘dissolved’’ fraction of contaminated sediment extracts. Environmental Science and Technology 39, 2478–2485. Janos, P., 2003. Separation methods in the chemistry of humic substances. Journal of Chromatography A 983, 1–18. Korshin G.V., Benedetti M.F., Claret F., Croue´ J-P., Fabbricino M., Gallard H., Scha¨fer T. Spectroscopic Study of the Protonation of Aquatic Natural Organic Matter from the Rio Negro Basin submitted to Chemical Geology.
231
Krusche, A.V., Martinelli, L.A., Victoria, R.L., Bernardes, M., de Camargo, P.B., Ballester, M.V., Trumbore, S.E., 2002. Composition of particulate and dissolved organic matter in a disturbed watershed of southeast Brazil (Piracicaba River basin). Water Research 36, 2743–2752. Litzen, A., Wahlund, K.G., 1991. Zone broadening and dilution in rectangular and trapezoidal asymmetrical flow field-flow fractionation channels. Analytical Chemistry 63, 1001–1007. McDonald, S., Bishop, A.G., Prenzler, P.D., Robards, K., 2004. Analytical chemistry of freshwater humic substances. Analytica Chimica Acta 527, 105–124. Molinier, M., Guyot, J.L., Callede, J., Guimaraes, V., Oliveira, E., Filizola, N., 1997. Hydrologie du bassin amazonien. In: Belin (Ed.), Environnement et de´veloppement en Amazonie bre´silienne, pp. 24–41. Moreira-Turcq, P.F., Seyler, P., Guyot, J.L., Etcheber, H., 2003. Characteristics of organic matter in the mixing zone of the Rio Negro and Rio Solimoes of the Amazon River. Hydrological Processes 17, 1393–1404. Ometto, J.P.H.B., Nobre, A.D., Rocha, H.R., Artaxo, P., Martinelli, L. A., 2005. Amazonia and the modern carbon cycle: lessons learned. Oecologia 143, 483–500. Pelekani, C., Newcombe, G., Snoeyink, V.L., Hepplewhite, C., Assemi, S., Beckett, R., 1999. Characterization of natural organic matter using high performance size exclusion chromatography. Environmental Science and Technology 33, 2807–2813. Perdue, E.M., Ritchie, J.D., Heinrich, D.H., Karl, K.T., 2003. Dissolved Organic Matter in Freshwaters, Treatise on Geochemistry. Pergamon, Oxford, pp. 273–318. Radambrazil, 1972–1978. Levamento de recursos naturais. Ministerio das minas e energia. Departamento Nacional de Produc¸ao Mineral, Rio de Janeiro. Reemtsma, T., These, A., 2003. On-line coupling of size exclusion chromatography with electrospray ionization-tandem mass spectrometry for the analysis of aquatic fulvic and humic acids. Analytical Chemistry 75, 1500–1507. Santschi, P.H., Roberts, K.A., Guo, L.D., 2002. Organic nature of colloidal actinides transported in surface water environments. Environmental Science and Technology 36, 3711–3719. Schimpf, M., Caldwell, K., Giddings, J.C., 2000. Field-flow Fractionation Handbook. John Wiley & Sons, Inc., Publication. Schmit, K.H., Wells, M.J.M., 2002. Preferential adsorption of fluorescing fulvic and humic acid components on activated carbon using flow field-flow fractionation analysis. Journal of Environmental Monitoring 4, 75–84. Tipping, E., 2005. Cation Binding By Humic Substances. Cambridge University Press. Town, R., Filella, M., 2002. Implications of natural organic matter binding heterogeneity on understanding lead(II) complexation in aquatic systems. Science of the Total Environment 300, 143–154. Wahlund, K.G., 2000. Asymmetrical flow field-flow fractionation. In: Schimpf, M., Caldwell, K., Giddings, J. (Eds.), Field-Flow Fractionation Handbook. John Wiley & Sons, Inc., Publication. Wilkinson, K.J., Lead, J.R., 2007. Environmental Colloids and Particles. IUPAC. Wyatt, P., 1993. Light-scattering and the absolute characterization of macromolecules. Analytica Chimica Acta 272, 1–40. Zhou, Q.H., Cabaniss, S.E., Maurice, P.A., 2000. Considerations in the use of high-pressure size exclusion chromatography (HPSEC) for determining molecular weights of aquatic humic substances. Water Research 34, 3505–3514.
water research 44 (2010) 232–242
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Disinfection by-products from halogenation of aqueous solutions of terpenoids Cynthia A. Joll*, Michael J. Alessandrino, Anna Heitz Curtin Water Quality Research Centre, Department of Applied Chemistry, Curtin University of Technology, GPO Box U1987, Perth WA 6845, Australia
article info
abstract
Article history:
We report the formation of trihalomethanes and other disinfection by-products from four
Received 1 April 2009
polyfunctional terpenoids during simulated chlorination of natural waters. Complex suites
Received in revised form
of products were identified by closed loop stripping analysis (CLSA)/gas chromatography–
30 August 2009
mass spectrometry (GC–MS) from halogenation of b-carotene and retinol. b-Ionone
Accepted 3 September 2009
appeared to be a key intermediate in the halogenation of b-carotene and retinol, reacting
Published online 8 September 2009
further under the reaction conditions to produce trans-b-ionone-5,6-epoxide and b-cyclocitral. Halogenation of the four terpenoids also produced trihalomethanes (THMs), most
Keywords:
likely through haloform reaction on methyl ketone groups within many of the interme-
Terpenoids
diates. Since halogenation of retinol produced a significant quantity of THMs at a slow
Disinfection by-products
reaction rate, retinol-based structures may possibly contribute to the slow reacting phase
Trihalomethanes
of THM formation in natural waters. Two polyhydroxyphenol model compounds were
Chlorine
halogenated for comparison. The only products identified by CLSA/GC–MS from
Polyhydroxyphenols
halogenation of 40 ,5,7-trihydroxyflavanone and ellagic acid were THMs. 40 ,5,7-Trihydroxy-
Drinking water
flavanone rapidly produced THMs, with an extremely high molar yield (94%) at pH 7. Terpenoids of the b-ionone and retinol type should be considered to be significant THM precursors, while 40 ,5,7-trihydroxyflavanone has been shown to be an extremely significant THM precursor, potentially present within natural organic matter in water treatment processes and distribution systems. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Terpenoids are produced by a wide variety of plants, animals and microorganisms for functional, defence and communication purposes. In plants, essential oils, latexes and resinous exudates are often composed mainly of terpenoids and terpenoids constitute important components of many wood extractives. Terpenoids, also referred to as isoprenoids, are defined as materials with molecular structures containing carbon backbones made up of isoprene (2-methylbuta-1,3diene) units and include degradation products of terpenoids in
which carbon atoms may have been lost through chemical and biochemical processes. While terpenoids occur widely in nature, only very recently have they been considered as precursors of natural organic matter (NOM). Terpenoids were earlier proposed to be incorporated into soil humus based on infrared spectral comparisons (Stepen and Korsunova, 1988). In 2003, Leenheer et al. (2003) provided cross polarization magic angle spinning (CPMAS) 13C NMR spectroscopic evidence of terpenoid precursor contribution to six dissolved organic matter (DOM) fractions derived from each of a river, a lake and its infiltrated
* Corresponding author. Tel.: þ61 8 9266 7229; fax: þ61 8 9266 2300. E-mail address: [email protected] (C.A. Joll). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.003
water research 44 (2010) 232–242
groundwater (three samples) and a separate groundwater source. Electrospray ionization mass spectrometry on the DOM sample from the groundwater source also supported the hypothesis of degraded terpenoid contribution to the sample. In further work, Leenheer et al. (2007) found that terpenoids constituted a large part of the DOM in recharge waters of the Santa Ana River Basin. McIntyre et al. (2005) proposed terpenoid input into a groundwater-derived hydrophobic acid sample, based on CPMAS 13C NMR spectroscopic evidence, consistent with the overlying vegetation (Eucalyptus) being rich in terpenoids. Terpenoids were also identified by gas chromatographic–mass spectrometric (GC–MS) analysis of the ester-based fractions of humic acid isolated from a volcanic soil (Fiorentino et al., 2006). Badin et al. (2008) observed triterpanes upon GC–MS analysis of dichloromethane extracts of fractions of natural and anthropogenic OM from urban storm water sediments. There have been many literature reports of direct identification of a variety of terpenoids, both polar and nonpolar, in water sources. The types of terpenoids identified primarily depend upon the plant, animal and microorganism input to the water source. For example, in 1974, Grob and Grob (1974) reported that terpenoids, in low ng L1 concentrations, including hydrocarbons, aldehydes, ketones and alcohols, were found in three surface water sources studied. Three naturally occurring terpenoids, reported to be resulting from normal biological processes taking place in the river, were identified in Delaware River water during summer and/or winter sampling by dichloromethane extraction, liquid chromatographic cleanup and GC–MS analysis (Sheldon and Hites, 1978). These terpenoids were 6,10,14-trimethyl-2-pentadecanone (probably resulting from oxidative degradation of phytol), a-terpineol and chlorophyll (which was observed in the spectra as phytadienes produced in the injection port by pyrolysis of the phytol ester part of chlorophyll) (Sheldon and Hites, 1978). A number of terpenoids, attributed to phytoplankton, and possibly conifer, sources, were found in Berlin’s surface water sources (Chorus et al., 1992). Monoterpenoids (e.g., indomyrmectin, limonene), sesquiterpenoids (e.g., farnesol) and their derivatives (e.g., di-epi-a-cedrene epoxide) were tentatively identified, through dichloromethane extraction and GC– MS analysis, as natural organic matter (NOM) constituents of the dissolved phase from surface runoff from individual irrigated agricultural fields, while a triterpenoid phytosterol from plant waxes was tentatively identified as being sorbed to suspended particles in the runoff samples (Pedersen et al., 2002). Geosmin, b-cyclocitral, b-ionone, geranylacetone, limonene and 2-methylisoborneol (MIB) have been detected in Australian water sources (Jones and Korth, 1995). Biological removal of terpenoids has been reported to be both effective and slow, depending on the structure of the terpenoid (e.g. Grob and Grob, 1974; Ju¨ttner, 1995a). Despite the potential for biodegradation / biological removal of some terpenoids, terpenoids have been detected in distributed drinking water. For example, Grob and Grob (1974) detected terpenoids in the drinking water derived from the three surface water sources, and terpenoids (including limonene, eucalyptol and 2 unidentified terpenes) have been identified in drinking water from Co´rdoba, Spain (Aramendı´a et al., 1998). We have observed terpenoids, such as farnesol acetate, geranyl acetate
233
and b-cyclocitral derivatives, in various chlorinated surface waters from towns in southern Western Australia (e.g. Heitz, 1997). These compounds are oxidative degradation products of carotenoids, such as b-carotene, and are believed to be indicators of considerable microbial activity in the water body (Ju¨ttner, 1992). The terpenoids, geosmin and 2-methylisoborneol (MIB), are responsible for the majority of unpleasant odours in potable water, with cyanobacteria and actinomycetes being their major source (Ju¨ttner, 1995b; Ho et al., 2004). Chlorination of terpenoids in aqueous solutions to simulate disinfection of drinking waters has not been well-studied to date. Reactions of halogens, such as chlorine and bromine, with aquatic organic matter are numerous and include oxidation, substitution and addition reactions, and the haloform reaction from enolizable carbonyl compounds. For reaction of halogens with terpenoids, the terpenoids must contain reactive functional groups, such as benzenes and phenols, alkenes, methyl ketones, b-diketones and primary and secondary alcohols. In fact, the haloform reaction was classically used as a degradative reaction to elucidate the structure of mono-, di- and sesqui-terpenoids through reaction of halogens with methylketone intermediates, derived from initial oxidation of the terpenoids, allowing identification of carboxylic acid derivatives of the original terpenoids (Fuson and Bull, 1934). Geosmin and MIB were found to be fairly resistant to removal by chlorine and chlorine dioxide treatment, because these terpenoids each contain only one functional group, a tertiary alcohol, which is not susceptible to oxidation (Lalezary et al., 1986). a-Terpineol, which contains an alkene and a tertiary alcohol group, when chlorinated in water, was reported to give mainly monochloro derivatives (chlorohydrins) at pH 2, with an epoxide becoming a major product at pH 10 (Kopperman et al., 1976). Larson and Marley (1988) separately treated camphene, limonene, a-pinene and b-pinene with a 10 molar excess of aqueous hypochlorite solution at pH 2, to simulate acid bleaching of paper pulp, and at pH 8, to simulate drinking water treatment. At pH 8, the major products were reported to be ring-opened, oxygenated products, with small amounts of mono- and di-chloro derivatives (Larson and Marley, 1988). Mono- and di-chlorocamphenes were produced from camphene and hypochlorous acid in aqueous acetone (Buchbauer et al., 1984). Hoehn et al. (1980) reported a correlation between the concentration of chlorophyll-a measured in a reservoir and the concentration of THMs, and found that high yields of THMs were produced from both algal biomass and algal extracellular products, possible sources of terpenoids. The objective of the present study was to investigate the role of some polyfunctional terpenoids in the formation of DBPs, particularly THMs, during halogenation reactions that occur upon chlorination of natural waters. The terpenoid model compounds, b-carotene, retinol, b-ionone and geranyl acetate (Fig. 1), were chosen as representatives of the carotenoid family and their oxidative degradation products, terpenoids which we have identified in local drinking waters. Closed loop stripping analysis followed by GC–MS was conducted for qualitative identification of halogenation products from the model compounds. Solid-phase microextraction followed by GC–MS was utilized for quantitative analysis of
234
water research 44 (2010) 232–242
a
Terpenoid Model Compounds H3C H3C
CH3 H3C CH3
H3C
CH3
beta-carotene
CH3
H3C
CH3
CH2OH
O
retinol
CH3
CH3
beta-ionone
O O geranyl acetate
b
Polyhydroxyphenol Model Compounds OH HO
O OH
O
O HO OH
O
4',5,7-trihydroxyflavanone
OH HO
O
ellagic acid O
Fig. 1 – Structures and names of the a) terpenoid and b) polyhydroxyphenol model compounds.
the THMs formed from halogenation of the model compounds. The reactivities of the terpenoid model compounds for THM formation were compared with those of a series of polyhydroxyphenols, containing functionalities which are well-known to form high concentrations of THMs (Rook, 1976; Boyce and Hornig, 1983; Gallard and von Gunten, 2002). These polyhydroxyphenols, 40 ,5,7-trihydroxyflavanone and ellagic acid (Fig. 1), are found free and/or combined in plants. Reaction pathways and mechanisms for the formation of a series of intermediates from the terpenoids are also proposed.
2.
Materials and methods
2.1.
Standard solutions of model compounds
Standard solutions (1 g L1) of 40 ,5,7-trihydroxyflavanone (naringenin; Sigma), ellagic acid (Fluka), b-carotene (Roche), b-ionone (Aldrich), all-trans-retinol (Aldrich; referred to subsequently as retinol) and geranyl acetate (Aldrich) in methanol (HPLC grade), and b-carotene and ellagic acid (for the THM quantification study) in a 50:50 mixture of methanol and redistilled dichloromethane (DCM), were prepared.
2.2. Aqueous halogenation of model compounds for closed loop stripping analysis Solutions of bromide ion (0.5 mg L1 (6 mM) as bromide; reagent used was KBr), phosphate buffer (pH 7, 100 mL of stock buffer solution: KH2PO4 þ Na2HPO4), sodium hypochlorite
(10 mg L1 (280 mM) as Cl2), followed by the model compound (1.0 mg L1; 1.9–5.2 mM), were prepared in redistilled Milli-Q water (2 L). The reactions were allowed to proceed at room temperature (25 C) in darkness with constant stirring. After 0.5, 1, 6 and 24 h, subsamples (200 mL) of the mixture were quenched with an aliquot (1 mL) of aqueous sodium thiosulfate solution (8 g L1 as S2O2 3 ). The thiosulfate quenching reagent was added at a molar ratio of thiosulfate to initial Cl2 concentration of 1.3:1. All model compound reactions were carried out in duplicate and with procedural blanks.
2.3. Closed loop stripping analysis followed by gas chromatography–mass spectrometry for identification of purgeable products from the reaction mixture A similar set-up to the closed loop stripping procedure developed by Grob and Zu¨rcher (1976) was used for the isolation and concentration of volatile analytes from the reaction mixtures. An aliquot (10 mL) of surrogate standard solution (1,2,3,5-tetramethylbenzene (Aldrich; 95%) in redistilled DCM; 55.3 ng mL1) was added directly into each quenched reaction mixture (100 mL). Closed loop stripping analysis (CLSA) was carried out using a sample purge temperature of 60 C (water bath), purge gas (air) at a flow of approximately 425 mL min1 and a Grob tube (carbon trap) containing activated carbon (1.5 mg) at a temperature of 65 C. Analytes were eluted from the Grob tube with DCM (total 30 mL). In separate extractions, aliquots of DCM (2 9 mL, 2 6 mL) were placed onto the carbon filter and extracted as described by Heitz (2002). The solvent extracts were combined in a microvial (100 mL) containing an aliquot (10 mL) of internal standard
water research 44 (2010) 232–242
(1-chlorohexadecane (Merck-Schuchardt) in hexane (AR HPLC grade; 7.0 ng mL1)). After all four individual extractions, an additional aliquot (20 mL) of DCM was added to the extracts. The reaction products in the extract were separated and identified by GC–MS using a Hewlett-Packard (HP) 6890 GC interfaced to a HP 5973 mass selective detector, operating in full scan mode. The sample extract aliquot (1 mL) was delivered directly into the column via a HP on-column injector. The column was a ZB-5 (Phenomenex; 60 m 0.25 mm i.d.; phase thickness 0.25 mm;), helium was the carrier gas, and the temperature program was: 30 C (6 min) to 230 C at a rate of 5 C min1, followed by 15 C min1 to 310 C (10 min).
2.4. Halogenation of aqueous solutions of model compounds for quantification of THM formation Reaction mixtures at pH 7 were prepared as described for CLSA and allowed to react for 168 h. To ensure there was a free chlorine residual after 168 h, an initial sodium hypochlorite concentration of 13 mg L1 (180 mM; as Cl2) was used for 40 ,5,7-trihydroxyflavanone and ellagic acid, and 6 mg L1 (85 mM; as Cl2) was used for retinol, b-carotene, b-ionone and geranyl acetate. Reactions were also conducted at pH 9 using a borax/HCl buffer. After 168 h, the residual free chlorine was measured and each mixture was quenched with a calculated aliquot of sodium thiosulfate solution (8 g L1 as S2O2 3 ) such that the molar ratio of thiosulfate to final chlorine concentration was 1:1. The free chlorine concentration was then measured to ensure no residual remained. The reactions were carried out in duplicate, with procedural blanks. In order to examine the THM formation potential of 40 ,5,7-trihydroxyflavanone and retinol, additional samples from these reaction mixtures were quenched at times 1, 24, 48, 96, 120, 144 and 168 h, at both pH 7 and 9.
2.5. Solid-phase microextraction / GC–MS analysis of THMs After quenching of the reaction mixture, THMs were recovered from the samples via manual headspace solid-phase microextraction (SPME). An aliquot (30 mL) of a surrogate standard solution (1,2-dibromopropane (Aldrich) in methanol (HPLC grade): 50 mg L1; 250 mM) was added directly into the sample (30 mL) contained in a 40 mL sample vial. A magnetic stirrer bar and sodium sulfate (5 g) were then added and the vial was capped. Headspace SPME using a 100 mm polydimethylsiloxane (PDMS) fibre (15 min) was followed immediately by GC–MS analysis. The THMs and the surrogate standard were separated using a HP 5890 gas chromatograph interfaced to a HP 5971 mass selective detector. Selected ions (m/z) were 83, 85, 96, 121, 123, 127, 129, 131, 173 and 175. The SPME fiber was injected manually via a split-splitless injector at 240 C. The column was a ZB-5 (Phenomenex; 30 m 0.25 mm i.d.; phase thickness 1 mm), helium was the carrier gas, and the temperature program was: 0 C (liquid CO2, 3 min) to 120 C at a rate of 8 C min1, followed by 15 C min1 until 305 C (5 min).
3.
235
Results and discussion
3.1. Closed loop stripping analysis of products from halogenation of terpenoid and polyhydroxyphenol model compounds Halogenation was conducted on aqueous solutions of a range of terpenoid and polyhydroxyphenol model compounds at a concentration of 1 mg L1 (1.9–5.2 mM), with the addition of bromide ion (0.5 mg L1 (6 mM) as bromide) and sodium hypochlorite (free chlorine; 10 mg L1 (280 mM)) at pH 6.9 at room temperature (25 C). Closed loop stripping analysis (CLSA) was performed on quenched subsamples of the reaction mixtures at reaction times of 0.5, 1, 6 and 24 h, but this paper focuses on the final reaction time (24 h). Quenching was conducted with thiosulfate solution, which is suitable for analysis of THMs and some other DBPs, although it may cause some decomposition of other DBPs, e.g. some haloacetonitriles (Clesceri et al., 1998). CLSA allowed analysis of volatile reaction products which could be purged from the reaction mixture; nonvolatile products can not be detected by this method. Chloroform was not analysed, since it co-eluted with the solvent (DCM). These experiments were primarily for qualitative purposes to identify reaction products; however, surrogate and internal standards were added for semiquantitative purposes. The reaction and analysis were carried out in duplicate for each model compound and typical chromatograms are presented here. The structures and names of the five model compounds, and geranyl acetate which was used in the subsequent quantitative study, are presented in Fig. 1. The presence and absence of six halogenation and oxidation products in the chromatograms from reaction times of 24 h for the five model compounds are indicated in Table 1. The concentrations used in these experiments were chosen with specific reference to our challenging local water treatment conditions and to allow detection and identification of reaction products from the less reactive model compounds. In water supplies in Western Australia, the dissolved organic carbon (DOC) concentration is often particularly high (5–40 mg L1) and the concentrations of bromide ion commonly range from 0.2 to 0.5 mg L1. Treatment of these waters varies from disinfection with chlorine only to conventional alum coagulation after magnetic ion exchange resin (MIEX) treatment, such that the DOC concentration in treated water being subjected to chlorination can range from 1 to 5 mg L1. While terpenoids or polyhydroxyphenols would form only a fraction of this total DOC in the water, the concentration of the model compounds in this study (1 mg L1) was chosen to model this overall DOC concentration in order to ensure sufficient concentrations of products to allow detection by CLSA. To ensure adequate disinfection throughout the distribution system, final chlorination doses can be up to 12 mg L1, sometimes resulting in total THM concentrations close to the Australian Drinking Water Guideline value of 250 mg L1. The concentrations of free chlorine and bromide used in these experiments were chosen to maximise the concentrations of products formed, while still representing concentrations which are sometimes present in the local drinking water systems.
236
water research 44 (2010) 232–242
Table 1 – Presence (O) or absence (3) of six products from CLSA followed by GC–MS analysis of the product mixtures from halogenation of the five model compounds after 24 h. Model Compound
CHBrCl2
CHBr2Cl
CHBr3
b-Cyclocitral
Trans-b-Ionone5,6-epoxide
b-Ionone
b-Carotene Retinol b-Ionone 40 ,5,7-Trihydroxyflavanone Ellagic Acid
U U U
U U U U U
U U U U U
U U U
U U U
U
Australia (e.g. Heitz, 1997), suggesting the possible presence of b-carotene-type terpenoid precursors in these surface waters. Three other products, b-cyclocitral, b-ionone and transb-ionone-5,6-epoxide (Fig. 3), were of particular interest because of their relative abundances over the 24 h reaction period (0.5, 1 and 6 h: results not shown). The relative abundances of b-cyclocitral and b-ionone were highest after 0.5 h, and decreased gradually to be present in low abundance after 24 h. The moderately low relative abundance of trans-b-ionone-5,6-epoxide was constant over the 24 h reaction period. These trends indicated that b-cyclocitral and b-ionone may have been rapidly formed intermediate products, which were then slowly converted into other products over time, in the halogenation of b-carotene. Trans-b-ionone-5,6-epoxide is a likely product of further reaction of b-ionone (Fig. 3). Its relatively constant abundance over the reaction period indicates that, while it appears to have been formed from b-ionone, it must itself be converted into other products. No possible products from b-cyclocitral, including THMs, were found to increase in abundance over the reaction period, suggesting that THMs were not degradation products of b-cyclocitral and that the degradation products of b-cyclocitral could not be detected by this analytical method.
10
Internal Std ß-Ionone
Trans-ß-ionone-5,6-epoxide
ß-Cyclocitral
Surrogate Std
3,5,5-Trimethyl-2-cyclohexen-1-one
1,8-Cineole
2,2,6-Trimethylcyclohexanone
3,3-Dimethylcyclohexanone
Styrene (vinyl benzene)
Bromoform
1,2-Dibromoethane
Dibromochloromethane
Contaminant in Solvent
The only products identified by CLSA/GC–MS from the polyhydroxyphenol model compounds (40 ,5,7-trihydroxyflavanone and ellagic acid) were three trihalomethanes (bromodichloromethane, dibromochloromethane and bromoform) (Table 1). Only a few, very minor, other peaks were observed in these chromatograms. 1,3-Dihydroxybenzenes are well-known for their prolific THM production upon halogenation (e.g. Rook, 1976; Boyce and Hornig, 1983). Phenols are also known to produce THMs, albeit in lower yields (e.g. Gallard and von Gunten, 2002). 1,3-Dihydroxybenzene and phenolic moieties are present in the structures of 40 ,5,7-trihydroxyflavanone and ellagic acid, accounting for the formation of THMs from these model compounds. In contrast, more complex suites of products were identified by CLSA/GC–MS from halogenation of the terpenoid model compounds, b-carotene and retinol. To illustrate this, the total ion chromatogram (TIC) of the mixture obtained from halogenation of b-carotene after 24 h is presented in Fig. 2, including the identified reaction products. Of the identified reaction products from b-carotene and retinol, 1,8-cineole, bcyclocitral and a variety of trimethylcyclohexanones and trimethylcyclohexenones have been found in various chlorinated surface waters from towns in southern Western
46 Retention time (mins)
Fig. 2 – Total ion chromatogram (TIC) from GC–MS analysis of the reaction products from halogenation of b-carotene after 24 h.
237
water research 44 (2010) 232–242
H H3C
CH3
CH3
H O H3C
CH3
CH3
CH3 H3C OH
H3C +
-H CH3
CH3
X OH
beta-carotene or retinol
X
H3C
CH3
CH3
X
CH3
H3C
CH3
CH3
+
OH OH
+
H
CH3
CH3
H3C CH3
H3C
-H
O
O H
CH3
H
CH3
CH3
O H CH3
H3C H3C
CH3
1. H2O + 2. -H CH3
OH
HOX / H2O
beta-ionone HOX / H2O H3C H3C
CH3
X
OH
H3C H3C
CH3
O
O X
OH CH3
CH3
CH3
HOX / H2O H3C H3C
X
OH
CH3
H3C
CH3 O
OH
CH3
OH OH
H3C
O CH3
1. H2O + 2. -H
trans-beta-ionone-5,6-epoxide
CH3
O
O
other DBPs (including THMs)
H3C
CH3
OH
CH3 O
OH CH3
OH CH3
H3C
CH3
CH3
CH3
1. epoxide formation 2. diol formation 3. oxidative cleavage H3C H3C
CH3
H3C CH3 O
CH3 O other DBPs (possibly THMs) CH3
2,2,6-trimethylcyclohexanone
other DBPs (THMs unlikely)
H CH3 beta-cyclocitral
Fig. 3 – Reaction pathways, including some mechanisms, for the formation of 2,2,6-trimethylcyclohexanone, b-ionone, trans-b-ionone-5,6-epoxide and b-cyclocitral from b-carotene and retinol precursors.
Halogenation of b-ionone itself produced a less complex mixture of products. A very high, relatively constant, abundance of trans-b-ionone-5,6-epoxide was produced, confirming the reaction pathway of b-carotene to b-ionone to the epoxide. b-Cyclocitral was also produced from b-ionone, with its relative abundance showing the same decreasing trend as in the bcarotene reaction. The only other identified products were bromodichloromethane and bromoform which were formed in very low, gradually increasing, abundance after 6 and 24 h. The structures of b-carotene and retinol are very similar (Fig. 1), with retinol being much more water soluble than b-carotene, since b-carotene can be enzymatically cleaved
into two molecules of retinol. Not surprisingly, halogenation of retinol yielded a similar suite of identified products, in different abundances, to halogenation of b-carotene, except that no 3,3-dimethylcyclohexanone was formed. b-Cyclocitral was initially formed in moderately high abundance, but decreased to a low abundance after 24 h, again with no degradation products evident. b-Ionone was initially present but its abundance decreased over time until it was all consumed after 6 h. Concurrently, one of the likely products from b-ionone, trans-b-ionone-5,6-epoxide, was again present in relatively constant, moderately low abundance over the 24 h period.
water research 44 (2010) 232–242
3.2. Quantification of trihalomethanes produced from halogenation of terpenoid and polyhydroxyphenol model compounds Each of the six terpenoid and polyhydroxyphenol model compounds (Fig. 1), at a concentration of 1 mg L1 (1.9–5.2 mM), were subjected to halogenation, with the addition of bromide ion (0.5 mg L1 (6 mM) as bromide) and sodium hypochlorite (free chlorine; 13 mg L1 (180 mM) for polyhydroxyphenols and 6 mg L1 (85 mM) for terpenoids), at pH 7 and 9. The
concentrations of individual and total THMs were determined over a 7 day period for 40 ,5,7-trihydroxyflavanone and retinol, and after 7 days for the other four model compounds, after quenching of the free chlorine equivalent residual with the exact concentration of sodium thiosulfate solution required. A time period of 7 days was chosen to be consistent with the methods standardised by Clesceri et al. (1998), other work in our laboratory and local water distribution times. Retinol was chosen for detailed study of the rate of THM formation from the terpenoids, since it appeared to produce the most THMs in the semi-quantitative study. These quantitative analyses were carried out in duplicate for each model compound and typical formation curves are presented here. Total THM formation curves for 40 ,5,7-trihydroxyflavanone and retinol at pH 7 and at pH 9 are presented in Fig. 4 a) and 4 b), respectively. The chlorine dose, 7 day oxidant demand, molar concentrations of individual and total THMs, and specific and conventional yields of total THMs for all halogenation experiments at pH 7 and pH 9 are presented in Tables 2 and 3, respectively. As expected for a substrate containing phenolic and 1,3dihydroxybenzene moieties, 40 ,5,7-trihydroxyflavanone produced THMs in high abundance (Fig. 4). THM formation from 40 ,5,7-trihydroxyflavanone appeared to include a period of very rapid production, followed by a period of slower production, corresponding most likely to the formation of
a
4000
Total THM C oncentration (nM)
b-Ionone appears to be the key intermediate in the halogenation of b-carotene and retinol, reacting further under the reaction conditions to produce trans-b-ionone-5,6-epoxide and, to a much lesser extent, b-cyclocitral. Another product, 2,2,6-trimethylcyclohexanone, identified in the reaction mixtures from b-carotene and retinol, still contains the structural features of the ring system of b-carotene, retinol and b-ionone. Reaction pathways, including reaction mechanisms, for formation of these products are proposed in Fig. 3. Electrophilic addition of hypochlorous or hypobromous acid (represented as HOX in Fig. 3) to the double bond of b-carotene or retinol, followed by nucleophilic attack of water on the intermediate carbocation, produces the halohydrin with Markovnikov-type regiochemistry. Internal SN2 attack of the nucleophilic hydroxide group results in an epoxide intermediate. Epoxide formation from the chlorohydrin was also observed at pH 10, and to a lesser extent at lower pH, in the aqueous chlorination of a-terpineol (Kopperman et al., 1976). Nucleophilic attack of water on the least hindered carbon of the epoxide produces the vicinal diol, which is subject to oxidative cleavage under the reaction conditions (March, 1992) to produce the key intermediate, b-ionone. Similar reactions could produce b-cyclocitral and trans-b-ionone5,6-epoxide from b-ionone (as depicted in Fig. 3). b-Cyclocitral could also be produced more directly from b-carotene, retinol or from another longer chain intermediate, rather than from b-ionone. Since 2,2,6-trimethylcyclohexanone was not observed in the b-ionone product mixture, it is likely to arise from reaction of b-carotene or retinol or from another longer chain intermediate, through acid-catalysed hydration of the cyclic double bond, followed by another similar series of halogenation reactions (Fig. 3). b-Ionone, trans-b-ionone-5,6-epoxide and many other possible reaction intermediates contain a methyl ketone functional group (Fig. 3), a moiety which is well-known to produce THMs (e.g. Morris and Baum, 1978) and a carboxylic acid via the haloform reaction. THMs were produced from halogenation of these terpenoids (Table 1). Of the three overall reaction pathways depicted in Fig. 3, the pathway from bionone was confirmed to produce THMs, the pathway including b-cyclocitral did not appear to produce THMs, while THM formation from the pathway including 2,2,6-trimethylcyclohexanone remains unclear. Having established that halogenation of b-carotene-based terpenoids can produce THMs, it was important to investigate the quantity of THMs produced, and the rate of their formation over 7 days, to determine if these terpenoid structures could be significant THM precursors within NOM in water treatment processes and distribution systems.
3500 3000 2500 Trihydroxyflavanone 2000
Retinol
1500 1000 500 0 0
50
100
150
Time (hrs)
b Total THM concentration (nM)
238
5000 4500 4000 3500 3000
Trihydroxyflavanone
2500
Retinol
2000 1500 1000 500 0 0
50
100
150
Time (hrs) Fig. 4 – Total THM formation curves for 40 ,5,7trihydroxyflavanone and retinol at a) pH 7 and b) pH 9.
239
water research 44 (2010) 232–242
Table 2 – Chlorine dose, oxidant demand, molar concentrations of individual and total THMs and specific and conventional yields of THMs produced after 7 days for all halogenation experiments at pH 7. Model compound
b-Carotene Retinol b-Ionone Geranyl Acetate 40 ,5,7-Trihydroxyflavanone Ellagic Acid
[CHCl3] (nM)
Chlorine dose (mM Cl2)
Oxidant demand (mM Cl2 equivalents)
85 85 85 85 180
36 61 34 60 150
<3 64 68 6 1640
180
150
26
[CHCl2Br] (nM)
[CHClBr2] (nM)
[CHBr3] (nM)
Total [THMs] (nM)
Specific yield total THMs (mmol/ mol C)
Yield total THMs (100 mol THMs/ mol model cpd) (%)
7 279 614 26 1270
66 245 307 22 520
11 64 28 6 53
84 652 1020 60 3480
1 9 15 1 84
5 19 20 1 94
204
597
162
989
21
30
THMs from the 1,3-dihydroxybenzene and phenolic moieties, respectively, within this model compound. In kinetic studies, Gallard and von Gunten (2002) found that THM precursors in NOM could be divided into fast and slow reacting fractions. The rapid production of THMs from 40 ,5,7-trihydroxyflavanone is consistent with their hypothesis that meta-dioxygenated (resorcinol-type) sites could be, at least in part, responsible for the fast reacting THM precursors in NOM. Halogenation of retinol produced a lower, but still very significant, quantity of THMs with an apparently slower reaction rate. These results are consistent with our semiquantitative CLSA study (Section 3.1) where THMs were formed from retinol, but various other intermediates, some of which were methyl ketones, and therefore recognised THM precursors, were also identified over the first 24 h. This was in contrast to the nearly exclusive production of THMs from 40 ,5,7-trihydroxyflavanone. The reactions to convert retinol (and the other related terpenoid structures) into the methyl ketone intermediates appeared to be slow reactions and methyl ketones are reported to react slowly to produce THMs (e.g. Gallard and von Gunten, 2002). Thus, it is proposed that halogenation of retinol-type terpenoids may possibly contribute to the slow reacting phase of THM formation in natural waters.
Increasing the pH from 7 to 9 resulted in an increase in total THM formation from both 40 ,5,7-trihydroxyflavanone and retinol, with increased rates of THM production in the initial fast phase (Fig. 4). In kinetic studies of the halogenation of phenols, apparent second-order rate constants for reaction of chlorine and bromine with the phenols were found to be at a maximum around pH 8-9 (e.g. Acero et al., 2005), with the rate constants for the reactions with bromine being much faster than for the reactions with chlorine (Acero et al., 2005). This pH range of maximum halogenation appears to correspond to the presence of the highest proportion of phenoxide ions (Acero et al., 2005) and of the major electrophilic species HOBr (pKa 8.6), with the weaker electrophile OCl- (pKa HOCl 7.5). The increase in total THM formation for 40 ,5,7-trihydroxyflavanone at pH 9 in the current study is consistent with the most favoured reaction pathway between HOBr and deprotonated forms of 40 ,5,7-trihydroxyflavanone at pH 9, with weak competition from the much less electrophilic OClspecies. At pH 9, retinol is unlikely to be deprotonated, but the basic conditions would also promote the haloform reaction on methyl ketone intermediates, leading to increased THM formation (Fuson and Bull, 1934). Comparison of the molar concentrations of total THMs produced after 7 days from the six model compounds (Tables 2
Table 3 – Chlorine dose, oxidant demand, molar concentrations of individual and total THMs and specific and conventional yields of THMs produced after 7 days for all halogenation experiments at pH 9. Oxidant demand [CHCl3] [CHCl2Br] [CHClBr2] [CHBr3] Total Specific Model compound Chlorine dose Yield (mM Cl2 equivalents) (nM) (nM) (nM) (nM) [THMs] yield (mM Cl2) total (nM) total THMs THMs (100 mol (mmol/ THMs/ mol mol C) model cpd) (%) b-Carotene Retinol b-Ionone Geranyl Acetate 40 ,5,7-Trihydroxyflavanone Ellagic Acid
85 85 85 85 180
25 35 31 53 100
180
74
<3 18 111 <3 2170 25
9 169 677 26 1460
253 483 770 65 808
172 438 256 60 243
434 1110 1810 151 4680
6 16 27 2 84
23 32 35 3 127
173
481
325
1000
22
30
240
water research 44 (2010) 232–242
and 3) shows that 40 ,5,7-trihydroxyflavanone produced 3.4 and 2.6 times the molar concentration of THMs of any other model compound. In the only previous study of THM formation from 40 ,5,7-trihydroxyflavanone (naringenin), under similar conditions of chlorination but without the addition of bromide ion, it was found that 270–298 mg THM/mg C, presumably all as chloroform, was formed after 24 h (Matsuo et al., 1989). In the current study, the 7 day total THM mass concentration was 526 mg L1, corresponding to 790 mg THM/mg C, where the THMs were comprised of the four chlorinated and brominated THMs. The current results are consistent with more THMs being formed in the presence of the more reactive halogenating agent HOBr over a longer reaction period than in the Matsuo et al. (1989) study. b-Ionone, ellagic acid, and retinol produced the next highest molar concentrations of THMs at pH 7, with significant concentrations formed from these model compounds. Ellagic acid is a cyclic ester dimer of gallic acid which we have found in another, unpublished, study to be a moderate producer of THMs. At pH 9, the terpenoids, b-ionone and retinol, produced higher molar concentrations of THMs than ellagic acid. At pH 7, only very low molar concentrations of THMs were formed from b-carotene and geranyl acetate, although the concentrations increased at pH 9. For the terpenoid model compounds, increasing the pH from 7 to 9 would again promote THM formation through base catalysis of the haloform reaction of methyl ketone intermediates. The specific yields of total THMs, calculated as mmol total THMs / mol carbon (C) of the model compound, are presented in Tables 2 and 3. The yields of total THMs (Tables 2 and 3) represent the conventional yield calculation of the conversion of the model compound to total THMs, calculated as the sum of the moles of total THMs produced divided by the initial moles of model compound present, as a percentage. This molar yield parameter does not take into account the number of carbons which are likely THM precursor sites within the model compound, since this number is not definitively known. The specific yields of total THMs and the yields of total THMs generally followed the same trends as the molar concentrations of total THMs formed after 7 days at both pH 7 and 9. Retinol and b-ionone, however, produced essentially the same yields of total THMs (approximately 20% at pH 7 and 33% at pH 9), possibly indicating the very similar reaction pathways to THMs from these model compounds. The pH did not affect the specific yields of total THMs (21–22%) or the yields of total THMs (30%) from ellagic acid, as was found with the molar concentrations of total THMs produced (989– 1000 nM). Extremely high yields of total THMs were observed for 40 ,5,7-trihydroxyflavanone, 94 and 127%, at pH 7 and 9, respectively. Reaction yields greater than 100% are possible if more than one carbon in the model compound reacts to form THMs. The greater than 100% conversion at pH 9 is consistent with the hypothesis that THMs are being formed from more than one carbon within the 40 ,5,7-trihydroxyflavanone structure, i.e. from the 1,3-dihydroxybenzene and phenolic moieties within the structure. Indeed, this may also occur at pH 7, with lower conversions from more than one carbon site, resulting in a total conversion of 94%. As comparison, Rook (1977) found that the yield of CHCl3 from chlorination of highly reactive resorcinol (meta-dihydroxybenzene) for 2 h at 15 C
was 85% at pH 7 and 100% at pH 9. Others (e.g. Boyce and Hornig, 1983) have found similar yields for resorcinol. Chlorination of the b-keto acid, 3-ketoglutaric acid (acetone dicarboxylic acid), with two carbons reactive to THM formation, was found to produce chloroform in yields 90% over time periods of 0.75–24 h (Larson and Rockwell, 1979; Hasegawa et al., 1983). The oxidant demands of the polyhydroxyphenol model compounds (mM Cl2 equivalents; Tables 2 and 3) were higher than those of the terpenoid model compounds. Ellagic acid had similar oxidant demands to 40 ,5,7-trihydroxyflavanone, with much lower yields of total THMs, indicating that oxidant was consumed in more non-THM forming reactions with ellagic acid. Similarly, oxidant appeared to be consumed in more non-THM forming reactions with b-carotene and geranyl acetate, as compared to retinol and b-ionone. The oxidant demands were lower at pH 9 than at pH 7, while the yields of total THMs were higher, indicating more effective conversion to THMs at the higher pH from all model compounds. This trend of lower oxidant demand with higher THM formation at higher pH (pH 6.8-10.7) has also been observed for chlorination of syringaldehyde after 6 and 24 h, although in a similar experiment, with a different mole ratio of chlorine to syringaldehyde and after a reaction time of 60 h, the oxidant demand did not change with pH (Morris and Baum, 1978). Reckhow and Singer (1985) also observed the trend of lower oxidant consumption with higher chloroform formation at higher pH (12 vs. 7) for chlorination of syringaldehyde, while both oxidant consumption and chloroform formation increased with increasing pH (7 vs. 12) for chlorination of pyruvic acid.
4.
Conclusions
This is the first study to demonstrate the formation of THMs from b-carotene-type terpenoids. Halogenation of b-ionone and retinol and, to a lesser extent, b-carotene and geranyl acetate produced THMs, most likely through haloform reactions on methyl ketone groups in the complex suites of reaction intermediates. Since halogenation of retinol produced a significant quantity of THMs at a slow reaction rate, retinolbased structures may possibly contribute to the slow reacting phase of THM formation in natural waters. b-Ionone appeared to be a key intermediate in the halogenation of b-carotene and retinol, reacting further under the reaction conditions to produce trans-b-ionone-5,6-epoxide and b-cyclocitral. The only identified products from halogenation of 40 ,5,7trihydroxyflavanone and ellagic acid were THMs. 40 ,5,7-Trihydroxyflavanone rapidly produced THMs at pH 7, with an extremely high molar yield of total THMs (94%), significantly higher than any other model compound studied. At pH 7, ellagic acid produced a similar 7 day molar concentration of total THMs to b-ionone, with a slightly higher specific and conventional yield of total THMs than b-ionone. Terpenoids of the b-ionone and retinol type could therefore be significant THM precursors, while 40 ,5,7-trihydroxyflavanone has been shown to be an extremely significant THM precursor, potentially present within NOM in water treatment processes and distribution systems. Since b-ionone can be
water research 44 (2010) 232–242
derived from microbial activity and 40 ,5,7-trihydroxyflavanone is a moiety in condensed tannins, management strategies for source waters containing high levels of microbial activity or high levels of tannin input should consider the role of these respective moieties in DBP formation from these source waters.
references
Acero, J.L., Piriou, P., von Gunten, U., 2005. Kinetics and mechanisms of formation of bromophenols during drinking water chlorination: assessment of taste and odor development. Water Research 39 (13), 2979–2993. Aramendı´a, M.A., Borau, V., Garcı´a, I., Jime´nez, C., Lafont, F., Marinas, J.M., Urbano, F.J., 1998. Determination of volatile organic compounds in drinking and waste water from Cordoba (Spain) by close-loop stripping analysis in combination with gas chromatography coupled with mass spectrometry. Toxicological and Environmental Chemistry 67, 9–25. Badin, A.-L., Faure, P., Bedell, J.-P., Delolme, C., 2008. Distribution of organic pollutants and natural organic matter in urban storm water sediments as a function of grain size. Science of the Total Environment 403, 178–187. Boyce, S.D., Hornig, J.F., 1983. Reaction pathways of trihalomethane formation from the halogenation of dihydroxyaromatic model compounds for humic acid. Environmental Science and Technology 17 (5), 202–211. Buchbauer, G., Freudenreich, S., Hampl, C., Haslinger, E., Robien, W., 1984. The reaction of camphene with hypochlorous acid. Chemical Monthly (Monatshefte fu¨r Chemie) 115, 509–517. Chorus, I., Klein, G., Fastner, F., Rotard, W., 1992. Off-flavours in surface waters – how efficient is bank filtration for their abatement in drinking water? Water Science and Technology 25 (2), 251–258. Clesceri, L.S., Greenberg, A.E., Eaton, A.D., 1998. Standard methods for the examination of water and wastewater, twentieth ed. American Public Health Association, American Water Works Association, Water Environment Federation, USA. Fiorentino, G., Spaccini, R., Piccolo, A., 2006. Separation of molecular constituents from a humic acid by solid-phase extraction following a transesterification reaction. Talanta 68, 1135–1142. Fuson, R.C., Bull, B.A., 1934. The haloform reaction. Chemical Reviews 15 (3), 275–309. Gallard, H., von Gunten, U., 2002. Chlorination of natural organic matter: kinetics of chlorination and of THM formation. Water Research 36 (1), 65–74. Grob, K., Grob, G., 1974. Organic substances in potable water and in its precursor. Part II. Applications in the area of Zu¨rich. Journal of Chromatography 90, 303–313. Grob, K., Zu¨rcher, F., 1976. Stripping of trace organic substances from water: Equipment and procedure. Journal of Chromatography 117 (2), 285–294. Hasegawa, K., Aizawa, T., Naito, S., Magara, Y., 1983. Characteristics for the formation of chloroform from organic compounds by aqueous chlorination. Suishitsu Odaku Kenkyu 6 (3), 151–160 (Chem. Abs. 99:218319). Heitz, A. (1997) Odour-causing Compounds in Water supplies from five towns in Southern Western Australia. A report prepared for the Water Corporation, July 1997. Heitz, A. (2002) Malodorous Dimethylpolysulfides in Perth Drinking Water, PhD Dissertation, Department of Applied Chemistry, Curtin University of Technology, Western
241
Australia. http://adt.curtin.edu.au/theses/available/adtWCU20030512.160534 Ho, L., Croue´, J.-P., Newcombe, G., 2004. The effect of water quality and NOM character on the ozonation of MIB and geosmin. Water Science and Technology 49 (9), 249–255. Hoehn, R.C., Barnes, D.B., Thompson, B.C., Randall, C.W., Grizzard, T.J., Shaffer, P.T.B., 1980. Algae as sources of trihalomethane precursors. Journal ofAmerican Water Works Association 72 (6), 344–350. Jones, G.J., Korth, W., 1995. In situ production of volatile odour compounds by river and reservoir phytoplankton populations in Australia. Water Science and Technology 31 (11), 145–151. Ju¨ttner, F., 1995a. Elimination of terpenoid odorous compounds by slow sand and river bank filtration of the Ruhr River, Germany. Water Science and Technology 31 (11), 211–217. Ju¨ttner, F., 1992. Flavour compounds in weakly polluted rivers as a means to differentiate pollution sources. Water Science and Technology 25, 155–164. Ju¨ttner, F., 1995b. Physiology and biochemistry of odorous compounds from freshwater cyanobacteria and algae. Water Science and Technology 31 (11), 69–78. Kopperman, H.L., Hallcher, R.C., Riehl, A., Carlson, R.A., Caple, R., 1976. Aqueous chlorination of a-terpineol. Tetrahedron 32, 1621–1626. Lalezary, S., Pirbazari, M., McGuire, M.J., 1986. Oxidation of five earthy-musty taste and odor compounds. Journal of the American Water Works Association 78 (3), 62–69. Larson, R.A., Marley, K.A., 1988. Sunlight photochlorination and dark chlorination of monoterpenes. Science of the Total Environment 77, 245–252. Larson, R.A., Rockwell, A.L., 1979. Chloroform and chlorophenol production by decarboxylation of natural acids during aqueous chlorination. Environmental Science and Technology 13 (3), 325–329. Leenheer, J.A., Nanny, M.A., McIntyre, C., 2003. Terpenoids as major precursors of dissolved organic matter in landfill leachates, surface water, and groundwater. Environmental Science and Technology 37 (11), 2323–2331. Leenheer, J.A., Aiken, G.R., Woodside, G., O’Connor-Patel, K., 2007. DOM in recharge waters of the Santa Ana River Basin. Journal of the American Water Works Association 99 (6), 118–131. March, J., 1992. Advanced Organic Chemistry: Reactions, Mechanisms, and Structure, fourth ed. John Wiley and Sons, New York, p 1174. Matsuo, H., Nagafuchi, Y., Nakamura, M., Minari, S., Kondoh, H., 1989. Studies on trihalomethanes precursors in industrial wastewaters. Flavanones in wastewaters of Satsuma mandarin cannery. Suishitsu Odaku Kenkyu 12 (9), 582–588 (Chem. Abs. 112:240016). McIntyre, C., McRae, C., Batts, B.D., Piccolo, A., 2005. Structural characterization of groundwater hydrophobic acids isolated from the Tomago Sand Beds, Australia. Organic Geochemistry 36, 385–397. Morris, J.C., Baum, B., 1978. In: Jolley, R.L., Gorchev, H., Hamilton Jr., D.H. (Eds.), Water Chlorination: Environmental Impact and Health Effects Volume 2. Ann Arbor Science Publishers, Ann Arbor, Michigan, pp. 29–48. Pedersen, J.A., Yeager, M.A., Suffet, I.H., 2002. Characterisation and mass load estimates of organic compounds in agricultural irrigation runoff. Water Science and Technology 45 (9), 103–110. Reckhow, D.A., Singer, P.C., 1985. Mechanisms of organic halide formation during fulvic acid chlorination and implications with respect to preozonation. In: Jolley, R.L., Bull, R.J., Davis, W.P., Katz, S., Roberts Jr., M.H., Jacobs, V.A. (Eds.), Water Chlorination: Environmental Impact and Health
242
water research 44 (2010) 232–242
Effects Volume 5. Lewis Publishers, Inc, Chelsea, MI, USA, pp. 1229–1257. Rook, J., 1977. Chlorination reactions of fulvic acids in natural waters. Environmental Science and Technology 11 (5), 478–482. Rook, J.J., 1976. Haloforms in drinking water. Journal of the American Water Works Association 68 (3), 168–172.
Sheldon, L.S., Hites, R.A., 1978. Organic compounds in the Delaware River. Environmental Science and Technology 12 (10), 1188–1194. Stepen, R.A., Korsunova, T.M., 1988. Participation of terpenic exometabolites in the formation of humus. Izvestiya Sibirskogo Otdeleniya Akademii Nauk SSSR, Seriya Biologicheskikh Nauk 2, 43–49 (Chem. Abs. 110:113627).
water research 44 (2010) 243–255
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Investigating the chlorination of acidic pharmaceuticals and by-product formation aided by an experimental design methodology Jose´ Benito Quintana a,b,*, Rosario Rodil a,c,*, Purificacio´n Lo´pez-Mahı´a b,c, Soledad Muniategui-Lorenzo c, Darı´o Prada-Rodrı´guez b,c a
Department of Analytical Chemistry, Nutrition and Food Sciences, IIAA – Institute for Food Analysis and Research, University of Santiago de Compostela, 15782 Santiago de Compostela, Spain b IUMA – University Institute of Environment, University of A Corun˜a, Pazo da Lo´ngora, Lia´ns, 15179 Oleiros (A Corun˜a), Spain c Department of Analytical Chemistry, Faculty of Sciences, University of A Corun˜a, Campus A Zapateira S.N., 15071 A Corun˜a, Spain
article info
abstract
Article history:
The degradation of seven acidic drugs and two metabolites during chlorination was
Received 19 May 2009
investigated by liquid chromatography–mass spectrometry (LC–MS). A triple-quadrupole
Received in revised form
(QqQ) system was used to follow the time course of the pharmaceuticals and by-products,
29 August 2009
while a quadrupole time-of-flight (Q-TOF) system was also used for the identification of the
Accepted 5 September 2009
by-products. Under strong chlorination conditions (10 mg/L Cl2, 24 h), only four of the
Published online 10 September 2009
target compounds were significantly degraded: salicylic acid, naproxen, diclofenac and indomethacine. The degradation kinetics of these four compounds were investigated at
Keywords:
different concentrations of chlorine, bromide and pH by means of a Box–Behnken exper-
Acidic pharmaceuticals
imental design. Depending on these factors, measured pseudo-first order half-lives were in
Non-steroidal anti-inflammatory
the ranges: 23–573 h for salicylic acid, 13–446 min for naproxen, 5–328 min for diclofenac
drugs (NSAIDs)
and 0.4–13.4 min for indomethacine. Also, it was observed that chlorine concentration was
Chlorination
the overall most significant factor, followed by the bromide concentration (except for
By-products
indomethacine), resulting in increased degradation kinetics as they are increased. The
Environmental fate
degradation path of salicylic acid, naproxen and diclofenac consisted of aromatic substi-
Liquid chromatography–mass
tution of one or two hydrogens by chlorine and/or bromide. Moreover, for diclofenac, two
spectrometry (LC–MS)
other by-products corresponding to a decarboxylation/hydroxylation pathway from the
Time-of-flight (TOF)
monohalogenated products were also identified. On the other hand, indomethacine degradation did not lead to halogenation products but to oxidation ones. The investigation of these by-products in real samples by LC–MS/MS (QqQ) showed that the halogenated derivates of salicylic acid occurred in all the drinking water and wastewater samples analysed. ª 2009 Elsevier Ltd. All rights reserved.
* Corresponding authors. Department of Analytical Chemistry, Nutrition and Food Sciences, IIAA – Institute for Food Analysis and Research, University of Santiago de Compostela, 15782 Santiago de Compostela, Spain. E-mail addresses: [email protected] (J.B. Quintana), [email protected] (R. Rodil). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.018
244
1.
water research 44 (2010) 243–255
Introduction
The occurrence of pharmaceuticals in the environment, mainly in water, is a topic that has attracted a strong attention since the last 10 years. In this time, several therapeutic classes of drugs and their human and animal metabolites have been found in the aqueous environment at levels reaching several mg/L in wastewater (Petrovic and Barcelo, 2007; Reemtsma and Jekel, 2006). The incomplete removal of these pollutants at wastewater treatment plants (WWTPs), not designed for this task, has permitted their spread through surface waters (Boyd et al., 2003; Carballa et al., 2004; Kim et al., 2007; Metcalfe et al., 2003; Okuda et al., 2008; Pa´xeus, 2004; Reemtsma et al., 2006; Tauxe-Wuersch et al., 2005; Ternes, 1998), which is actually a primary source of drinking water. Subsequently, some pharmaceuticals are again not completely removed during drinking water production and thus, they have been identified in drinkable water at the ng/L level (Drewes et al., 2002; Kim et al., 2007; Mompelat et al., 2009; Stackelberg et al., 2004, 2007; Ternes et al., 2002). Many research efforts have been put in the development of alternative wastewater and drinking water treatment technologies in order to decrease the concentration of pharmaceuticals. These include the use of membrane technologies, constructed wetlands, UV radiation, and advanced oxidation, including ozonation (Petrovic and Barcelo, 2007; Reemtsma and Jekel, 2006). On the other hand, the research on the possible transformation during conventional treatment has comparatively received less attention in spite of the fact that several transformation products of toxicological concern have been identified due to biodegradation of pharmaceuticals during activated sludge treatment at WWTPs (Quintana et al., 2005; Vanderford et al., 2003) or chlorination of pharmaceuticals and personal care products, both at WWTPs and drinking water treatment plants (DWTPs) (Bedner and Maccrehan, 2006; Canosa et al., 2006; Inaba et al., 2006; Negreira et al., 2008; Vanderford et al., 2008), and that current chromatography–mass spectrometry techniques are well suited for this purpose (Kosjek and Heath, 2008; Kosjek et al., 2007). The aim of this work was to study the chlorination of acidic drugs and some metabolites (Fig. 1), bearing in mind that this class of pharmaceuticals is among the most frequently found in the environment (Petrovic and Barcelo, 2007; Reemtsma and Jekel, 2006; Reemtsma et al., 2006) and that, according to the European Federation of Chlor-alkali Producers, 98% of the DWTPs in Europe use chlorination as one of the main disinfection steps (Euro Chlor URL). Moreover, residual chlorine in tap water can already react with organic pollutants producing unwanted by-products (Canosa et al., 2006; Negreira et al., 2008). Some previous works have already reported that acidic pharmaceuticals can be degraded upon chlorination (Pinkston and Sedlak, 2004; Westerhoff et al., 2005), but transformation pathways remained unclear. It is also worth mentioning that the presence of bromide has not been evaluated in those previous studies (Pinkston and Sedlak, 2004; Westerhoff et al., 2005), despite the fact that it has been proven to be a very relevant factor in the chlorination of several personal care products (Canosa et al., 2006; Negreira et al., 2008) and the
formation of trihalomethanes (Rodrigues et al., 2007). Thus, the reaction kinetics of acidic drugs were investigated in detail at different chlorine dose, pH and bromide concentrations by means of an experimental design methodology. Also, several transformation products were tentatively identified by liquid chromatography–(tandem) mass spectrometry (LC–MS(/MS)) with a quadrupole time-of-flight (Q-TOF) system and measured at different environmental conditions by LC–MS/MS with a triple-quadrupole (QqQ) instrument.
2.
Materials and methods
2.1.
Chemicals and stock solutions
The structures of the studied pharmaceuticals are presented in Fig. 1. Diclofenac sodium salt, bezafibrate, salicylic acid, clofibric acid, naproxen, indomethacine, ketoprofen, ibuprofen and fenoprofen calcium salt hydrate were obtained from Sigma–Aldrich (Steinheim, Germany). Stock solutions containing the compounds (ca. 2 mg/mL) were prepared in methanol (Romil, Barcelona, Spain) and diluted as necessary. Ultrapure water was obtained in the lab from a Milli-Q water generator (Millipore, Billerica, MA, USA). Ammonium acetate was from Fluka (Steinheim, Germany). Potassium bromide and glacial acetic acid were from Merck (Darmstadt, Germany), and ascorbic acid and sodium hypochlorite solution (w10%) from Sigma–Aldrich. Sodium hypochlorite stock solutions at the desired level were prepared daily by dilution in Milli-Q water and their concentration was iodometrically standardised.
2.2.
Chlorination experiments
Chlorination of pharmaceuticals was performed on 22 mL closed vials that were maintained in the dark. Preliminary experiments to determine the stability of drugs were done with 10 mL of Milli-Q water, adjusted to pH 7.1 with a phosphate buffer and spiked with the tested drug at the 1 mg/mL level and 10 mg/L Cl2. After 24 h, a 1 mL aliquot of the solution was taken and the reaction quenched with ascorbic acid (0.6 mg/mL). Further experiments to study chlorination kinetics were performed in a similar way, but with lower drug concentrations (50 mg/L) and different concentrations of chlorine (1–10 mg/L), bromide (0–100 mg/L) and pH of sample (5.7–8.3) being considered. In these experiments, five aliquots were taken at different reaction times and the reaction stopped with ascorbic acid. Parallel control samples (without chlorine) were also measured. For identification of chlorination by-products, experiments were done with Milli-Q water, adjusted to pH 7.1 with a phosphate buffer and spiked with the tested drug at the 2 mg/mL level, 5 mg/L Cl2, and either without bromide or with bromide added (50 mg/L). Five different aliquots were taken at different times depending on the degradation kinetics of the particular pharmaceutical. Degradation of drugs was measured by LC–MS/MS (QqQ) in the multiple-reaction monitoring (MRM) mode of acquisition (see Table 1). By-products were screened by LC–MS (scan) and
245
water research 44 (2010) 243–255
O OH
HO
Cl
H N
O
OH O
O
Cl
Ibuprofen
Diclofenac
Naproxen
NSAID
NSAID
NSAID O
O
Cl
N O
OH
O
OH
O
O
OH
O
Ketoprofen
Fenoprofen
Indomethacine
NSAID
NSAID
NSAID
O
OH
O
O
OH
O
H N
OH
Cl
Cl
O O
Salicylic acid
Clofibric acid
Bezafibrate
NSAID metabolite & skin-care product
Lipid regulator metabolite
Lipid regulator
OH
Fig. 1 – Structures of the pharmaceuticals and metabolites considered in this study.
LC–MS/MS (product ion scan) in both the QqQ and Q-TOF instruments.
2.3.
ammonium acetate. Gradient was as follows: 0 min, 5% B; 10 min, 100% B; 15 min, 100% B; 17 min, 5% B; 27 min, 5% B. Injection volume was set to 50 mL.
LC–MS/MS (triple quadrupole) 2.4.
The LC–MS/MS (QqQ) system consisted of an Agilent 1200 Series (Agilent Technologies, San Jose, CA, USA) liquid chromatograph comprising a membrane degasser, binary high-pressure gradient pump, autosampler, and column thermostat. The system was interfaced to an API-3200 QqQ mass spectrometer (Applied Biosystems, Foster City, CA, USA) equipped with a Turbo V electrospray ion source. Nitrogen was provided by a nitrogen generator (Peak Scientific, Bedford, MA, USA) and used as source and collision gas. Turbo V electrospray ion source and MS/MS parameters were as follows: curtain gas: 30 psi; collision gas: 5 psi; ion spray voltage: 4500 V (negative mode); temperature: 700 C; ion source gas 1: 40 psi; and ion source gas 2: 60 psi. Analytes and degradation products were determined in the electrospray negative mode and MRM (Table 1). Separation of analytes was carried out on a 3.5 mm SymmetryShield RP18 2.1 mm 150 mm column (Waters, Milford, MA, USA) at a flow rate of 0.2 mL/min at 30 C. Eluent A consisted of Milli-Q water and B MeOH, both containing 5 mM
LC–MS/MS (quadrupole time-of-flight)
The LC–MS/MS (Q-TOF) instrument was an Agilent 1200 Series liquid chromatographic system consisting of a membrane degasser, quaternary high-pressure gradient pump and autosampler; interfaced to a QSTAR Elite Q-TOF instrument (Applied Biosystems) equipped with a Turboionspray electrospray ion source. Nitrogen was provided by a nitrogen generator (Peak Scientific) and used as source and collision gas. Electrospray ion source and MS/MS parameters were as follows: curtain gas: 40 psi; collision gas: 5 psi; ion spray voltage: either 4500 (negative mode) or 5500 V (positive mode); temperature: 400 C; ion source gas 1: 35 psi; and ion source gas 2: 30 psi. By-products were screened in both positive and negative electrospray modes, but could only be detected in negative. Analytes and degradation products were determined in scan and product ion scan mode. Separation was performed as with the QqQ system but at room temperature (no column thermostat was available for the Q-TOF system).
246
water research 44 (2010) 243–255
Table 1 – LC–MS/MS parameters used for the MRM determination of pharmaceuticals and chlorinated by-products with the QqQ instrument, operated in the electrospray negative mode. Compound
tR (min)
MRM 1
DP/CEa
MRM 2
DP/CEa
Transitions ratiob
Pharmaceuticals Salicylic acid Clofibric acid Naproxen Ketoprofen Bezafibrate Fenoprofen Ibuprofen Diclofenac Indomethacine
9.4 11.9 12.6 12.9 14.0 15.2 16.2 18.9 20.9
137 > 93 213 > 127 229 > 169 253 > 209 360 > 274 241 > 93 205 > 159 294 > 250 356 > 312
25/20 20/20 15/40 15/8 35/22 5/46 20/8 25/12 15/10
137 > 65 215 > 129 229 > 185 253 > 197 360 > 154 241 > 197 205 > 161 296 > 252 358 > 314
25/38 20/22 15/12 15/10 35/40 5/12 20/6 20/12 5/10
13.8 6.9 2.8 0.7 1.7 0.3 5.8 1.7 1.6 0.3 1.5 0.3 3.8 1.0 1.0 0.2 2.3 0.6
Chlorination by-products Cl-SA Cl2-SA Br-SA BrCl-SA Cl-naproxen Br-naproxen Cl-diclo Cl-diclo-CO Br-diclo Br-diclo-CO OH-indo OH-indo-CO2 Indo-H2 Cl-BA
14.9 & 15.2 16.3 15.2 & 15.5 16.5 17.0 17.2 17.9 19.0 18.2 19.2 19.5 & 21.3 21.0 & 22.9 21.6 10.5
171 > 91 205 > 125 215 > 79 251 > 207 263 > 219 307 > 247 328 > 284 300 > 161 372 > 328 344 > 161 372 > 285 328 > 155 354 > 155 155 > 111
25/30 25/30 25/50 25/30 15/40 15/40 20/10 20/30 20/10 20/30 20/30 20/30 20/30 25/30
173 > 91 207 > 125 217 > 81 249 > 205 265 > 221 309 > 249 330 > 286 302 > 163 374 > 330 346 > 163 374 > 287 330 > 157 354 > 155 157 > 113
25/30 25/30 25/50 25/30 15/40 15/40 20/10 20/30 20/10 20/30 20/30 20/30 20/30 25/30
3.8 1.0 2.8 0.7 1.1 0.2 1.4 0.3 3.4 0.8 1.0 0.2 1.4 0.3 1.6 0.3 1.6 0.3 1.6 0.3 2.8 0.7 2.7 0.7 3.1 0.8 3.2 0.8
a Declustering potential (V)/collision energy (eV). b Ratio and tolerances between quantification and confirmation MRM transitions (EU, 2002).
2.5.
Samples
All samples were collected in amber glass bottles, previously washed with acetone, methanol and Milli-Q water, in September 2008. Subsequently after sampling, they were filtered through 0.45-mm nitrocellulose filters (Millipore) and stored at 4 C until analysed (within 48 h). Five tap water samples were collected at five different private homes distributed along the city of A Corun˜a (Galicia, NW Spain) within a 2-h time period. A further finished drinking water sample was collected from a DWTP producing chlorinated water for a population of ca. 10 thousand inhabitants. Ascorbic acid (0.6 mg/mL) was added to these samples in order to eliminate residual chlorine. Also, five surface water samples were collected along the Mero River basin in the area of A Corun ˜ a and another one was collected in Anllo´ns River near to the town of Ponteceso (Galicia, NW Spain). Finally, two wastewater influent samples were taken: WWTP A (ca. 400 thousand equivalent inhabitants) and WWTP B (ca. 1.5 thousand inhabitants). These samples were extracted by SPE and determined by LC–MS/MS as detailed elsewhere (Rodil et al., 2009). In brief, 200–500 mL of filtered sample were subjected to SPE on an Oasis HLB 200 mg (cartridge) and eluted with 3 10 mL of MeOH. This extract was evaporated and reconstituted in 1 mL of MeOH/water (1/1) and 10 mL were injected and analysed by LC–MS/MS in MRM mode (Rodil et al., 2009).
3.
Results and discussion
3.1.
Screening of degradable drugs
A first chlorination test of the seven acidic drugs and two metabolites was performed in order to assess their degradability upon chlorination. Thus, they were treated for 24 h with a 10 mg/L Cl2 concentration at neutral pH (7.1). These were considered as a strong chlorination dosage taking also into account that real water samples contain other organic chemicals that may compete with the pharmaceuticals. The degradation percentage of each drug at the end of this experiment is shown in the Supplementary data (Fig. S1). Only those compounds that were degraded over the cut-off level of 30% were further evaluated. These chemicals are: salicylic acid, degraded into a 68% extent, and the fully removed NSAIDs: diclofenac, naproxen and indomethacine. These results agree with the previous studies of Pinkston and Sedlak (2004) and Westerhoff et al. (2005), who performed batch test chlorination of some of the pharmaceuticals considered here. The only exception was ibuprofen, as no degradation (<10%) was observed in our test or by Pinkston and Sedlak (2004), whereas Westerhoff et al. (2005) found a 24 h removal in the ca. 30–80% range during 24 h chlorination experiments with real surface waters. This may point to a different removal mechanism in those experiments with real water samples, e.g. sorption to particulate matter and/or biodegradation by
247
water research 44 (2010) 243–255
chlorine resistant microorganisms, but remains unclear. Also, to our knowledge, there was no published data on chlorination batch tests of fenoprofen, clofibric acid, bezafibrate and particularly salicylic acid, which occurs often at high concentrations in water samples.
Table 3 – Standardised main effects (SMEs) obtained from the half-life analysis of the Box–Behnken design. Statistically significant factors are presented in italics. Statistical significance boundary for 95% confidence level: Student’s t [ 2.57. Salicylic Naproxen Diclofenac Indomethacine acid
3.2. Influence of pH, chlorine and bromide concentrations on chlorination kinetics A deeper study on the parameters influencing the chlorination of diclofenac, naproxen, indomethacine and salicylic acid was carried out by an experimental design methodology. This strategy permits evaluating not only the factors influencing the degradation of pharmaceuticals but also the interaction between these factors, fitted to a second order model (Lewis et al., 1999), over a range of values and it has, in fact, been successfully applied to the study of trihalomethanes’ formation during water chlorination (Rodrigues et al., 2007). As mentioned in Section 2.2, the factors considered here were pH (5.7–8.3), chlorine dose (1–10 mg/L as Cl2) and bromide concentration (0–100 mg/L as Br). The experimental levels were selected according to the environmental expected values and the typical chlorine concentrations employed during water treatment. This study was done by means of a Box– Behnken experimental design as it provides the best compromise between number of experiments and degrees of freedom for 3 factors (Ferreira et al., 2007). Thus, the final number of experiments was 15 (including 3 centre points), each experiment being sampled at five different times: between 0 and 20 min (indomethacine), 0 and 150 min (naproxen and diclofenac) and 0 and 144 h (salicylic acid). Then, empirical degradation half-lives (t1/2) were calculated from the pseudo-first order kinetic plots for each experiment and the design analysed for each pharmaceutical. The correlation coefficients (R) obtained from the logarithmic kinetic plots were usually higher than 0.9. The Box–Behnken experimental plan and calculated t1/2 values for each pharmaceutical are given in Table 2.
A: pH B: chlorine C: bromide AA AB AC BB BC CC
1.05 9.26 2.04 0.72 1.20 0.11 4.38 2.20 1.10
2.09 3.92 2.69 0.33 0.46 0.29 2.46 2.53 0.67
1.94 1.65 3.87 0.19 0.18 2.55 0.38 2.06 2.37
1.57 6.19 0.93 0.86 1.65 0.26 3.51 1.48 0.86
Table 3 compiles the standardised main effects (SMEs) obtained after experimental design analysis. It can be observed that chlorine concentration was the most significant factor for salicylic acid, naproxen and indomethacine, with a negative value for all pharmaceuticals. This negative value means that the t1/2 decreases as the concentration of chlorine increases, i.e. accelerating reaction kinetics, as it was expected. The concentration of bromide into the solution was also significant and negative for naproxen and diclofenac, and close to significance level for salicylic acid. This can be interpreted as these three compounds are halogenated much faster with HBrO (from the oxidation of Br) than with HClO. In turn, the bromide SME is far away from being statistically significant for indomethacine, so that bromination may not play an important role in its degradation upon chlorination. The effects of these two factors can be visualised in the response surface plots (Fig. 2). Particularly from the diclofenac plot (Fig. 2c) it can be observed that the presence of bromide speeds up the reaction kinetics (lower t1/2 in the figure) at low chlorine levels. These figures also show that even when the degradation of some pharmaceuticals is rather slow at low
Table 2 – Box–Behnken design experimental plan and measured pseudo-first order kinetics half-lives (t1/2). Exp. no.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
pH
8.3 5.7 7 7 7 5.7 8.3 7 8.3 5.7 8.3 7 5.7 7 7
Chlorine (mg/L)
1 10 5.5 5.5 10 5.5 5.5 1 10 5.5 5.5 1 1 10 5.5
Bromide (mg/L)
50 50 50 50 100 0 0 0 50 100 100 100 50 0 50
t1/2 Salicylic acid (h)
Naproxen (min)
Diclofenac (min)
Indomethacine (min)
402.5 23.4 62.1 70.0 22.8 95.5 121.4 573.5 40.0 39.4 48.7 323.5 254.2 31.6 60.6
255.6 13.1 52.6 48.4 19.2 46.8 124.0 446.0 111.7 16.6 57.2 94.4 98.5 51.3 48.4
31.2 8.1 18.9 17.9 16.9 5.4 328.3 325.2 23.8 7.5 15.4 15.7 38.2 72.6 20.6
13.38 0.80 1.64 1.40 1.08 1.25 2.65 4.15 0.92 0.73 1.19 10.21 7.27 0.38 1.32
248
water research 44 (2010) 243–255
Fig. 2 – Response surface plots for (a) salicylic acid, (b) naproxen, (c) diclofenac and (d) indomethacine. Response [ t1/2 in minutes, except salicylic acid in hours.
chlorine doses, it can be speeded at high chlorine doses; e.g. for salicylic acid t1/2 at pH 7 with a chlorine dose of 10 mg/L w1 day, while at the same pH but chlorine dose of 1 mg/L, t1/2 > 3 weeks (Table 2). It is also evident that indomethacine reacts very rapidly with chlorine even at low doses and bromide has no effect (Fig. 2d). The other factor considered, pH, although not being statistically significant was positive in all cases (Table 3). This was expected, because the oxidant specie, HClO, predominates at lower pH values, while the studied drugs would always be in their anionic form in the pH range considered (pKa values: ca. 3–4.5). Among the second order factors, the only one being statistically significant was BB (chlorine–chlorine), meaning that the response is quadratically correlated to chlorine. Moreover, the BC term (chlorine–bromide) is near the significance level for salicylic acid, naproxen and diclofenac. This fact is shown in the response plots for these three drugs (Fig. 2a–c) where it can be observed that at low chlorine doses the presence of bromide speeds up the reaction kinetics, while at high HClO levels bromide plays a less significant role.
3.3.
By-products identification
Chlorination by-products of each pharmaceutical were firstly investigated at the 2 mg/L level, either without or with bromide ion added (50 mg/L) with both the QqQ and Q-TOF instruments in scan and product ion scan modes (see Sections 2.3 and 2.4). A summary of the proposed structures of the detected by-products is shown in Fig. 3. Also, the
Q-TOF scan data of these products is compiled in Table 4 (spectra presented in the Supplementary data, Fig. S2) and the Q-TOF product ion scan spectra are shown in Fig. 4. Chlorination of salicylic acid yielded three intense peaks corresponding to halogenation in the ring activated positions 3 and 5, producing two isomers of chlorosalicylic acid (Cl-SA) and the third one being 3,5-dichlorosalicylic acid (Cl2-SA). Actually, the formation of 3-Cl-SA has already been postulated by Pru¨tz (1998). When bromide was added, then three brominated products were additionally produced, corresponding to the two bromosalicylic acid (Br-SA) isomers and bromochlorosalicylic acid (BrCl-SA). All these products could be easily identified by scan experiments, due to the halogenation patterns of the [MH] ions (Fig. S2) and by the exact masses obtained with the Q-TOF instrument (Table 4). Moreover, Q-TOF product ion spectra (Fig. 4a–d) also confirmed these structures, due to the characteristic succession of decarboxylation and hydrogen chloride/bromide losses, with the halide ion normally obtained as the final product. Dibromosalicylic acid was not observed in any tested circumstance, probably because the second bromination is sterically hindered by the first bromine. Similarly to salicylic acid, chlorination of naproxen produces chloronaproxen (Cl-naproxen) and bromonaproxen (Br-naproxen) when bromide is also added to the medium, as confirmed from the [MH] cluster pattern (Fig. S2) and the exact mass obtained from TOF scan data (Table 4). However, the low yield of these by-products permitted only obtaining the Q-TOF product ion spectra of Cl-naproxen (Fig. 4c), where only a product ion was obtained with enough intensity
249
water research 44 (2010) 243–255
O Cl
O OH
O Cl
OH
OH
OH
OH Cl
5-Cl-SA O
O
Br
OH
OH
OH
Br
5-Br-SA
3-Br-SA
Cl
H N
HO
HO
O
O
OH-indo-CO2 (2 isomers)
Cl
N O
OH
OH-indo (2 isomers)
O
Cl
N
OH
Br-diclo-CO
Br-diclo
O
Cl
N
Cl
Cl
Br
Cl
Br
H N
Cl
H N
OH
Cl-diclo-CO
O
O
Cl
Cl
Cl
Cl-diclo
HO
Br-Naproxen
BrCl-SA
O
Cl
H N
Cl
OH O
O
Cl
O HO
Br
OH
OH
OH
Cl-Naproxen
Cl2-SA
O
OH O
O
Cl
3-Cl-SA
Br
Cl
OH
O
Cl
HO
O OH
Cl-BA ?
Indo-H2 ?
Fig. 3 – Proposed structures of the detected by-products.
Table 4 – LC-Q-TOF scan data on by-product identification (spectra are given in the Supplementary data, Fig. S2). Proposed structures are compiled in Fig. 3. tR (min)
Experimental m/z
Cl-SA Cl2-SA Br-SA BrCl-SA
12.1 & 12.6 14.17 12.5 & 13.0 14.4
170.9865 204.9456 214.9322 248.8935
Cl-naproxen Br-naproxen
14.38 14.53
Cl-diclo Cl-diclo-CO Br-diclo Br-diclo-CO OH-indo OH-indo-CO2 Indo-H2 Cl-BA
By-product
Proposed formula
Calculated m/z
Difference (mDa)
Difference (ppm)
DBE
C7H4O3Cl C7H3O3Cl2 C7H4O3Br C7H3O3ClBr
170.9854 204.9464 214.9349 248.8959
1.1 0.9 2.7 2.4
6.1 4.3 12.7 9.8
5.5 5.5 5.5 5.5
263.0477 306.9981
C14H12O3Cl C14H12O3Br
263.0485 306.9969
0.8 1.1
3.2 3.6
8.5 8.5
15.3 17.5 15.5 17.7
327.9704 299.9762 371.9204 343.9263
C14H9NO2Cl3 C13H9NOCl3 C14H9NO2Cl2Br C13H9NOCl2Br
327.9704 299.9755 371.9199 343.9250
0.0 0.7 0.5 1.3
0.1 2.3 1.3 3.8
9.5 8.5 9.5 8.5
17.4 & 19.2 18.9 & 20.1 19.4 9.0
372.0642 328.0751 354.0549 154.9909
C19H15NO5Cl C18H15NO3Cl C19H13NO4Cl C7H4O2Cl
372.0644 328.0745 354.0538 154.9905
0.2 0.5 1.0 0.4
0.6 1.5 2.9 2.4
12.5 11.5 13.5 5.5
250
water research 44 (2010) 243–255
a
b
Cl (-0.8 mDa)
Products of 171 (Cl-SA)
65 55 45 35 C H O (-0.6 mDa)
25
91.0183
C H OCl (-2.4 mDa)
126.9932 15
-HCl
50
c
70
90
110 130 m/z, Da
150
170
30 20
C H OBr (-2.6 mDa)
0 30
50
70
90
-CO
110 130 150 m/z, Da
90
210
C H OClBr (+1.6 mDa)
204.9078
9.0 7.0 C H OCl (-0.3 mDa) 124.9797 C H OBr (+2.8 mDa)
5.0
168.9323 -HCl
-CO
[M-H]
-HBr
1.0 40
80
120
f
203.0276
190
Products of 249 (BrCl-SA)
11.0
110 130 150 170 190 210 230 m/z, Da C H OCl (+0.6 mDa)
170
[M-H]
Br (+1.6 mDa)
3.0
200 160 m/z, Da
280
240
C H NCl
(+3.1 mDa)
283.9837 6.5
Products of 263 (Cl-Naproxen)
7.0 6.0 5.0 4.0 3.0
Products of 328 (Cl-diclo)
5.5 Intensity, counts
Intensity, counts
70
50
-CO [M-H]
0.0
-HCl -HCl
13.0
170.9425 10
1.0
89.0033
15.0
Intensity, counts
Intensity, counts
40
2.0
C HOl (+0.01 mDa)
8
78.9205
50
8.0
12
d Products of 215 (Br-SA)
(+2.1 mDa)
160.9588
78.9184
60
9.0
C H OCl
16
0 30
190
Br (-0.5 mDa)
70
10.0
20
4
[M-H]
30
Products of 205 (Cl2-SA)
24
-CO
5
e
124.9805
28
Intensity, counts
Intensity, counts
75
C H OCl (+0.5 mDa)
34.9686
4.5 3.5
C H NCl
-HCl
1.5
-CO -CH
(-0.8 mDa)
248.0031
2.5
[M-H]
-CO
0.5 40 60 80 100 120 140 160 180 200 220 240 260 280 m/z, Da
[M-H]
40
80
120
160
200 240 m/z, Da
280
320
360
400
Fig. 4 – Q-TOF product ion scan spectra from the chlorination by-products.
(m/z 203.0276), corresponding to the loss of CO2 and methane, following a fragmentation pathway similar to that of naproxen (Quintana and Reemtsma, 2004). In the case of diclofenac, besides chloro-diclofenac (Cl-diclo) and bromo-diclofenac (Br-diclo), two other chemicals whose empirical formulae correspond to the loss of a CO group from these two monohalogenated products are also obtained (Table 4 and Fig. S2). These two products (Cl-diclo-CO and Br-diclo-CO) are believed to be formed through a lactone intermediate by decarboxylation and oxidation of the dichlorinated ring in position 4 following a mechanism similar to that described by Miyamoto et al. in the metabolic HClO oxidation of diclofenac (Miyamoto et al., 1997). Thus, the
resulting product is hydroxylated in position 4 of the dichlorinated benzenic ring (Fig. 3). These structures are confirmed by the Q-TOF product ion scan spectra. In the case of Cl-diclo, the spectrum shows the losses of CO2 and HCl (Fig. 4f). On the contrary for Cl-diclo-CO (Fig. 4g), the CO2 loss is not observed (meaning that there is no carboxylic group in the structure), whereas there is a CO loss (typical for alcohols), and a peak at m/z 160.9598 corresponding to the dichlorinated phenolate anion (C6H3OCl2; Dm/z ¼ þ3.1 mDa) is produced. Finally, the oxidation of indomethacine with HClO does not lead to further halogenation but to oxidation as the monochlorination pattern of the deprotonated products is maintained (Fig. S2) and confirmed from the exact masses obtained
251
water research 44 (2010) 243–255
C H OCl
g
(+3.1 mDa)
160.9598 Cl
2.2
Products of 300 (Cl-diclo-CO)
2.0 O
Cl
Intensity, counts
1.8 1.6 1.4 C H NOCl
1.2
(+3.1 mDa)
261.9863
1.0
C H NCl
(-0.9 mDa)
233.9874
0.8
C H NCl (-4.7 mDa)
0.6
198.0069
[M-H]
-HCl
0.2 0.0
-HCl-H -CO
0.4
40
80
120
160
200
240
280
320
360
400
m/z, Da C H NO Cl (-0.2 mDa)
h
285.0560 48
Products of 372 (OH-indo)
44
Intensity, counts
40 36 32 O
28
Cl O
20
N
300.0767
154.9897
Cl
12
C H NO
8 C H Cl (+2.4 mDa) 111.0032 4 100
i
120
C H NO Cl (+0.8 mDa)
- CH
313.0520
C H NO Cl (-0.2 mDa)
(+0.8 mDa)
174.0569
C H NO Cl (+3.4 mDa)
270.0325
328.0780
- CH
-CO [M-H]
140
160
180
C H NO
O Cl O
200
220 240 m/z, Da
260
280
300
320
340
(-2.5 mDa)
174.0535
HO
C H O Cl (-2.6 mDa)
1.4
-CO
C H NO Cl (-3.0 mDa)
O
C H O Cl (-0.8 mDa)
16
0
- CH
HO
24
154.9879
Products of 372 (OH-indo-CO2)
N
O
Intensity, counts
1.2 1.0
C H NO Cl (-2.4 mDa)
313.0487 0.8 Cl
0.6
C H Cl (+5.4 mDa)
0.4 111.0061 CH
0.2
[M-H]
0.0
100
120
140
160
180
200
220 240 m/z, Da
260
280
300
320
Fig. 4 – (continued).
from the TOF scan measurements (Table 4). The major products (Fig. 3) are two hydroxylated isomers (OH-indo), where the Q-TOF product ion spectrum (Fig. 4 h) shows the loss of CO2 and also the loss of CO (as mentioned, typical for alcohols), yielding finally the ions at m/z 174.0535 (whose postulated
structure is shown in Fig. 4 h), m/z 154.9897 (corresponding to 4-hydroxybenzoate and likely produced from the migration of the eOH group) and m/z 111.0032 (4-chlorophenide), proving that the eOH was not located in the chlorinated ring. Besides OH-indo, two decarboxylated hydroxylated products
252
water research 44 (2010) 243–255
(OH-indo-CO2) were also produced, which show product ion spectra similar to OH-indo (Fig. 4i). Also two other minor products, whose possible structures are shown in Fig. 3, were observed. The molecular formulae obtained from the Q-TOF scan measurements (Table 4) likely correspond to deshydroindomethacine (Indo-H2) and 4-chlorobenzoic acid (Cl-BA), probably produced by the final hydrolysis of the amide group. These two substances were not intense enough to permit accurate masses to be determined from product ion spectra and, thus, their identity could not be confirmed.
brominated compounds and negatively the chlorinated ones, while for BrCl-SA both factors are positive (Table 5). Brominated by-products replace the formation of the chlorinated ones when low amounts of bromide are present, as shown in the response surface obtained from the analysis of the experimental design (Supplementary data, Fig. S3). Thus, Br-SA is formed very rapidly and then is further transformed into BrCl-SA (Fig. 5). Also, as presented in Fig. 5, Cl-SA
a
Cl-SA
SA
Cl2-SA
BrCl-SA
3.4. Influence of pH, chlorine and bromide concentrations on by-products formation
Br-SA
1.00
After identification of by-products, the best MRM transition conditions were selected by varying the declustering potential and collision energy in the QqQ instrument as compiled in Table 1. So, their formation was confirmed at lower drug levels (50 mg/L) with the previously mentioned Box–Behnken design by measuring the products in MRM mode using the QqQ instrument. Unfortunately, the yield of the naproxen halogenation products was very low and they could not be detected in the Box–Behnken experiment at low naproxen concentration levels in spite of the fact that this drug was degraded. Moreover, other naproxen transformation by-products could not be detected in negative or positive scan modes. The analysis of the Box–Behnken design for the formed by-products was done by taking into account the MRM response at a fixed reaction time, as kinetic values cannot be calculated (Table 5). As expected, for salicylic acid and diclofenac, the degree on which any of their by-products is formed depends strongly on the concentrations of chorine and bromide in the medium. The analysis of the Box–Behnken design for the by-products of salicylic acid (Table 5) shows that chlorine and bromide concentrations are statistically significant factors. Obviously, bromide concentration affects positively the formation of the
0.80 0.60 0.40 0.20 0.00 0
20
60
40
80
100
time (h)
b
Diclo
Cl-diclo-CO
Cl-diclo
Br-diclo-CO
Br-diclo
1.00 0.80 0.60 0.40 0.20 0.00 0
50
100
150
time (min)
Table 5 – Standardised main effects (SMEs) obtained from the peak area analysis of the Box–Behnken design for the oxidation by-products. Statistically significant factors are presented in italics. Statistical significance boundary for 95% confidence level: Student’s t [ 2.57. Reaction times: salicylic acid: 24 h; diclofenac: 60 min; and indomethacine: 2 min. A: pH
B: chlorine
C: bromide
Cl-SA Cl2-SA Br-SA BrCl-SA
0.46 0.53 2.27 1.48
2.28 5.01 3.07 6.88
2.04 2.67 4.18 7.26
Cl-diclo Cl-diclo-CO Br-diclo Br-diclo-CO
9.00 4.66 2.73 0.34
0.11 3.61 0.35 0.51
3.05 2.66 4.53 4.27
OH-indo OH-indo-CO2 Indo-H2 Cl-BA
0.75 0.24 1.18 0.75
3.93 1.53 1.66 1.06
1.18 0.51 1.31 0.25
c
Indo
OH-Indo
OH-Indo-CO2
Indo-H2 Cl-BA
1.00 0.80 0.60 0.40 0.20 0.00 0
0.5
1
1.5 2 time (min)
2.5
3
Fig. 5 – By-products time profiles (pH 7; 10 mg/L Cl2 and 100 mg/L BrL) for (a) salicylic acid (SA), (b) diclofenac (diclo) and (c) indomethacine (indo). Results normalised, i.e. 1 is the maximum observed value in the set of experiments.
253
water research 44 (2010) 243–255
generation and its further transformation to Cl2-SA is slower than for the brominated analogues. Then, the dihalogenated products are stable for several days. Regarding diclofenac, the concentration of bromide is also statistically significant and positive for the formation of the brominated products and negative for Cl-diclo (Table 5). However, in the case of Cl-diclo-CO both the concentrations of chlorine and bromide play a significantly positive effect. This positive effect of Br can be attributed to the fastest formation of the bromoamine as compared to the chloroamine, which is a necessary intermediate for the further hydroxylation and decarboxylation reactions that lead to the final product (Miyamoto et al., 1997). The pH is also a significant factor, related to the higher reactivity of HClO vs its anion. Moreover, as presented in Fig. 5, although the by-products are rapidly formed at high Cl2 and Br levels, their further degradation is much slower. In the case of the fast reaction of indomethacine, the most important factor for products formation is the chlorine concentration, being only significant for OH-indo (Table 5), in agreement with the previous chemometrical analysis of indomethacine stability (Table 3). Furthermore, all metabolites are rapidly formed at high chlorine levels (Fig. 5), but at long reaction times, the only one remaining is the product tentatively identified as Cl-BA (data not shown).
16.5 1.8e5
249 > 205
1.6e5
251 > 207
1.0e5 8.0e4
16.3
6.0e4
205 > 125
Cl2-SA
207 > 125
5.0e4
Intensity, cps
1.2e5
Application to samples
Finally, once the by-products were identified, they were screened, together with the parent compounds, for their occurrence in water samples by LC–MS/MS in MRM mode after the SPE of the samples. Analytes were positively identified by comparing the retention time and the ratio of two MRM transitions, with the identification intervals suggested by the EU (EU, 2002), with those of the chlorination experiments (Table 1). The four pharmaceuticals were not detected in surface or in drinking water but occurred in wastewater at different levels: naproxen (300–1050 ng/L), diclofenac (320–460 ng/L), indomethacine (10–40 ng/L) and salicylic acid (630–8750 ng/L). Among the by-products, the only ones detected were the halogenated salicylic acid compounds. These halogenated products were found in the five tap water samples, the one DWTP finished drinking water and the two wastewater samples analysed, with the dihalogenated compounds being the most intense ones. A chromatogram of a tap water sample is exemplarily presented in Fig. 6. Although, their concentration could not be determined because of the lack of standards, they may occur at the mid-ng/L level in tap water, taking the salicylic acid response in LC–MS/MS (MRM) as a reference.
BrCl-SA
1.4e5
Intensity, cps
3.5.
4.0e4
3.0e4
2.0e4
6.0e4 4.0e4
1.0e4 2.0e4 0.0
0.0 14.0
14.4
14.8
15.2
15.6
16.0
16.4
16.8
17.2
17.6
18.0
14.0
14.4
14.8
15.2
15.6
2400
4.8e4
2200
215 > 79
Br-SA
15.2
17.2
17.6
18.0
3.2e4 2.8e4 2.4e4 2.0e4 1.6e4
171 > 91 173 > 91
15.2
1800
217 > 81
3.6e4
1600 1400 1200 1000 800 600
1.2e4 8000.0
400
4000.0
200
0.0
16.8
Cl-SA
2000
Intensity, cps
Intensity, cps
4.0e4
16.4
14.9
15.5 5.2e4 4.4e4
16.0
Time, min
Time, min
14.0
14.4
14.8
15.2
15.6
16.0
16.4
Time, min
16.8
17.2
17.6
18.0
0 14.0
14.4
14.8
15.2
15.6
16.0
16.4
16.8
17.2
Time, min
Fig. 6 – LC–MS/MS (MRM) chromatogram of a tap water sample after an SPE 500-fold concentration.
17.6
18.0
254
4.
water research 44 (2010) 243–255
Conclusions and outlook
Nine acidic pharmaceuticals have been studied and among them only salicylic acid, naproxen, diclofenac and indomethacine react with hypochlorous acid at significantly high reaction rates. The effect of pH, chlorine and bromide concentrations has been studied with a response surface methodology, showing the relevance of these factors in the degradation of the parent compounds and formation of byproducts, where chlorine plays the most important role and bromide is also an important factor leading to brominated byproducts for all drugs but indomethacine. Moreover, the by-products of these four pharmaceuticals have been tentatively identified by LC–MS/MS experiments. Halogenation by-products of salicylic acid have been detected in all analysed wastewater and (chlorinated) drinking water samples but not in surface water, confirming that they are produced during water chlorination. By-products of the other three pharmaceuticals have not been detected in real samples, but their relevance should not be neglected yet, since they could occur in more polluted water environments. In fact, parent drugs were not found in these samples but have been reported in other areas up to the mg/L level (Mompelat et al., 2009). Thus, an effort is required in the investigation of the (eco)toxicological effects of these by-products and even the parent compounds in future research.
Acknowledgements J.B.Q. and R.R. acknowledge Ministerio de Educacio´n y Ciencia (Ramo´n y Cajal research program). This work was financially supported by Xunta de Galicia (Project no. ‘‘PGIDIT06TAM16401PR’’ and ‘‘Programa de consolidacio´n e estruturacio´n de unidades de investigacio´n competitivas, en re´xime de concorrencia competitiva 2006’’), Ministerio de Educacio´n y Ciencia (Project no. ‘‘CTQ2007-63949/BQU’’) and EU FEDER funds. We are indebt to G. Ferna´ndez and J. Otero (UDC research support services) for their support on LC–MS/MS operation.
Appendix. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2009.09.018.
references
Bedner, M., Maccrehan, W.A., 2006. Transformation of acetaminophen by chlorination produces the toxicants 1,4-benzoquinone and N-acetyl-p-benzoquinone imine. Environ. Sci. Technol. 40 (2), 516–522.
Boyd, G.R., Reemtsma, H., Grimm, D.A., Mitra, S., 2003. Pharmaceuticals and personal care products (PPCPs) in surface and treated waters of Louisiana, USA and Ontario, Canada. Sci. Total Environ. 311 (1–3), 135–149. Canosa, P., Rodrı´guez, I., Rubı´, E., Negreira, N., Cela, R., 2006. Formation of halogenated by-products of parabens in chlorinated water. Anal. Chim. Acta 575 (1), 106–113. Carballa, M., Omil, F., Lema, J.M., Llompart, M., Garcia-Jares, C., Rodriguez, I., Gomez, M., Ternes, T., 2004. Behavior of pharmaceuticals, cosmetics and hormones in a sewage treatment plant. Water Res. 38 (12), 2918–2926. Drewes, J.E., Heberer, T., Reddersen, K., 2002. Fate of pharmaceuticals during indirect potable reuse. Water Sci. Technol. 46 (3), 73–80. Euro Chlor, European Federation of Chlor-alkali Producers. (accessed July 2009). EU, 2002. Commission Decision (2002/657/EC): implementing Council directive 96/23/EC concerning the performance of analytical methods and the interpretation of results. Off. J. Eur. Communities L067, 8–36. Ferreira, S.L.C., Bruns, R.E., Ferreira, H.S., Matos, G.D., David, J.M., Branda˜o, G.C., da Silva, E.G.P., Portugal, L.A., dos Reis, P.S., Souza, A.S., dos Santos, W.N.L., 2007. Box–Behnken design: an alternative for the optimization of analytical methods. Anal. Chim. Acta 597 (2), 179–186. Inaba, K., Doi, T., Isobe, N., Yamamoto, T., 2006. Formation of bromo-substituted triclosan during chlorination by chlorine in the presence of trace levels of bromide. Water Res. 40 (15), 2931–2937. Kim, S.D., Cho, J., Kim, I.S., Vanderford, B.J., Snyder, S.A., 2007. Occurrence and removal of pharmaceuticals and endocrine disruptors in South Korean surface, drinking, and waste waters. Water Res. 41 (5), 1013–1021. Kosjek, T., Heath, E., Petrovic, M., Barcelo, D., 2007. Mass spectrometry for identifying pharmaceutical biotransformation products in the environment. Trends Anal. Chem. 26 (11), 1076–1085. Kosjek, T., Heath, E., 2008. Applications of mass spectrometry to identifying pharmaceutical transformation products in water treatment. Trends Anal. Chem. 27 (10), 807–820. Lewis, G.A., Mathieu, D., Phan-Tan-Luu, R., 1999. Pharmaceutical Experimental Design in Drugs. Marcel Dekker, New York. Metcalfe, C.D., Koenig, B.G., Bennie, D.T., Servos, M., Ternes, T.A., Hirsch, R., 2003. Occurrence of neutral and acidic drugs in the effluents of Canadian sewage treatment plants. Environ. Toxicol. Chem. 22 (12), 2872–2880. Miyamoto, G., Zahid, N., Uetrecht, J.P., 1997. Oxidation of diclofenac to reactive intermediates by neutrophils, myeloperoxidase, and hypochlorous acid. Chem. Res. Toxicol. 10 (4), 414–419. Mompelat, S., Le Bot, B., Thomas, O., 2009. Occurrence and fate of pharmaceutical products and by-products, from resource to drinking water. Environ. Int. 35 (5), 803–814. Negreira, N., Canosa, P., Rodrı´guez, I., Ramil, M., Rubı´, E., Cela, R., 2008. Study of some UV filters stability in chlorinated water and identification of halogenated by-products by gas chromatography–mass spectrometry. J. Chromatogr. A 1178, 206–214. Okuda, T., Kobayashi, Y., Nagao, R., Yamashita, N., Tanaka, H., Tanaka, S., Fujii, S., Konishi, C., Houwa, I., 2008. Removal efficiency of 66 pharmaceuticals during wastewater treatment process in Japan. Water Sci. Technol. 57 (1), 65–71. Pa´xeus, N., 2004. Removal of selected non-steroidal antiinflammatory drugs (NSAIDs), gemfibrozil, carbamazepine, beta-blockers, trimethoprim and triclosan in conventional wastewater treatment plants in five EU countries and their discharge to the aquatic environment. Water Sci. Technol. 50 (5), 253–260.
water research 44 (2010) 243–255
Petrovic, M., Barcelo, D., 2007. Analysis, Fate and Removal of Pharmaceuticals in the Water Cycle. Elsevier, Amsterdam. Pinkston, K.E., Sedlak, D.L., 2004. Transformation of aromatic ether- and amine-containing pharmaceuticals during chlorine disinfection. Environ. Sci. Technol. 38 (14), 4019–4025. Pru¨tz, W.A., 1998. Reactions of hypochlorous acid with biological substrates are activated catalytically by tertiary amines. Arch. Biochem. Biophys. 357 (2), 265–273. Quintana, J.B., Reemtsma, T., 2004. Sensitive determination of acidic drugs and triclosan in surface and wastewater by ionpair reverse-phase liquid chromatography/tandem mass spectrometry. Rapid Commun. Mass Spectrom. 18 (7), 765–774. Quintana, J.B., Weiss, S., Reemtsma, T., 2005. Pathways and metabolites of microbial degradation of selected acidic pharmaceutical and their occurrence in municipal wastewater treated by a membrane bioreactor. Water Res. 39 (12), 2654–2664. Reemtsma, T., Jekel, M. (Eds.), 2006. Organic Pollutants in the Water Cycle. Wiley VCH, Weinheim. Reemtsma, T., Weiss, S., Mueller, J., Petrovic, M., Gonza´lez, S., Barcelo, D., Ventura, F., Knepper, T.P., 2006. Polar pollutants entry into the water cycle by municipal wastewater: a European perspective. Environ. Sci. Technol. 40 (17), 5451–5458. Rodil, R., Quintana, J.B., Lopez-Mahia, P., Muniategui-Lorenzo, S., Prada-Rodriguez, D., 2009. Multiresidue analytical method for the determination of emerging pollutants in water by solidphase extraction and liquid chromatography–tandem mass spectrometry. J. Chromatogr. A 1216 (14), 2958–2969. Rodrigues, P.M.S.M., Esteves da Silva, J.C.G., Antunes, M.C.G., 2007. Factorial analysis of the trihalomethanes formation in water disinfection using chlorine. Anal. Chim. Acta 595 (1–2), 266–274.
255
Stackelberg, P.E., Furlong, E.T., Meyer, M.T., Zaugg, S.D., Henderson, A.K., Reissman, D.B., 2004. Persistence of pharmaceutical compounds and other organic wastewater contaminants in a conventional drinking-water-treatment plant. Sci. Total Environ. 329 (1–3), 99–113. Stackelberg, P.E., Gibs, J., Furlong, E.T., Meyer, M.T., Zaugg, S.D., Lippincott, R.L., 2007. Efficiency of conventional drinkingwater-treatment processes in removal of pharmaceuticals and other organic compounds. Sci. Total Environ. 377 (2–3), 255–272. Tauxe-Wuersch, A., De Alencastro, L.F., Grandjean, D., Tarradellas, J., 2005. Occurrence of several acidic drugs in sewage treatment plants in Switzerland and risk assessment. Water Res. 39 (9), 1761–1772. Ternes, T.A., 1998. Occurrence of drugs in German sewage treatment plants and rivers. Water Res. 32 (11), 3245–3260. Ternes, T.A., Meisenheimer, M., McDowell, D., Sacher, F., Brauch, H.J., Gulde, B.H., Preuss, G., Wilme, U., Seibert, N.Z., 2002. Removal of pharmaceuticals during drinking water treatment. Environ. Sci. Technol. 36 (17), 3855–3863. Vanderford, B.J., Pearson, R.A., Rexing, D.J., Snyder, S.A., 2003. Analysis of endocrine disruptors, pharmaceuticals, and personal care products in water using liquid chromatography/tandem mass spectrometry. Anal. Chem. 75 (22), 6265–6274. Vanderford, B.J., Mawhinney, D.B., Rosario-Ortiz, F.L., Snyder, S.A., 2008. Real-time detection and identification of aqueous chlorine transformation products using QTOF MS. Anal. Chem. 80 (11), 4193–4199. Westerhoff, P., Yoon, Y., Snyder, S., Wert, E., 2005. Fate of endocrine-disruptor, pharmaceutical, and personal care product chemicals during simulated drinking water treatment processes. Environ. Sci. Technol. 39 (17), 6649–6663.
water research 44 (2010) 256–266
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Isotopic tracing of clear water sources in an urban sewer: A combined water and dissolved sulfate stable isotope approach J. Houhou a,*, B.S. Lartiges a,*, C. France-Lanord b, C. Guilmette b, S. Poix a,b, C. Mustin c a
Nancy University – LEM-ENSG/INPL-CNRS, Poˆle de l’Eau, 15 Avenue du Charmois, BP 40, 54501 Vandœuvre Cedex, France Nancy University – CRPG-CNRS, 15 rue notre dame des pauvres, BP 20, 54501 Vandoeuvre Cedex, France c Nancy University – LIMOS, BP 70239, 54506 Vandoeuvre Cedex, France b
article info
abstract
Article history:
This paper investigates the potential of stable isotopes of both water (dD and d18 OH2 O ) and
Received 16 February 2009
dissolved sulfate (d34S and d18 OSO4 ) for determining the origin and the amount of clear
Received in revised form
waters entering an urban sewer. The dynamics of various hydrological processes that
6 September 2009
commonly occur within the sewer system such as groundwater infiltration, rainwater
Accepted 8 September 2009
percolation, or stormwater release from retention basins, can be readily described using
Published online 12 September 2009
water isotope ratios. In particular, stable water isotopes indicate that the relative volumes of infiltrated groundwater and sewage remain approximately constant and independent of
Keywords:
wastewater flow rate during the day, thus demonstrating that the usual quantification of
Sewer system
parasitic discharge from minimal nocturnal flow measurements can lead to completely
Clear waters
erroneous results. The isotopic signature of dissolved sulfate can also provide valuable
Groundwater infiltration
information about the nature of water inputs to the sewage flow, but could not be used in
Water stable isotopes
our case to quantify the infiltrating water. Indeed, even though the microbial activity had
Sulfate isotope
a limited effect on the isotopic composition of dissolved sulfate at the sampling sites investigated, the dissolved sulfate concentration in sewage was regulated by the formation of barite and calcium-phosphate mineral species. Sulfate originating from urine was also detected as a source using the oxygen isotopic composition of sulfate, which suggests that d18 OSO4 might find use as a urine tracer. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
The reduction of groundwater infiltration and stormwater inflow in sewer systems is a major concern in urban water management (Marsalek et al., 2006). On a general basis, groundwater infiltration occurs through cracks or loose joints in sewer pipes, whereas in separate sewers, inflow is the result of improper connections between storm and sanitary sewers. Estimates of clear water discharge reported in the literature range from 30 to 72% of the sewage flow (Valiron and
Tabuchi, 1992; Verbanck, 1993; Kracht and Gujer, 2005; Kracht et al., 2007). Therefore, the extraneous flow drastically increases the costs associated with sewage treatment, the problem being magnified in wet weather especially as sudden inflows of clear rainwater often cause severe dysfunctioning of the treatment plant. The total amount of clear water entering the sewer system is classically inferred from the minimal nocturnal flow of wastewater (Chocat, 1997). Such an approach is rarely reliable, and further methods have been proposed to assess more
* Corresponding authors. E-mail addresses: [email protected] (J. Houhou), [email protected] (B.S. Lartiges). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.024
water research 44 (2010) 256–266
257
Fig. 1 – Main geological units of Greater Nancy and location of sampling sites for sewage (square), groundwater, (circle) and drinking water (black star) Detailed information on sampling sites can be found in Houhou et al. (2009) and in Table 1 (piezometers). WWTP (black diamond), LEM, and EMS, indicate the location of the sewage treatment plant, LEM laboratory where rainwater was collected, and Essey-le`s-Nancy weather station, respectively.
accurately the volume of extraneous flow. Thus, the fraction of clear water has been obtained by following both the wastewater flow and the concentration of specific sewage components such as borates (Verbanck, 1993), or chemical oxygen demand (Kracht and Gujer, 2005). In recent years, water isotope tracers (D and 18O) have become increasingly used in the context of the urban hydrological cycle, essentially to investigate the groundwater recharge from leaky sewers and water mains (Butler and Verhagen, 1997; Barrett et al., 1999; Navarro and Carbonell, 2007). Thus, Kracht et al. (2007) demonstrated that stable isotopes of water show promise for monitoring both the origin and the amount of extraneous waters in the sewer system. This method only requires that the isotopic signature of sewage be distinct from that of infiltrating water, which is the case in most cities where
drinking water – a proxy for sewage – is imported from outside the urban catchment area. However, the hydrological processes that occur in a sewer system can be quite complex, and the extraneous flow may involve more than two water sources. A multi-isotopic approach and/or supplementary hydrochemistry data are then required to quantify the various components of clear water (Barrett et al., 1999). Stable isotopes in dissolved sulfate (34S and 18 OSO4 ) have been considered for tracing water sources in urban aquifers (Barrett et al., 1999; Osenbru¨ck et al., 2007). Indeed, the wide range of sulfate isotopic signatures make those isotopes particularly sensitive indicators for investigating hydrological processes, provided that microbially mediated sulfate reduction does not alter the isotopic content of dissolved sulfate (Kno¨ller et al., 2008). The purpose of this
258
40.065 40.654 41.341 41.334 41.998 42.641 39.816
10.414 10.230 11.487 12.104 13.209 10.509 12.977
242 242 238 216 206 185 232
5 5 5 11.5 5 Well Well
0.9 7.6 3 – 1.1 4.7 1.5
study was then to explore the potential of stable isotopes of both water and dissolved sulfate, as tools to identify and to quantify the various sources that enter the sewer system in Greater Nancy urban community.
2.
Experimental section
2.1.
Study area
This study was conducted in the Greater Nancy urban catchment between November 2004 and December 2006. Nancy, a city of about 270,000 inhabitants located in north-eastern France, lies on both banks of the Meurthe river with a total catchment area of 193 km2 (144 km2 on left bank) (Fig. 1). As the Meurthe river water is brackish, the municipal water system is supplied with treated surface water originating from the Moselle river. The source of the Moselle river is located in the Vosges mountains, 250 km southeast of Nancy. The Moselle river mainly drains silicate formations in the upper part of the catchment, whereas it flows on carbonate and evaporite sediments downstream of Epinal (Brenot et al., 2007). The imported water is pumped at Messein and conveyed to the drinking water treatment plant through two 11 km aqueducts. Three reservoirs with a total storage capacity of 2.5 106 m3 secures the water supply in case of pollution or high turbidity level in the Moselle river. The drinking water distribution system delivers daily about 67,600 m3 via 950 km of pipes and 33 storage facilities (total capacity of 600,000 m3). Approximately 26% of drinking water is lost through leakage. The sewer system comprises about 1200 km of pipes, 250 km of them man-entry sewer (pipe diameter 1.2 m). As Nancy is established in basin position surrounded by hills dominating the Meurthe valley by about 200 m, the slope of the gravity sewer pipe may reach 5%. The central core of Nancy urban community is served by a combined sewer, whereas the peripheral areas are drained by separate systems. 25 detention basins (180,000 m3 storage capacity) have been constructed to limit discharges from combined sewer overflows in the Meurthe river. As depicted in Fig. 1, the geological subsurface of Nancy area is composed of three main units: the downtown area is underlain by Meurthe river alluvial deposits, Toarcian shales extend to the hillsides, whereas Bajocian recifal limestone caps the surrounding hills. Shallow aquifers are found in the three geological units, their hydraulic head being monitored
20
400
10 200
5
0
0
15
600
b
400 WWS 04/04/06
10 5
200 0
0 60
Discharge (m3/s)
48 –.–N
15
600 WWS 29/03/05
c
1000 GWS 25/10/06
750
40
500 20
WWS 26/10/06
250 0
0
Discharge (m3/s)
Piezometer Depth to diameter groundwater 006 Altitude (cm) (m) –.–E (m)
a
Discharge (m3/s)
CHAR VOI LE3 STD RG STEP CM
GPS coordinates
Rainfall (mm/day)
Sampling site
20
Rainfall (mm/day)
Table 1 – Piezometer characteristics.
Rainfall (mm/day)
water research 44 (2010) 256–266
Time (day)
Fig. 2 – Rainfall hyetographs and Moselle river hydrographs during the month preceding a given sampling campaign. (a) JB diurnal campaign on March 29–30, 2005; (b) PVP diurnal campaign on April 04–05, 2006; (c) Spatial sampling at the urban catchment scale on October 25–26, 2006. Dashed arrows indicate a snow event.
from 40 piezometers spread over the catchment area. However, the rather narrow diameter of most piezometers prevented groundwater sampling with the submersible pump in the vicinity of sewage sampling sites. Groundwater recharge occurs mainly from rainfall infiltration, leakage from water mains being negligible (<4% over the catchment area). The local climate is continental with an average annual precipitation of 760 mm occurring mostly during summer. Temperature and precipitation data were obtained from Nancy-Essey meteorological weather station, whereas Moselle river flow discharge data were measured at nearby gauge Tonnoy (Hydro Bank, Moselle Tonnoy Station, A5110610).
2.2.
Sample collection and analysis
Three sampling campaigns (29–30 March 2005, 4–5 April 2006, 25–26 October 2006) are reported in this paper. The location of sampling sites for sewage, drinking water, rainwater, and groundwater is shown in Fig. 1. A detailed description of those sampling sites has been provided in a previous publication (Houhou et al., 2009). Supplementary information on piezometers location and characteristics are given in Table 1. Fig. 2 shows the temporal evolution of rainfall amounts and Moselle river flow discharge during the month preceding each of three sampling campaigns. Sampling was generally conducted in dry weather, the minor precipitation event on March 29 not generating surface runoff. It should be noted that snowfalls occurred during the last two weeks of February 2005, which then led to the application of deicing salts on the roads. Grab-samples of sewage were taken from the sewer through open manholes using a polyethylene bucket, whereas
259
500
-48 -49 -50 -51 -52 -53 -54 -55 -56
400 300 200 100
Discharge (m3/s)
δD (‰/V-SMOW)
water research 44 (2010) 256–266
0 14/02/2007
04/02/2007
25/01/2007
15/01/2007
05/01/2007
26/12/2006
16/12/2006
06/12/2006
26/11/2006
16/11/2006
06/11/2006
27/10/2006
17/10/2006
07/10/2006
Fig. 3 – Temporal evolution of both drinking water dD and Moselle river discharge during the period October 07, 2006 to February 09, 2007.
-47 -48 -49 -50 -51 -52 -53 -54 -55 -56 -57
06h20
W
L
03h50
02h05
LM
CHAR 03h50 -7.0
06h20 02h05
-7.5
15h50 -8.0
2
VOI
δ OH O (‰ /V -SM OW )
-6.5
15h50
-8.5
18
δ D (‰ /V-SMOW)
a
-9.0
08:15 10:30 13:45 17:00
20:15 23:30
02:45 06:00
Tim e (H our)
-8.5
-8.0
-7.5
-7.0
-6.5
δ OH O (‰ /V-SMOW) 18
2
b
16 Moselle river ML Gypsum SW
12
34
δ S (‰ /CDT)
groundwater was collected from piezometers using a portable submersible pump (Mini-Twister (9.5 L/min) SDEC France). The piezometer was first purged for at least three well volumes before being sampled. Integrated rainwater samples were collected between September 2006 and November 2006 using a homemade sampler placed on the roof of LEM laboratory. Tap water was sampled at CRPG and LEM laboratories before and after the sewage sampling campaigns. Prior to sampling, the faucet was left open for 15 minutes. Sample characteristics such as temperature, pH, conductivity, dissolved oxygen concentration, redox potential (WTW, Multiline F/SET) were measured immediately after collection. The water sample was then filtered on-site through prewashed 0.22 mm pore-size cellulose-acetate membranes (Chromafil CA-20/25) and split into five aliquots for cations, anions, dissolved organic carbon (DOC), and isotopic analyses. The filtrates were stored in 65 mL polyethylene or glass bottles at 4 C until analysis. Soluble cation and trace element concentrations were measured on HNO3 acidified samples using a Jobin-Yvon JY70 ICP-AES (Inductively Coupled Plasma Atomic Emission Spectrometry) and Perkin–Elmer ELAN 6000 ICP-MS (Inductively Coupled Plasma Mass Spectrometry), respectively. Uncertainties were better than 2% for major cations and 5% for trace elements. Sulfate, chloride, nitrate, and phosphate anions were determined by ion chromatography using a Dionex ICS-3000 (AS9-HC column). The lower detection limits were 50 ppb for PO4 and 20 ppb for the other anions. DOC was measured with a Dohrman 190 analyzer. The oxygen isotope composition of water was measured by the conventional H2O–CO2 equilibration method (Epstein and Mayeda, 1953), using a modified VG 602D dual-inlet mass spectrometer. D/H isotopic ratio was obtained with a Isoprime mass spectrometer (GV Instruments, Manchester, UK) coupled with an elemental analyzer EA3000 after reducing the water to hydrogen gas on chromium at 1050 C (Morrison et al., 2001). The dissolved sulfate in water samples and the aluminum coagulant used at the drinking water plant, were precipitated as BaSO4 by adding a 5% BaCl2 solution (Brenot et al., 2007; Calmels et al., 2007). In order to prevent BaCO3 co-precipitation, the water samples were first acidified with HCl to pH 4.2 and then heated to boiling point to ensure a complete removal of CO2. BaSO4 precipitates were carefully washed and dried, and purity was checked by X-ray Diffraction (Bruker D8 diffractometer). The sulphur isotopic
06h20
03h50 02h05
8
Tap water 15h50
4
14h00
Urine
0
Al-Coagulant
5
6
7
8 9 10 11 12 δ18O SO (‰ /V-SMOW)
13
14
4
Fig. 4 – (a) dDL18 dOH2 O plot of sewage samples collected at JB sampling site on March 29–30, 2005. LMWL refers to the Local Meteoric Water Line (solid line); the dashed line represents the mixing line between the domestic sewage and the stormwater released from the retention basin. VOI and CHAR represent groundwaters collected at nearby piezometers. The inset indicates the variation of d18 OH2 O during the diurnal investigation. (b) d34 SLd18 OSO4 plot of dissolved sulfate contained in the previous sewage samples (-), the aluminum coagulant used at the drinking water treatment plant ( ), urine from three healthy people (6) and gypsum (:); the gray rectangle and the gray circle represent the range of d34S and d18 O values for Moselle river water (Brenot et al., 2007) and drinking water, respectively.
260
water research 44 (2010) 256–266
12 Sewage + Stormwater samples
11 10 9
15h50 02h05 03h50 06h20
5.7 3.7 3.5 2.6
44.1 44.7 38.3 30.5
25.2 64.2 43.6 25.1
0.6 1.1 0.5 0.2
7.3 12.9 11.0 15.6
188 38.9 39.0 22.1
Legret and Pagotto (1999)
4
18
34
2.6
19
0.33
147
18
4
Cl SO2 Pb Cu Cd Zn NO 3 4 (mg/ (mg/ (mg/ (mg/ (mg/ (mg/ (mg/ L) L) L) L) L) L) L)
a δ OSO (‰ /V-SMOW)
Table 2 – Comparison of major anions and trace elements concentrations of samples identified as a stormwater release from a retention basin with median analyses of filtered runoff waters from Legret and Pagotto (1999).
8 7 0
Rsample d¼ 1 1000 Rref mat_ _ The standards materials are V-SMOW (Vienna Standard Mean Ocean Water) for d18 OH2 O , d18 OSO4 and dD, and CDT (Canyon Diablo Troilite) for d34S. The accuracy of isotopic analyses was estimated by replicate analyses of internalstandards along with the sample series. Data presented in this study are given with uncertainty of 0.2& and 1& for d18 OH2 O and dD respectively, and the overall reproducibility (2s level) of d18 OSO4 and d34S analyses is better than 0.3& using the barium sulfate NBS 127 international reference.
3.
Results and discussion
3.1.
Suitability of isotopic systems
The identification of various water sources that may contribute to the flow of sewage requires, to the least, measurable isotopic differences between them. Previous successful studies using stable isotopes of water in an urban context relied on quasi-constant hydrological differences between the source of water supply (i.e. reservoir or lake water) and the local groundwater (Butler and Verhagen, 1997; Kracht et al., 2007). In our case, the water supply comes from treated Moselle river water, which implies that drinking water, and hence domestic sewage, possess a distinct isotope composition from local rainwater and groundwater. Indeed, the isotopic signature of meteoric waters depends on local geographic and climatic factors, whereas that of groundwater integrates over time the composition of water inputs (precipitation, mains and sewer leaks,.) that recharge the
50
80 Dissolved Sulfate (mg/L)
b composition was analyzed from the SO2 released from BaSO4 at 1100 C in presence of Sn and tungsten oxide (Giesemann et al., 1994), whereas the 18O/16O of SO4 was determined from the CO generated by mixtures of BaSO4 and glassy carbon at 1270 C in presence of Ni catalyst (Koziet, 1997). 34S/32S and 18 O/16O isotopic ratios of sulfate were measured using a GV Instruments Isoprime mass spectrometer coupled in continuous flow mode with a EuroVector elemental analyzer. The results are reported in d-notation (d18 OH2 O , dD, d34S, 18 d OSO4 ), i.e. the permil deviation of the measured isotopic ratio (Rsample) relative to a reference material (Rref.mat.):
10 30 40 20 Soluble P (orthophosphate) (mg/L)
02h05
60
Average tap water 03h50
40 15h50
06h20
20
0
Rainwater
08:15 10:30 13:45
17:00 20:15 23:30 02:45 Time (hour)
06:00
Fig. 5 – (a) Relationship between d18 O and soluble orthophosphate for sewage collected at JB sampling site; the circled data points correspond to the mixing of sewage and stormwater released from a nearby retention basin (INIST). (b) Temporal evolution of dissolved sulfate content in sewage collected at JB sampling site during March 29–30, 2005 diurnal campaign; the dashed lines represent the average concentration in dissolved sulfate in drinking water and rainwater, respectively.
aquifer (Mook, 2001). However, the hydrological regime of a river may strongly influence the isotopic signal of the surface water (Mook, 2001), and accordingly that of the corresponding treated water. As illustrated in Fig. 3, the temporal evolution of dD values from tap water collected at CRPG over a four month period, clearly follows the discharge of the Moselle river with a temporal shift of about 4 days. A decrease in dD that ranges from 2 to 5&, can be measured after each peak in flow discharge. The temporal shift between the two curves includes both the transport of surface water from the pumping facility to the treatment plant, the treatment of raw water, and the residence time within the drinking water distribution system. As a result, a drinking water of slightly varying isotopic composition may be supplied over the urban community. Nevertheless, the deviation in dD was not observed to exceed 2& over two consecutive days, which is the magnitude of analytical uncertainty for dD. The applicability of stable sulfate isotopes to differentiate water inputs in the sewer system is more arguable. On the one hand, dissolved sulfate may be derived from a variety of sources (evaporite dissolution, sulfide minerals oxidation, atmospheric precipitation, .), whose distinct isotopic signatures make d34S and d18 OH2 O effective indicators for
water research 44 (2010) 256–266
a
-46
δ (‰ /V-SMOW) δD
-48
LM
-50
W
L
Groundwater (RG)
Night-time
-52 00h00
Gremillon stream
-54
21h00
-56 -58
M
g ix i n
line
Day-time
Tap water
-60 -9.5
-9.0
-8.5 -8.0 -7.5 δ18O H O (‰ /V-SMOW)
-7.0
-6.5
2
b
16 Moselle river water
34
δ S (‰ /CDT)
12 8
Gypsum
Tap water Urine
4
Al-Coagulant
0
Day-time
-4
Gremillon stream
-8 Groundwater (RG)
-12 4
5
6
7
8
9
10
11
12
In the present study, the isotopic signatures of local water sources should allow a ready discrimination: the d34S values of dissolved sulfate in Moselle river water vary from 12.1& to 13.4& with little seasonal variation (Brenot et al., 2007), that of rainwater are about 5.8& (Brenot et al., 2007) which is consistent with the range of published values for atmospheric sulfate derived from anthropogenic sources (Yu et al., 2007), whereas groundwater from Toarcian aquifers should display a negative isotopic signal as dissolved sulfate originates in that case from sulfide oxidation (Goldhaber and Kaplan, 1980). Such isotopic contents should not be significantly modified during sewage transport as the sloped catchment maintains a relatively high oxygen content in most sewer pipes (4.06 3.29 mg/L), and thus restricts the activity of sulfate-reducing bacteria (SRB) to biofilms (Vollertsen et al., 2008 and references herein). Moreover, the SRB amount in sewage was found to be rather limited at our sampling sites (counts lower than 4033/mL) (see the Supporting information material).
3.2. Stability of water and sulfate isotopic signatures in the sewer
10:20
Night-time
261
13
14
δ O SO (‰ /V-SMOW) 18
4
18
Fig. 6 – (a) dDLd OH2 O plot of sewage samples collected at PVP sampling site on April 04–05, 2006 (-), water samples collected from Gremillon urban stream (:), RG groundwater (B), and drinking water (,). LMWL refers to the Local Meteoric Water Line (solid line); the dashed line represents the mixing line between drinking water and RG groundwater. (b) d34 SLd18 OSO4 plot of dissolved sulfate contained in the previous sewage samples (-), water samples collected from Gremillon urban stream (:), RG groundwater (B), drinking water (,), the aluminum coagulant used at the drinking water treatment plant ( ), urine from three healthy people (6) and gypsum (:); the gray rectangle represents the range of d34S and d18 O values for Moselle river water (Brenot et al., 2007); the dashed line represents the mixing line between drinking water and RG groundwater.
investigating hydrological processes (Berner et al., 2002; Osenbru¨ck et al., 2007). On the other hand, dissolved sulfate rarely behaves as a conservative tracer, both its concentration and its isotopic composition being most often affected by a combination of redox reactions and physico-chemical processes such as H2S release and mineral sulfide precipitation (Van Everdingen and Krouse, 1985; Kno¨ller et al., 2008). In the sewer system, microbially mediated sulfate reduction is likely to occur (Zhang et al., 2008). This would induce significant, but not easily predictable, isotopic fractionations essentially characterized by an enrichment in heavier isotopes in the remaining dissolved sulfate (Chambers and Trudinger, 1979; Rudnicki et al., 2001 and references herein). Such fractionation could then drastically alter the initial isotopic composition of sulfate.
Fig. 4a shows the dD d18 OH2 O plot of sewage samples collected during a diurnal investigation (29 March 2005) from a trunk in the separate domestic sewer that was unlikely to undergo groundwater infiltration. In that context, the temporal evolution of isotope signals is expected to be limited. Indeed, for stable isotopes of water, most of d18 OH2 O dD values fall into a small range that is close to the local meteoric water line (LMWL). However, four samples (sewage taken at 15h50, 02h05, 03h50, and 06h20) clearly diverge to the right of LMWL and show a fair correlation between dD and d18 OH2 O values. Such linear trend is reminiscent of a local evaporation line, the observed slope of about 4 being typical of waters undergoing secondary evaporative isotopic enrichment (Gibson et al., 2008). However, a significant increase in sewage flow rate was noted at these particular sampling times. This suggests that the observed correlation would rather correspond to a binary mixing line between two water sources. Groundwaters collected from two piezometers located nearby the study site present a distinct isotopic composition (Fig. 4a) and do not identify the unknown end-member source. The distribution of stable sulfate isotopes (d18 OSO4 and d34S) reveals a slightly different pattern (Fig. 4b). If the four previous samples are again characterized by a distinct trajectory denoted SWML hereafter (StormWater Mixing Line), the other d18 OSO4 values vary from 7.4& to 11.2& with an almost constant d34S at about 6.3&. Interestingly, the isotopic signature of gypsum from evaporite layers (d34S ¼ 14.1& and d18 OSO4 ¼ 12.6&) (Brenot et al., 2007), is situated on SWML, suggesting that this sulfate source contributes to the isotopic composition of four samples. The unknown end-member water source could then correspond to surface runoff, since some gypsum is present in the deicing salt used for melting snow and ice on roadways (Legret and Pagotto, 1999). Moreover, as shown in Table 2, the concentrations in anions and trace elements of samples collected at 3h50 and 06h20 are remarkably close to the median compositions of filtered runoff waters collected by Legret and Pagotto (1999). Indeed, the Greater Nancy Hydraulic Department later confirmed that
262
water research 44 (2010) 256–266
b
200 Groundwater (RG)
160 120 80 40
Tap water
0 07:50 10:45 13:45 16:45 19:45 22:45 01:45 04:45 06:20
Dissolved sulfate (mg/L)
Dissolved Ca (mg/L)
a
160 140 120 100 Groundwater (RG)
80 60 07:50 10:45 13:45 16:45 19:45 22:45 01:45 04:45 06:20
Time (Hour)
d
14 12 10 8 6 4
50 40
Groundwater (RG)
Tap water
DOC (mg/L)
Dissolved Ni (µg/l)
c
Time (Hour)
30 20 10 0
2 07:50 10:45 13:45 16:45 19:45 22:45 01:45 04:45 06:20
Time (Hour)
07:50 10:45 13:45 16:45 19:45 22:45 01:45 04:45 06:20
Time (Hour)
Fig. 7 – Temporal evolution of dissolved calcium (a), dissolved sulfate (b), dissolved Ni (c), and dissolved organic carbon DOC (d), at PVP sampling site on April 04–05, 2006.
a retention basin collecting stormwater runoff, was emptied in that section of separate domestic sewer during low-flow periods. Assuming that the 6h20 sample represents the endmember source, the deviation from LMWL observed for water isotopes suggests that the amount of rainwater evaporated during surface runoff was about 6% for that particular rain event (Craig and Gordon, 1965). Fig. 4b also reveals that the isotopic composition of sulphur in drinking water (d34S ¼ 6.70&) has undergone a significant depletion in 34S compared with that of Moselle river water (from 12.1& to 13.4&). Such variation is primarily due to the addition of aluminum sulfate (d34S w 0&) during the coagulation stage of raw water treatment. Furthermore, as indicated by the arrow in Fig. 4b, the measured d18 OSO4 is generally slightly shifted to the left of possible mixing lines between the aluminum coagulant and the Moselle river water end-member sources. This presumably originates from the oxidation of carbon-bonded sulphur contained in natural organic matter (Brenot et al., 2007), the raw water being chlorinated at the pumping facility and the clarified water undergoing ozonation at the treatment plant. The d34S value of carbon-bonded sulphur is generally similar to that of initial sulfate contained in rainwater, i.e. about 5.8& in the present study (Brenot et al., 2007), and it does not significantly change upon re-oxidation (Krouse and Grinenko, 1991). The isotopic signal of sulphur remains relatively constant in sewage samples during the diurnal investigation, whereas that of oxygen presents a slight enrichment in lighter 16O for most of them (Fig. 4b). Such variation in oxygen isotopic composition suggests that a supplementary source contributes to the dissolved sulfate pool in sewage. The fair correlation obtained between d18 OSO4 and phosphate concentration (Fig. 5a) identifies sulfate from human excreta as a potential source. Indeed, inorganic sulfate represents the main end-
product of sulphur metabolism in the human body, though other forms such as ester sulfate represent a 9–15% fraction of urinary sulfate (Hoffer et al., 2005). Sulfate excreted in urine derives from the oxidation of aminoacids such as cysteine and methionine. To our knowledge, the d18 OSO4 value generated by such a process has not been documented in the literature. Our own measurements obtained from the urine of three healthy people drinking only Greater Nancy tap water, gave values around 4.5& for d34S and between 5.9& and 7.5& for d18 OSO4 (Fig. 4b). It should be noted that sulfate esters are not precipitated upon BaCl2 addition (Lundquist et al., 1980), and that the oxygen isotopic composition of such compounds is not taken account in the d18 OSO4 of urinary sulfate. The fate of sulfate esters in the sewer has not been described in the literature, but they are likely to generate inorganic sulfate upon microbial hydrolysis (Fitzgerald, 1976). In principle, the relative contribution of dissolved sulfate from urine in sewage can be calculated from: d18 OSO4 ðsewageÞ ½SO4 sewage ¼ xDW d18 OSO4 ðDWÞ ½SO4 DW þxurine d18 OSO4 ðurineÞ ½SO4 urine
(1)
where xDW and xurine represent the percentages of drinking water and of urine contained in sewage, respectively. Average [SO4]urine is taken equal to 0.016 mol/L (Udert et al., 2003). However, negative values of xurine are then obtained from Eq. (1), thus implying a non-conservative behavior for dissolved sulfate. Solubility calculations reveal that the domestic sewage at JB is slightly supersaturated with regard to barite (BaSO4) – saturation indices ranging from 0.04 to 0.38 – except during the stormwater discharge episodes from the retention basin (Saturation as low as 2.2). Moreover, Houhou et al. (2009) have shown that brushite, a common calcium-
263
water research 44 (2010) 256–266
-48
PV
-50
JB RG
-52
27 oct
28 oct
-54
26 oct 25 oct
LM 24 oct
W
RA
L
LO
-8.0
-7.5
(2)
where xGW is the percentage of groundwater in sewage and xDW þ xGW ¼ 1 – about 63% of wastewater is composed of infiltrated groundwater, whereas the urban stream should contain approximately 20% of mains water. Such values are consistent with those calculated from soluble Ca and Ni concentrations in groundwater and in drinking water (Fig. 7a, b). Gremillon stream may receive sewage, i.e. drinking water, from leaks of the adjacent separate sanitary sewer. However, as several storm drains empty into the urban stream, it may also be contaminated through unintended cross-connections between sanitary sewer and storm sewer pipes. Interestingly, sewage samples collected during night-time and at 14h15 are characterized by dD values slightly lighter than those of daytime sewage. This indicates that a third end-member source contributes to the flow of sewage in this area. As these water isotope data plot close to the local meteoric water line, they might identify recently infiltrated rainwater. The d34 S d18 OSO4 diagram shows that samples from Gremillon stream, day-time and night-time sewage, occupy three fields that clearly overlap (Fig. 6b). The relatively low d34S value at 10.50& for sulfate in groundwater may be attributed to the oxidation of sulfide minerals contained in the Toarcian shales that form the shallow aquifer at PVP sampling site
-7.0
δ O H O (‰ /V-SMOW) 18
2
b
20 16 12
34
Fig. 6a shows the d18 OH2 O dD values of (i) sewage grab-samples collected over a 24 h sampling period at PVP, (ii) water samples taken the same day from the Gremillon urban stream which is adjacent to the sewer pipe, (iii) groundwater sampled at a nearby piezometer located at about 3 km downstream of the sampling site, and (iv) drinking water. The isotopic compositions of groundwater and drinking water differ by about 10 times the standard deviation of measurement, which ensures the detection of about 10% of infiltrated groundwater in sewage. Two outliers (sewage samples collected at 21h00 and 24h00) are shown in Fig. 6a but excluded from the discussion. Three main fields can be distinguished on the graph. Sewage samples taken during day-time are clustered around whereas Gremillon water d18 OH2 O ¼ 8&jdD ¼ 54.5&, samples occupy a field around d18 OH2 O ¼ 7.7&jdD ¼ 52.8&. Both groups are located near a mixing line between drinking water and groundwater. Assuming binary mixing – e.g.
GWL
PVP
-56 -8.5
Tracing parasitic sewer infiltration
dDsewage ¼ xDW dDDW þ xGW dDGW
-46 Rainwater 11-23 oct
δ S (‰ /CDT)
3.3.
a δ (‰ /V-SMOW) δD
phosphate mineral species that controls the dissolved phosphate concentration in Greater Nancy sewage, also incorporates small amounts of sulfate in its lattice. It is then likely that the activity of dissolved sulfate in sewage is regulated by the formation of barium sulfate and phosphate minerals (Fig. 5b). Therefore, even though urinary sulfate certainly contributes to the shift in d18 OSO4 , its relative proportion in sewage cannot simply be inferred from the oxygen isotopic composition of dissolved sulfate. It should also be pointed out that the involvement of inorganic sulfate from personal care products and detergents in the d18 OSO4 of sewage is not known either. The highest d18 OSO4 values at 10.6& and 11.2& are observed during and just after the afternoon low-flow period. No consistent explanation for those two values has yet been identified.
Moselle river Gypsum CL
8 4 0
LO
Tap water
Urine RA
-4 -8
Al-Coagulant
CM
-12 -16
RG
0
3
6
9
12
15
δ18O SO (‰ /V-SMOW) 4
Fig. 8 – (a) dDLd18 OH2 O plot of samples collected during the October 2006 campaign over Greater Nancy catchment area: (-) sewage, (B) groundwater, (,) drinking water, and (C) integrated sample of rainwater. LMWL and GW refer to the Local Meteoric Water Line and local groundwater line (solid lines); the dashed line represents the mixing line between sewage taken at JB sampling site and the integrated rainwater sample. (b) d34 SLd18 OSO4 plot of dissolved sulfate contained in the previous samples, aluminum coagulant used at the drinking water treatment plant ( ), urine from three healthy people (6) and gypsum (:); the gray rectangle represents the range of d34S and d18 O values for Moselle river water (Brenot et al., 2007); the dashed line represents the mixing line between drinking water and RG groundwater.
(Berner et al., 2002; Calmels et al., 2007). The isotopic composition of sulfates contained in sewage and in the urban stream, is intermediate between those of drinking water and of groundwater. However, the parasitic discharge in the sanitary sewer cannot be calculated from isotopic values because the content in dissolved sulfate found in sewage largely exceeds that measured in RG groundwater. Such discrepancy likely results from an incorrect identification of an end-member source, i.e. RG groundwater. Surprisingly, all day-time isotopic values are in the same close range. This implies that the relative volumes of sewage and groundwater remain almost constant in those samples, even though the amount of discharged wastewater obviously varies during the day. In their study of Ru¨mlang sewer system, Kracht et al. (2007) observed the same phenomenon and
264
water research 44 (2010) 256–266
a
b 240 200
mg/L
mg/L
160 120 80 40 0 CHU
JB
PV
RM
70 60 50 40 30 20 10 0 CHU
Sampling-site
JB
PV
RM
Sampling-Site
Fig. 9 – Variation of (a) Na and Cl concentrations (gray and white columns, respectively) and (b) Ca and Ba concentrations (gray and white columns, respectively) along CHU sewer section.
attributed the diurnal variation of the infiltration discharge to the presence of water pools and backwater zones in the sewer system. A similar explanation can certainly be invoked here. Furthermore, it is puzzling that the large amount of infiltrated groundwater becomes much lesser at night, the additional extraneous flow evidenced at that time being clearly minor during day-time. The presence of a nearby inverted siphon used to convey the sewage underneath a roadway, might explain such phenomenon: a decreased head at the upstream end of the siphon would stop its functioning during low-flow periods, and more particularly at night, the groundwater then infiltrating originating from a different sewer section located downstream of the siphon. These two observations clearly demonstrate that the quantification of clear water discharge from minimal nocturnal flow measurements can lead to erroneous results.
3.4.
Isotopic tracing in the sewer system at the city scale
Fig. 8a shows the water isotopic compositions of samples collected during the October 2006 campaign in the Greater Nancy catchment area. Drinking water and the majority of sewage samples plot close to the LMWL, whereas most groundwater define a straight line parallel but slightly offset to the right of LMWL. Unfortunately, the isotopic content of drinking water varied significantly during this sampling campaign, which may account for part of the scattering in sewage samples. Nevertheless, it is clear that most sewage samples are influenced by rainwater, their position being roughly intermediate between drinking water samples and the 11–23 October integrated rainwater end-member source. Such influence is also revealed by the change in sewage hydrochemistry at downstream sampling sites. Fig. 9 indicates that while sodium and chloride concentrations have been halved between JB and PV sampling sites, the sewage has acquired calcium and barium, the solute concentrations of those increasing by a factor of 2. Indeed, PV is positioned approximately halfway between rainwater and JB in the dD d18 OH2 O space. However, the increase in calcium concentration cannot be related to a release of stormwater from a detention basin. Instead, as rainwater infiltrates through the vadose zone, the weak acid formed from the solubilization of CO2 gas promotes the dissolution of minerals and the release of Ca2þ ions when
the ground is dominated by carbonates (Stumm and Morgan, 1996). The change in the isotopic composition of sewage is then consistent with a significant rainwater percolation in the sewer system. Groundwater infiltration may however dominate at RA and LO sampling sites, the isotopic characteristics of these sewage samples being located close to the groundwater line (GWL). The isotopic compositions of sulphur and oxygen in dissolved sulfate are presented in Fig. 8b. Drinking water shows a slight depletion in both 34S and 18O between 24 and 25 Oct. and 27 and 28 Oct. that can be interpreted as an intensification of the coagulation process applied to clarify Moselle river water. The isotope data from dissolved sulfate in groundwater appear to plot along a mixing line, thus suggesting the presence of two main local aquifers of differing water quality in the urban watershed. Most isotopic values of dissolved sulfate in sewage are clustered between drinking water and urine pools. Unlike water isotopes, rainwater percolation in the sewer system cannot be easily detected from the sulfate isotopic signature of sewage since the sulfate concentration in precipitations is quite low (Brenot et al., 2007). Two sewage samples show a drastically different behavior: (i) CL sewage, taken from a domestic sewer, is characterized by a dissolved sulfate enriched in 34S, which might be attributed to bacterially mediated SO4 reduction with removal of lighter H2S species; (ii) the relatively low d34S–SO4 value of RA sewage supports a significant groundwater infiltration at this sampling site.
4.
Conclusion
This study confirms the interest of a multi-component approach to unravel the complex nature of hydrologic phenomena that may occur in an urban sewer. Thus, typical processes such as (i) groundwater infiltration, (ii) stormwater release from a retention basin, and (iii) rainwater percolation, were readily identified from the combined use of isotopic and hydrochemistry data. Water stable isotopes were effective tools for determining the quantity and the nature of clear waters entering the sewer system. Isotopes from dissolved sulfate were also found appropriate for constraining the origin of water inputs, but could not be used to evaluate the
water research 44 (2010) 256–266
contributions of infiltrating water to the sewage flow due to the non-conservative behavior of dissolved sulfate. Diurnal sampling campaigns at a given sampling site always provided valuable information about the local functioning of the sewer system, provided that all end-member sources could be identified. In contrast, spatial sampling of sewage conducted at the urban catchment scale, showed various entries of clear waters and thus introduced supplementary difficulties in the interpretation of results.
Acknowledgements This study was mainly supported by a grant from CNRS-INSU (ECODYN). Lorraine Region also provided part of the funding through Zone Atelier Moselle. We are grateful to M. Pique´ and M. Villeroy (Service hydraulique CUGN) for logistical support during all field studies. We thank S. Bouly (ASGA) for providing hydrogeological information on CUGN piezometers. G. Frappier, J. Sieliechi, A. de Carvalho, and E. Montarge`s-Pelletier conducted sampling in the field, and this project would not have been possible without their precious help. We also wish to thank the staff of SARM (CRPG-UPR 80) where ICP analyses were carried out.
Appendix. Supplementary information Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2009.09.024.
references
Barrett, M.K., Hiscock, K.M., Pedley, S., Lerner, D.N., Tellam, J.H., French, M.J., 1999. Marker species for identifying urban groundwater recharge sources: a review and case study in Nottingham, UK. Water Res. 33 (14), 3083–3097. Berner, Z.A., Stu¨ben, D., Leosson, M.A., Klinge, H., 2002. S- and O-isotopic character of dissolved sulphate in the cover rock aquifers of Zechstein salt dome. Appl. Geochem. 17, 1515–1528. Brenot, A., Carignan, J., France-Lanord, C., Benoıˆt, M., 2007. Geological and land use control on d34S and d18O of river dissolved sulphate: the Moselle river basin, France. Chem. Geol. 244, 25–41. Butler, M.J., Verhagen, B.T., 1997. Environmental isotopic tracing of water in the urban environment of Pretoria, South Africa. In: Chilton, P.J., et al. (Eds.), Problems, Processes and Management. Groundwater in the Urban Environment, vol. 1. Balkema, pp. 101–106. Calmels, D., Gaillardet, J., Brenot, A., France-Lanord, C., 2007. Sustained sulfide oxidation by physical erosion processes in the Mackenzie river basin: climatic perspectives. Geology 35 (11), 1003–1006. Chambers, L.A., Trudinger, P.A., 1979. Microbial fractionation of stable sulfur isotopes: a review and critique. Geomicrobiol. J. 1, 64–71. Chocat, B., 1997. Encyclope´die de L’hydrologie Urbaine et de L’assainissement, Lavoisier, Technique et Documentation. Craig, H., Gordon, L.I., 1965. Deuterium and oxygen 18 variations in the ocean and the marine atmosphere. In: Togiorgi, E. (Ed.), Stable Isotope in Oceanographic Studies and
265
Paleotemperatures. Laboratorio di Geologia Nucleare, Pisa, Italy, pp. 9–130. Epstein, S., Mayeda, T., 1953. Variation of d18O content of waters from natural sources. Geochim. Cosmochim. Ac. 4, 213–224. Fitzgerald, J.W., 1976. Sulfate ester formation and hydrolysis: a potentially important yet often ignored aspect of the sulfur cycle of aerobic soils. Bacteriol. Rev. 40 (3), 698–721. Gibson, J.J., Birks, S.J., Edwards, T.W.D., 2008. Global prediction of da and d18O–d2H evaporation slopes for lakes and soil water accounting for seasonality. Global Biogeochem. Cycles 22, GB2031. Giesemann, A., Ja¨ger, H.-J., Norman, A.L., Krouse, H.R., Brand, W. A., 1994. On-line sulfur isotope determination using an elemental analyzer coupled to a mass spectrometer. Anal. Chem. 66, 2816–2819. Goldhaber, M.B., Kaplan, I.R., 1980. Mechanisms of sulfur incorporation and isotopic fractionation during early diagenesis in sediments of the Gulf of California. Mar. Chem. 9, 95–143. Hoffer, L.J., Hamadeh, M.J., Robitaille, L., Norwich, K.H., 2005. Human sulfate kinetics. Am. J. Physiol. Regul. Integr. Comp. Physiol. 289, R1372–R1380. Houhou, J., Lartiges, B.S., Hofmann, A., Frappier, G., Ghanbaja, J., Temgoua, A. 2009. Phosphate dynamics in an urban sewer: a case study of Nancy, France. Water Res., 43(4), 1088–1100. Kno¨ller, K., Vogt, C., Feisthauer, S., Weise, S., Weiss, H., Richnow, H.H., 2008. Sulfur cycling and biodegradation in contaminated aquifers: insights from stable isotope investigations. Environ. Sci. Technol. 42 (21), 7807–7812. Koziet, J.S., 1997. Isotope ratio mass spectrometric method for on-line determination of oxygen 18 in organic matter. J. Mass Spectrom. 32, 103–108. Kracht, O., Gresch, M., Gujer, W., 2007. A stable isotope approach for the quantification of sewer infiltration. Environ. Sci. Technol. 41 (16), 5839–5845. Kracht, O., Gujer, W., 2005. Quantification of infiltration into sewers based on time series of pollutant loads. Water Sci. Technol. 52 (3), 209–218. Krouse, H.R., Grinenko, V.A., 1991. Stable Isotopes. In: Natural and Anthropogenic Sulfur in the Environment. SCOPE, vol. 43. Wiley Chichester, U.K. 267–306. Legret, M., Pagotto, C., 1999. Evaluation of pollutant loadings in the runoff waters from a major rural highway. Sci. Total Environ. 235, 143–150. Lundquist, P., Martensson, J., So¨rbo, B., Ohman, S., 1980. Turbidimetry of inorganic sulfate, ester sulfate, and total sulfur in urine. Clin. Chem. 26 (8), 1178–1181. Marsalek, J., Jime´nez-Cisneros, B.E., Malmquist, P.-A., Karamouz, M., Goldenfum, J., Chocat, B., 2006. Urban Water Cycle Processes and Interactions. IHP-VI Technical Documents in Hydrology. No. 78. UNESCO, Paris. Mook, W.G., 2001. Environmental Isotopes in the Hydrological Cycle. Principles and Applications. IHP-V. Technical documents in hydrology. No. 39. UNESCO-IAEA.li. Morrison, J., Brockwell, T., Merren, T., Fourel, F., Phillips, A.M., 2001. On-line high-precision stable hydrogen isotopic analyses on nanoliter water samples. Anal. Chem. 73 (15), 3570–3575. Navarro, A., Carbonell, M., 2007. Evaluation of groundwater contamination beneath an urban environment: the Beso`s river basin (Barcelona, Spain). J. Environ. Manage. 85, 259–269. Osenbru¨ck, K., Gla¨ser, H., Kno¨ller, K., Weise, S.M., Mo¨der, M., Wennrich, R., Schirmer, M., Reinstorf, F., Busch, W., Strauch, G., 2007. Sources and transport of selected organic micropollutants in urban groundwater underlying the city of Halle (Saale), Germany. Water Res. 41, 3259–3270. Rudnicki, M.D., Elderfield, H., Spiro, B., 2001. Fractionation of sulfur isotopes during bacterial sulfate reduction in deep
266
water research 44 (2010) 256–266
ocean sediments at elevated temperatures. Geoch. Cosmochim. Acta 65, 777–789. Stumm, W., Morgan, J.J., 1996. Aquatic Chemistry. Wiley Intersciences. Udert, K.M., Larsen, T.A., Gujer, W., 2003. Estimating the precipitation potential in urine-collecting systems. Water Res. 37, 2667–2677. Valiron, F., Tabuchi, J.P., 1992. Maıˆtrise de la pollution urbaine par temps de pluie – Etat de l’art. Agence de l’Eau SeineNormandie/AGHTM. Tech & Doc, Lavoisier, p. 564. Van Everdingen, R.O., Krouse, H.R., 1985. Isotope composition of sulphates generated by bacterial and abiological oxidation. Nature 315, 395–396.
Verbanck, M.A., 1993. A new method for the identification of infiltration waters in sanitary flows. Water Sci. Technol. 27 (12), 227–230. Vollertsen, J., Nielsen, A.H., Jensen, H.S., Hvitved-Jacobsen, T., 2008. Modeling the formation and fate of odorous substances in collection systems. Water Environ. Res. 80 (2), 118–126. Yu, J.-Y., Park, Y., Mielke, R.E., Coleman, M.L., 2007. Sulfur and oxygen isotopic compositions of the dissolved sulphate in the meteoric water in Chuncheon, Korea. Geosci. J. 11 (4), 357–367. Zhang, L., De Schryver, P., De Gusseme, B., De Muynck, W., Boon, N., Verstraete, W., 2008. Chemical and biological technologies for hydrogen sulfide emission control in sewer systems: a review. Water Res. 42, 1–12.
water research 44 (2010) 267–277
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Impacts of residence time during storage on potential of water saving for grey water recycling system S. Liu a,*, D. Butler b, F.A. Memon b, C. Makropoulos b, L. Avery c, B. Jefferson d a
Department of Environmental Science & Engineering, Tsinghua University, Beijing 100084, China Centre for Water Systems, School of Engineering, Computer Science & Mathematics, University of Exeter, Exeter EX4 4QF, UK c Catchment Management, Macaulay Institute, Aberdeen AB15 8QH, UK d Centre for Water Science, Cranfield University, Cranfield MK43 0AL, UK b
article info
abstract
Article history:
Grey water recycling has been generally accepted and is about to move into practice in terms
Received 29 April 2009
of sustainable development. Previous research has revealed the bacteria re-growth in grey
Received in revised form
water and reclaimed municipal water during storage. However, in most present grey water
5 September 2009
recycling practices, impacts of water quality changes during storage on the system’s
Accepted 8 September 2009
performance and design regulation have not been addressed. In this paper, performance of
Available online 15 September 2009
a constructed wetland based grey water recycling system was analysed by taking the constraint of residence time during storage into account using an object based household
Keywords:
water cycle model. Two indicators, water saving efficiency (WSE ) and residence time index
Grey water recycling
(RTI ), are employed to reflect the system’s performance and residence time during storage
Residence time
respectively. Results show that WSE and RTI change with storage tank volumes oppositely. As
Storage tank
both high WSE and RTI cannot be achieved simultaneously, it is concluded that in order to
Water quality degradation
achieve the most cost-effective and safe solution, systems with both small grey and green
Water saving
tanks are needed, whilst accepting that only relatively modest water saving efficiency targets can be achieved. Higher efficiencies will only be practicable if water quality deterioration in the green water tank can be prevented by some means (e.g. disinfection). ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Grey water is defined as the water which is slightly contaminated by human activities and may possibly be reused after suitable treatment, for example, water from a washing machine, shower, bath etc. The reclaimed water or the treated grey water is termed as green water in this paper. Grey water recycling is emerging as an internal part of water demand management, promoting as it does the preservation of high quality fresh water supplies as well as potentially reducing the pollutant in the environment. The principle of domestic grey water reuse is to replace all or some of the non-potable water demand by reclaimed water. The general use of treated grey
water in a household context mainly includes toilet flushing and/or garden watering. This paper focuses on the toilet flushing. In the last decade, grey water recycling practices have been reported in many countries (Asano and Levine, 1996; Fittschen and Niemczynowicz, 1997; Kayaalp, 1996; Nolde, 2000; Smith et al., 2000; Yang et al., 2006).
1.1.
Grey water characterisation
As grey water arises from domestic washing operations, it varies in quality according to, amongst other things, geographical location, demographics and level of occupancy (Al-Jayyousi, 2003). Taking BOD as an indicator, its value has
* Corresponding author. E-mail address: [email protected] (S. Liu). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.023
268
water research 44 (2010) 267–277
been reported between 33–466 mg l1 in the literature (Al-Jayyousi, 2003; Prathapar et al., 2005; Gross et al., 2007). It has been noticed that the BOD in grey water from hand basin is slightly smaller than the one from combined sources (bath, shower, etc.) (Jefferson et al., 2000) and the quality of grey water varies with time during a day (Al-Jayyousi, 2003). Further to this, available evidences have shown changes of grey water quality during storage. Jefferson et al. (2000) reported that a 50% reduction in BOD over a 4 h period could be achieved. However, longer residence time in storage tank can encourage bacteria re-growth and lead to degradation of water quality. Dixon et al. (2000) conducted an experiment to examine the change of grey water quality during storage and observed obvious increases of BOD and DO after initial decrease in the sedimentation period. Therefore, the quality of grey water fed into treatment device is expected to be various not only because of the different sources and time of grey water generated, but also the changes of quality during storage. In terms of healthy concern, a key question or objective in domestic grey water recycling is to ensure the green water complies with relevant standards. This may be accomplished both by choosing robust and effective treatment and limiting grey water degradation during storage by rational design. This paper will mainly focus on the latter topic. On the other hand, the concept of grey water recycling is to reduce potable water demand by replacing non-potable demand with green water in term of water demand management and sustainable development. From this point of view, the objective of grey water recycling is to save as much potable water as possible. These two objectives may interact, or even conflict each other. Their interaction/confliction should be explored to increase the understanding and confidence in implementation of grey water recycling. Rational design can then be undertaken based on these understandings to ensure the great achievement of both objectives in recycling practices. This will be discussed in this paper.
1.2.
Treatment technologies
Researchers have reported the application of several technologies for grey water treatment. Both strengths and constraints in implementation of these technologies have been recognized. Sand filtration plus disinfection represents the most common technology used for domestic grey water recycling in the UK (Jefferson et al., 2000). The treated grey water from this kind of system has been noticed remaining high in organic load and turbidity, which thereby limit the effectiveness of the chemical disinfection process. Membrane systems offer a permanent barrier to suspended particles greater than the size of membrane material, which can range from 0.5 mm of microfiltration membranes down to molecular dimensions for reverse osmosis. The key factor constraining the viability of membrane systems is the fouling of the membrane surface by pollutant species. This has been reported by many researchers. For example, Nghiem et al. (2006) and Oschmann et al. (2005). Meanwhile, the energy demand for membrane systems is high (Jefferson et al., 2000). Biological treatment and physical treatment can effectively remove different species. The benefits of biological and physical treatments are combined in processes such as
membrane bioreactors (MBR). However, high cost implications have meant that this kind of treatment is more suitable for large-scale of recycling scheme than single house. As a low-cost technology, constructed wetland has recently gained much attention in grey water treatment. Experiences in Central America (Dallas et al., 2004), Middle East (Gross et al., 2007) and the UK (Frazer-Williams et al., 2008) showed that high averaged removal rate can be achieved provided appropriate hydraulic retention time is given. In this project, therefore, a constructed wetland based grey water recycling system is chosen to investigate the impact of residence time in storage tanks on the system’s performance.
1.3.
System configuration
Although different system configurations have been reported in practice, a grey water recycling system generally includes: a grey water storage tank, a treatment unit and a green water storage tank. For the system investigated in this project (Fig. 1), it also has the similar system configuration. The grey water tank is connected to appliances, which consumes potable water and produces grey water. By collecting the grey water, the grey water tank stores and feeds it to the constructed wetland, where the green water is produced. The constructed wetland was placed outside of house. The constructed wetland is linked with the green water tank. The green water tank then collects and serves green water to nonpotable water demand, for example, the toilet flushing. The design of grey water recycling system is a site-dependent problem. The storage tanks can be either placed underground or on the loft in terms of specific circumstance and the user’s preference. Pumps may be employed to facilitate the flows between treatment device, storage tanks and toilet cistern when gravity flow is not a choice. For the purpose of simplification, in this project, the grey water recycling system is simulated in a common sense, i.e. no specific implementation situation was considered, only the main parts of the system (storage tanks, treatment device and toilet cistern) and the dynamic flows among them were simulated. Although practices of grey water recycling system have been implemented widely, most published literatures mainly focused on reporting the performance of existing systems (for example, Jeppesen, 1996; Al-Jayyousi, 2003; March et al., 2004; Ghisi and Mengotti de Oliveira, 2007). Few attentions have been focused on the impacts of system configuration on potential of water saving. Especially, no attempt has been made to investigate this problem by taking the water quality degradation during storage into account. This paper concentrates on the analysis of potential of water saving from water quality point of view.
2.
Methodology
2.1.
The household water cycle model
The household water cycle model adopted in this project was developed on a MATLAB (Simulink) platform. It accounts for the production and storage of grey water and green water, and the water balance between compartments, for example, the water
269
water research 44 (2010) 267–277
Grey water flow Green water flow Potable water flow
Float switch: green water top up
Float switches and top ups
Mains Overflow
Green water tank
Float switch: mains top up Overflow Toilet T: Toilet W: Washing machine Hand basin, bath etc
Constructed wetland T
Grey water storage tank
W
Overflow
Fig. 1 – Illustration of grey water recycling system.
supply and the water demand. The allocation of water sources to water demands is facilitated by a ‘first comes, first served’ rule. This refers that the water request will be satisfied according to its appearing sequence. The model operates at 10 min time step which is determined by the data availability. The household water cycle model was designed with the capacity of coping with any type of treatment system operating modes. However, complicated management strategy is required to facilitate intermittent operating mode. Therefore, in this paper, for the purpose of simplification, continuous operating mode is assumed for the operation of constructed wetland.
2.2.
Input data
The input data required by the household water cycle model are the water use profile information for each appliance. To understand the performance of grey water recycling system at different situations (for example, peak and non-peak uses of toilet; different water use manners on weekdays and weekend), it is necessary to assess its behaviour over an extended period, ideally to cover its expected lifetime. However, in practice, it is hard to source this kind of data. Therefore, in this project, a Monte-Carlo method was adopted to generate water use profile time series data covering 10 years period at a time step of 10 min. The parent data uitilised in the Monte-Carlo method was derived from a large-scale survey conducted by Water Research Centre UK (WRc) to investigate water consumption trends in different parts of the UK. In this survey, flow meter and data logger were used to identify flow characteristics and classify water use events, which can be the use of toilets, showers, baths, internal and external taps, washing machines and dishwashers (Ton That, 2005). The system is capable of recording every 10 ml of water used at 1 s intervals for periods up to 2 week. In this research, water profile data from 100 three-person households was employed. Fig. 2 shows, by average, a threeperson household requires 369.11 l water per day, in which 103.99 l for toilet flushing, 39.91 l for washing machine, 55.56 l
for bath, 50.70 l for shower and 118.95 l for tap uses. Distributions of water use events in terms of time and household were examined. Spatial and temporal differences of water use event were found. Taking toilet flushing as an example, Fig. 3 shows the cumulative number of toilet use event in every 10 min interval during a day (144 intervals) for the 100 households. Except for the morning and evening peak uses, toilet flushing is featured as a randomized event. Fig. 4 displays the distribution of number of toilet use event and household numbers, which reveals that most households (79 households) use 10–14 times of toilet per day. It is also noticed from Fig. 4 that 8 households use less than 7 times of toilet per day, which might be because of less people living in. In generating water use profile time series data using Monte-Carlo method, spatial and temporal differences were taken into account to represent the differences of water use event in term of time and household.
2.3.
Residence time distribution
The residence time (RT ) in a storage tank is calculated according to the ‘first in first out’ (FIFO) algorithm (Walski et al., Water use profile(litres/day) 103.99 , 28% 118.95 , 32% Toilet Washing Machine Bath Shower Tap 39.91 , 11% 50.70 , 14% 55.56 , 15%
Fig. 2 – Water use profile data for three-person households.
270
water research 44 (2010) 267–277
48 h to avoid the significant water quality degradation (Dixon et al., 2000). No similar research has been conducted for green water. However, experience from reclaimed municipal water suggests green water quality degradation during storage is expected (Narasimhan et al., 2005). For simplification, therefore, a 0–48 h TR was adopted both for grey water and green water. To evaluate to what extent RT is within the TR (0–48 h), a residence time index (RTI ) is introduced, calculated as the ratio of the integral of RTD over the TR to the one over the whole range (Fig. 5). A RTI value of 0 means no RT is in the range of 0–48 h, while 1 indicates that all water leaving the storage tank as outflow stays in the tank less than 48 h. The greater the RTI is, the better the storage tank performs in terms of avoiding water quality degradation. R
RTD dRT
TR
RTITR ¼ R
2.4.
Performance indicators
Two indicators are employed to evaluate the system’s performance. One is from quality aspect, the residence time index. The other is from quantity aspect, the water savings efficiency. The time range of a RT for the question under discussion is defined as target range (TR). Previous research has recommended that the RT in a grey water tank should not be beyond
Number of household
30 25 20 15 10 5 0
1
2
3
4
5
6
7
8
9
10
11
12
13
Toilet use events per day
Fig. 4 – Distribution of toilet use event in terms of household.
14
15
(1)
WR
Fig. 3 – Distribution of toilet use event in terms of time.
2003). In the FIFO algorithm, the first volume of water to enter the storage tank as inflow is the first to leave as outflow. In the household water cycle model, each parcel of water is noted with the times entering and leaving the storage tank. The difference between these two times indicates the period the water staying in the tank and is therefore the RT, which is calculated at each time step and has a precision down to 10 min. The probability of a RT is accounted by dividing the number of its appearance over the whole running period (10 years in this case) with the total number of appearance for all RTs. Residence time distribution (RTD) refers to the curve of the probability against its corresponding RT (illustrated in Fig. 5). The RTD describes the probability and range of RT of water in the storage tank.
RTD dRT
in which RTITR refers to the RTI for the target period of TR; TR is the target range up to 48 h; WR is short for the whole range of retention time. Water saving efficiency (WSE ) is defined as the percentage of potable water saved by reusing grey water. It reflects to what extent the toilet demand is satisfied by non-potable water. A higher WSE means more potable water is saved. PT Wt WSE¼ 100 Pt¼1 T t¼1 Dt
(2)
where: T ¼ run duration; Wt ¼ amount of non-potable water used for toilet flushing; Dt ¼ toilet water demand
3.
Model simulation and discussions
By feeding the input data series into the household water cycle model, the water dynamics in household water cycle over 10 years time was simulated. The attention was paid on the impacts of storage tank and treatment capacity on the potential of water savings with consideration of limitation for residence time. Model simulation for a three-person household was taken as an example for demonstration purpose.
3.1.
The RT in grey and green water tanks
The function of grey and green water tanks is to deal with the synchronicity between water sources and demands. In this paper, the treatment is assumed to operate in a continuous mode, which implies that the outflow from the grey water tank and the inflow to the green water tank are continuous and at a constant rate. Meanwhile, the inflow to the grey water tank and the outflow from the green water tank are dependent on the grey water production and toilet water demand respectively, and they are at intermittent patterns. So, unlike a typical treatment reactor, which has an RT dictated solely by flow rate (for a given reactor volume), this RT will be more complex giving both varying tank volumes and intermittent supply/demand. The RTD is the reflection of the comprehensive interactions between inflow and outflow rates and patterns, and the volume of storage tank. In order to provide an insight to these
water research 44 (2010) 267–277
Fig. 5 – Illustration of residence time distribution.
interactions, the investigations of the impact of tank volume on RT for both grey and green water tanks are conducted in two types of analyses: offline analysis and online analysis. In the offline analysis, the inflow and outflow of the grey and green tanks keep unchanged while the tank volumes vary. The relationships of RTD and RTI0-48 with tank volume are focused. The interactions and impacts from other system components are not considered. Thus, the offline analysis offers a static snapshot understanding of the RT during storage. For the green water tank, the outflow is toilet water demand and the inflow is related to the treatment capacity. For the grey water tank, the inflow is grey water production from household water consumptions and the outflow is determined by the water request from treatment device, which is also related to the treatment capacity. An arbitrarily given treatment capacity value is adopted here. In the online analysis, the RT during storage is investigated by taking the system component interactions into account. For example, when the RT of a green water tank is under investigation, not only the impacts of volume of the green tank, but also the volume of grey water tank are considered. This reflects the situation in a real system. Therefore, the online analysis provides a more systematic understanding of the RT. It should be noted that the value of inflow to green water tank or the outflow from the grey water tank might not be the same as the treatment capacity because the grey water demand of the treatment device may not be always satisfied in the online analysis. The actual value, not the potential treatment capacity is adopted in the online analysis, while the potential treatment capacity is used in the offline analysis.
3.1.1.
271
The median value of the RTD is accounted as 1.88 days. The RTI0-48 is calculated to be 0.60 according to equation (1). For the 200 l tank, the median value of RTD and the RTI0-48 are 3.12 days and 0.43 respectively. It is observed that the increase of storage tank volume results in a longer residence time, and therefore, a reduced RTI0-48. This is also revealed in chart B, in which the RTI0-48 curve for tank size from 0 l to 1000 l with the same inflow rate is displayed. Chart C in Fig. 5 shows the result of the online analysis. It is clear that both the green and the grey water tank volumes impose impact on the RTI0-48 of the green water tank. It decreases with increasing of grey and green water tank volumes. However, the RTI0-48 is more sensitive to the green water tank volume. It is also noticed that very slight impact is imposed on the RTI0-48 by the size of grey water tank when the green water tank is relatively small (for example, less than 150 l). This is because the adoption of threshold treatment capacity values. The household model operates at a 10 min time scale and a ‘spill after yield’ assumption. In the model, the amount of grey water to spill is calculated after serving the treatment device in each time step. When the grey water tank volume is rather small, the grey water generated in the household is more prone to spill. The difference between available grey water tank capacity and the grey water production in each time step is termed as amount to potentially spill (APS) in the case of the former is smaller than the latter. When the grey water tank is small, its ‘buffer’ function in adjusting the inflow and the outflow is not significant. At this circumstance, the more the APS uptaked by the treatment device (i.e. the bigger the treatment capacity), the more grey water would be possibly reused. For a recycling system with a small grey water tank, the highest WSE might appear when the treatment capacity is big enough to uptake all APS. This results in a large threshold value of treatment capacity. The difference between grey water in APS and in the grey water tank is that the latter can last beyond the current time step in the tank, while the former will spill if it is not uptaked in the current time step. When a bigger grey water tank is employed (more grey water can then be supplied from the grey water tank), a relatively small treatment capacity may be required to produce the same amount of green water as the situation of small grey water tank with large treatment capacity. For both situations, when the green water tank is small, more than enough (compared to the green water tank volume) green water can be produced. Different from the grey water tank, in which outflow is continuous and the APS can be uptaked by the treatment device, the outflow from the green water tank is intermittent (determined by the toilet water demand) and APS will be more possible to spill rather than to be uptaked by toilet cistern. Therefore, the RTIs of small green water tanks, as shown in chart C, will be rather steady regardless the volume of grey water tank.
The green water tank
The results of the offline analysis for the green water tank are given in Fig. 5. It presents the RTD of a 50 l green water tank with an inflow of 0.7 l per 10 min, which corresponds to a treatment capacity of 100 l per day, and the RTD of a 200 l green water tank with the same inflow rate. As shown in Fig. 6 (chart A), for the 50 l tank, most water flowing out of the green water tank (excluding overflow) resided 0–10 days in the tank.
3.1.2.
The grey water tank
The results of the offline analysis for the grey water tank are shown in Fig. 7. Chart A in Fig. 7 presents the RTDs of 50 and 200 l grey water tanks with 100 l per day treatment capacity. The median value of the RTDs and the RTI0-48s for 50 l tank and 200 l tank are: 0.25 days and 1, 1.64 days and 0.95 respectively. It is clearly shown that the grey water is more prone to reside
272
water research 44 (2010) 267–277
Fig. 6 – Results for residence time of green water: online and offline analyses.
longer in a bigger grey water tank for a given treatment capacity. The RTI0-48 decreases with increasing grey water tank volume. This is also reflected by chart B in Fig. 7, which depicts the changes of RTI0-48 with various grey water tank volumes for a given treatment capacity. Chart B also suggests that, for a given treatment capacity, the RTI0-48 of grey water tank remains 1 for the grey water tank volume up to a specific threshold value (for example, for treatment capacity 100 l per day, the threshold value for grey water tank is about 190 l (chart B in Fig. 7)). The RTI0-48s for grey water tanks which are smaller than the threshold value are expected to be 1. A smaller RTI0-48 will be yielded for grey water tank which is bigger than this threshold. This turning point indicates the maximum grey water tank volume which a specific treatment capacity can ‘digest’ in terms of residence time up to 48 h. The turning point for a bigger treatment capacity is expected to be higher. The results of the online analysis are presented in chart C in Fig. 7. It is observed that both grey and green water tank volumes have impact on the RTI0-48 of the grey water tank. However, grey water tank volume is more influential on the value of RTI0-48. It should be noticed that similar to the investigation for the green water tank, the threshold treatment capacity values are adopted in the online analysis for the grey water tank. The contour for
RTI0-48 ¼ 1 indicates a front that any combination of grey and green water tank volumes below it can lead to the residence time of grey water during storage is statistically lower than 48 h given the adoption of threshold treatment capacity.
3.2. Relationship of potential WSE with grey and green water tanks Fig. 8 shows WSE versus treatment capacity for 200 l grey and green water tanks. It clearly indicates that WSE is maximised at a threshold treatment capacity of 200 l per day for this configuration. Beyond this point, efficiency slowly declines regardless the increasing of treatment capacity. This effect is produced by the complex interactions between water supply and demand in relation to the filling of the two tanks, remembering that the green water tank has the potential for mains top up if it cannot supply the requested demand. For given volumes of grey and green water tank volumes, a bigger treatment capacity means more grey water could be treated into green water. However, it might also result in less grey water to be actually reused for toilet flushing because a bigger treatment capacity can encourage overflow from the green water tank and deficit of grey water. By iterating this
water research 44 (2010) 267–277
273
Fig. 7 – Results for residence time of grey water: online and offline analyses.
calculation for any combinations of grey and green tanks in a reasonable range, the relationship of potential WSE with grey and green tank volumes can be explored. This is performed from two aspects: volume based analysis and quality based analysis. In the volume based analysis, the interaction between system performance and storage tank volumes are only investigated from water quantity aspect. The treatment device is assumed to be robust enough to cope with low quality grey water. It is also assumed that the quality of green water reaches the relevant standards and regulations before being consumed. In the quality based analysis, the relationship of grey and green water residence times during storage and their implications on water quality degradation are taken into account.
3.2.1.
water tank volume. However, the increase is not symmetrical. For a given green water tank size, impact of grey water tank volume changing on WSE is small. Whereas in the converse case, for a given grey water tank size, impact of changing green water tank volume on WSE is significant. The figure clearly indicates the relative importance of the green tank volume in terms of achieving high WSE. The figure clearly indicates the relative importance of the green tank volume in relation to the grey water tank. For the same total volume, a higher WSE is expected for the combination with a greater size of green water tank. For example, for 800 l total storage volume, the combination of 700 l grey þ 100 l green yields 60% WSE (point a in Fig. 9), 400 l grey þ 400 l green gives 76% WSE (point b), while 87% WSE is expected for the combination of 100 l grey þ 700 l green (point c).
Volume based analysis
In the volume based analysis, the system was assessed by taking just the quantity balance between water supply and demand into account. A range of different configurations was evaluated, based on both grey and green water tank volumes up to 1000 l, and treatment capacities up to 1000 l per day. Result of this analysis is presented in Fig. 9. It is observed that potential WSE increases with increasing total volume of grey and green water tank. For a given grey/green water tank volume, the WSE also increases with increasing green/grey
3.2.2.
Quality based analysis
For the quality based analysis, the same configurations as the ones in the volume based analysis are adopted. Results are shown in Fig. 10. Chart A in Fig. 10 presents the contours of WSE and RTI0-48 (for the grey tank) for different grey and green water tank volumes. It indicates that the residence time in the grey water tank is prone to be longer than 48 h when both grey and green water tanks are large (as shown in the top right area in chart A). This implies that special attention should be paid
274
water research 44 (2010) 267–277
volume based analysis implies that a high WSE (i.e. over 85%) can be achieved with reasonable configurations (sizes – grey: 150 l and green 650 l, point a in chart B). However, for this configuration, the RTI0-48 for the green water tank is 0.3. Currently, there are no standards for RTI0-48, but in the interim if a value of 0.5 is suggested as a reasonable target figure, chart B clearly indicates it is not possible to achieve both an RTI048 ¼ 0.5 and WSE ¼ 85%, whatever size tanks are used. If an RTI0-48 standard of 0.5 is needed, a WSE of no greater than 65% is possible based on a small green water tank (150 l).
4.
Fig. 8 – Relationship of WSE and treatment capacity.
in sizing storage tanks for a recycling system employing less robust treatment device like constructed wetland discussed in this paper, whose removal performance is sensitive to inflow grey water quality. Although the RTI0-48 of grey water tank imposes some influence on the system performance, a high WSE (i.e. over 85%) can still be accomplished theoretically by choosing rational grey and green water tanks given the RTI0-48 of the green water tank is also satisfactory. For example, a combination of relatively big green water tank and small grey water tank can promote both high WSE and low residence time (as shown in the bottom right area in chart A). Chart B in Fig. 10 shows the relationships of WSE and RTI048 (green) with grey and green water tank volumes. Opposing relationships are observed, such that higher grey and green water tank volumes lead to higher WSE but lower RTI0-48. In terms of water demand management, an objective to save as much potable water as possible is generally pursued. The
Discussion
A basic concern in grey water recycling is that the green water quality is good enough for non-potable purpose use and complies with relevant standards and regulations. Previous studies have revealed the water quality degradation of grey water and municipal reclaimed water during storage. This might decrease the effluent quality of treatment device. Furthermore, the quality of grey water produced varies with time due to different sources (for example, quality of grey water from bath is different from washing machine). This imposes more uncertainties and variabilities on quality of influent grey water to treatment device. Therefore, methods should be taken to ensure that the grey water is to be treated before its quality degrades to unacceptable level. In other words, residence time less than a certain value (48 h in this work) as a criterion should be taken into account in system design. Similar consideration applies to green water tank to ensure the green water to be supplied for toilet before quality degrades to unacceptable level. The RTI is introduced in this paper to assess the probability of water residing in a storage tank over a certain period. From the analysis for grey and green water tanks, it is observed that
Fig. 9 – Relationship of WSE and system configuration: volume based analysis.
water research 44 (2010) 267–277
275
Fig. 10 – Relationship of WSE and system configuration: quality based analysis.
the RTI0-48 is related to the volume of storage tank, inflow and outflow patterns. For the same volume of grey and green water tanks, differences in feature of RTD and value of RTI0-48 are observed. For example, the RTI0-48s for 200 l grey and green water tanks are 0.95 and 0.34 respectively (Fig. 6 and 7). The shapes of RTD also show opposite trends (charts A in Figs. 6 and 7). In the grey water tank, the probability of grey water flowing out (not spilling) immediately after flowing in (RT N 0 h) is rather small and it is more prone to reside for a while. However, in the green water tank, the green water is more prone to flow out immediately although the overall probability of residing over 48 h is bigger than the one for a grey water tank with the same volume (RTI0-48 grey 200 l ¼ 0.95 or RTIover 48 ¼ 0.05; RTI0-48 green 200 l ¼ 0.34 or RTIover 48 ¼ 0.66). This attributes to the different patterns of inflow and outflow of the storage tank. In the grey water tank, the inflow is the production of grey water in a household, which is intermittent and might be with high flow intensity over a short time period (for example, the use of bath and shower). The outflow is the water request from treatment device, which is continuous and with relatively low flow intensity. This results in that the grey water produced in the current time step is more possible to flow out afterwards. However, the situation for a green water tank is opposite, in which the inflow (green water production) is continuous and with relatively flow intensity and outflow is intermittent (toilet water request) and with relative high flow intensity (Assuming 1 toilet event with 9 l water request in a 10 min time step, the flow intensity is 9 l per 10 min. For comparison, the flow intensity for a 100 l per day treatment device is about 0.7 l per 10 min). This explains why the probability of RT N 0 h for
green water tank is high (chart A in Fig. 6) and the probability of RT N 0 h for grey water tank is low (chart A in Fig. 7). The residence times of water in grey and green water tanks are investigated both with online and offline situations in this paper. The former assumes the storage tank is isolated from the system except for the adoption of real inflow for grey water tank (grey water production) and outflow for green water tank (toilet water request) and ignores the interaction between grey and green water tanks, and the treatment device. It provides a static snapshot on the relationship of RT with the volume of storage tank and representative inflow and outflow patterns. The latter investigates the RTs by taking the dynamic interactions between treatment capacity, grey and green water tanks. The relationship of RT with grey and green water tank volumes for a system with optimal design is explored. It reveals the systematic influence on RT of grey and green water. The finding from the offline analysis provides an insight to the understanding of RT due to volume of storage tank and flow patterns. The finding from the online analysis offers a systematic understanding of relationship of RT with system configurations and assist in revealing the constraint of RT in system design and potential water saving. In the investigation for the impact of system configuration on potential of water saving, the quality based analysis reveals the impacts of RTs on the potential of water saving. Small values of RTI0-48 are observed for big grey and green water tanks in achieving high WSE. However, it is still possible to achieve both high WSE and RTI0-48 (for the grey tank) by rational system design for a recycling system with less robust treatment device theoretically given the RTI0-48 (for the green tank) is satisfactory. More significant influence of RTI0-48 (for the green tank) on
276
water research 44 (2010) 267–277
Fig. 11 – Impacts of target range on WSE.
WSE is noticed. Although no standard for RTI0-48 (for the green tank) value is currently available, results from quality based analysis show that the WSE is inevitably reduced to pursue to higher RTI0-48 (for the green tank). It is also suggested that a combination of ‘big grey’ and ‘small green’ be employed to achieve a higher RTI0-48 (for the green tank) without trading off the WSE (chart B in Fig. 7). This does not conflict with the finding from the volume based analysis, in which a combination of ‘big green’ and ‘small grey’ is recommended. In the volume based analysis, this conclusion is drawn on the condition of the same total volume of storage tank and in terms of water saving. However, in the quality based analysis, it is concluded to aim a better RTI0-48 (for the green tank) and the total volume of storage tanks are not the same. The discussion above indicates that the adoption of 0–48 h TR can significantly reduce the potential of potable water saved in a grey water recycling system. Fig. 11 demonstrates that a broader TR, RT less than 4 days (96 h) in the green water tank, can increase the potential of water saving by 16% for a grey water recycling system with the same configurations. In most present grey water recycling practices, the green water normally serves the toilet demand without further treatment. Therefore, target range of less than 2 days residence time is adopted in this project. The main concern of grey water recycling in terms of water demand management is to save as much potable water as possible, provided the water quality is satisfied. Result from above discussion shows that the main constraint to the percentage of potable water saved is the RT in the green water tank. Therefore, solutions should be sought to tolerate longer RT without compensate the quality of green water significant to achieve a higher water saving efficiency. A possible answer to this might be to introduce another treatment or disinfection option, between green water tank and toilet cistern.
5.
Conclusions
1. This paper explores the potential of water saving for a constructed wetland based grey water recycling system by taking the residence time of water during storage tank into account.
The dynamics of water cycle in a household over 10 years is simulated using an object based model at a 10 min time step. Results from the investigation of removal performance for different qualities of grey water suggest that attention should be paid in prohibiting degradation of grey water during storage for a constructed wetland based water recycling system. This conclusion may also apply to other grey water recycling system with less robust treatment technologies. 2. Analysis for the residence time in grey and green water tanks indicates that RTD and RTI are dependent on the volume of storage tank, inflow and outflow patterns. Results from the volume based analysis reveal that the WSE increases with increasing storage tank volumes. For a given total storage volume, greater WSE can be achieved by using greater volumes of green tank. Therefore, system configurations using larger green and smaller grey tanks are recommended in practice provided a suitable treatment strategy is employed. The quality based analysis has highlighted that although larger volume tanks produce higher water saving efficiencies, smaller volume tanks are needed to secure good water quality. Indeed water saving efficiencies of greater than approximately 60% cannot be safely achieved. 3. As both high WSE and RTI cannot be achieved simultaneously, it is concluded that in order to achieve the most cost-effective and safe solution, systems with both small grey and green tanks are needed, whilst accepting that only relatively modest water saving efficiency targets can be achieved. Higher efficiencies will only be practicable if water quality deterioration in the green water tank can be prevented by some means (e.g. disinfection). In this research, the effect of temperature on deterioration in storage tank is not considered. It is suggested that its impact should be included in future research.
Acknowledgement This work is developed by the ‘Water Cycle Management for New Developments’ (WaND) project [www.wand.uk.net] funded under the Engineering & Physical Science Research Council’s ‘‘Sustainable Urban Environment’’ Programme by EPSRC, UK government and industrial collaborators.
references
Al-Jayyousi, O.R., 2003. Grey water reuse: towards sustainable water management. Desalination 156 (1–3), 181–192. Asano, T., Levine, A.D., 1996. Wastewater reclamation, recycling and reuse: past, present and future. Water Science and Technology 33, 1–14. Dallas, S., Scheffe, B., Ho, G., 2004. Reedbeds for greywater treatment – case study in Santa Elena-Monteverde, Costa Rica. Central America Ecological Engineering 23 (1), 55–61. Dixon, A., Butler, D., Fewkes, A., Robinson, M., 2000. Measurement and modelling of quality changes in stored untreated grey water. Urban Water 1 (4), 293–306. Fittschen, I., Niemczynowicz, J., 1997. Experiences with dry sanitation and grey water treatment in the ecovillage toarp, Sweden. Water Science and Technology 35 (9), 161–170.
water research 44 (2010) 267–277
Frazer-Williams, R., Avery, L., Gideon, W., Jeffrey, P., ShirleySmith, C., Liu, S., Jefferson, B., 2008. Constructed wetlands for urban grey water recycling. The International Journal for Environment and Pollution 33 (1), 93–109. Ghisi, E., Mengotti de Oliveira, S., 2007. Potential for potable water savings by combining the use of rainwater and grey water in houses in southern Brazil. Building and Environment 42 (4), 1731–1742. Gross, A., Shmueli, O., Ronen, Z., Raveh, E., 2007. Recycled vertical flow constructed wetland (RVFCW) – a novel method of recycling grey water for irrigation in small communities and households. Chemosphere 66 (5), 916–923. Jefferson, B., Laine, A., Parsons, S., Stephenson, T., Judd, S., 2000. Technologies for domestic wastewater recycling. Urban Water 1 (4), 285–292. Jeppesen, B., 1996. Domestic grey water reuse: Australia’s challenge for the future. Desalination 106 (1–3), 311–315. Kayaalp, N.M., 1996. Regulatory framework in South Australia and reclaimed water reuse options and possibilities. Desalination 106 (1–3), 316–322. March, J.G., Gual, M., Orozco, F., 2004. Experiences on grey water re-use for toilet flushing in a hotel (Mallorca Island, Spain). Desalination 164 (3), 241–247. Narasimhan, R., Brereton, J., Abbaszadegan, M., Ryu, H., Butterfield, P., Thompson, K., Werth, H., 2005. Characterising
277
Microbial Water Quality in Reclaimed Water Distribution Systems. AWWA Research Foundation, USA. Nghiem, L.D., Oschmann, N., Scha¨fer, A.I., 2006. Fouling in grey water recycling by direct ultrafiltration. Desalination 187 (1–3), 283–290. Nolde, E., 2000. Grey water recycling systems for toilet flushing in multi-storey buildings – over ten years experience in Berlin. Urban Water 1 (4), 275–284. Oschmann, N., Nghiem, L.D., Schafer, A.I., 2005. Fouling mechanisms of submerged ultrafiltration membranes in greywater recycling. Desalination 179 (1–3), 215–223. Prathapar, S.A., Jamrah, A., Ahmed, M., Al Adawi, S., Al Sidairi, S., Al Harassi, A., 2005. Overcoming constraints in treated grey water recycling in Oman. Desalination 186, 177–186. Smith, A., Khow, J., Hills, S., Donn, A., 2000. Water reuse at the UK’s millennium dome. Membrane Technology 2000 (118), 5–8. Ton That, L., 2005. Investigation of Water Consumption and Wastewater Discharge Patterns using Domestic Water Data. Unpublished MSc dissertation of Imperial College London. Walski, T.M., Chase, D.V., Savic, D.A., Grayman, W., Beckwith, S., Koelle, E., 2003. Advanced Water Distribution Modeling and Management. Haestad Press, Waterbury, CT USA. Yang, W., Cicek, N., Ilg, J., 2006. State-of-the-art of membrane bioreactors: worldwide research and commercial applications in North America. Journal of Membrane Science 270 (1–2), 201–211.
water research 44 (2010) 278–286
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Anatoxin-a degradation by Advanced Oxidation Processes: Vacuum-UV at 172 nm, photolysis using medium pressure UV and UV/H2O2 Atefeh Afzal a, Thomas Oppenla¨nder b, James R. Bolton a, Mohamed Gamal El-Din a,* a
Department of Civil and Environmental Engineering, 3-133 Markin/CNRL Natural Resources Engineering Facility, University of Alberta, Edmonton, Alberta, Canada T6G 2W2 b Faculty of Mechanical and Process Engineering, Hochschule Furtwangen University, 78054 Villingen-Schwenningen, Germany
article info
abstract
Article history:
Two Advanced Oxidation Processes, namely vacuum-ultraviolet (VUV) photolysis at
Received 23 July 2009
172 nm and ultraviolet/hydrogen peroxide (UV/H2O2) were investigated for the degrada-
Received in revised form
tion of anatoxin-a in aqueous solutions. Solutions of anatoxin-a-fumarate were treated
7 September 2009
with VUV light at 172 nm with a UV dose of 200 mJ/cm2, where fumaric acid served as
Accepted 8 September 2009
a reference compound for a competition kinetics analysis. The second-order rate
Available online 17 September 2009
constant for the reaction between anatoxin-a and the hydroxyl radical was found to be (5.2 0.3) 109 M1 s 1 and was independent of pH, temperature, and initial concentra-
Keywords:
tion of anatoxin-a. The direct photolysis of anatoxin-a using a medium pressure (MP) UV
Anatoxin-a
lamp was also investigated, in which case a UV dose of 1285 mJ/cm2 was required to
Advanced Oxidation Processes
degrade anatoxin-a by 88% and 50% at concentrations of 0.6 mg/L and 1.8 mg/L of toxin,
Vacuum-UV
respectively. Treatment of anatoxin-a with a low pressure (LP) UV lamp in the presence of
Xenon excimer lamp
30 mg/L of H2O2 was examined, where it was found that more than 70% of toxin could be
Photolysis
degraded at a UV dose of 200 mJ/cm2. The degradation arises from the oxidation of the toxin by hydroxyl radicals. The addition of H2O2 clearly enhanced the degradation of
UV/H2O2
anatoxin-a, up to a concentration of 40 mg/L, after which addition of more H2O2 had little effect on the degradation kinetics of anatoxin-a. The effect of background constituents in the water on the degradation of anatoxin-a was also investigated using natural and synthetically produced model waters. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Blooms of cyanobacteria, also known as blue-green algae, in freshwaters can cause water quality problems, such as deoxygenation, bad taste and odor, color change, and production of cyanobacterial toxins (cyanotoxins) (Wood et al., 2007). Cyanotoxins include a range of chemical compounds, predominantly alkaloids and peptides that have been identified worldwide in
lakes and other raw water supplies (Falconer and Humpage, 2006). These toxins have been responsible for the poisoning and death of livestock (Kotak et al., 1993). Anatoxin-a is a neurotoxin that can be produced by Anabaena flos-aquae in water bodies. It was the first toxin identified from a freshwater cyanobacterium. Chemically, anatoxina (C10H15NO) is a bicyclic secondary amine alkaloid (1-[(1R,6R)9-azabicyclo[4.2.1]non-4-en-5-yl]ethanone) with the molecular
* Corresponding author. Tel.: þ1 780 492 5124; fax: þ1 780 492 8198 E-mail address: [email protected] (M.G. El-Din). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.021
water research 44 (2010) 278–286
279
In the present study, two types of AOPs, were investigated: the vacuum-UV (VUV) AOP using a xenon excimer lamp (excilamp) at 172 nm and the UV/H2O2 AOP in a bench-scale batch treatment of different water matrices spiked with anatoxin-a using a low pressure UV lamp. The direct photolysis of anatoxin-a using a medium pressure UV lamp was also investigated.
Fig. 1 – Chemical structure of anatoxin-a.
weight of 165.23 g mol1 (Carmichael, 1992). It is soluble in water with a pKa of 9.4. The chemical structure of anatoxin-a is presented in Fig. 1. Some studies have been carried out on the treatment of water contaminated with cyanotoxins using different technologies (Newcombe and Nicholson, 2004; Svrcek and Smith, 2004; Hitzfeld et al., 2000; Song et al., 2009); however, microcystins have been the focus of most of these studies, whereas the treatment of water containing anatoxin-a has not been investigated extensively. Conventional water treatment processes, including coagulation and flocculation, can be an efficient method for eliminating cyanobacterial cells, but soluble cyanotoxins, including anatoxin-a, cannot be removed effectively by such processes (Hitzfeld et al., 2000). Sand filtration and activated carbon may also remove the cells, but they are not capable of destroying or removing the soluble toxins (Newcombe and Nicholson, 2004). Oxidation of anatoxin-a by chlorine was shown not to be an appropriate option for the degradation of anatoxin-a. Only 16% of the toxin was removed after 30 min of contact with chlorine with a free chlorine residual of 4.5 mg/L. Chloramines and chlorine dioxide were also observed not to be good oxidants for anatoxin-a degradation (Newcombe and Nicholson, 2004; Rositano et al., 1998; Rodriguez et al., 2007a,b). Permanganate as an oxidizing reagent was investigated for the treatment of anatoxin-a, and it was shown that low permanganate doses (0.5 mg/L) can remove the toxin with a second-order rate constant of 2.3 104 M1 s1 and a half life of 4.8 s (Rodriguez et al., 2007a,b). Ozone appeared to be the most consistent and efficient process to attain complete elimination of anatoxin-a (Rositano et al., 1998, 2001; Rodriguez et al., 2007a,b; Onstad et al., 2007; Hall et al., 2000). While adding saturated ozone water to anatoxin-a at a concentration of 24 mg/L, 92% of the toxin was removed after 60 s with an ozone residual of approximately 0.11 mg/L (Rositano et al., 1998). To date, however, little investigation of the effect of other types of Advanced Oxidation Processes (AOPs), other than ozonation, on the degradation of anatoxin-a has been carried out. Ultraviolet (UV)-based AOPs are characterized by the generation of strong oxidizing species, principally hydroxyl radicals (OH), using combinations of strong oxidizing agents, such as ozone and hydrogen peroxide (H2O2) with UV light. Photolysis of anatoxin-a using a xenon UV lamp (spectral output 330–450 nm) in the presence of TiO2 led to the degradation of anatoxin-a by up to 95% after 30 min of exposure (Hall et al., 2000; Robertson et al., 1999).
2.
Materials and methods
2.1.
Chemicals and sample preparation
Anatoxin-a was obtained in the form of a 50/50 molar mixture of anatoxin-a and fumarate (Tocris Bioscience, Ellisville, MO, USA). Anatoxin-a–fumarate dissociates into anatoxin-a and fumaric acid in water. A nuclear magnetic resonance (1H NMR) analysis, at 400.4 MHz in D2O confirmed that the molar ratio of anatoxin-a and fumaric acid was 1:1. Since the rate constant for the reaction of fumaric acid with the OH radicals is known (Cabelli and Bielski, 1985), this is a suitable compound for carrying out competition kinetics. The 1H NMR spectrum also confirmed that the compound was pure; thus, anatoxin-afumarate was used as received without purification. A stock standard solution (500 mg/L) of anatoxin-a-fumarate was prepared by dissolving 1 mg anatoxin-a–fumarate into 2 mL of ultrapure water; this solution was stored in an amber glass vial at 20 C in the dark. Working solutions were prepared daily and stored at 4 C. Anatoxin-a should be used in a chemical fume hood, with air supplied by independent system. Measures should be taken to avoid inhalation, contact with eyes, skin or clothing, as well as prolonged or repeated exposure. Based on the material safety data sheet (MSDS) of the anatoxin-a–fumarate, this compound should be treated as extremely poisonous. All chemicals and reagents were of analytical reagent grade and were used as received. An H2O2 stock solution (Fisher Scientific Co., Canada, 30% w/w) was used for the experiments, and catalase from bovine liver (Sigma–Aldrich, 2950 units/mg solid) was used to quench any H2O2 residuals in the sample vials prior to analysis. Ultrapure water was supplied from an Elga and Millipore system equipped with an Elix UV lamp. To investigate the effect of background constituents of water, such as organic compounds, on the degradation of anatoxin-a, two other kinds of water were also used in this study: (1) a synthetically produced model water, which simulates the composition and properties of natural water, was prepared by spiking an exact amount of chemicals (Table 1) into ultrapure water (Liu et al., 2002); and (2) a natural water sample was obtained from the North Saskatchewan River after coagulation, flocculation and filtration in EPCOR Rossdale Water Treatment Plant (Edmonton, AB, Canada). The characteristics of the natural water were as follows: Turbidity of 0.1 NTU; COD of 7.5 mg O2/L; alkalinity of 105 mg CaCO3/L; free chlorine residual of less than 5 mg/L; and TOC of 3.94 mg C/L. The pHs of the synthetic and natural waters were 6.9 and 7.3, respectively. The waters were filtered through a 0.45 mm filter then kept at 4 C until use.
280
water research 44 (2010) 278–286
Table 1 – Synthetic water constituents and characteristics.
mg/L
2.2.
Ca2þ
Mg2þ
Naþ
NO 3
Cl
SO2 4
(CO3)TOT
Humic acid
Alginic acid
TOC (mg/L C)
29.1
9.99
36.2
3.0
40.0
55.0
90.0
2.56
5.32
6.63
Vacuum-UV apparatus
2.3.
A dielectric barrier discharge Xe2* excilamp (Model BD-Pbarrier discharge portable, Institute of High Current Electronics, Tomsk, Russia) with the dimensions 26.5 7.7 77 cm (length width height) including an air-cooling fan was used in this study. Photons emitted by the Xe2* excilamp are in the vacuum-UV region with the maximum emission at 172 nm and a full width at half maximum of about 14 nm. The emission spectra of the excilamp are presented in Fig. 2A. The VUV light is absorbed by water to produce OH radicals (reaction (1)) with a penetration depth of less than 100 mm. The photons emitted are also absorbed by oxygen of the air and produce ozone (reactions (2) and (3)) (Oppenla¨nder et al., 2005). Therefore, the Xe2* excilamp cannot be used in a regular collimated beam apparatus.
H2O þ hn (172 nm) / H2O* / H þ OH
(1)
O2 þ hn (172 nm) / O (3P) þ O (1D)
(2)
O2 þ O (3P) / O3
(3)
The UV exposure experiments were carried out using a collimated beam apparatus manufactured by Calgon Carbon Corporation (Pittsburgh, PA, USA). A 1 kW medium pressure (MP) UV lamp (Calgon Carbon, Pittsburgh, PA, USA) was used for the direct photolysis experiments, and a 10 W low pressure (LP) UV lamp (Calgon Carbon, Pittsburgh, PA, USA) was used for the UV/H2O2 advanced oxidation experiments. Fig. 2B shows the emission spectrum of the MP and LP UV lamps. A radiometer (International Light Inc. Model IL 1400A) equipped with a UV detector (International Light Inc. Model SED240) and a neutral density filter (Model QNDS2), calibrated at 5 nm intervals in the range of 200–400 nm, was used to measure the irradiance at the surface of the water sample. The average fluence rate in the solution and the delivered fluence (UV dose) were calculated based on the Bolton and Linden protocol (Bolton and Linden, 2003) and the separate Excel spreadsheets available at www.iuva.org for LP and MP UV lamps for shallow (<2 cm) solutions using parameters, such as solution volume, solution absorbance, water path length, distance from the lamp to the surface of the water, and exposure time.
2.4. A special collimated beam apparatus was designed for the experiments (Wang, 2008) in which the reactor was a 3 mL Suprasil quartz cuvette with a stopper and a small magnetic stir bar. An actinometry method using methanol was employed for calculating the fluence rate (irradiance) of the xenon excilamp at 172 nm (Oppenla¨nder and Schwarzwalder, 2002). The fluence rate of the lamp was calculated to be 0.67 5% mW/cm2 (Wang, 2008). The details of operation of the apparatus and the schematic (Fig. S-1) are presented in supporting information.
UV apparatus
Analytical equipment and methods
High-performance liquid chromatographic (HPLC) analysis was performed on a Shimadzu LC-10AT VP HPLC equipped with a UV–Vis detector. A Phenomenex Gemini 5 mm C18 column (250 mm 4.6 mm) was used as the analytical column. The UV detector was set at a wavelength of 225 nm, and an isocratic elution was employed for 15 min using 96.5% water and 3.5% acetonitrile both contained 0.05% trifluoroacetic acid. At a total flow rate of 0.9 mL/min, the retention times of anatoxin-a and fumaric acid were 9 and 10 min,
Fig. 2 – A: Emission spectrum of the Xe2* excilamp; B: emission spectra of the medium pressure (MP) and low pressure (LP) mercury lamps.
281
water research 44 (2010) 278–286
respectively. Each sample and standard was analyzed twice by the HPLC system. Using this analytical method, the detection limit of anatoxin-a was 33 mg/L.
2.5.
Competition kinetics
A competition kinetics method (Onstein et al., 1999) was used to determine the second-order rate constant of the reaction of anatoxin-a with the OH radicals. The reference compound was fumaric acid with the known rate constant of 6.0 109 M1 s1 (Cabelli and Bielski, 1985). To avoid high experimental errors, the reference compound and the target compound should have comparable reactivities (Huber et al., 2003). In general, the measurement of the OH radical rate constants using competition kinetics is conducted according to the following equation (Onstad et al., 2007; Huber et al., 2003) ½TðnÞ ½RðnÞ kOH ðTÞ ¼ ln ln ½Tð0Þ ½Rð0Þ kOH ðRÞ
(4)
where T is the target compound, R is the reference compound, and kOH (T ) and kOH (R) are the second-order rate constants of the target and reference compounds with the OH radicals, respectively. The different OH radical doses are presented by n. The target compound rate constant can be determined from a plot of ln([T(n)]/[T(0)]) versus ln([R(n)]/[R(0)]), where the factor k.OH(T )/k.OH(R) is the slope of the straight line. The Xe2* excilamp (with a maximum emission at 172 nm) was used to generate the OH radicals, and different exposure times represent different OH radical doses. Hydrogen atoms that are formed according to equation (1) are immediately scavenged by dissolved molecular oxygen with the formation of hydroperoxyl radicals (HO2). All the kinetic experiments were carried out using ultrapure water under an oxygen or nitrogen atmosphere. The effect of pH on the rate constant was investigated by using three pH values (4.5, 7, and 9.5), where the water pH was adjusted by addition of solutions of hydrochloric acid (0.1 N) or sodium hydroxide (0.1 N). The exposures were also carried out at different temperatures (8, 22, and 48 C) to investigate the effect of temperature on the rate constant. The concentration of anatoxin-a–fumarate was varied from 0.01 to 0.05 mM. A control experiment was carried out to check the effect of dissolved oxygen (details are available in the supporting information).
2.6.
Direct photolysis experiments
The direct photolysis experiments were carried out using the MP UV lamp in the collimated beam apparatus. Anatoxina (15 mL) in a glass Petri dish (60 mm diameter) as a reactor was exposed to the MP lamp under a completely mixed condition (using a magnetic stir bar). Two concentrations of the anatoxin-a (0.6 and 1.8 mg/L) were tested to determine the effect of concentration on the photodegradation process. The experiments were carried out at room temperature (20.5 1.0) C and without changing the initial pH of the solutions (approximately 5). The applied UV dose was varied up to 1300 mJ/cm2. Samples were taken at intervals of 20 s for 3 min and also at 5 and 10 min. The samples were analyzed for anatoxin-a using an HPLC system.
2.7.
UV/H2O2 experiments
A bench-scale UV/H2O2 process was carried out using the collimated beam apparatus equipped with an LP UV lamp. Hydrogen peroxide at various concentrations (30, 40 and 60 mg/L) was added to the anatoxin-a solution (0.6 or 1.8 mg/L), and this solution was exposed to the UV lamp. The applied UV doses were varied up to 185 mJ/cm2. Samples were taken at intervals of 5 min for 20 min and also at 40 and 60 min. To minimize the effect of the change of the volume of the solution during the exposure time, samples were taken for HPLC analysis in small amounts (less than 200 mL). Catalase (20 mL) was added into all of the HPLC vials before analyses for the destruction of H2O2 residuals. A control experiment was carried out to check whether or not catalase reacts with either anatoxin-a or fumaric acid, and no effect was found. For the determination of the effect of H2O2 alone, all the experiments were carried out once without UV exposure. Another control experiment was also carried out to check whether or not anatoxin-a and fumaric acid undergo direct photolysis using the LP UV lamp, and no direct photolysis was observed.
3.
Results and discussion
3.1. Rate constant for the reaction of anatoxin-a with OH radicals The rate constant of the reaction between OH radicals and anatoxin-a was determined using competition kinetics with fumaric acid as the reference compound. The rate constants determined at different concentrations, pHs, and temperatures are presented in Table 2. The 172 nm excilamp was used to generate OH radicals in these experiments. Under all experimental conditions, the rate constant kOH for the reaction of anatoxin-a with OH radicals was (5.2 0.3) 109 M1 s1. It is obvious from Table 2 that the rate constant is not dependent on the initial concentration of anatoxin-a.
Table 2 – Rate constants of reaction of anatoxin-a with OH radicals determined using the competition kinetics under O2 saturation (fumaric acid as a reference compound and excilamp as a source of OH radical production) at different concentrations, pHs, and temperatures. pH
Temperature ( C)
Concentration of anatoxin-a (mM)
Rate constant (109 M1 s1)
R2
4.5 4.5 7 7 7 9.5 9.5 7 7 7 7
22 22 22 22 22 22 22 8 8 48 48
0.04 0.01 0.05 0.03 0.01 0.05 0.01 0.03 0.01 0.03 0.01
4.8 5.8 5.3 5.3 5.0 5.2 5.4 5.1 5.1 5.0 5.5
0.99 0.99 0.98 0.98 0.97 0.99 0.96 0.97 0.99 0.98 0.97
282
water research 44 (2010) 278–286
The invariance of the rate constants at different temperatures confirmed that the reaction between the anatoxin-a and OH radicals, as well as that of fumaric acid with OH radicals, is independent of temperature. This means that when the toxin is exposed to OH radicals, no significant activation energy is required for the reaction to proceed, and therefore the reaction can proceed with a high rate independent of the temperature (in the temperature range used here). The results also showed that the reaction of anatoxin-a with OH radicals is virtually pH independent in the pH range 4.5–9.5. As mentioned earlier, a VUV lamp (172 nm) was used for the production of OH radicals. The penetration depth of VUV radiation into water is limited by the absorption coefficient of water being 550 cm1 at 172 nm. Thus under the condition of total absorbance of VUV radiation by water, the penetration depth at 172 nm is about 36 mm (Heit et al., 1998). The direct photolysis of anatoxin-a at low concentrations of 0.01– 0.05 mM does not compete with water photolysis under this condition, and the degradation reaction arises solely from OH radical attack (under oxygen atmosphere). Water is photolyzed by the VUV irradiation to produce hydrogen atoms (H) and hydroxyl radicals (reaction (1)). This reaction proceeds in liquid water with the quantum yield (F) of 0.42 (Heit et al., 1998). If the solution is saturated with oxygen (O2), hydrogen atoms are scavenged immediately to form hydroperoxyl radicals (HO2) (reaction (5)). Hydrogen atoms can also react with hydroperoxyl radicals to form H2O2. Furthermore, hydrogen atoms can react with the anionic form of fumaric acid (FA) and anatoxin-a in the water. The principal possible reactions of the hydrogen atoms are displayed in reactions (5)–(8). H þ O2 / HO2 k ¼ 2.0 1010 M1 s1 (Heit et al., 1998)
(5)
H þ HO2 / H2O2 k ¼ 2.0 1010 M1 s1 (Feng et al., 1970) (6) H þ FA / products k ¼ 7.0 109 M1 s1 (Hayon and Simic, 1973) (7)
H þ anatoxin-a / products k ¼ ?
(8)
On the other hand, hydroxyl radicals can react with OH and HO2 to produce H2O2 and water plus O2, respectively (reactions (9) and (10)). Hydroxyl radicals can react with fumaric acid and anatoxin-a to oxidize them. The rate constant of reaction between FA and OH is known, so the rate constant of the reaction between anatoxin-a and OH was calculated (reactions (11) and (12)). OH þ OH / H2O2 k ¼ 5.5 109 M1 s1 (Heit et al., 1998) (9) OH þ HO2 / H2O þ O2 k ¼ 1.0 1010 M1 s1 (Feng et al., 1970) (10)
OH þ FA / product k ¼ 6.0 109 M1 s1 (Cabelli and Bielski, 1985) (11)
OH þ anatoxin-a / products k ¼ 5.2 109 M1 s1
(12)
Scavenging of the hydrogen atoms by dissolved molecular O2 is very fast and also the reaction of H with HO2 is almost diffusion controlled. Therefore hydrogen atoms competing with OH radicals for anatoxin-a under the reaction conditions with dissolved molecular oxygen is not a significant process. A control experiment was carried out in which the dissolved molecular oxygen in the water was replaced by purging nitrogen into the solution. This O2-free condition can give an estimate of the rate constant of reaction between anatoxin-a and hydrogen atoms. If the rate of reaction of FA with H is: rH ¼ kH [FA] [H]
(13)
and the rate of the reaction of FA with OH is: rOH ¼ kOH [FA] [OH]
(14)
then roverall is: roverall ¼ kH [FA] [H] þ kOH [FA] [OH]
(15)
and [H] ¼ [OH] during 172 nm irradiation (reaction (1)), then
roverall ¼ (kH þ kOH) [FA] [OH]
(16)
Based on the calculations of the rate constants for the solutions with and without dissolved oxygen, the overall rate constant of the reaction of anatoxin-a with OH radicals and hydrogen atoms was determined to be 7.7 109 M1 s1 and therefore the rate constant of the reaction between anatoxina and hydrogen atoms was calculated to be 2.5 109 M1 s1.
3.2. Anatoxin-a degradation in synthetic and natural water using the VUV AOP Fig. 3 shows the degradation rate of anatoxin-a in different water matrices. DI, EP, and SY represent ultrapure water, natural water, and synthetic water, respectively. The figure illustrates the natural logarithm of the toxin concentration versus UV dose. The UV dose was calculated by multiplication of the fluence rate of the 172 nm excimer lamp (0.67 5.4% mW/cm2) and the exposure time in seconds. As seen from Fig. 3, the degradation in all water matrices followed first-order kinetics with an R2 value of >0.96 in all cases. Results showed that 0.6 mg/L of anatoxin-a in pure water was degraded to below the detection limit (<30 mg/L) after 150 s exposure to the VUV lamp, which is equivalent to a UV dose of 96 mJ/cm2. At higher concentrations, the rate of degradation did not change significantly, such that the firstorder rate constant was calculated to be 0.036 and 0.029 s1 for 1.8 and 0.6 mg/L of anatoxin-a, respectively. Results also showed that 0.6 mg/L anatoxin-a was degraded by 70% and 85% after 300 s exposure to the VUV lamp in natural water and synthetic water, respectively, corresponding to a UV dose of 193 mJ/cm2.
283
water research 44 (2010) 278–286
3.50 -2
k' = 3.6 × 10
-2
k' = 2.9 × 10
ln([atx-a] 0/[atx-a])
3.00 2.50 2.00
-3
k' = 9.0 × 10
-3
k' = 8.0 × 10
1.50 -3
k' = 7.0 × 10 1.00 -3
k' = 6.0 × 10
0.50 0.00
0
50
100
150
200
250
VUV dose (mJ/cm2) 0.6 mg/L atx-a - DI water
1.8 mg/L atx-a - EP water
1.8 mg/L atx-a - DI water
0.6 mg/L atx-a - SY water
0.6 mg/L atx-a - EP water
1.8 mg/L atx-a - SY water
Fig. 3 – Degradation of anatoxin-a in different water matrices using VUV AOP. DI: ultrapure water, EP: natural water, and SY: synthetically produced model water. The fluence-based rate constants are in cm2/mJ.
The degradation rate constants decreased dramatically when using natural and synthetic waters, compared to those obtained using DI water; however, there is no significant difference between natural water and synthetic water. This demonstrates that the background organic compounds in the water matrix can decrease the efficiency of the AOP towards anatoxin-a degradation using VUV photolysis at 172 nm. This arises from the competition between other constituents, primarily organic matter, in the water and anatoxin-a for reaction with OH radicals.
3.3.
Direct photolysis of anatoxin-a with a MP UV lamp
Because anatoxin-a does not absorb UV light significantly at 254 nm, the degradation was expected to be minimal for an LP UV lamp emitting at 254 nm; however, the use of a polychromatic MP UV lamp (emission throughout the 200–300 nm range – Fig. 2B) can be effective for degrading the anatoxina because the maximum absorbance of anatoxin-a is at 227 nm (James et al., 1998). The absorption spectrum of anatoxin-a and fumaric acid is presented in the supporting information (Fig. S-2-A, B). A control experiment was carried out using the LP UV lamp, which demonstrated that anatoxina (1.8 mg/L) and fumaric acid do not undergo photolysis using this UV lamp. The results of direct photolysis using the MP lamp showed that a fluence of 1285 mJ/cm2 was capable of degrading anatoxin-a by 88% at a concentration of 0.6 mg/L. The same UV dose could remove 50% of anatoxin-a at an initial concentration of 1.8 mg/L.
3.4.
Quantum yield
For direct photolysis to eliminate a compound, photons emitted from the UV lamp need to be absorbed and the
absorbed photons must be capable of transforming or degrading the compound. The quantum yield is a factor that can describe the efficiency of the absorbed UV light and is defined as the number of moles of compound removed divided by the number of moles (einsteins) of photons absorbed. The quantum yields for the direct photolysis of anatoxina with a polychromatic MP UV lamp at different concentrations of toxin were determined using the fluence-based rate constant method (Bolton and Stefan, 2002; Stefan and Bolton, 2005). For calculating the quantum yield, equation (17) (appropriate for broadband systems) was used (Stefan and Bolton, 2005). X ½atx-a0 Ni;l Flatx-a 3latx-a lnð10Þ ¼ F ln ½atx-aF 10 Ul l
(17)
where [atx-a]0 and [atx-a]F are the concentrations of anatoxina in the original solution and after applying fluence (UV dose) of 0 and F (J m2), respectively. Flatx-a is the quantum yield for the direct UV photolysis of anatoxin-a at wavelength l. 3latx-a (M1 cm1) and Ul (J einstein1) are the molar absorption coefficient of anatoxin-a and the energy per einstein, respectively, at wavelength l. Ni,l is the relative photon flow emitted by the light source at the wavelength l (einstein s1). The quantum yield can be calculated using the fluence-based firstorder rate constant k10 (m2 J1). The expression is given by equation (18). k01 ¼
X Ni;l Fl
l atx-a 3atx-a
l
lnð10Þ
(18)
10 Ul
Since the average fluence rate was calculated based on the Excel spreadsheets explained previously, parameters such as geometrical parameters, solution volume, solution absorbance, and radiometer sensor response in each wavelength band were considered in the fluence calculation. Thus, factors such as Petri factor and water factor are not present in the equations. Fig. S-3 illustrates the fluence-based first-order degradation of anatoxin-a in ultrapure water using the MP UV lamp. The quantum yields of the direct photolysis of anatoxin-a at concentration of 0.6 and 1.8 mg/L were calculated to be 0.15 and 0.05, respectively, over the wavelength range of 200– 300 nm. The results indicated that the rate of degradation of anatoxin-a is higher when the concentration is lower. The reason for this observation might be due to the formation of higher concentrations of intermediates at higher
Table 3 – Rate constants of direct photolysis of anatoxina using MP UV lamp in different water matrices DI: ultrapure water, EP: Natural water, and SY: synthetically produced model water. Water type DI water DI water EP water EP water SY water SY water
Anatoxin-a concentration (mg/L)
Rate constant (103 cm2 J1)
0.6 1.8 0.6 1.8 0.6 1.8
1.00 0.10 7.00 4.00 6.00 6.00
284
water research 44 (2010) 278–286
4.00
1.40
3.50
1.20
3.00
ln([atx-a] 0/[atx-a])
ln([atx-a] 0 /[atx-a])
-3
2.50 2.00 1.50 1.00
k' = 5.0 × 10 2 R = 0.99
SY water
-3
k' = 3.2 × 10 2
DI water EP water
1.00
R = 0.99
0.80 0.60
-3
k' = 2.5 × 10
0.40
2
R = 0.98
0.20
0.50
0.00
0
0.00 0
200
400
600
800
1000
1200
50
100
150
250
200
300
350
2
140 0
UV dose (mJ/cm )
2
UV dose (mJ/cm ) 0.6 mg/L atx-a - DI water
1.8 mg/L atx-a - EP water
1.8 mg/L atx-a - DI water
0.6 mg/L atx-a - SY water
0.6 mg/L atx-a - EP water
1.8 mg/L atx-a - SY water
Fig. 4 – Direct photolysis of anatoxin-a in different water matrices using the MP UV lamp; DI: ultrapure water, EP: natural water, and SY: synthetically produced model water.
concentrations of anatoxin-a. These intermediates can absorb UV light in the range of 200–300 nm, and therefore the fraction of light absorbed by the anatoxin-a itself decreases, resulting in lower quantum yields and lower degradation rates. The HPLC chromatograms confirmed the formation of by-products (data are not shown here). The fluence-based rate constants of the direct photolysis of anatoxin-a using the MP UV lamp in different water matrices are presented in Table 3. A comparison of the degradation rates among the three types of water is shown in Fig. 4.
Fig. 5 – Degradation of anatoxin-a using UV/H2O2 in different water matrices using the LP UV lamp; DI: ultrapure water, EP: Natural water, and SY: synthetically produced model water. The rate constants are in cm2/mJ.
higher concentrations of H2O2 on the degradation rate of anatoxin-a. The rate of degradation of anatoxin-a increased with increasing H2O2 concentration from 30 to 40 mg/L. The reason is that at higher concentration of H2O2, more UV is absorbed by the H2O2 and the rate of generation of OH radicals is higher, and this leads to a higher anatoxin-a degradation rate. Similar results were observed using natural and synthetic waters (the results are not shown here). At the higher concentration of H2O2 (60 mg/L), the rate of degradation did not change significantly as was expected by adding 20 mg/L more of H2O2. It has been shown in previous works that at high concentrations of H2O2, hydrogen peroxide itself becomes a competitor for OH radicals. A concentration of 60 mg/L of H2O2 is not considered as a large scavenger, but it can be the starting point of scavenging of OH radicals by H2O2.
3.5. UV/H2O2 degradation of anatoxin-a in different water matrices by an LP lamp
4. The UV/H2O2 AOP treatment of anatoxin-a in pure water showed that when 30 mg/L of H2O2 is added to 0.6 mg/L of anatoxin-a, a 45 min exposure to the LP UV lamp resulted in >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 (<24 h) of D. magna. The experiment was carried out in glass jars with 100 ml artificial RT-medium (P-free) in which Phoslock (batch 1 material) was tested at 0, 5, 50, 100, 500 and 5000 mg l1. These concentrations are centered on 100 mg l1. Based on a recommended Phoslock dose of 100 mg per mg of filterable reactive phosphorous FRP at neutral pH (http://www.phoslock.com.au/faqs.php#20),
3.
Results
3.1.
Phoslock leachate experiment
Aluminium, lanthanum and zinc were detected in the filtrates from the Phoslock suspensions, as were small amounts of copper (Table 1). Mercury and cadmium were not detected and in one batch a trace of lead was found (Table 1). The oldest batch 1 (obtained in August 2006) seemed to release more metals than the newer batch 2 (obtained in April 2008). For example, in the first batch 69–360 mg Al l1 was liberated, while in the second batch this was 15–19 mg Al l1. A similar pattern was observed for La: 12–43 mg l1 was measured in filtrate of suspension from the first batch, while La was considerable lower in filtrates from the second batch: 2.5–4.1 mg l1 (Table 1).
3.2. Effect of lanthanum on life-history traits of D. magna Measured lanthanum concentrations deviated from the nominal concentrations for the 33 and the 330 mg La l1 treatments. Measured concentrations in P-free water were 0, 12.1, 94.1, 98.9 and 1001 mg La l1 for the controls and nominal concentrations of 33, 100, 330 and 1000 mg La l1, respectively. In the absence of phosphate, somatic growth of D. magna expressed as the increase of the body length over time was influenced marginally by lanthanum (Fig. 1A). A repeated
312
water research 44 (2010) 309–319
Table 1 – Metals (in mg lL1, ±1 SD, N [ 3) measured in 0.45 mm filtrates from 5 g lL1 suspension of two different batches of Phoslock.
Control Batch 1 Batch 2
Al
Cd
Cu
Hg
0.7 (1.3) 218.1 (145.5) 16.2 (2.6)
0
0.17 (0.09)
<1
0
0
0
0.31 (0.09)
<1
22.9 (17.5)
0.02 (0.02)
30.3 (18.0)
0
0.21 (0.10)
<1
3.4 (0.8)
0 (0.01)
26.2 (15.0)
measures ANOVA indicated significant differences in body length over time (F10,420 ¼ 2728; P < 0.001) and a significant lanthanum treatment effect (F4,42 ¼ 6.47; P < 0.001). The differences among treatments were, however, small and a Tukey post-hoc comparison test revealed two homogenous groups: 1) 0, 33, 100, 330 mg La l1 and 2) 0, 33 and 1000 mg La l1. Animals in the 100, 330 mg La l1 treatments were significantly larger than those in the 1000 mg La l1 treatment, but none of the treatments differed from controls (Fig. 1A). By contrast, in P-containing medium a pronounced effect of lanthanum on the somatic growth of D. magna was observed (Fig. 1B). The repeated measures ANOVA indicated a significant
La
Pb
Zn 9.9 (2.1)
lanthanum treatment effect (F4,39 ¼ 116.1; P < 0.001) and Tukey’s test revealed four homogenous groups that were in order from the largest to the smallest animals: 1) 0 and 33 mg La l1, 2) 100 mg La l1, 3) 330 mg La l1 and 4) 1000 mg La l1. Comparisons of somatic growth of animals in P-free or P-containing medium at similar La exposures revealed that D. magna in 0 and 33 mg La l1 were equally sized (Table 2). However, for 100, 330 and 1000 mg La l1 exposures D. magna were significantly smaller in medium with P than those reared in P-free medium (Table 2). In P-free medium, age at first reproduction was similar in controls and treatments (F4,42 ¼ 0.56; P ¼ 0.694), which was also the case for size at first reproduction (F4,42 ¼ 2.40; P ¼ 0.065). Survival was 90% in controls and treatments (Table 3). In P-containing medium, age at first reproduction was similar (F4,42 ¼ 2.45; P ¼ 0.061), but size at first reproduction appeared significantly different (F4,42 ¼ 49.5; P < 0.001) with animals being the biggest in controls and the smallest in the highest dosage of lanthanum (Table 3). Survival was 80% in controls and treatments (Table 3). In P-free medium, the number of offspring in the first brood (F4,42 ¼ 1.44; P ¼ 0.240) and second brood (F4,42 ¼ 1.02; P ¼ 0.410) was similar among controls and treatments (Fig. 2A). In the third brood, however, significantly less offspring were produced in the 1000 mg La l1 treatment (F4,42 ¼ 15.0; P < 0.001). Also the total number of offspring produced per female was significantly (F4,42 ¼ 6.16; P < 0.001) lower in the 1000 mg La l1 treatment (Fig. 2A). When grown in P-containing medium and in the highest La treatment, D. magna produced significantly less offspring in all three broods (1st brood F4,42 ¼ 17.6; P < 0.001; 2nd brood F4,41 ¼ 41.3; P < 0.001; 3rd brood F4,41 ¼ 34.6; P < 0.001; Fig. 2B). Intrinsic rates of population increase were excellent for animals reared in P-free medium in all treatments and for
Table 2 – Average body-size of the animals during the experimental period (±1 SD) in P-free medium and P-containing where they were exposed to different concentrations of lanthanum (La), including F and P-values of the between-subject effects (presence/ absence of P in medium) of repeated measures ANOVAs. La (mg l1) Fig. 1 – Body length (mm ± 1 SD) of Daphnia magna exposed to different concentrations of lanthanum (nominal: 0, 33, 100, 330 and 1000 mg lL1) in P-free medium (upper panel A) and P-containing medium (phosphate 330 mg lL1, lower panel B) during a 14 d experimental period.
0 33 100 330 1000
Size (mm) P-free 2.94 2.95 3.03 3.00 2.87
(0.11) (0.04) (0.09) (0.04) (0.06)
Size (mm) P-containing 2.97 2.97 2.89 2.81 2.47
(0.03) (0.03) (0.04) (0.08) (0.07)
F
P
0.64 2.15 19.0 37.9 147.2
0.434 0.160 <0.001 <0.001 <0.001
313
water research 44 (2010) 309–319
Table 3 – Life-history characteristics, age at first reproduction (AFR, d), size at first reproduction (SFR, mm), survival (%) and population growth (dL1) of Daphnia magna exposed to different concentrations of lanthanum in P-free and P-containing medium including controls. P-free medium Treatment Control 33 mg La l1 100 mg La l1 330 mg La l1 1000 mg La l1
AFR (d) 7.9 (0.3)A 7.9 (0.3)A 7.6 (0.5)A 7.8 (0.4)A 7.8 (0.5)A
SFR (mm) 3.24 (0.08)A 3.27 (0.10)A 3.30 (0.07)A 3.29 (0.07)A 3.19 (0.12)A
Survival (%) 90 100 100 90 90
Population growth (dL1) 0.368 (0.026) 0.380 (0.031) 0.383 (0.039) 0.374 (0.032) 0.360 (0.027)
P-containing medium Treatment Control 33 mg La l1 100 mg La l1 330 mg La l1 1000 mg La l1
AFR (d) 7.8 (0.4)A 7.9 (0.3)A 8.0 (0.0)A 8.0 (0.0)A 8.2 (0.4)A
SFR (mm) 3.28 (0.09)AB 3.25 (0.07)BC 3.14 (0.08)CD 3.13 (0.12)D 2.71 (0.07)E
Survival (%) 100 100 80 90 80
Population growth (dL1) 0.376 (0.029) 0.381 (0.036) 0.359 (0.027) 0.348 (0.029) 0.288 (0.035)
Different symbols (A, .., E) indicate significant differences at the 95% level (Tukey’s post-hoc comparison test). Values for AFR and SFR are means 1 SD, population growth rates are means 1 SE.
animals raised in P-containing medium in controls and 33 mg La l1 treatments (Table 3). However, in P-containing medium compared to P-free medium intrinsic rates of population increase were 6% and 7% lower in 100 and 330 mg La l1,
respectively, while in the highest dosage of 1000 mg La l1 it was 20% lower (Table 3). A significantly reduced reproduction in the P-containing medium caused the lower intrinsic rates of population increase at the highest dose of 1000 mg La l1.
3.3.
Effect of lanthanum on S. obliquus
Because precipitation was observed in medium containing P and lanthanum, the effect of this on the food availability was examined in a short assay (Fig. 3). A one-way ANOVA indicated that biovolumes after two hours were similar in all treatments (F3,8 ¼ 1.23; P ¼ 0.362). However, after 18 and 25 h significant differences were observed (F3,8 ¼ 371.4; P < 0.001 and F3,8 ¼ 371.4; P < 0.001, respectively), and where the biovolumes were similar in the La-free P-containing medium and in the P-free treatments (both with and without La), biovolumes were significantly lower in the La treatment in P-containing medium (Fig. 3).
3.4.
Fig. 2 – Number of neonates per female (±1 SD) Daphnia magna exposed to different concentrations of lanthanum (nominal: 0, 33, 100, 330 and 1000 mg lL1) in P-free medium (upper panel A) and P-containing medium (phosphate 330 mg lL1, lower panel B) for the first three consecutive broods.
Effect of Phoslock on growth of Daphnia
The juveniles at the start of the experiment had a body length of 0.91 (0.05) mm and a body weight of 12.3 (1.1) mg (N ¼ 15). All animals survived during the experimental period in controls and treatments up to 100 mg Phoslock l1; in 500 mg Phoslock l1 89% of the animals survived and in the highest dosage (5000 mg Phoslock l1) all animals had died. After 5 days, animals reached 2.59 (0.25) mm in controls, were slightly smaller in the 500 mg Phoslock l1 treatment (2.28 0.17) mm, but had not grown in the highest dosage of 5000 mg Phoslock l1 (Fig. 4). Here animals were 0.91 (0.13) mm. Growth rates, based on the increase in body length over time, were similar in controls and Phoslock concentrations up to 500 mg l1, but were significantly lower (F5,12 ¼ 128.7; P < 0.001) in the 5000 mg Phoslock l1 treatment (Fig. 4). The animals reached a body weight of 144 (12) mg in controls. They were lighter in the 100 mg Phoslock l1 treatment (126 4 mg), significantly (F5,12 ¼ 189.4; P < 0.001) lighter in the 500 mg Phoslock l1 treatment (87 8 mg) and with
314
water research 44 (2010) 309–319
Fig. 3 – Biovolume of the green alga Scenedesmus obliquus (in mm3 mlL1), used as food for Daphnia, in experimental tubes initially and after 2, 18 and 25 h in P-free and P-containing medium in the absence (No La) and presence of lanthanum (1000 mg lL1). Error bars indicate one standard deviation (N [ 3).
a body weight of 5 2 mg they had lost weight in the 5000 mg Phoslock l1 treatment (Fig. 4). Estimated EC50 values were 871 and 1557 mg Phoslock l1 for weight and length based growth rates, respectively.
4.
Discussion
4.1.
Phoslock leachate
Phoslock is applied to the degraded waters as a solid/water slurry through a spray manifold. Settling through the water column it will permanently bind orthophosphates and settle on the sediment. Phoslock will efficiently intercept P released from the sediments (Douglas et al., 1999; Robb et al., 2003; Akhurst et al., 2004; Ross et al., 2008). The main carrier of the product is bentonite clay, which may release metals during the application (Guy and Chakrabarti, 1976). Indeed, some metals were detected in the Phoslock leachate. A small amount of copper was found. Based on a worst-case scenario, dosage of 0.25 g Phoslock l1 (Haghseresht, 2006) would equal 0–0.8% of the Maximum Allowable Concentration (MTR) of 1.5 mg l1 for dissolved copper (RIVM, 2008). For Zinc it fluctuates between 0.5 and 22% of the MTR (9.4 mg l1) and for lead between 0 and 0.01% of the MTR (11 mg l1). For aluminium an ad hoc MTR of 45 mg l1 exists (Van de Plassche, 2002) meaning that leakage during a worst-case dosing equals 1.5–40% of this MTR. However, the MTR is based on soluble aluminium, which is strongly pH dependent; between pH 5.2 and 8.8 the solid Al(OH)3 predominates (Martel and Motekaitis, 1989; Driscoll and Letterman, 1995). Hence, no environmental risks are expected from metals leaking from Phoslock during an application. During the preparation of Phoslock, Rare Earth lanthanum ions are exchanged with surface adsorbed exchangeable cations in the bentonite. However, not all
Fig. 4 – Juvenile growth rates of Daphnia magna, based on the increase in weight (black bars) and on the increase in body length (white bars), exposed for 5 days to different concentrations of Phoslock (0–5000 mg lL1). Error bars indicate one standard deviation (N [ 3). Different symbols indicate significant differences (a, b for length; A, B, C for weight).
lanthanum is locked permanently into the clay and some may re-equilibrate on dispersion in the water. Here about 0.001% of the lanthanum in the clay was released from it, which is in the same order of magnitude as the 0.02% reported by NICNAS (2001). Although the amount of free lanthanum released from the clay is only a very small fraction of the total lanthanum contained in the clay, the potential leakage is of importance in The Netherlands, because the maximum permissible concentration of lanthanum in Dutch surface water is 10.1 mg l1 (Sneller et al., 2000). A worst-case scenario dosage of 0.25 g Phoslock l1 (Haghseresht, 2006), yields dissolved lanthanum concentrations between 0.13 and 2.13 mg l1 based on the results of this study. Lanthanum ion concentrations measured after various Phoslock applications in Australia were 12 mg l1 in two waters (Edith Cowan University, after 24 h; and Queanbeyan STP Lagoon, after 48 h), but dropped rapidly to below 1 mg l1 (Haghseresht, 2006). In four other Australian waters, the La3þ concentrations were below 10 mg l1 (Flapper, 2003; Haghseresht, 2006). Although this implies a short-term, minor exceedance of the Dutch lanthanum standard, total lanthanum concentrations in the water might reach as high as 400 mg l1 straight after an application (Flapper, 2003; McIntosh, 2007; Anonymous, 2008a,b). Undoubtedly, a great portion of this lanthanum will be imbedded in suspended clay particles, or biologically unavailable through rhabdophane formation, but these values are exceeding the NOEC of 100 mg La l1 found in a 21 d Daphnia reproduction test that forms the basis for the Dutch standard (Sneller et al., 2000).
4.2. Effect of lanthanum on life-history traits of D. magna The outcomes of this study are not consistent with our hypothesis that in the presence of phosphate the toxicity of
315
water research 44 (2010) 309–319
lanthanum to Daphnia would be mitigated through the formation of the insoluble mineral rhabdophane as the underlying hypothesis that there is observable statistically significant La toxicity was not supported. In fact, in P-free medium no evidence for lanthanum toxicity to Daphnia was found. When grown in P-free medium, somatic growth in all lanthanum treatments was similar to the controls; age and size at first reproduction were similar in controls and treatments, and survival was 90% in controls and treatments. Only in the highest dosage of 1000 mg La l1 animals produced less offspring in the third brood and also the overall reproduction was significantly lower in these treatments. The measured lanthanum concentration in the (nominal concentration) 330 mg l1 treatments was nearly 100 mg l1. Based on reproduction this implies a NOEC of 100 mg l1, which is similar to the result found in a 21 d Daphnia reproduction test (Sneller et al., 2000). It should be noted, however, that the reduction in reproduction was only 9% compared to controls and that it hardly influenced the intrinsic rate of population increase, which was only 2.2% lower in the highest La treatments. The per capita rate of increase has often been used as a measure of Daphnia fitness (e.g. McCauley et al., 1990; Lu¨rling, 2003) and represents a relevant measure of ecological consequences of stressors (Forbes and Calow, 1999). Here, population growth rates were 0.36 d1 that reflect excellent growth of D. magna and do not support the view that lanthanum is toxic to particularly Daphnia in both acute and chronic tests (Akhurst et al., 2004). A possible explanation for the absence of La toxicity in our experiment is that the pH of around 7.6 in our artificial medium is such that only a fraction of La might be present as the free La3þ ion, while most La will have formed inorganic 2 complexes with anions such as HCO 3 , SO4 , Cl , NO3 (Lee and Byrne, 1992; Moermond et al., 2001; Kamble et al., 2007; Tian et al., 2009) that are present in the RT-medium (Tollrian, 1993). In P-containing medium, significant impacts were observed from the presence of lanthanum. With increasing amounts of lanthanum, animals grew less; matured at a smaller size, showed a tendency to mature slightly later and reproduced less. Consequently, population growth rates were reduced. Precipitation was observed in the P-containing
medium dosed with 100 mg La l1. Unexpectedly, this precipitation also drastically reduced the amount of food algae. For example, after 25 h in the highest La dosage sedimentation reduced the amount of suspended food particles to 7% of the initial concentration, while in the absence of La food amounts remained at 93% of the initial value. This reduction in food availability at elevated La concentration in the P-containing medium provides a very likely explanation for the reduced growth and reproduction of the daphnids (e.g. Lu¨rling, 2003). A direct toxic effect of La on the food alga is not expected as La can promote growth of phytoplankton (Yang et al., 1999; Zhou et al., 2004), or is toxic only at high concentrations (Jin et al., 2009). Because precipitation was only observed in the treatments with P in the medium, the most likely explanation is lanthanum forming complexes with the and HPO2 main phosphate oxyanions (H2PO 4 4 ). These complexes formed larger aggregates (Herna´ndez and Martin, 2007; Lucas et al., 2004) which on attaching to the algae were subjected to enhanced sedimentation due to the increased ballast. The literature reports a huge variation in the lanthanum toxicity data. When we exposed D. magna to lanthanum in RTmedium a LC50 of 14 mg l1 was found (unpublished data), which is lower than the 160 mg l1 reported by USEPA (2002). Acute assays have revealed EC50 values between 40 and 103 000 mg l1 (Table 4), where the lowest values were obtained running assays in deionised water or tap water thereby including additional stress to the animals as indicated by high mortality in controls (Barry and Meehan, 2000; Stauber, 2000; NICNAS, 2001). The use of different media, container size, modified protocols and organisms makes it rather difficult to delineate causality. However, also within those studies some inconsistencies appeared. For example, a 7 d Ceriodaphnia reproduction assay yielded a 7 d LC50 value of 824 mg La l1, while the corresponding acute 48 h toxicity assay gave an LC50 of only 80 mg La l1 (Stauber, 2000). This LC50 value deviated notably from the 48 h EC50 of 5000 mg La l1 for Ceriodaphnia immobilization in another study (Stauber and Binet, 2000). In contrast, 7 d reproduction EC50 values were similar in those studies, i.e. 281 and 430 mg La l1 (Stauber, 2000; Stauber and Binet, 2000).
Table 4 – Summary of acute toxicity assays (following OECD protocol 202) with cladocerans (Dc [ Daphnia carinata, Dm [ Daphnia magna, Cd [ Ceriodaphnia dubia) exposed for 48 h to lanthanum or leachate from Phoslock, including hardness of the water used (in mg CaCO3 lL1). Water type
Hardness
Species
EC50 (mg La l1)
Reference
Tap water Art. medium ASTM Art. medium Art. medium Art. medium Art. medium Milli-Q Art. medium
22 98 160 40–48 210 – 40–48 <10 40–50
Dc Dc Dc Cd Dm Dm Cd Cd Dm
43 49 1180 5000a 23000 103 000 80b 40b >50c g Phoslock l1
Barry and Meehan (2000) Barry and Meehan (2000) Barry and Meehan (2000) NICNAS (2001) Sneller et al. (2000) Anonymous (2008b) Stauber (2000); NICNAS (2001) NICNAS (2001) Martin and Hickey (2004)
a Total lanthanum. b Animals exposed to leachate; EC50 corresponding to La in leachate. c No lanthanum concentration given. Art. medium ¼ artificial medium.
316
water research 44 (2010) 309–319
There are some indications that Ceriodaphnia might be more sensitive than D. magna (Table 4). Although several studies have employed a so-called Toxicity Characteristic Leach Protocol (Stauber, 2000; NICNAS, 2001; Martin and Hickey, 2004), there are differences in the procedure for preparation of the filtered fraction to which experimental organisms have been exposed: Ceriodaphnia to 0.45 mm filtered leachate (Stauber, 2000) and D. magna to 40 mm filtered leachate (Martin and Hickey, 2004). The results were extremely different with Daphnia being unaffected and Ceriodaphnia influenced in general by leachate of 25% (Stauber, 2000; NICNAS, 2001; Martin and Hickey, 2004). The negative effects on Ceriodaphnia could have been even stronger as Ceriodaphnia is much more sensitive to fine suspended clay particles than Daphnia (Kirk and Gilbert, 1990). Hence, different sensitivities might explain some of the variable results in the lanthanum toxicity to water fleas with D. magna being less sensitive. Another explanation for the large differences in results between studies might be caused by factors affecting the availability of lanthanum. Glass containers might absorb up to 25% of the total La to the glass (Weltje et al., 2002). However, employing daily renewal of the medium probably makes this absorption loss much less significant. In many studies, lanthanum rapidly precipitated with oxyanions, such as orthophosphate, due to its extreme low solubility product Ksp ¼ 1024.7 to 1025.7 mol2 l2 at 25 C and infinite dilution (Johannesson and Lyons, 1994; Liu and Byrne, 1997). Also in the current study, in the presence of phosphate clear precipitation was observed. Finally, it has been shown that organic ligands, which can form Rare Earth Elements – Organic complex species led to a great reduction of the REE bioconcentration in algae (Sun et al., 1997). In natural waters, humic substances are typically non-specific complexing ligands that maybe found in concentrations from less than 1 mg l1 to hundreds of mg l1 (Steinberg et al., 2006) that also form complexes with lanthanum (Moermond et al., 2001). The artificial RT-medium used in this study contains 5 mg Na2– EDTA l1 (Tollrian, 1993), which will influence the bioavailability as the EDTA is supposed to compete with cellmembrane ligands (Sun et al., 1997). Moreover, the La3þ-ion is a substitute or antagonist of Ca2þ in biological systems and it has been postulated that La3þ can replace Ca2þ at well-defined tissue loci or sites (Weiss, 1974). Inasmuch as Daphnia actively absorbs Ca2þ from the water after moulting for hardening the carapace (Porcella et al., 1969), La binding to the carapace might be another way of reducing the concentrations in the water and might explain the high concentrations of La in Daphnia after exposure (Yang et al. 1999).
4.3.
Effect of Phoslock on growth of Daphnia
Growth of juvenile Daphnia was not affected up to 100 mg Phoslock l1. Nevertheless, at this concentration animals showed a tendency of being lighter (7–13%) than their conspecifics in controls and treatments up to 50 mg Phoslock l1. Probably feeding inhibition by suspended clay particles had reduced the feeding rate of the animals (Kirk, 1991). The NOEC concentration of 100 mg Phoslock l1 is close to the average application dosage. Data on four different
applications yield an estimated average application dose of 84 (24) mg Phoslock l1 (Anonymous, 2008a; Ross and Cloete, 2006). Application usually takes places through a spray manifold where the slurry is brought into the upper water layer (Robb et al., 2003). Consequently, Phoslock concentrations in these moments in the top layers will be much higher than the ones calculated over the whole water body. However, rapid sedimentation and dispersal of the clay particles is expected lowering the turbidity to normal values within a few days (Haghseresht, 2005a; Ross and Cloete, 2006). In this study, a juvenile growth assay was employed without refreshment of water to avoid continuous exposure of the animals to high concentrations of suspended material and without adding additional food to mimic the presumed phytoplankton growth reduction through P-depletion. The relationship between juvenile growth rates and population growth rates, either calculated from individual biomass or length at successive times, is highly significant in a wide range of natural and stress conditions (Lampert and Trubetskova, 1996; Hanazato, 1998). The Phoslock application is developed for restoring degraded or eutrophied systems by immobilizing phosphate and thereby reducing blooms of nuisance cyanobacteria. It is generally accepted that cyanobacteria may cause major disruptions of the aquatic ecosystem and that cyanobacteria have strong negative effects on Daphnia (Lampert, 1987; Christoffersen, 1996; DeMott, 1999; Lu¨rling, 2003). Hence, improving water quality and reducing the vulnerability to the threats of harmful cyanobacterial blooms is a key issue. Despite possible short-term reductions in juvenile Daphnia survival and growth by higher clay concentrations (Kirk and Gilbert, 1990), such minor effects will be outweighed by significant reductions in the amount of cyanobacteria. This study did not reveal any major detrimental effects on Daphnia when using Phoslock or its active ingredient lanthanum. Further studies may be warranted to monitor the zooplankton community during the application of Phoslock in real situations as there are inconsistent responses by zooplankton to the secondary impacts of turbidity on feeding. A re-assessment of the turbidity impacts on bottom-dwelling organisms that might experience the highest exposure to the modified clay should be undertaken although these are known to tolerate higher suspended solids concentrations.
5.
Conclusions
The purpose of this study was: 1) to establish a dose response relationship between Phoslock and the growth of D. 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. From the results presented here it can be concluded that: C
C
About 0.001% of the lanthanum embedded in the clay was released from it. In medium without phosphorous, lanthanum up to the 1000 mg l1 tested had no effect on life-history traits of the zooplankton grazer D. magna.
water research 44 (2010) 309–319
C
C
C
In the presence of phosphorous, rhabdophane (LaPO4 $ nH2O) formation resulted in significant precipitation of the food algae and consequently affected lifehistory traits of D. magna. A juvenile growth assay with 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.
Acknowledgements Mr Nigel Traill and Patrick van Goethem are cordially thanked for delivering the batches of Phoslock. Dr Said Yasseri (Limnological Institute Dr Nowak, Germany), Dr Sarah Groves (Phoslock Water Solutions Ltd., Australia) and Dr David Garman (Environmental Biotechnology CRC Pty Ltd, Australia) are thanked for comments on an earlier draft of the manuscript.
references
Akhurst, D., Jones, G.B., McConchie, D.M., 2004. The application of sediment capping agents on phosphorus speciation and mobility in a sub-tropical dunal lake. Mar. Freshwater Res. 55, 715–725. Anonymous, 2008a. Measures for Lake Restoration at the Ba¨rensee, City of Bruchko¨bel – Application Report and Results. Institut Dr. Nowak, 7 pp. Anonymous, 2008b. Report on the Application of Bentophos on the Silbersee. Institut Dr. Nowak, 7 pp. Barry, M.J., Meehan, B.J., 2000. The acute and chronic toxicity of lanthanum to Daphnia carinata. Chemosphere 41, 1669–1674. Biswas, B.K., Inoue, K., Ghimire, K.N., Ohta, S., Harada, H., Ohto, K., Kawakita, H., 2007. The adsorption of phosphate from an aquatic environment using metal-loaded orange waste. J. Colloid Interface Sci. 312, 214–223. Carpenter, S.R., 2008. Phosphorus control is critical to mitigating eutrophication. Proc. Natl. Acad. Sci. USA 105, 11039–11040. Carpenter, S.R., Caraco, N.F., Correll, D.L., Howarth, R.W., Sharpley, A.N., Smith, V.H., 1998. Nonpoint pollution of surface waters with phosphorus and nitrogen. Ecol. Appl. 8, 559–568. Christoffersen, K., 1996. Ecological implications of cyanobacterial toxins in aquatic food webs. Phycologia 35, 42–50. Cleuvers, M., 2003. Aquatic ecotoxicity of pharmaceuticals including the assessment of combination effects. Toxicol. Lett. 142, 185–194. Codd, G.A., Morrison, L.F., Metcalf, J.S., 2005. Cyanobacterial toxins: risk management for health protection. Toxicol. Appl. Pharmacol. 203, 264–272. Correll, D.L., 1998. The role of phosphorus in the eutrophication of receiving waters: a review. J. Environ. Qual. 27, 261–266. DeMott, W.R., 1999. Foraging strategies and growth inhibition in five daphnids feeding on mixtures of a toxic cyanobacterium and a green alga. Freshwater Biol. 42, 263–274.
317
Dittmann, E., Wiegand, C., 2006. Cyanobacterial toxins – occurrence, biosynthesis and impact on human affairs. Mol. Nutr. Food Res. 50, 7–17. Douglas, G.B., Adeney, J.A., Robb, M.S., 1999. A novel technique for reducing bioavailable phosphorus in water and sediments. In: International Association Water Quality Conference on Diffuse Pollution, pp. 517–523. Driscoll, C.T., Letterman, R.D., 1995. Factors regulating residual aluminium concentrations in treated waters. Environmetrics 3, 287–309. Flapper, T., 2003. Research and Development Program. Project completion report Phoslock trial at Fyshwick STP Lagoon. Ecowise number A011389, 18 pp. Forbes, V.E., Calow, P., 1999. Is the per capita rate of increase a good measure of population-level effects in ecotoxicology? Environ. Toxicol. Chem. 18, 1544–1556. Fogg, G.E., 1969. The physiology of an algal nuisance. Proc. Royal Soc. London B 173, 175–189. Gulati, R.D., Van Donk, E., 2002. Lakes in the Netherlands, their origin, eutrophication and restoration: state-of-the-art review. Hydrobiologia 478, 73–106. Guy, R.D., Chakrabarti, C.L., 1976. Studies of metal–organic interactions in model systems pertaining to natural waters. Can. J. Chem. 54, 2600–2611. Haghseresht, F., 2005a. A Revolution in Phosphorous Removal. Phoslock Water Solutions Ltd. http://www.phoslock.com.au, 21 pp. Haghseresht, F., 2005b. Protocol for the Determination of Filterable Reactive Phosphorous (FRP) Uptake of Phoslock. Report Integrated Mineral Technology Pty Ltd. http://www. phoslock.com.au. Haghseresht, F., 2006. The Availability of Lanthanum Ions from Phoslock: Residual Concentrations Following Phoslock Applications & Implications for Consumption Report No. PLT06-2. Hanazato, T., 1998. Growth analysis of Daphnia early juvenile stages as an alternative method to test the chronic effect of chemicals. Chemosphere 36, 1903–1909. Herna´ndez, T., Martin, P., 2007. Microstructural and optical features of a Eu-monazite. J. Eur. Ceramic Soc. 27, 109–114. Jin, X., Chu, Z., Yan, F., Zen, Q., 2009. Effects of lanthanum(III) and EDTA on the growth and competition of Microcystis aeruginosa and Scenedesmus quadricauda. Limnologica 39, 86–93. Johannesson, K.H., Lyons, W.B., 1994. The rare earth element geochemistry of mono lake water and the importance of carbonate complexing. Limnol. Oceanogr. 39, 1141–1154. Kamble, S.P., Jagtap, S., Labhsetwar, N.K., Thakare, D., Godfrey, S., Devotta, S., Rayalu, S.S., 2007. Defluoridation of drinking water using chitin, chitosan and lanthanum-modified chitosan. Chem. Eng. J. 129, 173–180. Kirk, K.L., 1991. Suspended day reduces Daphnia feeding rate: behavioural mechanisms. Freshwater Biol. 25, 357–365. Kirk, K.L., Gilbert, J.J., 1990. Suspended clay and the population dynamics of planktonic rotifers and cladocerans. Ecology 71, 1741–1755. Lampert, W., 1987. Feeding and nutrition in Daphnia. In: Peters, R.H., De Bernardi, R. (Eds.), ‘‘Daphnia’’ memorie dell’Istituto Italiano di Idrobiologia, vol. 45, pp. 143–192. Lampert, W., Trubetskova, I., 1996. Juvenile growth rate as a measure of fitness in Daphnia. Funct. Ecol. 10, 631–635. Lee, J.H., Byrne, R.H., 1992. Examination of comparative rare earth element complexation behavior using linear free-energy relationships. Geochim. Cosmochim. Acta 56, 1127–1137. Liu, X., Byrne, R.H., 1997. Rare earth and yttrium phosphate solubilities in aqueous solution. Geochim. Cosmochim. Acta 61, 1625–1633.
318
water research 44 (2010) 309–319
Lucas, S., Champion, E., Bregiroux, D., Bernache-Assolant, D., Audubert, F., 2004. Rare earth phosphate powders RePO4 $ nH2O (Re ¼ La, Ce or Y) – Part I. Synthesis and characterization. J. Solid State Chem. 177, 1302–1311. Lu¨rling, M., 2003. Effects of microcystin-free and microcystincontaining strains of the cyanobacterium Microcystis aeruginosa on growth of the grazer Daphnia magna. Environ. Toxicol. 18, 202–210. Lu¨rling, M., Beekman, W., 1999. Grazer-induced defenses in Scenedesmus (Chlorococcales; Chlorophyceae): coenobium and spine formation. Phycologia 38, 368–376. Martel, A.E., Motekaitis, R.J., 1989. Coordination chemistry and speciation of Al(III) in aqueous solution. In: Lewis, T.E. (Ed.), Environmental Chemistry and Toxicology of Aluminum. Lewis Publishers, Chelsea, MI, pp. 3–17. Martin, M.L., Hickey, C.W., 2004. Determination of HSNO Ecotoxic Thresholds for Granular Phoslock (Eureka 1 Formulation) Phase 1: Acute Toxicity. NIWA Project PXL 05201. NIWA, New Zealand. Martı´nez-Jero´nimo, F., Espinoza-Cha´vez, F., Villasen˜or, R., 2000. Effect of culture volume and adult density on the neonate production of Daphnia magna, as a test organism for aquatic toxicity tests. Environ. Toxicol. 15, 155–159. McCauley, E., Murdoch, W.W., Nisbet, R.M., 1990. Growth, reproduction, and mortality of Daphnia pulex Leydig: life at low food. Funct. Ecol. 4, 505–514. Mehner, T., Diekmann, M., Gonsiorczyk, T., Kasprzak, P., Koschel, R., Krienitz, L., Rumpf, M., Schulz, M., Wauer, G., 2008. Rapid recovery from eutrophication of a stratified lake by disruption of internal nutrient load. Ecosystems 11, 1142–1156. Meyer, J.S., Ingersoll, C.G., McDonald, L.L., Boyce, M.S., 1986. Estimating uncertainty in population growth rates: jackknife vs. bootstrap techniques. Ecology 67, 1156–1166. Moermond, C.T.A., Tijnk, J., van Wezel, A.P., Koelmans, A.A., 2001. Distribution, speciation, and bioavailability of lanthanides in the Rhine-Meuse estuary, The Netherlands. Environ. Toxicol. Chem. 20, 1916–1926. McIntosh, J., 2007. Phoslock application – Lake Okareka Final Report. Environment Bay of Plenty Environmental Publication 2007/23, ISSN: 1175 9372. NICNAS, 2001. National Industrial Chemicals Notification and Assessment Scheme, Full Public Report – Lanthanum Modified Clay, File No: NA/899, 28 pp. Paerl, H.W., 1988. Nuisance phytoplankton blooms in coastal, estuarine, and inland waters. Limnol. Oceanogr. 33, 823–847. Paerl, H.W., 2008. Chapter 10: nutrient and other environmental controls of harmful cyanobacterial blooms along the freshwater–marine continuum. In: Hudnell, K.E. (Ed.), Cyanobacterial Harmful Algal Blooms: State of the Science and Research Needs. Adv. Exp. Med. Biol, vol. 619, pp. 217–237. Paerl, H.W., Huisman, J., 2008. Blooms like it hot. Science 320, 57–58. Porcella, D.B., Rixford, C.E., Slater, J.W., 1969. Molting and calcification in Daphnia magna. Physiol. Zool. 42, 148–159. Reynolds, C.S., 1987. Cyanobacterial water blooms. Adv. Bot. Res. 13, 67–143. Reynolds, C.S., Walsby, A.E., 1975. Water blooms. Biol. Rev. 50, 437–481. RIVM, 2008. Digitale Documenten: Zink en zinkverbindingen, Lood en loodverbindingen en Koper en koperverbindingen. http://www.rivm.nl (accessed 07.01.09) (in Dutch). Robb, M., Greenop, B., Goss, Z., Douglas, G., Adeney, J., 2003. Application of Phoslock, an innovative phosphorus binding clay, to two Western Australian waterways: preliminary findings. Hydrobiologia 494, 237–243. Ross, G., Cloete, T.M., 2006. Phoslock Field Trial at K’shani Lake Lodge, Hartbeespoort Dam. Report University of Pretoria, 13 pp.
Ross, G., Haghseresht, F., Cloete, T.M., 2008. The effect of pH and anoxia on the performance of Phoslock, a phosphorus binding clay. Harmful Algae 7, 545–550. Schindler, D.W., 1974. Eutrophication and recovery in experimental lakes: implications for lake management. Science 184, 897–899. Schindler, D.W., 1975. Whole-lake fertilization experiments with phosphorus, nitrogen, and carbon. Int. Ver. Theor. Angew. Limnol. Verh. 19, 3221–3231. Schindler, D.W., 1977. Evolution of phosphorus in lakes. Science 195, 260–262. Schindler, D.W., 2006. Recent advances in the understanding and management of eutrophication. Limnol. Oceanogr. 51, 356–363. Schindler, D.W., Hecky, R.E., Findlay, D.L., Stainton, M.P., Parker, B.R., Paterson, M.J., Beaty, K.G., Lyng, M., Kasian, S.E.M., 2008. Eutrophication of lakes cannot be controlled by reducing nitrogen input: results of a 37-year whole-ecosystem experiment. Proc. Natl. Acad. Sci. 105, 11254–11258. Sharpley, A.N., Tunney, H., 2000. Phosphorus research strategies to meet agricultural and environmental challenges of the 21st century. J. Environ. Qual. 29, 176–181. Søndergaard, M., Jensen, A., Jeppesen, E., 2003. Role of sediment and internal loading of phosphorus in shallow lakes. Hydrobiologia 506/509, 135–145. Sneller, F.E.C., Kalf, D.F., Weltje, L., van Wezel, A.P., 2000. Maximum permissible concentrations and negligible concentrations for rare earth elements (REEs). RIVM report 601501011. Stauber, J.L., 2000. Toxicity Testing of Modified Clay Leachates Using Freshwater Organisms. CSIRO Centre for Advanced Analytical Chemistry Energy Technology. Report No: ET/ IR267R. March 2000. Stauber, J.L., Binet, M.T., 2000. Canning River Phoslock Field Trial – Ecotoxicity Testing Final Report. CSIRO Centre for Advanced Analytical Chemistry Energy Technology. Report No: ET317R. August 2000. Steffensen, D.A., 2008. Chapter 37: economic cost of cyanobacterial blooms. In: Hudnell, K.E. (Ed.), Cyanobacterial Harmful Algal Blooms: State of the Science and Research Needs. Adv. Exp. Med. Biol, vol. 619, pp. 855–865. Steinberg, C.E.W., Kamara, S., Prokhotskaya, V.Y., et al., 2006. Dissolved humic substances – ecological driving forces from the individual to the ecosystem level? (Special review). Freshwater Biol. 51, 1189–1210. Sun, H., Wang, X.-R., Wang, L.-S., 1997. Bioconcentration of rare earth elements lanthanum, gadolinium and yttrium in algae Chlorella vulgaris Beijerinck: influence of chemical species. Chemosphere 34, 1753–1760. Tian, S., Jiang, P., Ning, P., Su, Y., 2009. Enhanced adsorption removal of phosphate from water by mixed lanthanum/ aluminum pillared montmorillonite. Chem. Eng. J. 151, 141–148. Tollrian, R., 1993. Neckteeth formation in Daphnia pulex as an example of continuous phenotypic plasticity: morphological effects of Chaoborus kairomone concentration and their quantification. J. Plankton Res. 15, 1309–1318. USEPA, 2002. TSCA New Chemicals Program (NCP) Chemical Categories. URL: http://www.epa.gov/oppt/newchems/pubs/ cat02.htm#Lanthanides or Rare Earth Metals. Van de Plassche, E.J., 2002. Een ad-hoc Maximaal Toelaatbaar Risiconiveau (MTR) voor aluminium in oppervlaktewater. Report: 4L1574.A1/R0001/EVDP/Nijm, 7 pp. (in Dutch). Vanni, M.J., Lampert, W., 1992. Food quality effects on life history traits and fitness in the generalist herbivore Daphnia. Oecologia 92, 48–57. Welch, E.D., Cooke, G.D., 1995. Internal phosphorus loading in shallow lakes: importance and control. Lake. Reserv. Manage. 11, 273–281.
water research 44 (2010) 309–319
Weltje, L., Brouwer, A.H., Verburg, T.G., Wolterbeek, H.Th., De Goeij, J.J.M., 2002. Accumulation and elimination of lanthanum by duckweed (Lemna minor L.) as influenced by organism growth and lanthanum sorption to glass. Environ. Toxicol. Chem. 21, 1483–1489. Weiss, G.B., 1974. Cellular pharmacology of lanthanum. Annu. Rev. Pharmacol. 14, 343–354.
319
Yang, X.-Y., Ying, D.-Y., Sun, H., 1999. Distribution and bioavailability of rare earth elements in aquatic microcosm. Chemosphere 39, 2443–2450. Zhou, P.-J., Lin, J., Shen, H., Li, T., Song, L.-R., Shen, Y.-W., Liu, Y.D., 2004. Kinetic studies on the combined effects of lanthanum and cerium on the growth of Microcystis aeruginosa and their accumulation by M. aeruginosa. Bull. Environ. Contam. Toxicol. 72, 711–716.
water research 44 (2010) 320–330
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Complementary methods to investigate the development of clogging within a horizontal sub-surface flow tertiary treatment wetland P.R. Knowles a,*, P. Griffin b, P.A. Davies a a b
Sustainable Environment Research Group, Aston University, Birmingham B4 7ET, UK Severn Trent Water Ltd, Avon House, St.Martins Road, Coventry CV3 6PR, UK
article info
abstract
Article history:
A combination of experimental methods was applied at a clogged, horizontal subsurface
Received 25 July 2009
flow (HSSF) municipal wastewater tertiary treatment wetland (TW) in the UK, to quantify
Received in revised form
the extent of surface and subsurface clogging which had resulted in undesirable surface
7 September 2009
flow. The three dimensional hydraulic conductivity profile was determined, using
Accepted 10 September 2009
a purpose made device which recreates the constant head permeameter test in-situ. The
Published online 12 September 2009
hydrodynamic pathways were investigated by performing dye tracing tests with Rhodamine WT and a novel multi-channel, data-logging, flow through Fluorimeter which allows
Keywords:
synchronous measurements to be taken from a matrix of sampling points. Hydraulic
Treatment wetlands
conductivity varied in all planes, with the lowest measurement of 0.1 m d1 corresponding
Horizontal sub-surface flow
to the surface layer at the inlet, and the maximum measurement of 1550 m d1 located at
Hydraulic conductivity
a 0.4 m depth at the outlet. According to dye tracing results, the region where the overland
Clogging
flow ceased received five times the average flow, which then vertically short-circuited
Tracer tests
below the rhizosphere. The tracer break-through curve obtained from the outlet showed
Hydrodynamics
that this preferential flow-path accounted for approximately 80% of the flow overall and arrived 8 h before a distinctly separate secondary flow-path. The overall volumetric efficiency of the clogged system was 71% and the hydrology was simulated using a dual-path, dead-zone storage model. It is concluded that uneven inlet distribution, continuous surface loading and high rhizosphere resistance is responsible for the clog formation observed in this system. The average inlet hydraulic conductivity was 2 m d1, suggesting that current European design guidelines, which predict that the system will reach an equilibrium hydraulic conductivity of 86 m d1, do not adequately describe the hydrology of mature systems. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Treatment Wetlands (TWs) have become an established technology choice in the UK for the treatment of wastewaters in remote locations. Over 1200 systems have been installed over the last 20 years – predominantly horizontal, sub-surface
flow (HSSF) systems used for the tertiary treatment (i.e. polishing) of wastewater (Cooper, 2007). Such HSSF TWs consist of a bed of porous gravel, in which the wastewater is subject to the correct conditions for final purification. However, the cleaning process results in the gradual clogging of the subsurface, and thus deterioration of hydraulic conductivity
* Corresponding author. Tel.: þ44(0)121 204 3724. E-mail address: [email protected] (P.R. Knowles). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.028
321
water research 44 (2010) 320–330
Table 1 – Hydraulic conductivity measurements and system details reported in previous studies of gravel bed HSSF TWs. Reference
System name
Treatment type
Conductivity (inlet)a (m d1)
Conductivity (outlet)a (m d1)
Method
Fisher, 1990
Scripus Typha Control Benton Cell 3 Bed 4 Cell 1 Cell 2 Jones Gray Terrell Snelling Verdu 1 Verdu 2 Alfes Corbins Almaret N Almaret S Verdu Corbins
Secondary Secondary Secondary Secondary Landfill Leachate Acid drainage Acid drainage Secondary Secondary Secondary Secondary Secondary Tertiary Secondary Primary Secondary Secondary Secondary Secondary
1800 2500 2500 2500 4150 6 6 1000 10200 4900 85 2 25 7 2 1 1 20 3
25000 25000 25000 27500 3370 3500 3500 5400 8100 4700 325 12 61 2 200 87 82 45 55
Level survey Level survey Level survey Level survey Level survey Level survey Level survey Level survey Level survey Level survey Level survey Falling head Falling head Falling head Falling head Falling head Falling head Falling head Falling head
Kadlec and Watson, 1993 Sanford et al., 1995a Drury and Mainzhausen, 2000 Watson and Choate, 2001
Caselles-Osorio et al., 2007
a
Pedescoll et al., 2009
a Values are width averaged.
over time (Wallace and Knight, 2006). The various symptoms that accompany clogging in HSSF TWs are widely documented and include surface sludge accumulations, overland flow, poor reed growth and weed infestation (Batchelor and Loots, 1997; Caselles-Osorio et al., 2007; Cooper et al., 2005; Fisher, 1990; Kadlec and Watson, 1993; Maloszewski et al., 2006; Rousseau et al., 2005; Watson et al., 1990). Indeed, such occurrences have been reported for many HSSF TWs in the UK, with Cooper et al. (2008) reporting that from a survey of 255 SSF TWs (82% of which were HSSF tertiary treatment systems) almost 30% exhibited surface flow over most of the bed, and over 50% displayed weed colonisation across at least one quarter of the bed surface. Additionally, it was not unusual to find surface sludge accumulations exceeding 150 mm at the inlet and 40 mm at the outlet. Rousseau et al. (2005) made similar observations in their survey of 12 UK
Fig. 1 – Frequency Distribution of the Volumetric Efficiency measured in 37 HSSF TWs, by comparison of observed and design Hydraulic Residence Times. Adapted from Kadlec and Wallace (2008).
based HSSF stormwater TWs, reporting that the vast majority of them had experienced sludge build-up over the entire surface of the bed. Despite vast anecdotal evidence the exact mechanisms of clogging are not wholly understood, Blazejewski and Murat-Blazejewska (1997) and Platzer and Mauch (1997) summarised the numerous physical design parameters and operational variables responsible. These include wastewater solids characteristics, subsurface media characteristics and choice of batch or continuous operating mode. According to European Design Guidelines (Cooper, 1990), HSSF TWs which employ gravels with sizes of 3–12 mm will mature to reach an equilibrium hydraulic conductivity of about 86.4 m d1. However, the non-cohesive nature of gravel makes it difficult to remove representative samples from the field to test this hypothesis using a laboratory permeameter. Ranieri (2003) used an in-situ permeameter called the Guelph Permeameter (Reynolds and Elrick, 1986) to survey a gravel bed HSSF TW in Italy finding values ranged from 190 to 610 m d1, although these values are above the practical measurement range of the device (1–90 m d1). Though Langergraber et al.(2003) used the Guelph Permeameter to measure hydraulic conductivity within a Vertical Flow Subsurface Treatment Wetland, their system contained sand in addition to gravel resulting in a hydraulic conductivity low enough for measurement with this instrument. Indeed, most in-situ permeability tests are intended for cohesive, unsaturated geological material with hydraulic conductivities far below those usually measured in gravel beds (ASTM-D5126, 2004). Consequently, the hydraulic conductivity profile of the wetland has usually been estimated from Darcy’s Law, by measuring the corresponding water table height at different points in the bed. Conclusions generally agree that the inlet zone becomes more clogged than the remainder of the system
322
water research 44 (2010) 320–330
Table 2 – Details of the site used to trial the proposed combination of experimental methods. Moreton Morrell Latitude./Longitude. Test Date Age at test (months) Treatment Type Dimensions (m) Substrate HLR (m3 d1) Inlet Arrangement Effective gravel diameter d10 (mm) Gravel median diameter d50 (mm) Gravel uniformity coefficient Particle size distribution gradation coefficient Gravel clean hydraulic conductivity (m d1)
52.20/-1.55 June 2008 177 Tertiary municipal 15 L 15 W 0.6D 3–9 mm gravel 175 m3 d1 4* Vertical risers 2.6 5.2 2.11 1.48 2100
due to the high degree of solids removal in this region (Table 1). However, this method may not accurately determine the extent of clogging, as Darcy’s Law cannot take into account the varying thickness of the water table resulting from the groundwater energy balance and also requires that the flow field is saturated (Bear, 1979). For small hydraulic gradients, the accuracy with which the water table can be measured is limiting (Sanford et al., 1995a). Further, although this method may estimate the bulk resistance offered by the wetland media in the direction orthogonal to flow, it does not detect whether clogging varies vertically (Dittrich, 2006). Recently, methods to directly measure substrate permeability have been employed by Caselles-Osorio et al.(2007) and Pedescoll et al. (2009) on full-scale HSSF TWs in Spain, by utilising an in-situ falling head permeability test (NAVFAC, 1986). This method removes the uncertainty involved in estimating substrate conductivity from water-table surveys. Knowles and Davies (2009) have proposed a new method for the in-situ measurement of high conductivity materials, which recreates the constant head permeability test (BS-ISO-17313, 2004) in-situ. This allows both the magnitude and position of clogging within the subsurface to be determined. The hydrodynamics that correspond to the developed clogging profile can be determined by tracer studies. Through
this several authors have established that multiple preferential flow paths can exist through the wetland subsurface (Batchelor and Loots, 1997, Maloszewski et al., 2006). Regarding the UK condition, systems highly loaded with solids may develop separate overland and sub-surface flow fractions due to a high degree of surface clogging (Batchelor and Loots, 1997; Christian, 1990; Sanford et al., 1995b). Secondly, several authors have observed that sub-surface flow will often shortcircuit across the bottom of the bed concluding that the greater resistance to flow offered by the rhizosphere is responsible for this (Breen and Chick, 1995; Fisher, 1990; Waters et al., 1993). This likelihood is supported by Tanner et al. (1998), who confirmed during a 5 year study of a dairy effluent HSSF TW that the clogging contribution from vegetation could be 1–2 times more significant than that from wastewater. Further studies of this system identified that the organic matter was highly refractory in nature and mainly accumulated in the top 100 mm of media (Nguyen, 2001), possibly explaining the 50% reduction in observed hydraulic retention time. Indeed, hydraulic inefficiency caused by clogging and short-circuiting is a common facet of TW performance. This is emphasised in Fig. 1 which shows that from a survey of tracer tests performed on 37 HSSF TWs, 29 underperformed with an average volumetric efficiency of 91% (Kadlec and Wallace, 2008). The two tracer materials which have been used predominantly in previous wetland studies are bromide salts and Rhodamine Water Tracer, although each noted to have certain limitations (Flury and Wai, 2003). For example, Rhodamine dye is only said to behave conservatively if the TW is small, shallow (less than 0.6 m) and with a hydraulic residence time of less than one week (Lin et al., 2003). However, Rhodamine has obvious advantages over more conservative tracers, such as radioactive or biologically based alternatives, in that it has low eco-toxicity and will eventually photochemically destabilise, making it a good choice for ecological applications. Another advantage of Rhodamine is that it is easy to measure in-situ with relatively inexpensive equipment, making it the choice for this study and many previous TW hydraulics studies (Bhattarai and Griffin Jr, 1999; Holland et al., 2004; Shilton and Prasad, 1996; Simi and Mitchell, 1999). This paper details complementary methods that have been developed to assess the impact of clogging at a mature HSSF TW located in Moreton Morrell, Warwickshire, UK, under the jurisdiction of Severn Trent Water plc. One method allows the
Fig. 2 – Photographs detailing the more prominent findings at the site, including a) Completely clogged inlet distributors, b) Surface sludge layers up to 200 mm thick at the inlet, c) two water levels at certain points in the system, representing disconnected overland and sub-surface flows.
water research 44 (2010) 320–330
323
Fig. 3 – Plan view of Moreton Morrell tertiary treatment TW in Warwickshire (UK), showing the major design features, operational observations, and the locations of sampling points used during the experiments (drawing not to scale).
in-situ measurement of substrate hydraulic conductivity at different longitudinal, transverse and vertical locations within the bed, so that the exact position of clogging within the subsurface can be located. The second method is a multipoint tracer study to confirm that flow follows the path of least resistance. The Rhodamine break-through curve is also monitored at the outlet and levelling surveys are performed to explore the physical characteristics of the system. It is intended that the tests performed will elucidate the mature hydrology of this type of Treatment Wetland, and reveal details of the internal clogging mechanism.
2.
Methods
2.1.
Site description
Information regarding the site is given in Table 2. The site is almost 15 years old and as of June 2008 had a fully established
macrophyte root network (Phragmites australis). Symptoms of hydraulic maturity were evident including inlet distributor pipes with variable degrees of clogging such that width distribution was uneven (Fig. 2a), and a thick surface sludge layer towards the inlet (Fig. 2b). It was assumed these factors were responsible for overland flow across a quarter of the bed surface (Cooper et al., 2008). As shown in Fig. 2c, the water table height was often below the level of any surface waters that were present at that point, suggesting a vertical head loss existed between the two water levels. Fig. 3 details the major design features of the site, and shows the location of the sampling points from which results were collected. Using water levelling techniques a survey of the test site was performed, measuring the relative heights of the surface sludge layer, gravel surface, and water table surface, at each sampling point. Kadlec and Watson (1993) quote the accuracy of their standard water survey technique to 10 mm. To try and improve the accuracy of the method for this study, the location of the water surface was established using a digital
Fig. 4 – Experimental set-up (left) for in-situ determination of the hydraulic conductivity of highly porous media (not to scale) and the apparatus installed at a HSSF TW in the UK (right). Reservoir stands approx. 1 m off the ground. Reproduced from Knowles and Davies (2009).
324
water research 44 (2010) 320–330
Fig. 5 – Schematic representation of one channel from the portable, multi-channel fluorimeter, showing the major optoelectronic and hydraulic modules involved in the design (left) and the multi-channel, in-situ fluorimeter installed at a HSSF TW in the UK (right).
manometer (Kane 3100-1, Kane International, Herts., UK). Through this, the associated error with each measurement is approximated as 5 mm, mainly arising from human error in misreading graduated scales. All readings were made relative to an arbitrary datum which represents the bottom of the bed at the outlet. To set this it was assumed that the gravel depth was 0.6 m at the outlet and that the system was built with a bottom slope of 0.5%, following the European Design Guidelines being employed when it was constructed (Cooper, 1990).
2.2.
Hydraulic conductivity
The test monitors the discharge of a constant head of water through a 400 mm (LCELL) core of substrate, contained within
the walls of a 160 mm diameter permeameter cell with crosssectional area ACELL. The level of water in the cell is kept constant by a Mariotte Siphon activated water reservoir, similar in principle to the Guelph Permeameter (Reynolds and Elrick, 1986) but enlarged to make it suitable for applications in higher conductivity media such as gravel (86.4 m d1 to 86400 m d1). From this point on, all numbers in square brackets refer to those labels indicated on Fig. 4 (left), which illustrates the major components of the experiment. Four manometer take off tubes [19] (11.5 mm ID tubes ranging from 200 to 500 mm length in 100 mm increments), are inserted into the gravel core prior to the experiment commencing, to provide takeoff points at 100, 200, 300 and 400 mm depths into the gravel core. Using a digital differential manometer [13]
Fig. 6 – The longitudinally varying profile of the four transverse sampling transects A–D, detailing the relative height of the surface, water, gravel and base levels.
water research 44 (2010) 320–330
325
Fig. 7 – (Left) The variation in hydraulic conductivity over the surface of the TW surveyed at Moreton Morell. The darker regions represent those of lower hydraullo conductivity (greater clogging) and the lighter regions those of higher conductivities (lesser clogging). A possible primary preferential flow path is indicated with a white arrow , based on the equipotential lines across the permeability contours. (Right). The corresponding flow fraction measured at the 500 mm depth plane within the TW subsurface as a result of the clogging. The darker regions represent those that receive more flow (Figures produced using COMSOL Multiphysics 3.4-COMSOLAB).
with a 500 mm graduated depth gauge [12] in each take off tube, it is possible to determine the static [17] and dynamic [14] water levels, thus allowing the vertical head loss across each 100 mm section (hn) to be measured. By monitoring the
_ from the reservoir [5,6], in keeping the discharge of water ðqÞ permeameter head (hT) constant, it is possible to calculate the permeability of the gravel core (kT) using Darcy’s Law Eq. (1). kT ¼
q_ LCELL ACELL hT
(1)
Subsequently the permeability of each 100 mm section (kn) can be found Eq. (2). kn ¼
hT kT 4hn
(2)
The test provides vertical conductivity profiles, although by interpolating between sample points it would be possible to predict a horizontal conductivity profile. This is based on the assumption that flow would behave identically in all directions. Fig. 4 (right) gives a photo of the experimental setup at a TW in South Warwickshire. The apparatus cost just over £1000 (V1270) in 2008; approximately £100 (V130) to construct the Mariotte siphon reservoir and £900 (V1070) for the purchase of four digital manometers (Knowles and Davies, 2009). Experience has shown that obtaining one set of results can take between 20 min and one hour, depending on the degree of clogging. The test was performed at each sampling point indicated on Fig. 3. Fig. 8 – Linearly interpolated hydraulic conductivity values for five transverse cross sections at Moreton Morell TW. The darker regions represent those of lower hydraulic conductivity (greater clogging) and the lighter regions those of higher conductivities (lesser clogging). A possible primary preferential flow path, which snakes vertically and transversely, is indicated with a white circles, based on the equipotential lines across the permeability contours. (Figures produced using COMSOL Multiphysics 3.4-COMSOLAB).
2.3.
Tracer studies
In response to there being no affordable proprietary fluorimeter for synchronous measurement from a matrix of sampling points, a novel multi-channel fluorimeter was created. The fluorimeter contains 20 separate opto-electronic modules, each based around a 570 nm high pass filter, photodiode with peak sensitivity at 580 nm, and an 18000 mCd LED with peak wavelength at 525 nm. These components correspond to the
326
water research 44 (2010) 320–330
Table 3 – Parameters derived from the breakthrough curves of each sampling point, studied using the multi-point fluorimeter. The hydraulic efficiency factor and short-circuiting factor are defined according to Persson et al.(1999) and Ta and Brignal (1998) respectively.
Relative Flow Fraction Peak Time (d) Centroid Time (d) Theoretical Centroid Time (d) Volumetric Efficiency Hydraulic Efficiency Factor 16% Breakthrough Curve (d) 50% Breakthrough Curve (d) 84% Breakthrough Curve (d) Short-Circuiting Factor
A1
A2
A3
A4
B1
B2
B3
B4
C1
C2
C3
C4
D1
D2
D3
D4
0.57 0.11 0.08 0.07
0.01 0.03 0.17 0.14
5.40 0.03 0.17 0.22
0.30 0.09 0.13 0.29
0.24 0.01 0.14 0.07
0.71 0.03 0.16 0.14
0.67 0.02 0.05 0.22
1.20 0.03 0.19 0.29
0.46 0.10 0.09 0.07
0.74 0.07 0.16 0.14
0.00 0.05 0.09 0.22
2.22 0.19 0.17 0.29
0.00 0.02 0.04 0.07
NR NR NR NR
1.23 0.33 0.31 0.22
1.25 0.13 0.17 0.29
1.06 1.45 0.03
1.17 0.16 0.04
0.76 0.21 0.07
0.45 0.72 0.08
1.92 0.08 0.02
1.14 0.17 0.06
0.24 0.33 0.01
0.64 0.15 0.09
1.22 1.18 0.06
1.11 0.41 0.07
0.41 0.51 0.04
0.57 1.18 0.09
0.61 0.47 0.02
NR NR NR
1.41 1.08 0.31
0.58 0.74 0.09
0.07
0.16
0.16
0.11
0.13
0.16
0.03
0.18
0.09
0.15
0.10
0.16
0.03
NR
0.33
0.16
0.10
0.29
0.25
0.19
0.25
0.25
0.09
0.27
0.10
0.24
0.13
0.22
0.07
NR
0.35
0.24
0.45
0.23
0.44
0.71
0.13
0.38
0.40
0.47
0.64
0.43
0.43
0.57
0.50
NR
0.94
0.57
fluorescing properties of the Rhodamine WT used in the experiments (Tolbest, UK), which has excitation and emission wavelengths at 556 nm and 580 nm respectively. Each module produces a voltage proportional to the concentration of Rhodamine passing through a flow-through PVC cuvette. The sensitivity of each channel is adjustable, and the practical resolution, as imposed by the signal to noise ratio, would allow a useful detection range of about 10 parts per billion (ppb) of Rhodamine in water. The sample fluid is drawn from different points within the TW aquatic environment using a 20 channel peristaltic pump; each channel connected to a different opto-electronic module. Along with data logging equipment, all of these components are mounted in a modified IP66 rated, weatherproof enclosure (Sarel, Italy) which can be wheeled onto the monitoring site of interest. More information regarding these components is included in Fig. 5 (left), which gives a schematic representation of one channel. Fig. 5 (right) is a photograph of the interior of the multi-channel fluorimeter at a TW in South Warwickshire. A 5 ml single-shot impulse of concentrated Rhodamine WT solution was added to the inlet manifold, upstream of the wetland cell. Lengths of 20 m silicon tubing with an ID of 3 mm (Fisher Scientific, UK) were attached to 16 of the fluorimeter channels, with one tube originating from each sampling point. The tube openings were submersed to 0.5 m below the gravel surface to investigate the nature of the flow field at this height. Each peristaltic pump channel was set to deliver 0.2 ml s1 through the fluorimeter. The machine discharged the analysed
water directly onto the centre of the bed surface as it was considered the total discharge (0.003 l s1) was significantly small compared with the flow though the wetland (2 l s1) to have negligible impact on downstream fluorescence measurements. The Rhodamine breakthrough curve was measured at the outlet using a Cyclops 7 submersible fluorimeter (Turner Designs, USA).
3.
Results
3.1.
Site survey
Fig. 6 shows the longitudinally varying profile of the four transverse sampling transects A–D. Variations in hydraulic gradient across the four transects were between 109 mm (transect A) and 7 mm (and transect D) relating to the uneven flow distribution through the inlet risers. The equilibrium saturated water table height of the system is consistently below the bed surface, despite the appearance of overland flow. The water table is located below the gravel surface at all points apart from in the first 5 m of transects A and B, where it is located within the sludge layer. It can be seen from Fig. 6b that towards the inlet of transect B, the depth of sludge that has accumulated on the surface of the gravel exceeds 200 mm. The depth of surface sludge is generally greatest at the inlet and becomes progressively shallower towards the outlet.
Fig. 9 – The obtained Rhodamine WT Residence Time Distribution Curve (dotted line) and the fitted Aggregated Dead-Zone Model (solid line).
327
water research 44 (2010) 320–330
Table 4 – Hydraulic parameters for the various flow paths derived from the outlet residence time distribution, and parameters used to fit the dead-zone storage model. The hydraulic efficiency factor and short-circuiting factor are defined according to Persson et al. (1999) and Ta and Brignal (1998) respectively. Hydraulic loading rate Theoretical residence time Total tracer injected
m3 d1 d ml
Mean Residence Time Peak Time 16% Tracer Recovery Time 84% Tracer Recovery Time Total Tracer Recovered
d d d
Single path 0.26 0.04 0.06
Overland path 0.21 0.04 0.05
Subsurface path 0.46 0.59 0.39
d
0.45
0.40
0.55
ml
3.83
3.15
0.68
1 0.71 1.20 0.17 0.27
0.82 0.59 1.20 0.21 0.31
0.18 1.27 0.85 1.29 0.90
4.E-03 8.E-04 0.69 40000
8.E-05 4.E-04 0.10 30000
175 0.36 5
Flow-Split Volumetric Efficiency Number of CSTRs (N) Hydraulic Efficiency (l) ShortCircuiting (S) ADZ Model parameters, *Estimated from method of moments *Dispersion *Velocity *Effective area Storage transfer coefficient
3.2.
m2 s1 m s1 s
Hydraulic conductivity
Fig. 7 (left) shows how the hydraulic conductivity through each 0.4 m long gravel core varies across the bed surface (note the shading bar is a logarithmic scale). The darker regions represent those of lower hydraulic conductivity (greater clogging) and the lighter regions those of higher conductivities (lesser clogging). The minimum value of 0.1 m d1 was measured at inlet point C1 and the maximum of 1550 m d1 measured at outlet point D3. Width average conductivities for the four transects orthogonal to flow were 2, 343, 650 and 800 m d1, from inlet to outlet respectively. The order of magnitude of the inlet values are in good agreement (Table 1) with other studies that have directly measured substrate conductivity (Caselles-Osorio et al., 2007; Pedescoll et al., 2009). According to the contours on Fig. 7 (left), the overland flow in the system seems to correspond to wherever the hydraulic conductivity is less than 100 m d1. Fig. 8 demonstrates four transverse slices through the TW, developed from the 3D permeability results, which elucidates the locality of clogging. Values at the surface of the inlet zone across transects A–C were below 0.5 m d1. This explains the poor percolation of the influent waters at the inlet region and the resulting overland flow. Beyond the inlet the effect of the rhizosphere is to keep the conductivity between 100 and 1000 m d1, whilst below the rhizosphere (especially towards the outlet), values are generally greater than 1000 m d1. Overall, values varied by 6 orders of magnitude, from a minimum of 0.04 m d1 at a 200 mm depth below point B1, to a maximum of 40,000 m d1 at a 400 mm depth below point C4. The effect of poor inlet distribution is also evident in Fig. 8. Values along transect D are at least twice as high as values measured elsewhere, corresponding to the negligible influent distribution along this transect.
3.3.
Flow pattern visualisation
Fig. 7 (right) compares the total dye detected at each point with the average detected over the 500 mm vertical plane. This indicates the relative location of preferential flow-paths and dead-zones as values greater and less than one respectively. An instrument error meant no information was collected for point D2. The biggest preferential flow-path is detected at point A3 (5.4 times the average flow) which roughly coincides with the area where overland flow ceases. In contrast, upstream points A1 and A2 have very little involvement in the flow field, confirming observations that the majority of the flow short-circuits over the surface sludge until it can infiltrate into the subsurface and follow the path of least resistance below the rhizosphere. In accordance with the hydraulic conductivity results, flow follows the path of least resistance downstream from point A3, by steering towards points B3, B4, C4 and D4; perhaps avoiding a clogged outlet collector at point A4 (Cooper et al., 2008). Calculation of the volumetric efficiency emphasises the inefficiency created by this flow regime. Downstream points 3 and 4 at transects A B and C have values between 24% and 76% efficiency (Table 3) reflecting the premature passage of the RTD centroid where the overland flow secedes. Upstream points 1 and 2 at these transects consistently have volumetric efficiencies greater than 100% which is attributable to the low infiltration rates through the sludge layer at the inlet. The outlet RTD very clearly indicates the existence of two major flow-paths through the HSSF TW (Fig. 9) which have been partitioned according to the parameters reported in Table 4. The first path to arrive at the outlet corresponds to the overland flow that short-circuits to point A3, carrying 82% of the flow and arriving after only 1 h. The second represents the highly retarded flow-path from upstream percolations into
328
water research 44 (2010) 320–330
Table 5 – Suggested design guidelines applicable to the gravel media considered in this study, for equilibrium hydraulic conductivity at the inlet zone. Source
Range of applicable media (mm)
Predicted hydraulic conductivity (m d1)
5–10 3–6 4–8
86 2600 518
3–16
1000
EC/EWPCA (1990) TVA (1993) ¨ NORM B O l2505(1997) IWA (2000)
the subsurface which arrives almost 8 h later than the overland flow path. The mean residence times of these two paths were 0.21 and 0.46 days respectively, giving an overall volumetric efficiency of 71% when compared with the nominal residence time of 0.36 days. Tracer recovery was 77% with the loss mainly attributed to adsorption.
4.
Discussion
According to the water-table survey, a flow-rate of 2 l s1 through the wetland produced a hydraulic gradient of 0.9%. If calculated using Darcy’s Law, this would correspond to a hydraulic conductivity of 140 m d1. Despite the clogging evident from overland flow, this would suggest that system conductivity is still greater than the equilibrium guideline of 86.4 m d1 which was employed when it was designed (EC/ EWPCA, 1990). However, this figure represents the overall hydrology of the system and does not describe the six orders of magnitude variation in substrate hydraulic conductivity and corresponding short-circuiting that was measured between the inlet and outlet. Table 5 indicates that the guideline value used is the most conservative available when compared to equivalent guidelines published in various manuals and regulatory documents from around the world. This calls into question the current basis for the hydraulic design of HSSF TWs in the UK. A dual-path, dead-zone storage model was used to describe the shape of the outlet RTD. Multiple flow-path models
Fig. 10 – Representation of the hydrological regime prevalent in UK, HSSF TWs for the tertiary treatment of municipal wastewater. Diagram depicts a transverse cross-section and is not to scale.
(Chazarenc et al., 2003; Maloszewski et al., 2006; Wang and Jawitz, 2006) and dead-zone storage models (Martinez and Wise, 2003; Wang and Jawitz, 2006; Werner and Kadlec, 2000) have been used previously to describe the multiple peaks and weighted significance of the tail observed in HSSF TW breakthrough curves (Kadlec and Wallace, 2008). However, there is no knowledge of them being combined to describe flow through these systems. This was achieved using the MS Excel based HUSKY 1 code as formulated by Hellweger (2005), and the modelling parameters indicated in Table 4. Three of the model parameters were estimated by analysing the RTDs using the method of moments, leaving only the storage transfer coefficient to be optimised. The successfully fit transfer coefficients are similar to values used by Wang and Jawitz (2006) of 40000 s. The model is thought to adequately describe the obtained breakthrough curve (Fig. 9), despite the slight mistiming of the 1st peak. Fig. 10 describes the hydrological regime within a mature HSSF TW with surface loading, based on the observations made during the study. Major features include a low permeability surface sludge layer which is deepest at the inlet due to preferential solids accumulation in this region, a low permeability region in the top 200–300 mm of media associated with the bulk of the rhizosphere, separate equilibrium water levels in the system related to overland flow across a vertical percolation region and the water table resulting from horizontal flow respectively, and finally a nonlinearly varying water table surface. Inlet clogging to the point where surface flow ensues may be unavoidable given the current design configuration of UK systems. The main factor responsible for the way that clogging has developed at Moreton Morrell is poor inlet distribution causing preferential sludge and detritus accumulation and overland flow down one side of the bed. Systems built in the UK since Moreton Morrell have incorporated improved inlet distributors (Griffin et al., 2008), such as troughs with numerous v-notch weirs, to try and reduce the problems associated with riser pipes (Cooper et al., 2008) However, poor inlet distribution is not the only condition responsible for clogging and ponding of water on the surface of the bed. Tertiary systems in the UK have a higher areal hydraulic loading rate than secondary systems and the municipal wastewater that they receive is a combination of domestic and urban diffuse sources. During wet weather events the larger flows can overwhelm upstream processes causing washout of biological flocs and accumulated solids. These factors mean UK tertiary systems often receive a higher solids load than their secondary counterparts elsewhere in the world. Loading solids onto the bed surface increases the likelihood that the surface will seal; especially when the rhizosphere develops and reduces the pore diameter available for particle infiltration. American HSSF TW systems, such as those reported by Wallace and Knight (2006), utilise submersed inlet distributors which load the influent wastewater over the entire crosssectional area orthogonal to flow, reducing the tendency of the surface to seal near the inlet. The continuous application of secondary wastewater will also promote surface clogging. In French Vertical Flow systems, primary wastewater is intermittently dosed onto the bed surface, giving opportunity for the surface sludge layer to dry out. This prevents terminal
water research 44 (2010) 320–330
329
ponding despite the comparatively high solids loading in French systems (Chazarenc and Merlin, 2005).
University of Catalonia, Spain for her indispensable assistance whilst performing the site survey.
5.
references
Conclusions
Complementary in-situ hydraulic conductivity and Rhodamine dye tracing experiments have been successfully employed to model the hydraulic conditions in a clogged, gravel bed, tertiary wastewater treatment wetland. The in-situ constant head permeameter method revealed that the hydraulic conductivity at the inlet was of the order of 1 m d1, compared to 2–3 orders of magnitude greater elsewhere. The ability to measure vertical variations in subsurface hydraulic conductivity established that the greatest resistance corresponded to the surface sludge and rhizosphere regions. The multi-point dye tracing technique confirmed that sub-surface flow vertically short-circuits below the rhizosphere following the path of least resistance. Almost 36% of the total flow detected at a 500 mm depth occurred at one point which roughly coincided with where overland flow terminated. The outlet RTD confirmed that two distinctly separate flow paths exist through the system: a primary overland path which constituted almost 80% of the flow field, and a secondary subsurface path that was heavily retarded in comparison to the overflow path. The nature of the clogging resulted in tracer detection at the outlet within one hour, and an overall system volumetric efficiency of 71%. A dual-path, dead-zone storage model was used to adequately describe the observed flow dynamics. It can be concluded that observations of overland flow across one quarter of the bed are attributable to poor surface infiltration rates at the inlet and uneven influent distribution across the width. The hydraulic conductivity of the system generally decreases from inlet to outlet and from surface to base. Even the most conservative design guidelines overestimated the hydraulic conductivity of the fouled inlet media by a factor of 40. If, however, design guidelines were based on a hydraulic conductivity of 1 m d1, as measured here, this would result in impractically large wetland areas. Instead, it is recommend to design beds with large width to length ratios, such that the influent is distributed over a wider area, thus delaying terminal clogging. Subsequent work will compare a large cross-section of operational wetlands with a variety of ages, operating conditions and design characteristics that may help explain the clogging mechanism and allow guidelines to be stipulated which help maximise the longevity of these systems. A predictive finite element model will also be developed which will allow the correlation between the permeability and dye tracing results to be analysed in more detail.
Acknowledgements This work was made possible thanks to joint funding from Severn Trent Water Plc. (UK) and a CASE studentship granted by the ESPRC UK (ref. CASE/CNA/06/28). The first author gratefully acknowledges Anna Pedescoll from the Technical
ASTM-D5126, 2004. Standard Guide for Comparison of Field Methods for Determining Hydraulic Conductivity in the Vadose Zone. American Society for the Testing of Materials, West Conshohocken, PA, USA. Batchelor, A., Loots, P., 1997. A critical evaluation of a pilot scale subsurface flow wetland: 10 years after commissioning. Water Science and Technology 35 (5), 337–343. Bear, J., 1979. Hydraulics of Groundwater. McGraw-Hill, New York; London. Bhattarai, R.R., Griffin Jr., D.M., 1999. Results of tracer tests in rock-plant filters. Journal of Environmental Engineering 125 (2), 117–125. Blazejewski, R., Murat-Blazejewska, S., 1997. Soil clogging phenomena in constructed wetlands with subsurface flow. Water Science and Technology 35 (5), 183–188. Breen, P.F., Chick, A.J., 1995. Rootzone dynamics in constructed wetlands receiving wastewater: a comparison of vertical and horizontal flow systems. Water Science and Technology 32 (3), 281–290. BS-ISO-17313, 2004. Soil Quality. Determination of Hydraulic Conductivity of Saturated Porous Materials Using a Flexible Wall Permeameter. British Standards Institute. Caselles-Osorio, A., Puigagut, J., Segu, E., Vaello, N., Grane´s, F., Garcı´a, D., Garcı´a, J., 2007. Solids accumulation in six full-scale subsurface flow constructed wetlands. Water Research 41 (6), 1388–1398. Chazarenc, F., Merlin, G., 2005. Influence of surface layer on hydrology and biology of gravel bed vertical flow constructed wetlands. Water Science and Technology 51 (9), 91–97. Chazarenc, F., Merlin, G., Gonthier, Y., 2003. Hydrodynamics of horizontal subsurface flow constructed wetlands. Ecological Engineering 21 (2–3), 165–173. Christian, J.N.W., 1990. Reed bed treatment systems: experimental gravel beds at Gravesend – the southern water experience. In: Cooper, P.F., Findlater, B.C. (Eds.), Constructed Wetlands in Water Pollution Control. Cambridge, UK: Pergamon Press, Oxford, UK, pp. 309–319. Cooper, P.F. (Ed.), 1990. European Design and Operations Guidelines for Reed Bed Treatment Systems. WRc, Swindon, UK. Cooper, P., 2007. The Constructed Wetland Association UK database of constructed wetland systems. Water Science and Technology 56 (3), 1–6. Cooper, D., Griffin, P., Cooper, P., 2005. Factors affecting the longevity of sub-surface horizontal flow systems operating as tertiary treatment for sewage effluent. Water Science and Technology 51 (9), 127–135. Cooper, D., Griffin, P., Cooper, P.F., 2008. Factors affecting the longevity of sub-surface horizontal flow systems operating as tertiary treatment for sewage effluent. In: Vymazal, J. (Ed.), Wastewater Treatment, Plant Dynamics and Management in Constructed and Natural Wetlands. Springer, Dordrecht, The Netherlands, pp. 191–198. Dittrich, E., 2006. Experiences on hydraulic performance of subsurface flow constructed wetlands. Pollack Periodica 1 (1), 53–66. Drury, W.J., Mainzhausen, K., 2000. Hydraulic characteristics of subsurface flow wetlands, 2000 Billings Land Reclamation Symposium, Billings, MT, USA EC/EWPCA, 1990. European Design and Operations Guidelines for Reed Bed Treatment Systems. WRc, Swindon, UK.
330
water research 44 (2010) 320–330
Fisher, P.J., 1990. Hydraulic Characteristics of Constructed Wetlands at Richmond, NSW, Australia. In: Cooper, P.F., Findlater, B.C. (Eds.), Constructed Wetlands in Water Pollution Control. Cambridge, UK: Pergamon Press, Oxford, UK, pp. 21–31. Flury, M., Wai, N.N., 2003. Dyes as tracers for vadose zone hydrology. Reviews of Geophysics 41 (1), 1–2. Griffin, P., Wilson, L., Cooper, D., 2008. Changes in the use, operation and design of sub-surface flow constructed wetlands in a major UK water utility. 11th International Conference on Wetland Systems for Water Pollution Control, Indore, India, 419–426. Hellweger, F.L., 2005. Measuring and modeling large-scale pollutant dispersion in surface waters. Collection Systems Alexandria, Virginia, 812–835. Holland, J.F., Martin, J.F., Granata, T., Bouchard, V., Quigley, M., Brown, L., 2004. Effects of wetland depth and flow rate on residence time distribution characteristics. Ecological Engineering 23 (3), 189–203. Kadlec, R.H., Wallace, S., 2008. Treatment wetlands. In: Boca Raton, Fla.: CRC, second ed. Taylor & Francis, London [distributor]. Kadlec, R.H., Watson, J.T., 1993. Hydraulics and solids accumulation in a gravel bed treatment wetland. In: Moshiri, G.A.E. (Ed.), Constructed Wetlands for Water Quality Improvement: Conference: Selected Papers. Boca Raton: Lewis Pub, pp. 227–235. Knowles, P.R., Davies, P.A. (2009) A method for the in-situ determination of the hydraulic conductivity of gravels as used in constructed wetlands for wastewater treatment. Desalination and water treatment, 1 (5), 257–266. Langergraber, G., Haberl, R., Laber, J., Pressl, A., 2003. Evaluation of substrate clogging processes in vertical flow constructed wetlands. Water Science and Technology 48 (5), 25–34. Lin, A.Y.C., Debroux, J.F., Cunningham, J.A., Reinhard, M., 2003. Comparison of rhodamine WT and bromide in the determination of hydraulic characteristics of constructed wetlands. Ecological Engineering 20 (1), 75–88. Maloszewski, P., Wachniew, P., Czuprynski, P., 2006. Study of hydraulic parameters in heterogeneous gravel beds: constructed wetland in Nowa S1upia (Poland). Journal of Hydrology 331 (3–4), 630–642. Martinez, C.J., Wise, W.R., 2003. Analysis of constructed treatment wetland hydraulics with the transient storage model OTIS. Ecological Engineering 20 (3), 211–222. NAVFAC, 1986. Soil Mechanics. Design Manual 7.01. Naval Facilities Engineering Command, Alexandria, Virginia, USA. Nguyen, L., 2001. Accumulation of organic matter fractions in a gravel-bed constructed wetland. Water Science and Technology 44 (11–12), 281–287. Pedescoll, A., Uggetti, E., Llorens, E., Grane´s, F., Garcı´a, D., Garcı´a, J. (2009) Practical method based on saturated hydraulic conductivity used to assess clogging in subsurface flow constructed wetlands. Ecological Engineering, 35 (8), 1216–1224.
Platzer, C., Mauch, K., 1997. Soil clogging in vertical flow reed beds – mechanisms, parameters, consequences and solutions? Water Science and Technology 35 (5), 175–181. Ranieri, E., 2003. Hydraulics of sub-superficial flow constructed wetlands in semi arid climate conditions. Water Science and Technology 47 (7–8), 49–55. Reynolds, W.D., Elrick, D.E., 1986. A method for simultaneous in situ measurement in the vadose zone of field-saturated hydraulic conductivity, sorptivity and the conductivitypressure head relationship. Ground Water Monitoring Review 6 (1), 84–95. Rousseau, D.P.L., Horton, D., Griffin, P., Vanrolleghem, P.A., De Pauw, N., 2005. Impact of operational maintenance on the asset life of storm reed beds. Water Science and Technology 51 (9), 243–250. Sanford, W.E., Steenhuis, T.S., Parlange, J.Y., Surface, J.M., Peverly, J.H., 1995a. Hydraulic conductivity of gravel and sand as substrates in rock-reed filters. Ecological Engineering 4 (4), 321–326. Sanford, W.E., Steenhuis, T.S., Surface, J.M., Peverly, J.H., 1995b. Flow characteristics of rock-reed filters for treatment of landfill leachate. Ecological Engineering 5 (1), 37–50. Shilton, A.N., Prasad, J.N., 1996. Tracer studies of a gravel bed wetland. Water Science and Technology 34 (4), 421–425. Simi, A.L., Mitchell, C.A., 1999. Design and hydraulic performance of a constructed wetland treating oil refinery wastewater. Water Science and Technology 40 (3), 301–307. Tanner, C.C., Sukias, J.P.S., Upsdell, M.P., 1998. Organic matter accumulation during maturation of gravel-bed constructed wetlands treating farm dairy wastewaters. Water Research 32 (10), 3046–3054. Wallace, S.D., Knight, R.L., 2006. Small-Scale Constructed Wetland Treatment Systems: Feasibility, Design Criteria and O & M Requirements. Water Environment Research Foundation (WERF), Virginia, Alexandria. Wang, H., Jawitz, J.W., 2006. Hydraulic analysis of cell-network treatment wetlands. Journal of Hydrology 330 (3–4), 721–734. Waters, M.T., Pilgrim, D.H., Schulz, T.J., Pilgrim, I.D., 1993. Variability of hydraulic response of constructed wetlands. Proceedings – National Conference on Hydraulic Engineering, 406–411. Watson, J.T., Choate, K.D., 2001. Hydraulic conductivity of onsite constructed wetlands. In: On-Site Wastewater Treatment, Proceedings of the Ninth National Symposium on Individual and Small Community Sewage Systems. Michigan, St. Joseph, pp. 631–648. Watson, J.T., Choate, K.D., Steiner, G.R., 1990. Performance of constructed wetland treatment systems at Benton, Hardin and Pembroke, Kentucky, during the early vegetation establishment. In: Cooper, P.F., Findlater, B.C. (Eds.), Constructed Wetlands in Water Pollution Control. Cambridge, UK: Pergamon Press, Oxford, UK, pp. 171–182. Werner, T.M., Kadlec, R.H., 2000. Wetland residence time distribution modeling. Ecological Engineering 15 (1–2), 77–90.
water research 44 (2010) 331–339
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Dechlorination kinetics of TCE at toxic TCE concentrations: Assessment of different models P.J. Haest, D. Springael, E. Smolders* Division Soil and Water Management, Katholieke Universiteit Leuven, Kasteelpark Arenberg 20, 3001 Heverlee, Belgium
article info
abstract
Article history:
The reductive dechlorination of trichloroethene (TCE) in a TCE source zone can be self-
Received 3 March 2009
inhibited by TCE toxicity. A study was set up to examine the toxicity of TCE in terms of
Received in revised form
species specific degradation kinetics and microbial growth and to evaluate models that
15 July 2009
describe this self-inhibition. A batch experiment was performed using the TCE dechlori-
Accepted 13 September 2009
nating KB-1 culture at initial TCE concentrations ranging from 0.04 mM to saturation
Published online 17 September 2009
(8.4 mM). Biodegradation activity was highest at 0.3 mM TCE and no activity was found at
Keywords:
degradation rates and Dehalococcoides numbers were modeled with Monod kinetics
Dehalococcoides
combined with either Haldane inhibition or a log-logistic dose-response inhibition on these
Monod kinetics
rates. The log-logistic toxicity model appeared the most appropriate model and predicts
Optimization
that the species specific degradation activities are reduced by a factor 2 at about 1 mM TCE,
Reductive dechlorination
respectively cis-DCE. However, the model showed that the inhibitive effects on the time for
Self-inhibition
TCE to ethene degradation are a complex function of degradation kinetics and the initial
Trichloroethene
cell densities of the dechlorinating species. Our analysis suggests that the self-inhibition
concentrations from 4 to 8 mM. Species specific TCE and cis-DCE (cis-dichloroethene)
on biodegradation cannot be predicted by a single concentration threshold without information on the cell densities. ª 2009 Elsevier Ltd. All rights reserved.
1.
Introduction
Groundwater contamination by Chlorinated Aliphatic Hydrocarbons (CAHs), such as trichloroethene (TCE), is often found near dry cleaning facilities or metal processing plants. The sanitation of such a site is a difficult and time consuming process when the TCE is present as a free phase. Clean-up is therefore evolving to a phased treatment where bioremediation is considered a valuable polishing step (Christ et al., 2005). TCE can be biodegraded to cis-dichloroethene (cis-DCE), vinylchloride (VC) and eventually to the harmless ethene (ETH) through sequential reductive dechlorination reactions which occur under anaerobic conditions. Several bacterial species are able to metabolically convert TCE to cis-DCE but up to now only Dehalococcoides has been found to perform the
final step from VC to ETH. Batch degradation experiments in which chlorinated ethene concentrations were applied up to the aqueous saturation revealed a self-inhibition of the dechlorination reaction (Yu and Semprini, 2004). Yang and McCarty (2000) showed that the lag-phase associated with the TCE degradation reaction increased above 1 mM TCE and that the degradation is inhibited at the TCE saturation of 8.4 mM. A stronger inhibition was observed by Duhamel et al. (2002) and Haest et al. (2006) using the KB-1 culture, a culture able to dechlorinate TCE to ethene. A complete inhibition was observed at TCE concentrations >w2 mM. A similar abrupt stop of the dechlorination reaction was observed by Amos et al. (2007) in a batch experiment where all pure cultures tested ceased dechlorinating at w0.54 mM perchloroethene (PCE).
* Corresponding author. Tel.: þ32 16329677; fax: þ32 16321997. E-mail address: [email protected] (E. Smolders). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.033
332
water research 44 (2010) 331–339
Nomenclature b Ci EC50 Hc KCI kd
describes the slope of the dose-response curve [–] aqueous concentration of compound i [mM] the concentration at which kmax is half of the uninhibited level [mM] dimensionless Henry constant [–] Competitive inhibition constant [mM] decay coefficient [d1]
A model describing the kinetics of TCE degradation including the observed self-inhibition is required to describe near-source bioremediation. Initially, Michaelis-Menten type kinetics were used to describe the reductive dechlorination reaction at low to moderate chlorinated ethene concentrations (Fennell and Gossett, 1998; Garant and Lynd, 1998; Haston and McCarty, 1999; Tandoi et al., 1994). Competitive inhibition terms were included following enzyme kinetics to account for the observed inhibition by chlorinated ethenes on the dechlorination of daughter compounds and microbial growth was described using Monod kinetics (Cupples et al., 2004; Yu and Semprini, 2004). However, no model has yet been calibrated to the dechlorination kinetics at a TCE concentration range up to the aqueous saturation. Yu and Semprini (2004) could describe the TCE dechlorination kinetics at 4 mM TCE by means of a model developed for PCE self-inhibition applying Haldane inhibition. A Haldane inhibition term cannot describe an abrupt stop of the dechlorination activity as observed in other studies (Amos et al., 2007; Duhamel et al., 2002; Haest et al., 2006). In addition, the parameters describing microbial growth in the model presented by Yu and Semprini (2004) were selected from literature (Fennell and Gossett, 1998; Maymo-Gatell et al., 1997) and are based on a general biomass indicator, i.e. total protein content. The effect of a high TCE concentration on the yield of the dechlorinating species was not examined. Moreover, the general biomass indicator precludes a comparison of the dechlorination kinetics determined in different studies as the degradation activity in a mixed culture relates to the number of dechlorinators. The objective of this study was to assess the self-inhibition of TCE in terms of species specific growth and degradation activity. Two different equations were compared to empirically describe the dose-response relationship. Real-time quantitative PCR (qPCR) enabled a determination of species specific activity and growth rates. Advanced approaches are required to fit the growth and degradation in all treatments with one model. A novel multi-objective global optimization algorithm allowed the identification of species specific parameters describing the TCE degradation at high TCE concentrations by the KB-1 culture.
2.
Material and methods
2.1.
Culture and medium preparation
The KB-1 culture was kindly provided by SiREM (Ontario, Canada). This culture has been intensively studied and contains
KHI kmax Ks RATE Vaq Vg Xm Ym
Haldane inhibition constant [mM] maximal degradation rate [mmol cell1 d1] half velocity constant [mM] Degradation rate [mM d1] liquid volume [L] gaseous volume [L] cell number of species m [cell# L1] yield coefficient of species m [cell# mmol1]
Dehalococcoides spp., among a wide variety of other microorganisms (Duhamel et al., 2002; Duhamel and Edwards, 2006, 2007). The culture reductively dechlorinates TCE to ethene (Duhamel et al., 2004). The inoculum for the batch degradation experiment was grown on 1 mM TCE and 1.5 mM butyrate at 20 C in an anaerobic mineral medium containing: 2.88 g/L (NH4)H2PO4, 0.1 g/L MgSO4$7H2O, 0.05 g/L Ca(NO3)2$H2O, 0.1 g/L yeast extract, 1% resazurin, 1 g/L KOH, 2.4 g/ L NaHCO3, 0.25 g/L Na2S.9H2O, 1 ml/L trace elements (stock solution: 0.5 g/L EDTA, 0.1 g/L ZnSO4$7H2O, 0.3 g/L H3BO3, 0.01 g/L CuCl2$2H2O, 0.03 g/L Na2MoO4, 0.033 g/L Na2WO4$2H2O, 0.2 g/L CoCl2.6H2O, 0.01 g/L AlCl3$6H2O, 1 ml HClc (37 %)) and 1 ml vitamin solution (stock solution: 100 mg/L p-aminobenzoic acid, 50 mg/L folic acid, 100 mg/L lipoic acid, 100 mg/L riboflavic acid, 200 mg/L thiamine, 200 mg/L nicotic acid, 500 mg/L pyridoxamine, 100 mg/L pantotheic acid, 100 mg/L cobalamine, 20 mg/L biotine). The presence of Dehalococcoides spp. in the inoculum was confirmed by a nested 16S rDNA PCR-DGGE analysis with Dehalococcoides specific primers DeF (50 -gca att aag ata gtg gc-30 ) DER (50 -act tcg tcc caa tta cc-30 ) (Cupples et al., 2003) and semi-specific primers 968-GC-F (50 -cgc ccg ggg cgc gcc ccg ggc ggg gcg ggg gca cgg ggg gaa cgc gaa gaa cct tac-30 ) DHC 1350-R (50 -cac ctt gct gat atg cgg-30 ) (He et al., 2003). Triplicate batches for the degradation experiment were set-up in 120 mL vials with an N2/CO2 80/20 atmosphere. The vials were inoculated with 6.5 vol% of the culture grown on 1 mM TCE (see above) in a total volume of 80 mL. TCE was added using a gastight glass syringe from an anaerobic pure stock solution at final concentrations of 0.04–0.3–0.9–1.3–1.8–4–6–8 mM TCE. Butyrate was provided as carbon and electron donor because butyrate fermentation yields low H2-concentrations promoting Dehalococcoides over competitors (Aulenta et al., 2005; Fennell et al., 1997; Yang and McCarty, 1998). It was provided in a 5-fold electron equivalents (eeq) excess taking into account a complete fermentation of butyrate to CO2 and dechlorination of TCE to ethene. The vials were sealed with Viton stoppers and aluminum crimp caps and incubated in darkness on a horizontal shaker at 100 rpm at 20 C. Duhamel and Edwards (2006) recently found that a Geobacter strain degraded up to 80% of the TCE in the KB-1 culture. Unfortunately, this information was not yet available during the time that this experiment was performed. As such, Geobacter was not monitored in this experiment. Our recent data in other batches confirmed the observations made by Duhamel and Edwards (2006) by showing an increase in Geobacter numbers during TCE degradation and an increase in Dehalococcoides numbers during cis-DCE and VC degradation in the KB-1 culture (see Fig. S1, supporting information).
water research 44 (2010) 331–339
2.2.
Analytical methods
TCE, cis-DCE, VC, ETH and methane were measured in 1 mL aqueous samples taken from the vials at each sampling time. The extracted liquid volume was replenished with sterile N2 gas. The aqueous sample was analyzed by means of headspace analysis using a TriPlus autosampler and a Focus GC-FID (Thermo-Electron Corporation) equipped with an Rt-QPLOT column (0.53 mm 30 m). The samples were heated
k X C max;i m i RATEi ¼ Ciþ1 Ciþ2 Ci Ks;i 1 þ þ Ci 1 þ exp bi log þ KCI;iþ1 KCI;iþ2 EC50;i
at 75 C for 30 min before a headspace sample was injected in the GC. Helium was used as carrier gas at 5 mL/min. The oven temperature program started at 50 C with a ramp of 20 C/ min to 180 C and a final ramp of 30 C/min to 220 C for 5.5 min. Calibration curves were obtained from external standards. DNA was extracted from 400 mL of aqueous sample as described by Uyttebroek et al. (2006). Numbers of Dehalococcoides spp. 16S rRNA gene copies were quantified by realtime PCR (qPCR) as described by Dijk et al. (2008) using the Dehalococcoides specific primers Dco728F (50 -aag gcg gtt ttc tag gtt gtc ac-30 ) and Dco944R (50 -ctt cat gca tgt caa at-30 ) (Smits et al., 2004). The cycling program consisted of 15 min of initial denaturation at 95 C, followed by 40 cycles of 10 s of denaturation at 95 C, 20 s of annealing at 50 C and 20 s of extension at 72 C with a final extension step at 72 C for 5 min. One 16Sr RNA gene copy was assumed per Dehalococcoides cell (Klappenbach et al., 2001).
3.
Model development
3.1.
Batch degradation model
The reductive dechlorination reaction was modeled using Monod kinetics. The maximal degradation rate kmax was expressed on a unit cell basis [mmol cell1 day1] with cell growth related to the degradation activity and assuming that yield and biomass decay were unaffected by the CAH concentrations. The degradation rate of the CAHs was calculated as described by Yu and Semprini (2004). The self-inhibition of CAHs was embedded in the kmax parameter that decreases as the CAH concentration increases. The Haldane inhibition model as applied by Yu and Semprini (2004) was contrasted with a log-logistic dose-response model as described by Doelman and Haanstra (1989) (see Eqs. (1) and (2)). The log-logistic dose-response model is frequently used in ecotoxicological studies and empirically describes the inhibition of biological processes by toxic substances. It will be referred to as EC50 model in this study. Fig. 1 illustrates the features of the different models where the degradation rate described by the Haldane inhibition model slowly and asymptotically approaches zero while the EC50 model can
333
predict a sharp decrease of the degradation rate in a narrow concentration range. The Haldane inhibition model for a given compound ‘i’ (e.g. VC) reads
RATEi ¼ KS;i
kmax;i Xm Ci Ciþ1 Ciþ2 Ci 1þ þ Ci 1 þ þ KCI;iþ1 KCI;iþ2 KHI;i
(1)
while the EC50 inhibition model for that compound reads
(2)
with RATE the degradation rate in solution of the respective compound [mM d1], Ci [mM] the aqueous concentration and Ks [mM] the half velocity constant. Compounds ‘i þ 1’ and ‘i þ 2’ are the parent compounds of ‘i’ (e.g. cis-DCE respectively TCE) that are included in Eq. (1) and Eq. (2) to account for the competitive inhibition using the competitive inhibition constants KCI,iþ1 and KCI,iþ2 [mM] as reported previously (Cupples et al., 2004; Yu and Semprini, 2004). If not applicable, these terms (C/KCI) were omitted from the equations. KHI,i is the Haldane inhibition constant for compound ‘i’ [mM]. Variable Xm [cell# L1] represents the cell number of the dechlorinating species ‘m’. The parameter EC50,i [mM] describes the concentration at which kmax,i is half of the uninhibited level while bi is the parameter that describes the slope of the dose-response curve. Experiments were performed in vials containing a gas and a liquid phase. Therefore, Monod equations were modified
Fig. 1 – Dechlorination kinetics according to the Haldane inhibition model, Eq. (1) (- -), and the log-logistic EC50 inhibition model, Eq. (2) (–), at non-limiting biomass concentration. The concentration of the chlorinated ethene is shown on the x-axis in a logarithmic scale. The dotted lines indicate the concentrations corresponding to a given percentage inhibition compared to the maximal degradation rate as predicted by the EC50 model in terms of parameters EC50 and b (Eq. (2)). * KHI [ EC50 if KHI >> Ks.
334
water research 44 (2010) 331–339
assuming degradation only takes place in solution. Vg and Vl [L] are the gaseous, respectively liquid volume and Hc the dimensionless Henry’s constant (Gossett, 1987). The total molar degradation rates in the vials were determined using the mass balance equation with Mi the total mass of compound ‘i’ in the vial and Ci,l the concentration of i in the aqueous phase: Mi ¼ Ci;l ðVl þ Vg Hc;i Þ. The Monod equations hence read: dXm X RATEi ¼ Ym V dt 1 þ Vgl Hc;i
described by Vrugt and Robinson (2007) is a multialgorithm, genetically adaptive multiobjective method (AMALGAM). It incorporates multiple objectives by looking for the globally optimal solution of the trade-off problem between different objectives, the so-called Pareto optimal solution. It could be especially useful in environmental research where difficulties exist in determining a specific microbial activity from the large amount of microbial processes taking place. AMALGAM was kindly provided by Dr. Vrugt as a Matlab code. Only treatments where degradation was observed were included in the parameter optimization, i.e. treatments containing initial TCE concentrations of 0.3, 0.9, 1.3 and 1.8 mM. The treatment containing 0.04 mM was not included for reasons given below. For each of these treatments 4 different observations, termed variables, were fitted at each sampling occasion, i.e. concentrations of TCE, cis-DCE, VC and the growth of Dehalococcoides. In total, there were 4 4 RMSE values, each one defined as
! kd;m Xm
(3)
with Ym [cell# mmol1] the yield coefficient of species m and kd,m [d1] the decay coefficient, with the mass balance: dCi 1 ¼ ð RATEi þ RATEiþ1 Þ V dt 1 þ Vgl Hc;i
! (4)
The findings of Duhamel and Edwards (2006) suggested a split of the biomass into 2 different actors for the KB-1 culture. Geobacter was found to degrade 80% of the TCE to cisDCE while Dehalococcoides converted cis-DCE to ethene. Both species couple dechlorination to growth. We confirmed this observation in a later experiment (Fig. S1, supporting information). Dechlorination kinetics were therefore adjusted for microbial growth of both species: Geobacter was assumed to grow on the expense of TCE degradation while Dehalococcoides was assumed to grow only on the expense of cis-DCE and VC degradation. Experimental observations largely confirmed this assumption (see further). The obtained set of differential equations was solved in Matlab using a variable order solver based on numerical differentiation formulas.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 Pn i¼1 qi;obs qi;sim RMSEj ¼ n
with qi,obs the observed and qi,sim the simulated variable at time i from a total of n observations. The model equations were solved simultaneously for all 4 treatments. This allows AMALGAM to search for one globally optimal solution, including the inhibitive effect of a high TCE concentration on the degradation activity and on the concurrent growth of the dechlorinating species. For each treatment, the 4 (m) calculated RMSEs were aggregated in one objective function (OF) yielding a total of 4 OFs (Madsen, 2003). Pm
3.2.
(5)
OF ¼
Optimization algorithm
j¼1
gj RMSEj
(6)
m
with gj a weight factor to compensate for differences in absolute values of RMSE terms of CAH concentrations and Dehalococcoides spp. numbers, defined as
The Monod model describing either TCE or cis-DCE degradation kinetics (Eqs. (1) and (2)) shows 11 or 13 adjustable parameters for the Haldane, respectively EC50 approach (see Table 1). These model parameters are correlated (Liu and Zachara, 2001; Robinson and Tiedje, 1983). As such, the indirect parameter determination of these equations using data from a degradation experiment is not straightforward. An evolutionary optimization algorithm, i.e. AMALGAM (Vrugt and Robinson, 2007) was used in this study. The algorithm
RMSEj þ ej ; j ¼ 1 : m gj RMSEj ¼ sj
(7)
with sj the standard deviation of variable RMSEj for the p model solutions from a preliminary model evaluation and ej a transformation constant given by
Table 1 – The tested parameter intervals for the Haldane and the EC50 inhibition models in the AMALGAM optimization algorithm.
Min Max
kmax,tce [mmol cell1 d1]
KS,tce [mM]
KCI,tce [mM]
KHI,tce [mM]
EC50,tce [mM]
btce [–]
Ytcea [cell# mmol1]
kd [d1]
6.57E13 6.57E09
1.40E3 4.39E2
1.40E3 1.00
0.5 5
1 4
4.39 13.17
– –
0.024 0.050
KS,dce [mM]
KCI,dceb [mM]
KHI,dce [mM]
EC50,dce [mM]
kmax,dce [mmol cell1 d1] Min Max
1.00E14 1.00E10
1.00E3 1.00E1
– –
1 8
1 8
bdce [–]
Ydce [cell# mmol1]
kd [d1]
4.39 13.17
7.00Eþ08 7.00Eþ12
0.024 0.050
a The yield parameter for the TCE degrading species was determined in a later experiment with the KB-1 culture at 1 mM TCE. b The reductive dechlorination kinetics describing VC degradation were not optimized using AMALGAM. KCI,dce was set to 4.79E3 mM, i.e. the average from the literature values presented in Table 3.
water research 44 (2010) 331–339
) RMSEj RMSE min 3j ¼ max min sj sj i¼1:p (
(8)
Parameters were sampled from a log-transformed interval of the minimum and maximum values given in Table 1. The interval was selected from estimates based on literature data. The parameters were optimized by minimizing the aggregated objective functions per treatment for all treatments simultaneously. For reasons mentioned above, there were no data about the TCE degrading species, i.e. Geobacter, in this experiment. To constrain its numbers, a penalty term was implemented when Geobacter numbers where lower, respectively higher than 1E6–1E11 cell# L1. The initial Geobacter cell numbers were set to 5E8 cell# L1, an amount frequently measured in later experiments (see 4.2). The absence of a detectable lag-time for TCE degradation at non-inhibitive TCE concentrations justifies the assumption of a high initial number of Geobacter cells. The yield coefficient of Geobacter on the expense of TCE degradation was set to 9E8 cell# L1, a value determined in a later experiment with a TCE concentration of 1 mM (Fig. S1, supporting information). It was not the aim of this study to optimize the parameters describing VC degradation by the KB-1 culture. Therefore, parameters describing VC degradation were manually fitted to the experimental outcome starting from the parameters presented by Yu and Semprini (2004) and recalculated to degradation per cell (Duhamel et al., 2004). VC concentrations were included as a fitted variable in AMALGAM because Dehalococcoides grows on the expense of VC degradation.
4.
Results and discussion
4.1.
Batch test
The observed TCE degradation and concurrent microbial growth for an initial TCE concentration of 0.3, 0.9, 1.3 and 1.8 mM is presented in Fig. 2. The dechlorination of TCE started without a detectable lag period in treatments containing TCE concentrations <1.3 mM beyond which the lag-time increased with increasing TCE concentrations. Dechlorination activity stopped at 4 mM TCE and above (not shown). Relative concentration changes in the treatment containing 0.04 mM TCE (Fig. S3, supporting information) were similar to those recorded for an initial TCE concentration of 0.3 mM. However, the species specific cis-DCE degradation rate was 25% lower in the former than in the latter treatment, likely due to an effect of substrate limitation, reflected by a high KS,dce. For that reason, the 0.3 mM treatment was defined as the uninhibited control treatment. Dechlorination of the daughter products (cis-DCE or VC) only started when the parent product (TCE respectively cis-DCE) was almost depleted. All these results are consistent with previous data (Fennell and Gossett, 1998; Yang and McCarty, 1998; Yu and Semprini, 2004) and follow degradation models which include self-inhibition and competitive inhibition (see Eq. (1)). Microbial data showed that the 16S rRNA gene copy numbers of Dehalococcoides significantly increased with an average of 1.1E11 copies L1 on the expense of cis-DCE and VC degradation by the end of the experiment. No marked trend
335
was observed for the yield of Dehalococcoides at higher CAH concentrations. The increase of 16S rRNA gene copy numbers of Dehalococcoides after the degradation of TCE to cis-DCE was on average only 5.5E8 copies L1. This implies that Dehalococcoides in the KB-1 culture mainly grows on the expense of cis-DCE and VC degradation as suspected by Duhamel and Edwards (2006).
4.2.
Parameter optimization
The fitted model output is compared to the measured CAH concentrations in Fig. 2. The AMALGAM algorithm could not find an optimal solution for the complete parameter set if all variables were included in the objective functions. Indeed, AMALGAM cannot cope with too many objectives (Vrugt, personal communication). Therefore, the model was fitted for the subsequent reactions. The parameters describing TCE degradation were determined in a first step and adequately described TCE degradation (Fig. 2). The yield parameter (Ytce) was adopted from a later experiment at 1 mM TCE (Fig. S1, supporting information) and kmax,tce was optimized using AMALGAM assuming X0,geo 5E8 cell# L1 (see 3.2). Alternatively, kmax,tce could also be adopted from Fig. S1, i.e. 3.1E10 mmol cell1 d1, with X0,geo fitted to the observations. This alternative approach yielded an optimal value for X0,geo of 1.8E8 cell# L1 in the EC50 inhibition model. The difference between the assumed and optimized Geobacter cell numbers is smaller than a factor of 3 which is the analytical uncertainty of our microbial quantification protocol, hence either fitting approaches are equivalent. The parameters describing cis-DCE degradation and growth of Dehalococcoides were determined in a second step with optimized parameters for TCE degradation from the first step. No cis-DCE was degraded in treatment 4 with 1.8 mM initial TCE concentration and this treatment was omitted from the cis-DCE optimization step. The resulting 9 objective functions for the 3 remaining treatments were aggregated per treatment yielding 3 aggregated objective functions in step 2. In each step, the model fit of the EC50 model was better than the Haldane model as the latter has a more rigid structure of the dose response curve (Fig. 1). Overall, the EC50 model appeared more flexible than the Haldane inhibition model and this fitted model is illustrated in Fig. 2. The Haldane inhibition model was unable to describe the observed lag-phase for TCE and cis-DCE degradation and the obtained parameters were therefore rejected (Fig. S2, supporting information). The optimized parameters for the EC50 model describing TCE and cisDCE degradation are presented in Tables 2 and 3. The parameters describing VC degradation were manually fitted to the observations with kmax,vc 5E-13 mmol cell1 d1, KS,vc 2.6 mM, KCI,dce 4.79 mM and Yvc 2E11 cell# mmol1 VC. The obtained parameter set describing the EC50 model predicts that the maximal TCE species specific degradation activity is reduced by a factor 2 at 1.01 mM TCE. The maximal cis-DCE species specific degradation activity is reduced by a factor 2 at 1.27 mM cis-DCE. The numbers of Dehalococcoides were underestimated at the end of the experiment suggesting that the parameters for VC degradation are not optimal. In addition, the degradation data recorded at an initial TCE concentration of 0.04 mM slightly differed from those
336
water research 44 (2010) 331–339
Fig. 2 – Experimental and fitted data of the EC50 inhibition model for the 4 batch treatments with increasing initial TCE concentrations from 0.3, 0.9, 1.3 to 1.8 mM. The upper graph per treatment shows in dots the measured amount of CAHs with error bars indicating the variation between replicate treatments (C TCE; V cis-DCE; - VC; > ETH). If error bars are not visible the variation between the replicate treatments was negligible. The lines represent the model output (– TCE; . cis-DCE; – – VC; (–.. ETH). The lower graph per treatment shows the measured numbers of Dehalococcoides (V) and the lines represent the model species for the degradation of TCE (–) and cis-DCE (.). Microbial numbers were only determined for one of the triplicate tests.
recorded at 0.3 mM (Fig. S3, supporting information). Surprisingly, if the EC50 model was fitted to all data 0.04– 1.8 mM, it could not adequately approximate the observations at the larger initial TCE concentrations. Along the same lines, we note that the EC50 model was able to describe the long lagtime at 1.8 mM TCE but failed to predict that TCE degradation finally started after day 100. It illustrates that even a parameter rich model as presented here is not sufficiently flexible to describe all data over that large concentration interval. The underlying idea of the EC50 model is that self inhibition affects activity but not yield. An alternative approach would be to incorporate the self-inhibition in the yield coefficient. Toxicity could not only inhibit the species specific activity but could also reduce yield by diverting gained energy to survival instead of growth. However, yields of Dehalococcoides on the expense of cis-DCE and VC degradation differed by less than a factor 3 among treatments with no significant effect of initial concentrations. The combined concept with inhibitive effects
on growth and activity is probably most realistic, however requires additional parameter fitting. The absence of data on the growth of Geobacter at other TCE concentrations than 1 mM did not allow to verify if yield varies with TCE concentrations.
4.3.
Monod model implications
The EC50 model incorporates a self-inhibition of the species specific degradation activity and, hence, inhibition on total degradation rate cannot be described by a single concentration threshold without information about the number of dechlorinating cells. In addition, when considering the entire degradation pathway of TCE to ethene the self-inhibition is a complex function of initial cell numbers for each degradation step and the inhibition of the species specific degradation activity. For example, in case of a low initial cell density (X0,geo ¼ 5E6 cell# L1 and X0,dcoc ¼ 1.5E7 cell# L1) the predicted reaction
337
water research 44 (2010) 331–339
Table 2 – The optimized parameters of the EC50 model describing TCE degradation versus parameters published in literature. Literature data were recalculated to identical units. Reference EC50 modela Duhamel and Edwards (2007) He et al. (2005) Maymo-Gatell et al. (1995)b,c Yu and Semprini (2004) b Yu and Semprini (2004) b Cupples et al. (2004) Garant and Lynd (1998) b Haston and McCarty (1999) b
kmax,tce [mmol cell1 d1]
Ks,tce [mM]
KCI,tce [mM]
KHI,tce [mM]
EC50,tce [mM]
btce [–]
Ytce [cell# mmol1]
kd [d1]
1.56E10 – – 2.09E13 2.60E13 2.63E13 7.45E13 0.39E13 0.07E13
4.19E3 – – – 2.76E3 1.80E3 12.4E3 17.4E3 1.4E3
37E2 – – – 0.28E2 0.18E2 0.68E2 1.74E2 –
– – – – 0.9 0.9 – – –
1.01 – – – – – – – –
8.83 – – – – – – – –
9.00Eþ8d 1.00Eþ11 7.80Eþ10 22.9Eþ11 28.6Eþ11 28.6Eþ11 4.70Eþ11 – –
0.029 – – – 0.024 0.024 0.050 – –
a Parameters obtained through inverse optimization with AMALGAM. b Recalculated values according to Duhamel et al. (2004) assuming a conversion factor of 4.2E–15 g dry weight of cell material per 16S rRNA gene copy and a protein content of 50%. c One yield constant was reported for the complete degradation of PCE to VC and ethene. d The yield parameter for the TCE degrading species was determined in a later experiment with the KB-1 culture at 1 mM TCE (Fig. S1, supporting information).
time for TCE degradation to cis-DCE increases by a factor 2 at an initial TCE concentration of 1 mM compared to the reaction time for the uninhibited control (0.3 mM). However, the time required for the degradation of 1 mM TCE to VC is less than 2 times the reaction time in the uninhibited control. The increase in reaction time due to the self-inhibition of TCE is not linearly passed on to the next steps of the sequential degradation reaction as those are mediated by a different species, i.e. Dehalococcoides, with its own growth kinetics. The determination of species specific degradation kinetics allows to compare data between different studies and experimental scales. For example: Azizian et al. (2008) performed a continuous-flow study and used an inoculum described by Yu et al. (2005). They measured species specific dechlorination rates that were orders of magnitude higher than those presented by Yu et al. (2005) rescaled to degradation per cell using a conversion factor of 4.2E15 g dry weight of cell material per
16S rRNA gene copy, assuming 1 copy per cell and a protein content of 50% (Duhamel et al., 2004). However, the average species specific VC degradation rate derived from the results presented by Azizian et al. (2008), 3E13 mmol cell1 d1, is in line with the optimized kmax,vc that we obtained in this study (5E13 mmol cell1 d1). It shows how the normalization of the degradation parameters strongly determines a correct comparison of the dechlorination kinetics. In addition, if yield constants from a mixed culture are expressed on a total protein content basis, the actual yield of the dechlorinating species cannot be estimated, especially at low CAH concentrations where other reactions are not yet inhibited (<0.3 mM (Yang and McCarty, 2000)). The determination of species specific degradation kinetics is not straightforward. Sung et al. (2006) found that growth yield estimates obtained by qPCR can vary by up to 1 order of magnitude with differing DNA extraction protocols. The qPCR
Table 3 – The optimized parameters of the EC50 model describing cis-DCE degradation versus parameters published in literature. Literature data were recalculated to identical units. Reference EC50 modela Schaefer et al. (2009) Duhamel and Edwards (2007) He et al. (2005) Maymo-Gatell et al. (1995) b,e Yu and Semprini (2004) b Yu and Semprini (2004) b Cupples et al. (2004) c,d Garant and Lynd (1998) b Haston and McCarty (1999) b
kmax,dce [mmol cell1 d1]
KS,dce [mM]
KCI,dce [mM]
KHI,dce [mM]
EC50,dce [mM]
bdce [–]
Ydce [cell# mmol1]
kd [d1]
2.08E11 1.25E11 – – 2.09E13 0.46E13 0.29E13 8.85E13 0.25E13 0.002E13
99.7E3 2.00E3 – – – 1.90E3 1.76E3 3.30E3 11.9E3 3.30E3
4.79E3 5.20E3 – – – 1.90E3 1.76E3 3.60E3 11.9E3 –
– – – – – 6 0.75 – – –
1.27 – – – – – – – – –
10.4 – – – – – – – – –
1.56Eþ10 0.44Eþ10 1.80Eþ11 8.40Eþ10 22.9Eþ11 28.6Eþ11 28.6Eþ11 5.20Eþ11 – –
0.050 – – – – 0.024 0.024 0.050 – –
a Parameters obtained through inverse optimization with AMALGAM. KCI,dce was arbitrarily set to the average of reported constants in this table. b Recalculated values according to Duhamel et al. (2004) assuming a conversion factor of 4.2E15 g dry weight of cell material per 16S rRNA gene copy and a protein content of 50%. c Values for the enriched culture were adopted from the non-enriched source culture. d The yield constant for cis-DCE degradation was adopted from the reported yield constant on VC degradation. e One yield constant was reported for the complete degradation of PCE to VC and ethene.
338
water research 44 (2010) 331–339
quantification protocol in this experiment was compared to another quantification technique, i.e. Catalyzed Reporter Deposition-Fluorescent In Situ Hybridization (CARD-FISH) in a previous study (Dijk et al., 2008). The quantification of Dehalococcoides was found to differ by less than a factor 4 between both methods. As such, the obtained yield estimates for Dehalococcoides in the current study were assumed to be sufficiently accurate. The yield values in this study were at the lower limit of the values reported in literature (Table 2 and 3). However, most yield coefficients reported in other studies were obtained from highly enriched lab-cultures under optimal conditions at non-inhibitive chlorinated ethene concentrations or had to be rescaled to represent cellular growth. A standardized microbial quantification would be beneficial for a comparison of yield coefficients between different studies (Cupples, 2008).
5.
Conclusion
Batch results showed a self-inhibition of TCE at concentrations above 1 mM and a complete inhibition at 4 mM and more. This strong inhibition could limit the potential benefits of bioremediation in a TCE source zone. Microbial data indicated that cis-dichloroethene (cis-DCE) was dechlorinated by Dehalococcoides spp. while another organism dechlorinated TCE in the KB-1 culture. The EC50 model rather than the Haldane inhibition model most accurately simulated the observations. Although the parameter determination of and modeling with species specific Monod kinetics is a demanding process, its use allows a better prediction of reactions taking place in a CAH source zone. First order or Michaelis-Menten models cannot incorporate observed lag-times and inhibitive effects as found and fitted in this study.
Acknowledgements We thank Dr. R. Richardson of Cornell University (USA), the SiREM company, M. Duhamel of the University of Toronto (Canada) and H. Smidt and M. Sturme of Wageningen University (The Netherlands) for providing their cultures and clones. We also thank J. Dijk, J. Mertens, J.A. Vrugt and S. Ruymen for their advice and kind assistance in the experimental work and in the model development, and the 2 anonymous reviewers for their helpful comments. This research was funded by a Ph.D. grant of the Institute for the Promotion of Innovation through Science and Technology in Flanders (IWT-Vlaanderen).
Appendix. Supplementary information Supplementary information associated with this article can be found, in the online version, at doi:10.1016/j.watres.2009. 09.033
references
Amos, B.K., Christ, J.A., Abriola, L.M., Pennell, K.D., Loffler, F.E., 2007. Experimental evaluation and mathematical modeling of microbially enhanced tetrachloroethene (PCE) dissolution. Environmental Science & Technology 41 (3), 963–970. Aulenta, F., Gossett, J.M., Papini, M.P., Rossetti, S., Majone, M., 2005. Comparative study of methanol, butyrate, and hydrogen as electron donors for long-term dechlorination of tetrachloroethene in mixed anaerobic cultures. Biotechnology and Bioengineering 91 (6), 743–753. Azizian, M.F., Behrens, S., Sabalowsky, A., Dolan, M.E., Spormann, A.M., Semprini, L., 2008. Continuous-flow column study of reductive dehalogenation of PCE upon bioaugmentation with the Evanite enrichment culture. Journal of Contaminant Hydrology 100 (1–2), 11–21. Christ, J.A., Ramsburg, C.A., Abriola, L.M., Pennell, K.D., Loffler, F.E., 2005. Coupling aggressive mass removal with microbial reductive dechlorination for remediation of DNAPL source zones: a review and assessment. Environmental Health Perspectives 113 (4), 465–477. Cupples, A.M., Spormann, A.M., McCarty, P.L., 2003. Growth of a Dehalococcoides-like microorganism on vinyl chloride and cisdichloroethene as electron acceptors as determined by competitive PCR. Applied and Environmental Microbiology 69 (2), 953–959. Cupples, A.M., Spormann, A.M., McCarty, P.L., 2004. Comparative evaluation of chloroethene dechlorination to ethene by Dehalococcoides-like microorganisms. Environmental Science & Technology 38 (18), 4768–4774. Cupples, A.M., 2008. Real-time PCR quantification of Dehalococcoides populations: methods and applications. Journal of Microbiological Methods 72 (1), 1–11. Dijk, J.A., Breugelmans, P., Philips, J., Haest, P.J., Smolders, E., Springael, D., 2008. Catalyzed reporter deposition-fluorescent in situ hybridization (card-fish) detection of Dehalococcoides. Journal of Microbiological Methods 73 (2), 142–147. Doelman, P., Haanstra, L., 1989. Short-term and long-term effects of heavy-metals on phosphatase-activity in soils – an ecological dose-response model approach. Biology and Fertility of Soils 8 (3), 235–241. Duhamel, M., Wehr, S.D., Yu, L., Rizvi, H., Seepersad, D., Dworatzek, S., Cox, E.E., Edwards, E.A., 2002. Comparison of anaerobic dechlorinating enrichment cultures maintained on tetrachloroethene, trichloroethene, cis-dichloroethene and vinyl chloride. Water Research 36 (17), 4193–4202. Duhamel, M., Mo, K., Edwards, E.A., 2004. Characterization of a highly enriched Dehalococcoides-containing culture that grows on vinyl chloride and trichloroethene. Applied and Environmental Microbiology 70, 5538–5545. Duhamel, M., Edwards, E.A., 2006. Microbial composition of chlorinated ethene-degrading cultures dominated by Dehalococcoides. FEMS Microbiol Ecology 58 (3), 538–549. Duhamel, M., Edwards, E.A., 2007. Growth and yields of dechlorinators, acetogens, and methanogens during reductive dechlorination of chlorinated ethenes and dihaloelimination of 1,2-dichloroethane. Environmental Science & Technology 41 (7), 2303–2310. Fennell, D.E., Gossett, J.M., Zinder, S.H., 1997. Comparison of butyric kid, ethanol, lactic acid, and propionic acid as hydrogen donors for the reductive dechlorination of tetrachloroethene. Environmental Science & Technology 31 (3), 918–926. Fennell, D.E., Gossett, J.M., 1998. Modeling the production of and competition for hydrogen in a dechlorinating culture. Environmental Science & Technology 32 (16), 2450–2460.
water research 44 (2010) 331–339
Garant, H., Lynd, L., 1998. Applicability of competitive and noncompetitive kinetics to the reductive dechlorination of chlorinated ethenes. Biotechnology and Bioengineering 57 (6), 751–755. Gossett, J.M., 1987. Measurement of Henrys law constants for C1 and C2 chlorinated hydrocarbons. Environmental Science & Technology 21 (2), 202–208. Haest, P.J., Ruymen, S., Springael, D., Smolders, E., 2006. Reductive dechlorination at high aqueous TCE concentrations. Communications in Applied and Biological Sciences 71 (1), 165–169. Haston, Z.C., McCarty, P.L., 1999. Chlorinated ethene half-velocity coefficients (K–S) for reductive dehalogenation. Environmental Science & Technology 33 (2), 223–226. He, J., Ritalahti, K.M., Aiello, M.R., Loffler, F.E., 2003. Complete detoxification of vinyl chloride by an anaerobic enrichment culture and identification of the reductively dechlorinating population as a Dehalococcoides species. Applied and Environmental Microbiology 69 (2), 996–1003. He, J., Sung, Y., Krajmalnik-Brown, R., Ritalahti, K.M., Loffler, F.E., 2005. Isolation and characterization of Dehalococcoides sp. strain FL2, a trichloroethene (TCE)- and 1,2-dichloroethenerespiring anaerobe. Environmental Microbiology 7 (9), 1442–1450. Klappenbach, J.A., Saxman, P.R., Cole, J.R., Schmidt, T.M., 2001. rrndb: The ribosomal RNA operon copy number database. Nucleic Acids Research 29 (1), 181–184. Liu, C.X., Zachara, J.M., 2001. Uncertainties of Monod kinetic parameters nonlinearly estimated from batch experiments. Environmental Science & Technology 35 (1), 133–141. Madsen, H., 2003. Parameter estimation in distributed hydrological catchment modelling using automatic calibration with multiple objectives. Advances in Water Resources 26 (2), 205–216. Maymo-Gatell, X., Tandol, V., Gossett, J.M., Zinder, S.H., 1995. Characterization of an H2-utilizing enrichment culture that reductively dechlorinates tetrachloroethene to vinyl chloride and ethene in the absence of methanogenesis and acetogenesis. Applied and Environmental Microbiology 61 (11), 3928–3933. Maymo-Gatell, X., Chien, Y., Gossett, J.M., Zinder, S.H., 1997. Isolation of a bacterium that reductively dechlorinates tetrachloroethene to ethene. Science 276 (5318), 1568–1571. Robinson, J.A., Tiedje, J.M., 1983. Non-linear estimation of Monod growth kinetic-parameters from a single substrate depletion
339
curve. Applied and Environmental Microbiology 45 (5), 1453– 1458. Schaefer, C.E., Condee, C.W., Vainberg, S., Steffan, R.J., 2009. Bioaugmentation for chlorinated ethenes using Dehalococcoides sp.: Comparison between batch and column experiments. Chemosphere 57 (2), 141–148. Smits, T.H.M., Devenoges, C., Szynalski, K., Maillard, J., Holliger, C., 2004. Development of a real-time PCR method for quantification of the three genera dehalobacter, dehalococcoides, and desulfitobacterium in microbial communities. Journal of Microbiological Methods 57 (3), 369–378. Sung, Y., Ritalahti, K.M., Apkarian, R.P., Loffler, F.E., 2006. Quantitative PCR confirms purity of strain GT, a novel trichloroethene-to-ethene-respiring Dehalococcoides isolate. Applied and Environmental Microbiology 72 (3), 1980–1987. Tandoi, V., Distefano, T.D., Bowser, P.A., Gossett, J.M., Zinder, S. H., 1994. Reductive dehalogenation of chlorinated ethenes and halogenated ethanes by a high-rate anaerobic enrichment culture. Environmental Science & Technology 28 (5), 973–979. Uyttebroek, M., Breugelmans, P., Janssen, M., Wattiau, P., Joffe, B., Karlson, U., Ortega-Calvo, J.J., Bastiaens, L., Ryngaert, A., Hausner, M., Springael, D., 2006. Distribution of the Mycobacterium community and polycyclic aromatic hydrocarbons (PAHs) among different size fractions of a longterm PAH-contaminated soil. Environmental Microbiology 8 (5), 836–847. Vrugt, J.A., Robinson, B.A., 2007. Improved evolutionary optimization from genetically adaptive multimethod search. Proceedings of the National Academy of Sciences of the United States of America 104 (3), 708–711. Yang, Y.R., McCarty, P.L., 1998. Competition for hydrogen within a chlorinated solvent dehalogenating anaerobic mixed culture. Environmental Science & Technology 32, 3591–3597. Yang, Y.R., McCarty, P.L., 2000. Biologically enhanced dissolution of tetrachloroethene DNAPL. Environmental Science & Technology 34 (14), 2979–2984. Yu, S., Semprini, L., 2004. Kinetics and modeling of reductive dechlorination at high PCE and TCE concentrations. Biotechnology and Bioengineering 88 (4), 451–464. Yu, S., Dolan, M.E., Semprini, L., 2005. Kinetics and inhibition of reductive dechlorination of chlorinated ethylenes by two different mixed cultures. Environmental Science & Technology 39 (1), 195–205.
water research 44 (2010) 340–350
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Colloidal organic matter from wastewater treatment plant effluents: Characterization and role in metal distribution Isabelle A.M. Worms a, Zsofia Al-Gorani Szigeti a, Stephane Dubascoux b, Gaetane Lespes b, Jacqueline Traber c, Laura Sigg c, Vera I. Slaveykova a,* a
Environmental Biophysical Chemistry, GR-SLV-IIE-ENAC, Ecole Polytechnique Fe´de´rale de Lausanne (EPFL), Station 2, 1015 Lausanne, Switzerland b LCABIE CNRS UMR 5254 IPREM, Universite´ de Pau and des Pays de l’Adour, Helioparc, Avenue du Pre´sident Pierre Angot, 64053 Pau, France c Eawag, Swiss Federal Institute of Aquatic Science and Technology, Ueberlandstrasse 133, PO Box 611, CH-8600 Duebendorf, Switzerland
article info
abstract
Article history:
Colloidal organic matter from wastewater treatment plants was characterized and exam-
Received 25 March 2009
ined with respect to its role in metal distribution by using tangential flow ultrafiltration,
Received in revised form
liquid chromatography coupled with organic carbon and UV detectors, and an asymmet-
10 September 2009
rical flow field-flow fractionation (AFlFFF) multidetection platform. Results revealed that
Accepted 15 September 2009
a humic-like fraction of low aromaticity with an average molar mass ranging from 1600 to
Published online 30 September 2009
2600 Da was the main colloidal component. High molar mass fractions (HMM), with molar mass ranges between 20 and 200 kDa, were present in lower proportions. Ag, Cd, Cu, Cr, Mn
Keywords:
and Zn were found mainly in the dissolved phase (<0.45 mm) and their distribution
Colloids
between colloidal and truly dissolved fractions was strongly influenced by the distribution
Metal binding
of dissolved organic carbon. AFlFFF coupled to ICP-MS showed that Ag, Cd, Cu, Cr, Mn and
Wastewater effluent
Zn associate to the low molar mass fraction of the colloidal pool, whereas Al, Fe and Pb
Asymmetrical flow field-flow
were equally bound to low and high molar mass fractions. ª 2009 Elsevier Ltd. All rights reserved.
fractionation LC-OCD Tangential flow ultrafiltration
1.
Introduction
Colloids are macromolecules or assemblages, with sizes operationally defined as ranging from 1 nm to 1 mm (Lead and Wilkinson, 2006). The colloidal pool in natural systems consists of a heterogeneous mixture of polydisperse inorganic and organic components. Major inorganic colloids are often dominated by iron oxides, whereas the colloidal organic matter (COM) is mainly composed of humic substances (HS) and extracellular polymeric substances. Colloidal stability is controlled by coagulation and flocculation processes,
depending on the size, the shape and the relative concentration of colloidal components (Buffle et al., 1998).Within the dissolved phase, colloids can be transported through natural waters and for long distances, as is the case for the truly dissolved species (Gustafsson and Gschwend, 1997). COM is highly reactive regarding metal complexation due to its polyfunctional character and large binding site density. Consequently, COM is recognized to control metal distribution, speciation and bioavailability in freshwater systems (Warren and Haack, 2001; Town and Filella, 2002) and has been shown to be important in processes such as metal mobilization and
* Corresponding author. Tel.: þ41 21 693 63 31; fax: þ41 21 693 80 70. E-mail address: [email protected] (V.I. Slaveykova). 0043-1354/$ – see front matter ª 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2009.09.037
water research 44 (2010) 340–350
Nomenclature COM Cdiss Cisol Ctd Ccol EfCOM EfOM HMM
colloidal organic matter organic carbon concentration in the dissolved (0.45 mm filterable) fraction organic carbon concentration in the colloidal isolates organic carbon concentration in the truly dissolved (1 kDa filterable) fraction organic carbon concentration in the colloidal fraction (between 1 kDa and 0.45 mm) effluent colloidal organic matter effluent organic matter high molar mass fraction
dispersion (Weng et al., 2002; Schafer et al., 2003; Ren and Packman, 2004, 2005; Pedrot et al., 2008). COM behavior depends on its nature (Buffle et al., 1998). In the case of humic substances both aromaticity and metal binding capacity have been shown to depend on molar mass or size fractions (Cabaniss et al., 2000). Recent technological advances, with the coupling of size fractionation techniques such as sizeexclusion chromatography (SEC) or flow field-flow fractionation (FlFFF) to highly sensitive metal detectors such as inductively coupled plasma-mass spectrometry (ICP-MS), have confirmed the differences in the molar mass and size distribution of HS– metal complexes (Hassellov et al., 1999; Schmitt et al., 2001; Wu et al., 2004; Bolea et al., 2006; Dubascoux et al., 2008). In freshwater systems, evidence for the preferential binding of some metals with inorganic colloids of larger size, mainly composed of iron or manganese with a low carbon content, suggests that both HS and mineral oxides of colloidal size could be potential carriers for trace metal dispersion (Lyven et al., 2003; Stolpe et al., 2005; Baalousha et al., 2006; Stolpe and Hassellov, 2007). Although wastewater treatment plant (WWTP) effluents discharged in surface waters rise in volume continuously and are considered as one of major sources of organic matter in urbanized areas (Dignac et al., 2000; Imai et al., 2002; Shon et al., 2006a), little is known about the influence of wastewater treatment plant COM on the distribution and dispersion of metals. In contrast to the considerable work on the characterization of the composition of organic matter from effluents (EfOM), which is important for wastewater treatment optimization (Dignac et al., 2000; Shon et al., 2004, 2006b), its role in the distribution of metals in effluents, and in their potential transfer through the receiving ecosystem, is not well understood. The present study therefore aims to improve our understanding of the role of the colloidal organic matter from wastewater treatment plant effluents in trace metal transfer and distribution. More specifically, it has an emphasis on the characterization of the composition, chemical properties and molar mass distribution of the effluent colloidal organic matter (EfCOM) isolates and associated metals. Results obtained for three Swiss municipal WWTP effluents are compared with those found for Lake Geneva water or standard fulvic and humic acids. The influence of chemical properties and mass distributions of EfCOM on metal transport, and the implications for the receiving waters, are discussed.
LMM [M]diss [M]td [M]isol [M]col [M]p [M]raw Mp Mn Mw
341
low molar mass fraction total dissolved metal (0.45 mm filterable) concentration truly dissolved metal concentration (1 kDa filterable) metal concentration in the isolate metal concentration in the colloidal fraction (between 1 kDa and 0.45 mm) metal concentrations in the particulate phase metal concentration of the raw samples molar mass corresponding to the peak maximum number-average molar mass weight-average molar mass
2.
Materials and methods
2.1.
Sampling and isolation of colloidal components
The effluents of Duebendorf (DB), Hinwil (HW) and Zurich (ZH) wastewater treatment plants were sampled in February 2007 and January 2008. The treatment processes of each WWTP include gravitational sedimentation and activated sludge treatments, followed by chemical treatment with iron addition for phosphate removal. Samples were collected in 20 L plastic bottles that had previously been washed with 0.1 M suprapure HNO3 (Baker) and rinsed with MilliQ water to avoid metal contamination. Samples were transported to the laboratory and filtered through 0.45 mm pore size cellulose acetate membrane (Millipore). A tangential flow ultrafiltration system (TFF; Pellicon, Millipore), holding a regenerated cellulose membrane with 1 kDa nominal molar mass cut-off was used to concentrate the colloidal matter of each sample, in a sampling mode, in which the retentate flow was directed back to the feed bottle. The permeate flow was directed to a container to be collected. The colloidal isolate was obtained after 5–7 h of circulation, which corresponded to a volumic concentration factor (cf) varying between 12.6 and 20. The colloidal fraction from Lake Geneva water (Vidy Bay (VD) near Lausanne) was isolated in January 2008 using the same procedure as for WWTP effluents. Vidy Bay is known to be impacted by the main WWTP of Lausanne (Pote et al., 2008) especially near the outlet that is located at 127 m offshore at 30 m below the surface. Standard humic and fulvic acids from Suwannee River (SRFA and SRHA), used for comparative purposes, were obtained from the International Humic Substances Society (St. Paul, MN).
2.2.
Organic carbon and metals distributions
Organic carbon (OC) concentrations were measured, using a Shimadzu TOC-V analyzer after filtration through 0.45 mm pore size membranes (dissolved, Cdiss), in the colloidal isolate (Cisol) and in the TFF permeate (truly dissolved, Ctd). The colloidal organic carbon content in the collected water samples was estimated by Eq. (1): Ccol ¼ ðCisol Ctd Þ=cf
(1)
342
water research 44 (2010) 340–350
where Ccol is the OC concentration in the colloidal fraction and cf the volumetric concentration factor obtained at the end of the TFF procedure. Total dissolved metal concentrations ([M]diss), the metal concentration in the colloidal isolate ([M]isol) and in the metal concentrations in the TFF permeate (truly dissolved, [M]td) were determined by high resolution (HR) ICP-MS (Thermo). The metal concentrations in the colloidal fraction ([M]col) were estimated by Eq. (2): ½Mcol ¼ ½Misol ½Mtd cf
(2)
Metal concentrations in the particulate phase ([M]p) were determined using Eq. (3): ½Mp ¼ ½Mraw ½Mdiss
(3)
where [M]raw is the metal concentrations measured in the unfiltered collected water samples and [M]diss the concentrations after filtration through 0.45 mm pore size membranes.
2.3. Characterization of the colloidal isolates by fluorescence spectrometry Emission-excitation matrixes (EEM) were used to gain information on the quantity of colloidal isolates (e.g. TFF concentration procedure efficiency), as well as to characterize their fluorogenic constituents (e.g. HS-like or protein-like fluorophores) (Baker, 2001; Baker et al., 2003). Fluorescence spectra were recorded on an LS 55 luminescence spectrometer using a 500 mL, 1 cm path length quartz cuvette. EEMs were generated by recording emission spectra from 300 to 550 nm at 0.5 nm steps for an excitation wavelength between 200 and 400 nm. The fluorescence index (FI), which provides information about the extent of condensation of humic substances, and thus their origin, was derived from EEM fingerprints according to Eq. (4) (McKnight et al., 2001): FI ¼ ðIlem ¼500 nm =Ilem ¼450 nm Þ
(4)
C L1) was done after deconvolution of the OC signal using the FIFFIKUS software (DOC-Labor Dr Huber, Germany).
2.5. Fractionation and characterization of colloidal matter by asymmetrical flow field-flow fractionation – UV – multiangle light scattering detection (AFlFFF-UV-MALS) Fractionation and molar mass distribution characterization of colloidal matter were carried out with AF2000 Focus (Postnova Analytics, Landsberg, Germany) coupled on-line to a sevenangle laser light scattering (Postnova Analytics) and UV detection at 254 and 280 nm. System control and data collection were performed using the AFlFFF 2000 Control software (version 1.1.011 Postnova Analytics). Trapezoidal channels of 350 mm thickness, as well as a 1 kDa cut-off regenerated cellulose membrane (RC, Postnova Analytics) for the accumulation wall were used for all experiments. Colloidal concentrates were injected in the AFlFFF channel using a 1 mL sample loop at injection flow, Vinj ¼ 0.09 mL min1 for 13 min. During this step the sample was focused using a focus flow, Vfoc ¼ 4.00 mL min1, a cross flow, Vxf ¼ 3.09 mL min1 and a channel flow, Vout ¼ 1 mL min1. After 1 min transition time, elution steps of 18 min were recorded at constant cross-flow values, comprising 1 between 3 mL min1 and 0.1 mL min1 and Vout ¼ 1 mL min (Fig. 1). The runs ended with a 5 min elution at Vxf ¼ 0 mL min1. The carrier solution was 10 mM NaNO3 at pH 5.4. For separation of low molar mass fraction (LMM), a cross flow Vxf ¼ 3 mL min1 was applied and for fractionation of higher molar mass (HMM) fraction, Vxf ¼ 0.25 mL min1 was chosen.
2.6. Characterization of metals associated with colloidal matter by AFlFFF-UV-MALS coupled to ICP-MS The association of the metals to the COM was characterized using a second AFlFFF setup (Eclipse 3 AFlFFF system, Wyatt Technology, Dernbach, Germany) coupled with a quadrupole ICP-MS 7500ce model (Agilent Technology, Tokyo, Japan) by
for excitation wavelength lex ¼ 370 nm.
2.4. Characterization of organic matter composition by liquid chromatography – organic carbon and UV detection The composition of the EfCOM was characterized by liquid chromatography coupled on-line to UV, total organic carbon and nitrogen detectors (LC-OCD (DOC-Labor, Dr Huber, Germany)). WWTPs colloidal isolates were diluted five times before measurement and 2 mL injection volumes were used. Samples were eluted at 1 mL min1 using a 24 mM phosphate buffer at pH 6.6. The gel filtration column Toyopearl HW-50S (Tosoh Bioscience) had a hydrophilic steric separation range of 20–0.1 kDa followed by an amphiphilic elution of small molecules. The LC-OCD measurable organic carbon represents 80% of the total organic carbon for DB, ZH and VD and 50% for HW colloidal isolates. Based on their elution time, specific UV absorbance (SUVA ¼ absorbance at 254 nm/mg C L1) and C/N ratio, the EfCOM components were classified in biopolymers, humic-like substances, building blocks, small organic acids and neutral compounds as previously described (Huber and Frimmel, 1992, 1994). Quantification of each component (in mg
Fig. 1 – Cross-flow rate decrease effect on colloidal component fractionation followed by the 908 light scattered signal. Values of the cross flow rate (Vxf) applied are indicated for corresponding elution profiles. The end of the fractionation is always finished by an elution without cross flow (Vxf [ 0).
343
water research 44 (2010) 340–350
connecting the outlet of the MALS detector to the ICP-MS nebulization chamber using a peek tube. Channel dimensions were 26.5 cm length, and width ranging from 2.1 to 0.6 cm. For the LMM fractionation the channel thickness was defined by a spacer of 490 mm, and a Vxf ¼ 3 mL min1 and a polyestersulfone membrane with 1 kDa cut-off (PES, Wyatt Technology) were used. For the HMM fraction, a spacer of 350 mm and a regenerate cellulose membrane with 10 kDa cut-off (RC, Postnova Analytics) were used, and a gradient cross flow decreasing from 1 to 0 mL min1 was applied over 30 min. Cross-flow gradients were used in order to obtain a maximum mass range in a minimum time of analysis (Prestel et al., 2005). Flows were controlled by an isocratic pump equipped with a micro vacuum degasser and an UV detector at 280 nm (Agilent 1100 series controlled by Chemstation for LC systems software). The carrier used was composed of 10 mM NaNO3 for which pH ¼ 5.4 was measured. All colloidal isolates were analyzed with or without addition of 50 nM Cd, Cu and Pb, and allowed to equilibrate for 24 h before fractionation.
2.7. Determination of the molar mass distribution characteristics of colloidal components Molar mass distribution of EfCOM was determined following AFlFFF fractionation and calibration with polystyrenesulfonate standards (PSS). For this purpose, 1 mL of 5 mg L1 PSS standards with Mw values ranging from 1360 Da to 979 kDa (Postnova Analytics) was injected. The relationship between log P P weight-average retention time (trw ¼ (hitri)/ hi) and log weight-average molar mass (Reszat and Hendry, 2005) was used as an external calibration in order to convert the raw UV254 nm fractograms to molar mass profiles. After baseline correction of the UV signals, the molar mass distribution P P parameters: weight-average molar mass (Mw ¼ (hiMi)/ (hi)), P P number-average molar mass (Mn ¼ (hi)/ (hi/Mi)) and polydispersity index (PDI ¼ Mw/Mn) were derived. No further correction was made to account for the non-linear relationship between tr and Mw since this has only a minor influence on the molar mass distribution parameters. For high molar mass components (HMM), the UV signal was too low in intensity to be used together with light scattering for evaluation of the gyration radius and molar masses according to MALS theory. PSS and sulfate polystyrene latex standards (22 nm, 58 nm, 97 nm, Postnova Analytics) were used to determine molar mass and diameters, respectively, at the maximum intensity of the light scattered peak (Mp and dp) after conversion of the 90 angle signal to molar mass profiles. In that case, no distribution parameters were estimated since the scattered light depends on the shape, size and concentration of colloids.
2.8.
Table 1 – Organic carbon concentrations obtained in the different fractions before and at the end of the TFF for Duebendorf (DB), Hinwil (HW) and Zurich (ZH) WWTP effluents from two sampling dates (February 2007 and January 2008) and Vidy Bay lake water (VD) sampled in January 2008. Samples
DB, Feb 07 HW, Feb 07 ZH, Feb 07 DB, Jan 08 HW, Jan 08 ZH, Jan 08 VD, Jan 08
cf
14.3 12.6 14.3 20 16 14 16.6
OC concentrations (mg L1) Cdiss
Ctd
Cisol
Ccol
4.1 3.7 4.3 5.7 3.4 4.8 1.1
2.7 2.7 3.6 5.3 3.1 4.3 1.1
21.5 14.6 13.3 11.2 9.3 9.7 2.5
1.3 1.0 0.7 0.3 0.4 0.4 0.1
COC (%)
31 24 16 5 11 8 8
cf, volumic concentration factor; OC, organic carbon; Cdiss, dissolved organic carbon concentration (0.45 mm filterable); Ctd, organic carbon concentration in truly dissolved fraction (1 kDa membrane filterable); Cisol, organic carbon concentration in the colloidal isolate (1 kDa, 0.45 mm); Ccol, colloidal organic carbon (COC) concentration in the original samples further expressed as percentage of dissolved organic concentration.
concentrations obtained by LC-OCD. When metals are detected in both LMM and HMM components, colloidal metal partitioning between those two components was evaluated as percentages of the total colloidal signal intensity P P P ( (hi)COL ¼ (hi)LMM þ (hi)HMM).
3.
Results and discussion
3.1.
Characterization of EfOM composition
Dissolved organic carbon concentrations determined in EfOM were higher than those in the lake water from Vidy Bay and
Metal–colloidal organic matter distributions
The overall intensity of each individual metal peak, obtained by ICP-MS after AFlFFF separation, was calculated after baseline corrections in Origin 7 (OriginalLab), by summing the signal intensity (hi, in counts per second (CPS)) obtained for P each Mi (total CPS ¼ (hi)LMM for the LMM fractions and P P (hi)HMM for the HMM fractions). (hi)LMM was then compared to humic substance absorbance or concentration obtained by P LC-OCD and (hi)HMM was compared to biopolymer
Fig. 2 – Organic carbon concentrations distribution in the dissolved phase (DOC < 0.45 mm) between truly dissolved (Ctd < 1 kDa, white) and colloidal (Ccol, gray) fractions of Duebendorf (DB), Hinwil (HW) and Zurich (ZH) WWTP effluents sampled in February 2007 or in January 2008, and Vidy Bay lake water (VD) sampled in January 2008. Ccol was obtained from mass balance of the TFF.
344
water research 44 (2010) 340–350
Fig. 3 – Excitation-emission matrixes of colloidal isolates (ISOL) and truly dissolved (TD) fractions obtained at the end of the TFF procedure for Duebendorf (DB), Hinwil (HW) and Zurich (ZH) WWTP effluents, Vidy Bay lake water (VD) and 20 mg LL1 Suwannee River fulvic or humic acids (SRFA and SRHA). A and B, fulvic-like; C, humic-like; T, tryptophan-like fluorophores.
comparable to the 2007 and 2008 samples (Table 1 and Fig. 2). The colloidal organic carbon (COC) content in the effluents sampled in 2007 was slightly higher than that obtained in 2008. The EEM fingerprints obtained for the WWTP effluent isolates were more intense than those obtained from the ultrafiltered samples, indicating the efficiency of the concentration procedure for the fluorogenic COM ((Liu et al., 2007),
(Fig. 3). For all WWTP effluent colloidal isolates, EEM fingerprints showed the presence of fulvic-like (lex/em ¼ 340 nm/ 420 nm), humic-like (lex/em ¼ 380 nm/470 nm), and tryptophan-like (lex/em ¼ 280 nm/355 nm) fluorophores (Table 2). For DB and ZH WWTP samples, similar emission intensities were obtained for these fluorophores, whereas, for the HW sample, their intensities were low considering that the COC content in
Table 2 – Fluorescence properties and composition of Duebendorf, Hinwil, Zurich and Vidy Bay colloidal isolates (2008 sampling) characterized by excitation-emission matrixes (EEM) fluorescence and LC-OCD analysis. Sampling site
Fluorescence (EEM)
FI
SUVA
LC-OCD
Intensity
Duebendorf Hinwil Zurich Vidy Bay
Humic substances
B
C
T
25.5 15 25 2.5
16 9 15 1.2
12 8 14 3.2
1.9 2.0 2.2 1.9
2.1 1.6 1.8 1.7
Biopolymers
% OC
SUVA
C/N
% OC
C/N
36.8 27.7 35.8 38.3
2.85 2.18 1.76 2.38
11 14 14 16
8.1 17.6 14.3 13.7
8 8 6 18
B, fulvic-like (lex/em ¼ 340 nm/400–460 nm); C, humic-like (lex/em ¼ 380 nm/470 nm); T, tryptophan-like fluorophores (lex/em ¼ 280 nm/310– 410 nm); FI, fluorescence index; SUVA, specific UV absorbance (L mg1 m1); C/N, carbon to nitrogen ratio; % OC, humic substance or biopolymer contents expressed as a percentage of the organic carbon quantifiable by LC-OCD.
345
water research 44 (2010) 340–350
this isolate is equivalent to that obtained for DB. All isolates have fluorescence indexes larger than 1.8 (Table 2), as compared with those typically found for standard humic substances (FI values of 1.4 for SRFA and 1.2 for SRHA), indicating their potential microbial origin (McKnight et al., 2001). Furthermore, EEM fingerprints of the WWTP isolates exhibited a shift in their fulvic-like fluorophore peaks to higher wavelengths (SRHS: lex/em ¼ 310 nm/445 nm) (Fig. 3). LC-OCD measurements revealed that humic-like substances, biopolymers and small organic acids represent the major components of the EfOM. Humic-like substances represented between 4% and 22% of the DOC while biopolymers represented only 1% to 9% of the EfCOM. These results are consistent with literature data obtained by resin extraction, showing that humic-like substances represent between 3% and 28% and small organic acids from 32% to 84% of DOC in WWTP effluents (Imai et al., 2002; Pernet-Coudrier et al., 2008). Differences in light-absorbing properties (SUVA) and C to N ratio were observable for humic-like substances and biopolymers in the EfCOM from different sampling locations (Table 2). Humic-like components represented 36.8%, 27.7%, 35.8% and 38.3% of carbon measured in the colloidal isolates for DB, HW, ZH and VD samples, while the biopolymers represented about 8.1%, 17.6%, 14.3% and 13.7%. SUVA values decreased in the order DB > VD > ZH > HW for the whole isolate and DB > VD > HW > 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.
